This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
~(B
SO
=
exp[*[So(r} + SI(r)]]
exp[* So(r}]exp [* SI(r}]
IA)
you see that the external field produces a change obtained simply
by multiplying the original amplitude with the gauge phase faatop:
=
J ¢(x)
where 1/J,
= 0). to vanish. It must necessarily "vibrate a little". A rough estimate of the energy is given by and minimizing the resultin~ expression with respect to variation in produces the minimum <x> = = P ~ = Po ' will however cause us trouble. When we square the derivative, we get additional terms which are linear in aj.l<pa.' These terms will prevent the displaced action from being identical to the original action. All we can hope is therefore that they differ by a surface term, i.e. that the displaced Lagrangian is essentially equivalent to the original Lagrangian. The condition for f being a symmetry transformation is consequently axV a[Tf
Furthermore we may generally assume that the set of eigenfunctions {ljJn(x)} is complete i.e. that we may expand an arbitrary wavefunction as a superposition of eigenfunctions: 1/J(x)
=
~
anljJn (x)
Here the coefficient an is determined by the relation am
provided the set of eigenfunctions is normalized, i.e.
Worked exercise: 2.11.2 Let {1/Jnix)} be a complet: orthonormal set of eigenfunctions of the Hamil. tonlaJl operator H. Show that. Pro~lem:
(2.69)
~ 1/J (xd 1/J
n
n (x,) =0 (Xl- xz)
Consider an eigenfunction of H,i.e.a wavefunction 1/Jn(x) with the property:
This eigenfunction can immediately be extended to a solution of the Schrodinger equation:
I
71
In accordance with the Einstein-de Broglie rule this is interpreted as tl1e state of a particle with energy En. Observe that the probability distribution is time independent: l¢n(x,t)1
2
= jlj!n(x)i
2
and we therefore say that the wave function represents a
stationmy
state. Let lj!(x) be an arbitrary wave function. Then we can decompose it as a superposition of eigenfunctions: lj!(x) But the Schrodinger equation is linear so we can immediately extend this to the solution ¢(x,t) =
~
an1/Jn(x) • exp [-
~
Ent]
which reduces to 1/J(x) for t=O. Once we know a complete set of eigenfunctions we therefore
controll the dynamical evolution of Schr6din-
ger wave functions. This suggest that the Feynman-propagator it self can be expressed through a complete set of eigenfunctions:
Worked exercise: 2.11.:3 Problem: Let {lj! (x)} be a complete orthonormal set of eigenfunctions to the Hermitian operator ft . Show that the Feynman propagator can be expanded as: (2.70)
Let us make a final comment about the canonical quantization. In the preceding descussion the Schrodinger wavefunction has played a central role. However it is possible to avoid it. The physical quantities are then represented by Hermitian operators that do not necessarily operate on the space of Schrodinger wave functions. These operators cannot be chosen arbitrarily: .In the classical context the physical quantities are represented as functions A(q';p.) of the generalised coordinates and their conjugate momenta. In the quantum cditext these functions are rep raced by Hermitian operators in such a way that their Poisson brc:cket is rep raced by the commutator: (2.71)
{A;B}+-1-[~,~1 if!
Exercise: 2.11.4 Problem: Show that the commutator [ ~;~ ] ding to exercise 2.10.1.
= ~~ -
~ satisfies the rules correspon-
Especially the generalised coordinates qi and their canonical momenta p. must be replaced by Hermitian operators satisfying the socalled Heisenberg commlltation rures: (2.72)
Compare (2.60).
I\i ; I\j] 1\ 1 [q q = [1\ p.;p. = 0 1 J
ih6~
J
I
72
Exercise: 2.11 . .s . P:oblem: Show x~at if we replace the generalised coordinate q" by the multiplicatIon operator q = q • and the canonical momentum Pi by the differential operator
~.
= _ ih _a_.
I
then
and
aql.
satisfy the Heisenberg's commutation rules.
Exercise: 2.11.6 . Problem: Show that if ~l. and satisfy Heisenberg's commutation rules then their commutator with other operator§ satisfies rules corresponding to exercise 2.10.4.
£.
ILLUSTRATIVE EXAMPLE:
2.12
SUPERCONDUCTORS AND FLUX QUANTIZATION We finish this chapter by reviewing another experiment which shows how gauge potentials may produce unexpected quantum mechanical effects~
Consider a suitable piece of metal say aluminum. If we cool
it down it becomes superconducting and interesting things happen. (Feynman
[1964], de Gennes [1966]). Suppose we originally had an ex-
ternal magnetic field. This would penetrate into t!1e metal when the temperature is high, but it turns out that when the temperature falls down below a certain critical temperature Tc' then the external field is expelled. This is the famous Meissner effect.
Fig. 32
T>T
c
T
c
It turns out that there is produced currents in the outher layers of the metal and these currents prevent the magnetic field from penetrating into the metal. This is an interesting situation. We have a magnetic field
B which
stays entirely outside the -lump of metal, but there is a gauge potential
A too,
and it may very well penetrate into the metal!
73
I
We now change the experiment a little. We take a ring whose width is great compared to the penetration depth of the magneT>T
c
tic field.
i~e
place it in an ex-
ternal magnetic field
B
at room
temperature. Nhat happens? The magnetic field is spread throughout the whole space and especially it penetrates into the metal. Then we cool dm'Tn the ring, and when we pass below the critical temperature Tc' the r-ieissner effect occurs. The magnetic T
c
field is expelled due to surface currents in the superconducting ring. Hence some of the field lines pass outside the ring and some of them pass through the hole in the ring. But now we remove the external field. However, this does not mean that the magnetic field disappears completely. Part of the field passing through the hole is trapped by the surface
currents. So the superconducting ring now acts much like a solenoid!
B , only B which pass
Observe, that inside the metal there is no magnetic field a gauge potential A.Since we have trapped a magnetic field through the ring, there is a magnetic flux
~
through the ring. It is
this flux we are going to examine! Let us try to get a qualitative understanding of the situation. According to the BCS-theory, the electrons in the metal will form pairs in the superconducting state. They are called Cooper pairs. Now the electrons are fermions, but the Cooper pairs act like bosons! This has the important consequence that the Pauli principle no longer applies. Two electrons cannot
occ~py
the same state, but
74
I
there is nothing to prevent two Cooper pairs from occupying the same state! Actually hosons have a strong tendency to occupy the same state. Suppose ~(r,t) denotes the Schr5dinger wave fUnction for a Cooper pair. We may actually assume that there is a macroscopic number of Cooper pairs all described by the same wave function
~(r,t).
This has important consequences. The absolute square I~(;,t) 12
is
the probability for finding a specific Cooper pair at the point r . If N
is the total number of Cooper pairs, then I~(r,t) 12
N
is simply the dEnsity q
=
-+
of Cooper pairs at the pOint r. If furthermore
2e is the charge of a Cooper pair, we conclude that:
is the cha:t'ge dEnsity of the Cooper pairs. Since a solution to the Schrodinger
~tion
is only determined up to a constant we may redefine it:
~(r,t)
-+
~ ~(r,t)
=
~'(r,t)
So rcw we have a macroscopic number of Cooper pairs described by a
single wave function:
~' (r,t) and
-I ~
l
(r, t)
12
simply denotes the charge density at r . Observe
that although I~I
2
in general only has a statistical meaning,
I~'
12
will in this case denote a macroscopic physical quantity! In what follows we will drop the prime and simply denote the wave function by ~(r,t). We may decompose it in the following way: (2.73)
Here p (r,t) is the charge density, which has a direct physical meaning, and
-+ ~(r,t)
•
is the phase which has no physical meaning as we
can change it by performing a gauge transformation! In the superconducting state the number of Cooper pairs are conserved and so is the charge. If we denote the Cooper current by we therefore conclude that p and (1.17)
J
3
obey an equation of continuity:
75
I
We can use this to find an expression for the Cooper current. p
= -
ljilji
an __ a,,, at - at lji
implies that
-
aij;
at
lji
Using the Schrodinger equation we can rearrange this as:
which may be rewritten as
ap
at
in
~
~
(v +
= 2m v {lji
~ ~
n
-
A) lji -
-lji (v~
-
!g
n
~
A) lji }
From this expression we immediately read off the Cooper current: (2.74)
This show us that the Cooper current is a gauge scalar as it ought to be, and that it is a real quantity. In the following we will always assume that the density p of Cooper pairs is constant thoughout our piece of metal! This is not strictly correct at the boundary where it falls rapidly to zero. Except for that,it is a very reasonable assumption that the Cooper pairs do not "crowd" together but are evenly spaced.
Exercise: 2.12.1 Problem: Show that the expression for the Cooper current can be rearranged as: ~pn ~ qA] (2.75) [ v \1>- 11
J=m
provided pis a constant.
Using exercise 2.12.1 we can now derive the Meissner effect. Since there is no external electric field involved, the Maxwell equations (1.3) and (1.6) reduce to (2.76)
v. B
~
-t
J =~
= 0
But from (2.75) we then get Vx (VxB)
=~ EoC
Vx.... J
= - Eomc ~
VxA
where we have used (1.12) to get rid of the term
= - ~2 B Eomc Vx(V~).
Furthermore using (1.14) and (1.3) the left hand side can be rearranged as
I
76
Hence we end up with the simple equation (the London equation) (2.77)
ai3
=
1
I"f
it
where (2.78) is the socalled London-length. That equation (2.77) indeed implies the Meissner effect is left as an exercise, see below. Exercise: 2.12.2 Problem: Consider a semi-infinite superconductor occupying the half-space: z>O. Let us apply a constant field = Eo (1,0,0) parallel to the surface. Show that inside the superconductor the solution to (2.78) is given by: ... [ z B (z) = Bo (exp - ~1,0,0) z > 0
B
so that the magnetic field vanishes exponentially for z > AL.(A is typically a L few hundred Angstrom.)
Observe that (2.76) implies that not only
B vanishes
inside the
superconductor, the same thing holds for the Cooper current
3-
Consider now a superconducting ring as shown on fig. 34. Inside the ring at the curve r the Cooper current vanishes, as does the magnetic field!
Us~
(2.75) we
therefore get:
o
=
~ m
(Vc.p-
9. Ii.
lA)
But now we can compute the magnetic flux inside the ring beccluse: (2.79)
pit. r
dr
Ii..! q!ll
r
V,n' '"
dr
You might be tempted to say that it is zero! But let us look a little closer at the phase lP. The wave function ljJ(r,t)
=
jl p(r,t) I exp[
ilP(r,t) 1
is only non-trivial in the ring, and all we can demand is
that it
is single valued. But then nothing can prevent lP from making a jump of 2nn. If that is the case, VlP will contain a 6 -like singularity
77
and the integral need not vanish!
I
(Compare the discussion in sec. 1.4) •
TO compute the line integral we assume that V~ makes the jump at the pOint B. If B+ and B- are pOints extremely close to B on each side of B we get:
q
P A· d r
B+
h 0<
[~(B+) - tp(B - ) 1
-+
V~
L
q
dr
B h
q
21Tn
+ as B ,B
+
B
consequently we get
(2.80)
where
n<1>o
So the trapped flux is
quantized~
<1>0
21Th
Iqf
This remarkable effect has been
established experimentally (Deaver and Fairbanks [l96l1)~)The quantity
m
1
provided the Cooper density p is constant. (b) Use this expression and exercise 2.8.3to prove directly that a vanishing Cooper current implies a vanishing magnetic field.
In the previous discussion we have assumed that the superconducting state of a metal only depends on the temperature. Actually it also depends on the strength of an external magnetic field. Let us discuss this in some detail for a special kind of superconductor known as a type II superconductor. Consider a superconduction cylinder. Outside the cylinder we have a solenoid. Suppose a weak current flows in the coil. Then it will produce a magnetic field of strength B which is expelled from the cylinder due to the Meissner effect. When we increase the current, B will reach a critical value BCl where the superconducting state starts breaking
U
~t
~l:--~ ---------------- Fig. 35
*) "Experimental evidence for quantized flux in superconducting cylinders", Phys. Rev.
Lett. 7 (1961) 43
I
78
down.
Thin vortices are formed where the normal state of the
metal is reestablished. The magnetic field starts penetrating into the metal through these vortices. As the magnetic field strength increases more and more vortices are formed and when we approach another critical field strength BC2 only small superconducting regions are still destributed throughout the cylinder. When we finally pass BC2 the superconducting state breaks down completely and the cylinder is now back in its normal state. If we decrease the current again then the same things happen in the reversed order. Now let us concentrate on a single vortex. Let us inclose it by a great circle
r as shown on the fi-
gure. Far away from the vortex the magnetic field and
thus
the Coo-
per current vanishes. The magnetic flux through the vortex is given by:
1111
Fig. 36
due to the same arguments as for the superconducting ring. But then we see that the magnetic flux q~tize~
~
through the vortex is necessarily
The existence of quantized vortices was predicted by Abri-
kosov (Abrikosov [1956 tland they are therefore ~ye
call~d
Abrikosov vortices.
have previously stated that charged particles may interact
with the gauge potential A in a spacetime region Q, even if the field )J strength F)JV vanishes identically throughout this region. This interaction was a pure quantum mechanical effect, the Bohm-Aharonov effect (sec. 2.6). We may throw light on this using our results concerning the flux quantization: Consider two identical rings: A and B. In the following experiment the two rings are placed at room temperature. Inside ring B we also place a tiny solenoid. In
L~is
solenoid we have a current
which produces exactly one half of a flux quantum. Hence in the beginning of the experiment the flux through A is 0, while the flux
through B is
%~o.
*) Soviet Phys. Jetp 5 (1957) 1174
79
I
A
Fig.37
Now we cool down the two rings and they become superconducting. !ihat happens? In the first ring nothing happens. But in the second ring the preceeding analysis concerning flux quantization is clearly valid. The superconducting ring will only allow a quantized flux through the ring.
Thus
a Cooper current is produced in the outer
layers and this contributes to the flux, so that the total flux becomes 0 or
~o
•
But what is the origin of this Cooper current? If the Cooper pair only interacted with
L~e
field strength , this would be a mystery
because there is only a nonvanishing field strength inside the solenoid. But as we have seen, there is a non-trivial gauge potential outside the solenoid and this penetrates into the ring. This suggests that it is the interaction between the gauge potential and the Cooper pairs
that is responsible for the Cooper current.
This is also in accordance with the schrodinger equation: 2 (2.65) ih( d~ + ~)~(r,t)= A)2o/(r,t)
:m (V-ir
ir
which shows us that the Schrodinger wave function interacts directly with the gauge potential.
SOLUTIONS OF WORKED EXERCISES: No. 2.4.1 Now let x = Xcl(t) be the cla~sical path. Then we can write an arbitrary path in the follow1ng form: X=XC1(t) + yet)
where
y(tll = y(tz) = 0
Let us expand L using Taylor's theorem: •• • dL dL· L(xcl +y;x cl +y;t) =L(xcixcht ) + dX ·y+ax y+
1. '2 [
d1:. •. d1:. • d1. ·21 dxdX y ... 2 dXdX YY + m~ J
This expansion is exact because L is qUadratic! Then we can rewrite the action:
I
80
s
t 2 • L(x l'x l,t)dt tl c C
=f
~2aL aL· (-- y+ - r y)dt tl ax ax
=J
where the term in the miMle vanishes t 2 dL
f (3"
tl
x
dL •
y +
ar y)dt x
t 2 =
f (
tl
aL ax -
d aL dt ax )y dt
o
because x satisfies Euler's equation! cl The last term can easily be computed explicitly:
substituting this into the path integral we get:
D[y(t)J
But Xl and X2 are not at all involved in this last path integral, so it can depend only on tl and t2 ! Consequently we have shown:
If furthermore the coefficients we can easily show that
Observe that:
a(t), b(t) and c(t) do not depend on time, then
t2
t2+At t2+At [ay2 (t )+by 2 (t )+cy:y.Jdt = f [ay2 (t -lIt)+ ... Jdt = f lay ,2 (t)+ ... Jdt t I t I +At t I + At
f
where we have introduced
y'(t) = y(t - At) . Thus
from which the desired result follows immediately when we put: At
-tl·
n
I
81 No. 2.8.2
A (X B+ f OX B)
A (x) +
B
2 £ oA
2
~ox + •••
o
axP
c
A
B
In the end of the calculation we are going to let £ + 0 hence we need only to compute the lowest order terms. In this approximation we get:
Interchanging the dummy indices a and
~
a
AadX
B in
the last term we finally get:
B a
= £2 [oaAB- "BAal Ax 6x
Taking the limit we obtain the exact result:
No. 2.11.1
(a)
Let E be an eigenvalue and let tion:
~
be the corresponding normalised eigenfun-
-E = -E~I~> = <~I~ = < H~I~> " Using the hermiticity of
Q this
is rearranged as
= ~~~> = ~IE~> = ~1~>E = E Thus
E = E and therefore E is real
82
(b)
Let E"Ez be different eigenvalues and let eigenfunctions. Then we get:
I
~"
~z
be the corresponding
No. 2.11.2
Let ~(x) be an arbitrary wavefunction. We can decompose it on the complete set of eigenfunctions:
Using this decomposition we get:
But this clearly shows that:
No. 2.11.3
For fixed (x, ,t I) we know that the Feynman propagator is a solution to the Schrodinger equation. Hence we may decompose it as i
K'X2; t 40
n
~.n (x,)exp[- hi
E t,J n
This forces us to put: an(xl;t,J = an(x,)exp[
~
Ent,l
since the left hand side is independent of t,. The above formula then reduces to:
but a comparison with exercise 2.11.2 then gives us:
and we are through.
n
83
I
ehapter 3 DYNAMICS OF CLASSICAL FIELDS 3,1
ILLUSTRATIVE EXAMPLE: LAGRANGIAN FORMALISM FOR A STRING
We have already discussed the Lagrangian formulation of the dynamics of a system with a finite number of degrees of freedom, say a finite number of particles moving in an external field. Now we want to include the dynamics of ~ields. We will start by considering a one-dimensional string. It has an important property: If you disturb the string at one place, then this disturbance may propagate along the string. We can understand this in an intuitive way. The string consists of "atoms". Each atom interacts with its nearest neighbours. Hence, if you disturb one atom, this disturbance has influence on the neighbour. But this disturbance of the neighbour has influence on the neighbour of the neighbour, etc! In this way a travelling wave is created which propagates along the string!
x
a
- axis
....-"-
• • • • • • • • • • • • • • • • •~ • -4a -3a -2a -a
Fig. 38
0
a 2a 3a 4a
xn
)-
na
In our model the string is composed of "atoms", which in the equilibrium state are evenly spaced throughout the x-axis. The important assumption is that the "atoms" are coupled to each other through forces proportional to their relative displacements (Hooke-forces). We will enumerate the "atoms" with an integer n so that the equilibrium position of the n'th "atom" is Xu = an. If we set the "atoms" in motion, then the n'th "atom" will be displaced an amount q from its equilibrium position. When the string is put into vibration~, its dynamical evolution is described by the functions: n = ••• , -2,-1,0,1,2 ... qn = qn(t) The kinetic energy of the
n'th "atom" is:
1
.2
2 mqn and the potential energy associated with the separation of atom 1
2
nand
2
[qn+l - qn 1 Hence the total Lagrangian for the system is:
2 mv
(3.1)
....
L =
L
n=-
Let us determine the equations of motion. Using (2.6) we get 2 (3.2) mqi = mv (qi+l - 2qi + qi-l)
n + 1
is:
84
I
It is easy to check that the equations of motion actually allow wave solutions. we put:
If
(3.3) then this will represent a travelling wave. motion, we get:
Inserting this into the equation of 2
mv Cos (kx -wt) (2Cos(k· a) -1) n or
(3.4) Thus (3.3) is a solution to the equation of motion provided (3.4) is satisfied. Relation (3.4) is called a dispersion relation. Observe that for small k, i.e. in the long wave length limit we may expand the cosine, getting: (3.5)
which is a linear dispersion relation. Now we want to investigate a continuous string, where there is "atoms" everywhere. We can do this by letting a ~ 0 in our discrete model. We say that we pass to the continuum limit. Now, instead of describing the displacements by the infinite set of numbers: qj(t), we will represent it by a smooth function q(x,t) giving the displacement of the "atom" with equilibrium position x .
Fig. 39 In the discrete model the mass density is: limit, we suppose that it approaches a constant
m When we pass to the continuum ~ , the mass density of the conti-
nuous string, i.e. m " .... p
for
a
~
0
Finally it will be necessary to make an assumption about v. In the disctete model the velocity of a wave in the long wave length limit is (compare (3.3) and (3.5» w k = va We assume that it approaches a constant c, the velocity of a travelling wave in the continuous string:
va ~ c for a ~ 0 With these preliminaries we can investigate what happens to the Lagrangian in the continuum limit. Let us take a look at the kinetic energy:
x= +1:0
! Xn=-OO
'! [q("n,t)]2axn .... J
x= -ao We can treat the potential energy in a similar way.
From the observation:
85
I
we get: Xn=+co
2
!mv [q n=-oo
l:
1 (t) - q (t)]2
n+
n
x=-c:o n x=+ao
Xn=+CO
l:
xn=-otI
l!! (va)2 [oq (xn,t)] 2 lIx a
ax
n
-+
Jl x=-c:o
p
c
2 [~ ax (X,t) ]2 dx
Thus we get for the total Lagrangian:
(3.6) showing that in the continuum limit the Lagrangian is expressed as an integraZ over The integrand is called the Lagrangian density.
space.
-
(3.7)
~ but also
Observe that it contains not only vative
~x come from? o
It came from the term
~ pc 2(~)2 ax
~~
Where did the space deri-
( qn+1-q)2 n
in the discrete model.
Hence, it reflects the property of ZocaZ interactions. Each point in space interacts with its nearest neighbours. In a similar way we may analyse the equations of motion. In the discrete model we have: We may rearrange the term on the right side:
= [q(x n+l't:}
qn+1 - 2qn + qn-1
- q(x n , t)] - [q(x n , t) - q(x n, t)]
a[~xq(xn+ ~,t) - 19. (x - g t)] o
ax
""
2'
n
Inserting this we get: (3.8)
-q(x ,t) n
~
2
a q (x ,t) v 2 a 2 --ax2
i. e.
-+
n
.1....Gt=is 2 2 c
at
ax2
As before, we may look for a solution representing a travelling wave:
(3.9)
= A Cos(kx-wt)
q(x,t)
If we insert this into the equation of motion (3.8), we get: -w 2A Cos(kx -wt) = -c 2k 2A Cos(kx-wt) i.e.
(3.10)
w
= ±
ck
Hence, (3.9) is a solution to the equation of motion provided ear dispersion relation (3.10).
w satisfies the lin-
I
86
3.2
LAGRANGIAN FORMALISM FOR RELATIVISTIC FIELDS You should now be motivated for the abstract field theory.
start with a field
~(t,~)
defined throughout spacetime.
of the field at a particular point stretching
q(t,~o)
We
The value
~o ' ~(t,~o)' corresponds to the The dynamics of the L, which we in analogy with the
in the preceding example.
field is governed by a Lagrangian preceding example write as: (3.11)
The Lagrangian density
depends not only on the time derivative
L
but on the space derivatives, as well:
L
= L(~,a\l~)
The presence of space derivatives
ai~
reflects the principle of local
interactions. If we choose two times
tl
and
t2 ,we may specify the field at these times.
Any smooth function
~(t,~)
which satisfies the boundary conditions:
~(tl'~) = ~l(~) and ~(t2,i) = ~2(i) represents a possible history of the field.
To each such history we asso-
ciate the action: t2 S
i.e. (3.12)
S
J Ldt tl
JL
t2 3 J JLd 1Cdt tl
(
d"x
n n
where
is the four-dimensional region between the hyperplanes: t=tl
As usual, we want to determine a history ~(t,~) which 2 extremizes the action. This, of course, leads to the equation of moand
t=t
tion for the field. action.
Now suppose that
~(t,~) where
~o(t,~)
really extremizes the
Consider another history:
n(t,i}
= ~ o (t,~)
+
£n(t,x}
satisfies the boundary conditions:
n(~,i)
I
87
Then the action:
I
SeE)
n has an extremal value, when
L
4
(o+£n, dll
£=0 . Consequently we get
where we have neglected the surface terms due to the boundary conditions on n. But n (t,~) was arbitrarily chosen. 'llierefore the above result is consistent only if <1>0 satisfies the differential equation:
o
(3.13)
This generalizes the Euler-Lagrange equation for a system with a finite number of degrees of freedom. Observe that all the derivatives occur! This has an important consequence: The equation of motion is Lorentzinvariant provided
L
is a Lorentz-scaLar.
We would also like to discuss fields with several components:
o
(3.14 )
a=l, ... ,n
We may also discuss the energy-momentum corresponding to our field a A direct generalization of the Hamiltonian method suggests that the energy density is given by the HamiLtonian density:
(3.15)
H
(Compare with (2.56».
We expect energy to propagate throughout space,
so we must also have an energy current
5
associated with the field.
I
88
Since the total energy: E =
J Hd 3~
should be conserved, we must demand that
H
and
-+
s
(
satisfy an equa-
tion of continuity:
We can use this to determine an expression for the energy current
dL
doa (do
dL • dotb a + d(do
-+
s:
dL d
Using the Euler-Lagrange equations (3.14) we get:
Inserting this you find:
From this we can immediately read off the energy current:
(3.16)
We know that the energy density corresponds to the
TOO-component of
the energy momentum tensor and that the energy current si corresponds to the T Oi component (compare (l.3~), but then the above expressions for
Hand
s1
suggest that we put
(3.17)
If
L
is a Lorentz scalar, this is a tensor which reproduces
Hand
-+
s .
Exercise 3.2.2 Problem:
Use the equations of motion to show that the above energy-momentum tensor is conserved:
From exercise 3.2.2 we learn that momentum!
~aB
produces a conserved energy-
So everything should be all right:
~aB
is called the
89 canonica~
~aS
you should be careful:
I
However, there is one point where
energy-momentum tensor.
need not
be symmetric, contrary to our
previous demand (sec. 1.6) based upon conservation of angular momentum. Consider the expression (3.17) for the canonical energy-momentum tensor:
Obviously, the last term is symmetric, but the first term need not be so~
What can we do if the canonical energy-momentum tensor is not
symmetric?
We must repair it:
Suppose we can find a tensor
eaB
with the following properties: is symmetric (3.18)
Then
TaB
+
2)
dSe
aB
3)
Je
od
eaB
a
= 0 3
i =
0
is a symmetric tensor which is conserved and it.re-
~aS
produces the same energy-momentum as T
aS = ~aB + eaB
Hence, we might call
the true energy-momentum tensor.
eaS
.systematic method to construct linfante, see for instance:
There exists a
(the method of Rosenfeld and Be-
Barut [1964], ch. 3, sec. 4).
But the
method is very complicated and we will not have to use it. Later on (ch. 11) we shall see how it can be const"xucted fran a dEferent point of view. Okay, suppose we want to construct a free field theory. What should we demand about
?
L
That depends on what we expect of it!
Although we will not quantize the fields,
it might help to look at the
following very naive discussion: A particle of mass
m, energy
by the wave function:
~(PX-Et) e
(where we have put
E
1).
~
and momentum
p
is represented
iP~xP = e
When we quantize a field theory of a free
field, we expect that the field represents particles (quanta)
.~efore
we might look for solutions to the equations of motion on the form:
(3.19)
~(x)
=e
ip
P
x~
where
and we might interpret these solutions as the wave functions of the quanta of the field.
For instance, if we quantize the Maxwell field,
we expect it to represent massless particles
=
photons.
(You might be
worried about the fact that we allow comp lex solut'ions in a classical theory of fields. of the form
If you are very worried, you may look for solutions
.
Cost p ~XP)I)
90
I
Let us summarize our expectations:
We want to construct a free
field theory based upon a Lagrangian density with the following properties:
(3. 20)
11
L
is a Lorentz scalar
21
H
is positive definite
3}
The equations of motion allow solutions on satisfies the the form: e ip xlJ where P]J lJ 2 dispersion relation p plJ =_m • )l
Property 1) quarantees that we are constructing a relativistic theory, 2) guarantees that the energy density is positive, and 3) guarantees that if we quantize the field, it will represent free particles with a mass
3.3
m, where
HAMILTONIAN FORMALISM FOR RELATIVISTIC FIELDS
In analogy with the Hamiltonian formalism for a particle we can now associate to each component of the field a conjugate momentum defined by: a _ aL " - a(ao~a) Here "a should be considered as a function of ~a' ao~a' ai~a' If we can invert tne relation and obtain ao~a as a function of ~a' "a, ai~a we say that the system is canonicaL (3.21)
Exercise J. 3 • 1 Problem: Consider a canonical system. to the following HamiLton equations:
~ a"a
(3.22)
=
a~ 0
a
Show that the Euler equations are equivalent
aH
a~a
_ a "a =
0
+
a
aH
i~ ~ a
Consider a suitable function space M. A functionaL F is a map, F : M -+ R • that maps a function into a real number. If M is the set of all smooth functions f : [a,b]-+R then we can for instance consider the following functiona1s: b
F1[f] =
Jf(x)dx a
b
Fz[f] =
J!(f'(x»Zdx a
If the functional F is sufficiently nice, we can introduce a jUnctionaZ derivative. Consider first the case of an ordinary smooth function , f : Rn -+ R .
I
91
Let
Xo
be given.
Then we can expand f(x
O
f
in a neighhourhood of
Xo
+ y)
To formalise this we consider the map: y
This is a linear map.
Thus
J
af
aE
(xo + £Y)[£=O we can write it in the form: +
+ £y)
a. (x 1
depend on
Here the coefficients rivatives of f:
xQ
0
)yi
and actually they are the partial de-
at a.(x o) ~ -.(xo) 1 axl This motivates the following definition: consider a functional F. Let fa be given. Then we can expand bourhood of fa F [fa + g] = F£f o ] + ~: (x) g(x}dx + ••• If=fo To formalise it, we consider the map:
F
in a neigh-
J
of
This is a linear map.
g + aE [fa + £g] Thus, we canl~~te it in the form:
aa
J k(x)g(x)dx
F[fo+ £g]
£1£=0 The function k(x) depends on of F at the function fO :
and we define it to be the funationaL dePivative
fa
= k(x) 5F Oflf=f
o
N.B. When we perform the variation: g vanishes on the boundary.
fo
+
fo + £g
then we will always assume that
Exez>aise 3.3.2 Problem:
Consider the following functionals: t2 df 2 F} [f] = [jm(dt) - V(f]dt , F2 [f] = f(x O)
J t}
Show that the functional derivatives are given by:
Exeraise
of 2
_
Of
- 5(x-xO)
5F
,If =
-
5'(x-xO)
3.3.3
Problem:a)Show that the Euler-Lagrange equation can be written in the form: 5S _ 0 oq,a
(3.23) where
=
J
92
I
b) Show that Hamilton's equations can be written in the form: oH
(3.24)
o~a = -
6H a
alia
at
where
Let
F
and
G be two functionals of the field components and their conjugate
momenta, i.e.
f F(<pa;ai<Pa;lI)a
F[
d 3-+ x
R3 and a similar expression for logy with (2.59):
We may then define their Poisson Bracket in ana-
G.
{F;G}
Exercise J. J. 4 Problem: Show that the Poisson Bracket of two functionals satisfies the properties listed in exercise 2.10.1.
Exercise J. J.5 Problem: Consider a field theory. through the formula:
For fixed (t,lt}) we can define a functional
a (t,~})
'S
This functional which depends on t and will simply be denoted as In a similar way we can define a functional:
~a(t';;:l)
lib -+ lIb(t;~ ) b
1
+
simply denote as 11 (t,x 1 ). Show that the functionals satisfy the rules:
{
{1I
6~ Hint:
Rearrange the functionals
b .... 11 (t,x ) 2 o<pa(t,xl ) 6~b
b -+ 611 (t,x)
-+
(t,x 6
3
l
);
b-+
11 (t,x )} = 0 2
(it} - ~2)
and
a
and
in the form
-+3-+
-+
(x - Xl) f b -+3-+-+ f 11 (t,x)6 (x - x 2 ) ~a(t,x)o
d3~
d 3;t
+
ob 63(;t_~) a 1
0
o
0
61!b 6l1 b (t,; ) 2
6
611 -------
a
b 3 6 6 a
(x - it2 )
93
Exeraise
I
3.3.6
Problem: Consider a canonical system. Let F be a functional of the fields and their conjugate momenta. We will also allow F to depend explicitly on time:
The field components $ and their conjugate momenta themselves evolve in time according to Hamilton's e~uations (3.20). Show that the total time derivative along a field history is given by: dF of dt = at + {FIH} Let us make a short comment on how to quantize a field theory. quantum mechanics we can use two different strategies: (a)
As in the elementary
Path-integraL formaLism: This is closely connected to the Lagrangian formalism. configuration
~ (2)(~)
$a (])(;)
at time
Consider a field
t1 and another field configuration
We are interested in the transition ampLitude
at time t2
<: (2)1$ (1», i.e. the probability amplitude for finding the field in the a
state
a
$a (2)
at time
t2
when we know that it was in the state
To come from the configuration
at time
must have developed according to some history between $ (1) and $ (2):
$ (1) a
$a(t,~)
to
$ (2) a
$a (I) the field
which interpolates
a
a
~ ( +) "'a t 2 ,x
$ (1) 6~) a
= "'a ~ (2)(x+)
To each such history we have associated the action: t2 S[$a] = L ($a,oj.l$a}d3~ dt
f J tl R3
A straightforward generalization of Feynman's principle (2.21) then gives the result: ~a(t2,i)=~~21(~) (3.26)
J
<$a(2)1
exp{~S[~a]} D[~a(t,i)l
~a (tloi)~~ 1) (~) where we sum over all histories interpolating between (b)
$ (1) a
and
$ (2). a
CanoniaaL quantization: This is closely connected to the Hamiltonian formalism. Consider a canonical field theory. The physical quantities are represented as funationaLs F[$a,~b] of the field components and their conjugate momenta. In the quantum context these functiona1s are replaced by Hermitian operators in such a way that their Poisson Brackets are replaced by commutators: (3.27)
{F ,G}
-+
.." if!1 [~,G]
As we have seen the field components ~a(t'~l) and their conjugate momenta n b (t,i2 ) can themselves be interpreted as functiona1s. According to exercise 3.2.7 they must, therefore, be replaced by Hermitian operators satisfying the Heisenberg Commutation RuLes: ['a(t'~l) ~b(t'!2l1 = [wa(t,xll ; h (t,x2 )] - 0
n
[~a(t'~l}
;b(t'!2)] -
iflo~o3(;2-il)
I
94 As an illustration of the preceding ideas we take a look at the following Lagrangian density
(3.28) This is obviously a Lorentz-scalar and if we write out the first term explicitly, L(>
, a)l <j»
= ~[(l!)2_ at
(l!)2j - U(<j» ax
we see that it is a direct generalization of the string-Lagrangian (3.7). The last term can be interpreted as a potential energy density. It can easily be included in the string model too if we simply assume that each »atom« in the string has a potential energy U(x) arising, e.g., from the gravitational potential. Notice too that the above Lagrangian has the same form as the non-relativistic Lagrangian for a pointparticle (2.5). With this choise of the Lagrangian the conjugate momentum is given by
1T=~ at
i.e. it coincides with the kinematical momentum. Furthermore the Hamiltonian density reduces to
(3.29) This is positive definite, as it ought to be, provided the potential energy-density is positive definite. In the following we will assume that the minimum of U is zero, and that this minimum is attained only when ~ vanishes identically. ~e Lagrangian therefore satisfies the first two conditions in (3.30) which must be valid for any field theory if we are going to make sense out of it. The third condition is specifically related to free-field theories. In our case the equation of motion is given by )l a2~ a2~ (3.30) a~a <j> = U'(<j» i.e. at2 - ax 2 = -U'(<j» which should be compared with Newton's equation of motion. If we make a Taylor eXpansion of the potential energy density, using that U(O) and U,(O) vanish by assumption, the equation of motion reduces to a a)l<j> = U"(O). + )l
~ U'"(O).2 + •••
This will only allow solutions of the form <j>(x) = Eexp[ip x)lj provide~the potential is purely quadratic. That follows immediately from the fo~ulas: a a)l<j> _p p)lEexp[ip x)lj )l )l )l U'(<j»
{U' '(0) + W"'(O)Eexp[ip x)lj +.;.}Eexp[ip x)lj )l )l
In the case of a quadratic potential we get the dispersion relation
_p p)l = U' '(O)l )l which shows that U',(O) is the square of the mass of the particle in the theory. In the general case Eexp[ip x)lj will never be an exact solution, but it will be an approximative solution pro~ided € is so small, that we can neglect the higher order terms.In the weak field limit we can therefore treat the field theory with a general potential as a free field theory, where the mass-square of the particle in the theory is still given by U·'(O). As the field quanta in the general case are no longer free they must exert forces on each other. We therefore say that they are selfinteracting. Notice that the free field case,where U is quadratic, corresponds to a linear equation of motion, while the self-interacting case corresponds to a non-linear equation of motion.
95
3.4
I
THE KLEIN-GORDON FIELD. Now we are in a position where we can construct our first explicit
field theory.
We start by constructing the equations of motion.
sider the energy momentum relation: p~p~ =_m 2 . stitution
p~ =-ia~
Con-
If we perform the sub-
, this leads to the equation of motion
(3.31) This is known as the invariant.
K~ein-Gordon
equation and it is obviously Lorentz
To check that it has the solutions we want, we observe that
(a~a~ - m2 ) eXp[ip~x~l = (- p~p~- m2)exp[iP~X~1 and this shows us that Gordon equation provided
>(x) = exp[ip~x~l is a solution to the Klein2 p satisfies: p p~ =_m • ~
~
Now we must find the Lagrangian density
L.
We know that the
equation of motion
must reproduce From this
we immediately read off: a
(~~» =-a~>
.... L
=-~ (a~» (a~»
+ terms involving
~=_m2> .... L = _1 m2 >2 + terms involving o> 2 So we have reconstructed the Lagrangian density:
> a~>
(3.32) Since
>
is a Lorentz scalar, the Lagrangian density is a Lorentz
scalar, too.
Finally, we should check that the energy density is po-
sitive definite:
Le.
(3.33) This is obviously positive definite, so everything is okay! Let us look a little closer at the Lagrangian density. dratic in
>
and
a~>
It is qua-
, which is typical for a free fietd theory.
I
96
Furthermore, it consists of two terms: a) A term,- ~ (a » (a l1 », which is quadratic in the derivatives and which acEs ~ike a kinetia energy term. Observe, however, that it includes the space derivatives, too: -1(a » (a).l>l = 1 (1.1.) 2 _ 1 (V» 2 2 ).l 2 at 2 If you compare with the model for the continuous string (sec. 3.1), you see that (v<j» 2 should be counted as a potential energy term, due to tfie local interaction. A term, - ~ m2 <j>2 ,which is quadratic in <j>. It is called the mass term, because m gives the mass of the field quanta. Observe the sign! It will be crucial later on.
3
b)
E:x:eraise
3.4.1
Problem: Determine the conjugate momentum of the Klein-Gordon field and show that the theory of the Klein-Gordon field is a canonical field theory. Determine the Hamiltonian functional
H[>,n] and verify by an explicit calculation that Hamilton's equations reproduce the KleinGordon equation.
At this pOint we have only looked for solutions on the form: >(x)
= exp[ip 11 x).l]
that had an important significance on the quantum mechanical level. We might look for other solutions. The simplest possible classical solution is a static spherically symmetric solution > (t,x). '= > (r) . Using this ansatz, the Klein-Gordon equation reduces to: 2
m >(r)
1 d
2
= r dr 2 (r»
i.e.
From this we find the solution: (3.34 )
= >(r)
>(t,i)
±mr
£: e
r
This solution is singular at the origin. Only the decreasing solution is physically acceptable because the contribution to energy from infinity otherwise explodes. How can we interpret the solution (3.34)? In electromagnetism we have a similar solution, the Coulomb solution. There we use the ansatz:
o
2
=
1: ..i!..2
r dr
[rep]
+
e
r
This corresponds to a pure spherically symmetric electric field, the
I
97
Coulomb field: The field
aA E=-VCP-at
of singularity, r
!1 =
r
b r2
-+
r
r
is interpreted in the following way:
At the point
0, we have an electrically charged particle, say
a proton, acting as a source for the electromagnetic field.
The field
itself is interpreted as a potential for the electromagnetic force
E = - Vcp
which other charged particles will experience.
On the quan-
tum mechanical level two charged particles will interact by exchanging photons.
The photons,
i.e. the quanta of the electromagnetic field, will transfer momentum and when the momentum of particle changes, it experience a force. Let us make the same interpretation of the static spherically symmetric solution of the Klein-Gordon equation.
The electromagnetic field
is responsible for the electromagnetic interaction between protons.
Let us assume that the
Klein-Gordon field is responsible for the strong interaction between protons.
When you quantize the electromagnetic field, you get mass-
less photons.
In the same way we assume that you get
you quantize the Klein-Gordon field.
neutron) then interact strongly by exchanging
This exchange produces the strong forces. derived from the potential: cp(r) In other words:
n-mesons when
Two protons (or a proton and a
c r
e
w-mesons:
The strong force is then
-mr
At the position of the singularity, we have a par-
ticle, say a proton, acting as the source of the Klein-Gordon field. It produces the potential cp(r) = ~ e-mr and hence the force which r
other strongly interacting particles will experience. mr cp(r) = ~ eis called the Yuka~a potentiaL.
The potential
Contrary to the Coulomb potential we see that the Yukawa potential is exponentially damped! Hence, it has only a finite range. Let us estimate this range.
If
y
is a length measured in meters and
x
is
98
I
the mass measured in MeV, then in our units where
=~
c
1
it can
be shown that
x The observed mass of the
~-meson
is
140 MeV, hence, the typical
m
range of the force is:
2-10- 13 140 meter
= 1,4-10 -15
meter
Now that is exactly the typical distance between protons and neutrons in the atanic nucleus
am
the Yukawa potential can therefore very well account for the
strong force binding the nucleus together. force was never observed classically.
It also explains why this
In fact, the extranely short range
means that the force operates on spacetime regions so small that the quantum mechanical effects dominate completely.
3.5
THE MAXWELL FIELD. The next field we attack is the Maxwell field
the equation
Au
Here we know
of motion:
(1.33)
Can we find a simple Lagrangian density which leads to this equation? Since the equation of motion only involves derivatives of,' Au expect that
only involves
L
' we
a~Av
L
and we should determine it so that: all
3L _ U v v II II v v II a(aA}-olldA - a (oA)=a (aA - o A ) II v II II
This is a mess, so we will use a trick. gauge invariant,_
The equations of motion are
Now the Simplest way to construct a gauge-invariant
theory is to use a gauge invariant action. ever
All
All + o~X
This guaranties that when-
extremizes the action, then the same thing will be true for Note, however, that it is not the only possibility when
you want to construct a gauge invariant theory (compare the discussion in section 2.2), but let us try it! The starting point is a gauge invariant Lagrangian density Since
L
only involves derivatives of
Aa
a gauge invariant quantity out of the derivatives. the field strength:
L.
we must try to construct But this must be
99
Thus
we expect that
L
I
only depends on L = L(P
CtB
F CtB
)
But the simplest quantity you can construct out of
P
aB
which is gauge
and Lorentz invariant is the square: P
CtB
p CtB
So we expect (3.35) where
k
is an arbitrary constant.
Worked Exercise Problem:
3.5.1
Show that
4F\l\l
Using exercise 3.4.1, we see that the ansatz (3.35) leads to the correct equations of motion:
To determine ¥CtB
k
we investigate the energy-momentum tensor:
a(~~A\I)
aCtA\I
+
But here something is wrong:
Ct nCtBL =-4k[p B\la Av -
~nCtBpyOpYO}
It is not symmetric, and in fact it is
not gauge invariant, due to the term:
aCtA \I
This suggests that we
should make the replacement: aCtA\I
aCtA\I - avACt = pet
+
v
because the new term is gauge invariant. Therefore we try to repair the canonical energy momentum tensor a CtB = Is this legal?
Remember that
properties (3.18).
¥CtB
with the correction term:
4kFB\l a ACt
v
aCtB
should possess three characteristic
Pirst we observe that
¥CtB+aCtB
is symmetric:
~CtB + aCtB =_4k[pBVpCt Secondly we observe that
a CtB
_ ~ netBp pYa} v 4 YO is conserved:
4ka a{ pBVavACt]
..
4k{(a pav)a ACt \I
a
+
pBva
a
B v
ACt}
o
I
100
because the field's equation of motion kills the first term, and the antisymmetry of the field strength kills the symmetric tensor
e aB
Pinally we observe that
0Sov Aa !
does not contribute to the energy-
momentum of the Maxwell field:
4kJ povovACld3x
4kJ F
Oi
ai Aa d 3 x
4kJ (opFo~}Aad3X = 0
4kJ (oiFoi)Aad3X
where we have neglected the surface terms because the field strength vanishes at infinity.
The last integral vanishes due to the field's
equation of motion. So we have succeeded in repairing the canonical energy-momentum tensor.
Compairing the energy-momentum tensor:
¥a B +
pYa} e aB =-4k[p BV p Cl V _ :1411 aBp Yo·
with the true energy-momentum tensor for the Maxwell field: (1.38)
we finally obtain the result
k
~!
Thus we have constructed the Lagrangian density for the Maxwell field:
L
(3.36)
Exercise
EM
3.5.2
Problem: Determine the conjugate momenta of the Maxwell field A~ the theory of the Maxwell field is nota canonical field theory.
As we have seen
LEM
and show that
is Lorentz and gauge invariant, and it pro-
duces a positive definite energy density:
(Compare with (1.41)).
It remains to investigate the equations of
motion:
o
(1.33)
We look for solutions of the form: (3.37)
Here
£~
is a Lorentz vector, called the poZarization vector.
insert the wave function into the equations of motion, we get:
If we
I
101
o=
(-P)Jp)J£v + pv(p)J£)J»
exp[ip)Jx)J]
£)Jexp[ip)Jx)J] is a solution of the equations of motion pro-
Hence A)J vided:
(P)Jp)J)£v = (p)J£)J)Pv . This algebraic condition has two kinds of solutions: and Let us examine the condition
al.
b)
If
p )J p)J ~ 0 , we conclude:
(p)J£)Jl (p)JP)J)
Pv
Consequently £v is proportional to Pv and we see that the equations of motion allow solutions of the form: (3.38)
They are not observable because we
But these solutions are trivial. can gauge them away! AV' We therefore
=
Choosing
i.cL exp[ip)Jx~],we get:
X
Av + avx
disregard them!
Then we are left with the case
b).
If
P)JP)J
o ,
we conclude
but this is only consistent if p)J£j1 = 0 Thus , we conclude that the equations of motion allow solutions of the form:
A
(3.39)
Since
P)JP)J
with
v
= 0,
the photons are massZess!
o
mation. X
=
However, it
£)J,)J = 0,1,2,3 , which carry physical infor-
To see this, we make a gauge transformation choosing
iaexp[ipjJx)J]
o
These solutions are
non-trivial, i.e. we cannot completely gauge them away. is not all the components
and
• This leads to the transformed potential:
102
I
+ a x = E exp[ip xll] - ap expUp xll] = (£ - ap lexp[ip xll] v v v II v II v v II So if we perform a gauge transformation, we can change the polarization vector according to the rule: A' v
= A
(3.40l
To investigate the consequences of this gauge freedom, we consider a wave travelling along the z-axis. pll where
w
= -k
=0
pllEll
Then the four-momentum is given by (w,O,O,kl
because the photon is massless.
Due to the condition
, we conclude: £
o
= £ 3
Thus , the wave function of the photon has the form:
=
All
[E 3 ,E 1 '£2'£3] exp[iw(z-tl]
But we are still allowed to make gauge transformations:
Hence, i f we choose pletely! A =£ eiw(z-t) II
A =~ q,
pf
II
eiw(z-tl
=£Oeiw(z-t)
longitudinal
r...
part
=
£3/W
,
we
£11
tion vector into its different components, then:
+
t
£1 £2
£.1.
~=~
y-axJ.S
is called the scalar part
£0
0£ ,+
£ll= (£o,~ l
a
EO and £3 comIf we split the polariza-
have gaugEd away
z-axis
}
is called the transverse part is called the longitudinal part
£3
The above result then shows that we can gauge away the scalar photons and
Fig. 41
the longitudinal photons, but we cannot change the transverse part. Consequently all the physical infoonation is con-
tained in the transverse parts
E:r:eraise
£1
and
£2 !
3.4.3
Problem: Compute the complex field strengths corresponding to the above travelling wave and show explicitly that they only depend on £1 and £2
I
103
3.6 SPIN OF THE PHOTON - POLARIZATION OF ELECTROMAGNETIC WAVES Let us investigate the physical meaning of closer.
Each polarization vector
£1 and £2 a little may be decomposed in the follow-
£11
ing way: [£0'£1'£2'£3]
£0[1,0,0,0] + £+[O,l,i,O] + £_[O,l,-i,O] + £3[0,0,0,1]
where £ ± = b(£ 2 1
+ i£2 l
We may decompose the wave function All
£ll exp[ iw (z-t) ]
in exactly the same way.
We want to show that this decomposition is
closely related to spin!
To see this we must investigate what happens
if we perform a rotation about an axis, say the z-axis. Now a rotation is nothing but a special kind of Lorentz-transformation: X,ll
= aY
v
XV
represented by the matrix (allvl where
all
Here
e
o o o
v
o
o
1
o
Cos 8
Sin 8
o
-Sin 8
Cos 8
a
o
is the angle of rotation.
a
1
In the new coordinate system we
have a new coordinate representation of the wave:
with
£'=£aY v II v
and (~llvl the reciprocal matrix of (all v}. Hence, the rotation only attacks the polarization vector. We now get,
[1,0,0,0] ... [1,0,0,0]
: 0 [0
o
o
Cos 8
-Sin 8
Sin 8
Cos 8
°
°
~1
[1,0,0,0]
104
I
and similarly for [O,O,O,lJ. Finally [O,l,±i,OJ transforms into
1 [O,l,±i,OJ
o
o
o
o
Cos 8
-Sin 8
o
o o
Sin 8
Cos 8
a
[0,Cos8±iSine,-Sin8±iCose,OJ
o
o
1
= exp(±ie).[O,l,±i,O] Thus, if we let
RS
denote the rotation operator, then:
[1, 0 , 0 , a ] exp[ ihl (z-t)] and
[ 0,0,0,1] exp[ihl (z-t) ]
are eigenfunctions with eigenvalue 1, and [0,1, i, 0 J exp[ihl (z-t)] and
[0,1, -i, 0 Jexp[ihl (z-t)]
are eigenfunctions with eigenvalues:
e
±is is
But the rotation operator is given by: the operator of angular momentum around the z-axis.
You then see that
the wave function for a scalar or a longitudinal photon carries spin projection
0 , while the wave functions for the transverse photons
carry spin projection ±l.
You should, however, remember that only the transverse photons are observable!
Thus, a
photon is a spin l-particle, and i t has
spinpro-
spinp')ojection +1 either spin projection 1 in the direction of the momentum or it has spin projection -1 ~ p photon in the direction of momentum, but i t never
------.. jecti~ I
...
has spin projection O!
We therefore say
that the photon has heZicity ±l.
The preceding discussion may seem to bear little resemblance to what you have previously learnt about electromagnetism. Maybe the following remarks will clarify this;. We have looked for solutions on the form: (3.37)
~(x) = £~ exp[ip~x~]
The above solution is complex valued and hence, it can have relevance only on the quantum mechanical level. Rut clearly the real and imaginary part will solve the Maxwell equations too, and they are real solutions so that they may have relevance On the classical level. We want to find out how we can interpret the above solution classically and quantum mechanically. Let us discuss the classical interpretation first. We know that Maxwell's equations allow plane waves as solutions If we introduce a coordinate system, where the 3-dimensional wave vector along the z-axis, this formula reduces to: A~(X)
= £~Cos(wz-wt)
k
points
105
I
We may decompose it into a scalar part and a vector part: AP (~,it) ~ gOCos(wz-wt)
it
=
;:Cos (wz-wt)
But as we have already seen, we can gauge gO and g3 away! If we work in the special gauge, where gU = g3 = 0, the scalar potential drops out, and the vector potential is represented by:
A=
(3.41)
;Cos(wz-wt)
k.
where ;: = [,1,,2,0] is orthogonal to the wave vector We can also calculate the field ,strengths. Observe that since it is time dependent, this wave will also represent an electric field:
R= g x it - -(;Xk) Sin(wz-wt) E = -~~ =;w Sin(wz-wt)
(3.42)
Here the electric field is pointing along the polarization vector and the magnetic field is orthogonal to bot~ the wale vector and the polarization vector. As B and E are pointing in constant directions in space, we speak about linearly E polarized light. ' The density of momentum g is pointing along E Poynting's vector (C=EO=l): 2 ~~~~~~~~~~~~ g = Sin (wz-wt) Hence, there is a simple connection between the momentum density g and the wave vector k which represents the momentum of a single photon (P--hk). Similarly, the energy density is given by (c='o=l) ~ ±2 ~2 dx,t) = i(l> +B )=w2 Sin 2 (wz-wt) Let us consider a big box at a certain time, say t=O. Assume that the box contains N photOns. Each has an energy given by ~w and a momentum given by ~k. i.e. in relativistic units, where 'o=c=~ = I, we see that the total energy and momentum is given by: E = Nw , P =Nk tot tot Now let us compare this with the similar classical calculation. Here the total energy is represented by: w2V E = Sin 2wzdxdydz w2 ,(to)d 3x -2tot
...
EXR kW
~
f
15tot
Jg(~,t)d3x
J wk J S'in2wzdxdydz ~
~
wkV -2-
where V is the volume of the box. Thus, you see that the classical and quantum mechanical calculations are consistent provided you put: N = wV 2 We have also found waVes representing spin: [O,l,i,O]exp[i(wz-wt)] We extract the real part and get the classical wave: (3.43)
o
A- [
;:i~:::l ]
106
I
which solves the Maxwell equations. In this case the vector potential is circulating around the wave vector with a frequency w • Let us compute the field strengths:
....
(3.44)
E -
....
aA
- -
at
= w
l;!~i~:::~
x
A=
-cos(wz-wt)] w
Sin(~z-wt)
[
Clearly they are rotating too, and we speak about circularly polarized light. Then we turn to the quantum mechanical interpretation of the complex~wave solution: A]:I(x) = £Pexp[i(~-wt)] which we interpret as the quantum mechanical wave function for a photon with the momentum =~k and energy E·= ~w • Again we work in the special gauge where £0 and E3 disappear:
p
A' (xl -
[l~
"",(i (ki-ll
We may decompose this wave function along the x-axis and the y-axis:
A' (x l
• "
[i 1
,,,,wki_ll'"
[! ],""
(i
(ki_) l
As the solutions are only determined up to a factor, we may normalize the solution so that: (£1)2 + (£2)2 = 1 i.e. £1 = Cos a ;£2= Sin e, where 9 is the angle between £ and the x-axis. Since all states can be obtained as a linear combination of A~ll) and AP(2) , these states represent a frame for the linearly polarized state~. AP (1 ) repr~sents a photon polarized along the x-axis, and AP(2) represents a photon polarized along the y-axis. In the general state, A).! (x) = cose A).! (1) (x) + Sin a A).! (2) (x), Cos e will give the probability amplitude for finding the photon polarized along the x-axis and Sin e will give the probability amplitude for finding the photon polarized along the y-axis. An experiment may he performed in the following way: If we have a beam of photons, a light beam, which passes through a polarizer then some of the photons will be going through the polarizerwhile some of them will be absorbed. After the passage, the light is polarized. All photons are now in the same state, they are polarized along the characteristic axis of thepolarize~ Now, suppose we insert a second polarizer which is rotated an angle a relative to the first polarizer. What is the intensity I of the beam after the passage of the second polarizer in terms of the intensity I of the polarized beam? Let us introduce a coordinate system whePe the x-axis points in the direction of the axis of the second polarizer: Then the polarized beam is characterized by a polarization vector:
; =
I :::: I
and the wave is decomposed according to
A~ - Cos
e
AP(I) + Sin
e
AP (2)
107
I
Detector
I~I
Unp:>1arized ligth beam
Fig. 43 Polarizer 2
Polarizer 1
(Observe that we have a macroscopic number of photons all occupying the same state. This is possible because photons are bosons. Compare with the discussion of superconductivity). Consider a single photon: The probability amplitude that it is polarized along the x-axis is Cos a. Thus, there is a probability cos 2 a that a given photon goes through the second polarizerl But as there is a macroscopic number of photons in the same state, we may actually interpret cos 2 a as a macroscopic physical quantity. If Ni is the total number of photons before the passage of the second polarize~ and Nf is the number of photons after the passage of the secondpolarize~ then: Nf ~ Ni cos 2a . As the number of photons are proportional to the classical intensity of the light beam, we may rearrange this as, (3.45)
and that is a formula which you Can test very simply in a classical experiment.
3.7
THE MASSIVE VECTOR FIELD We have seen that the Lagrangian density of the Maxwell field only
contains a kinetic energy term and no mass term. photons are massless!
Correspondingly, the
Now, suppose we add a mass term to this Lagran-
gian density and consider the following field theory:
(3.46)
L
The first thing we observe is that we have spoiled the gauge invariance.Therefore ACt
this field
ACt
is not a gauge field.
and
So in what follows:
ACt + 0Ct X represent different physical situations. the equations of motion. Using (3.14) we get:
m2 A f! These equations imply that:
m2aSAa = aadetFetS= 0 •
Let us find
108
Hence, the Lorenz condition, for the massive field.
l
=
aaAa
0, is automatically fulfilled
But then the equations of motion simplify con-
siderably: (3.47)
(a)Ja)J -
m2 )Aa = 0
and
aaAa = 0
Each of the components Ail thus satisfies the Klein-Gordon equation, and the field Aa will clearly represent free particles with mass
m.
We call Aa
the components the Loren<: pendent. Ao
All a massive vector fieZd.
Observe, however, that
are not independent, as they are connected through
condition. Therefore only three of the components are indeAs we shall see in a moment, this means that the scalar part
can be eliminated completely.
dom correspond to spin.
The reamining three degrees of free-
Using exactly the same method as we did in
connection with the Maxwell field, we can show that the massive vector particle is a spin
1 particle.
The projection of the spin onto the
z-axis assumes the values -1, 0, +1 part also contributes!
Worked exercise
and this time the longitudinal
The massive vector field is not a gauge field.
3.7.1
Problem: (a) Calculate the canonical energy-momentum tensor for the massive vector field and show that it is not symmetric. (b) Show that we can repair the canonical energy-momentum tensor by adding the following correction term: aaS =-0 (FSPAa ) P
leading to the true energy momentum-tensor (3.48)
(c) Show that the energy-density momentum tensor is positive definite. 'lb conclude
we have presented
TOO
corresponding to the true energy-
very naive semi-classical arguments to mo-
tivate three kinds of fields:
,
Name
Lagrangian density - ~ (0)Jq,) (allq,) _ ~ m2q,2
(3.49)
Klein-Gordon field
(3.50)
Maxwell field
- '4
(3.51)
Massive vector field
- 4
1
1
F
F
Ila
F aS
as
Equation of motion (O)J all - m2 )q,
=
0
(alldll)Av- dv(OpAll)
FaS _ 1 m2AaAa
'2
(0
II
=0
all - m2 )A Il = 0 Il
aaA
=0
and we have seen that they represent spin 0 particles, massless spin 1 particles (photons) and massive spin 1 particles (vector particles).
109
3;8
I
THE CAUCHY PROBLEM In what follows we will treat our fields as classical fields, i.e.
we will only admit reaL solutions to the field equations. The first thing we will study is the Cauchy problem. If we have a single particle moving in a one-dimensional space, it has the equation of motion:
d 2x m dt 2
Now, fix a time change of
x
tl
= -v
1
(xl
and prescribe the value of
x , and the rate of
at this particular moment: x(tll
=
dx dtlt
xl
=
VI
l What can we say about the dynamical evolution of the system once we have prescribed these initial data?
This is the Cauchy problem.
In this case there is a simple answer.
The above differential e-
quation is an ordinary second order equation, hence, it has a unique solution once we have specified the value of of
x
at a given moment.
x
and the rate of change
It is because of this that we say that
x
is a dynamical quantity and that the system in question is uniquely characterized by this single dynamical t-axis
quantity. Now let us look at the Klein-Gor-
SPACE-TIME DL~l
don field. Again we fix a time tl and we prescribe the value of the
'
field and the rate of change at that moment:
y-axis
)r Fig. 44
(3.49)
What can we say about the dynamical evolution? Well, let us write out the equa tions
o
or
0
f motion :
110
This
I
is a second order equation anl thus it has a unique solution once
we have prescribed
$
and
O
at a given moment.
In fact, it is easy
to see that we may construct the dynamical evolution directly by using an iteration procedure. We know the value of $ and a~ at the time at t) We may then find them at time t = t) + at using the formulas: + + a$ ..$ (t) + at,x) "" $ (tj,x) + at at (t),x) a$(t + at )
at,~)
at~(t ,x) at2 )
"" aa$t(tj,X) +
+ a$ 2 at (t),x) + ot[M (t),x) _m $ (t),x) 1 a$ But when we know $ and at at the time t = t) + at , we may repeat a$ at the time t the procedure to find $ and t) + 2at, etc. Thus at we can reconstruct the dynamical evolution of the Klein-Gordon field.
This shows that the field variable
$
is a dynamical quantity complete-
ly characterizing the system. Then we look at the Maxwell field and prescribe the values of Au and
Au
~
.
We select a time t = tj at this time. Now, what can
we say about the dynamical evolution of the Maxwell field? Nothing! And here is the reason why: We know that if we find a solution
Au
to the equations of motion, then the gauge transformation: Act
produces another solution.
+
Act
+
auX
X so that it is zero
Thus, if we choose
in a neighbourhood of t t ) , this will not disturb the boundary conditiond and you see that there is infinitely many solutions to the equations of motion, all satisfying the boundary conditions. On the other hand we have seen that the equations of motion are second order differential equations (avaV)A V - aV(avAV)
0
so you might wonder what is Wrong? There is a subtle reason for this.
Remember that the full equations
for a Maxwell field with a source look as follows (1. 33)
These equations automatically guaranteed the conservation of charge:
But this conservation has a price. rator
It means that the differential ope-
111
I
o beys a special identity:
o
(3.52)
00
Suppose we isolate
o [(0
°
on the left side:
all)AO II
This equation is interesting.
On the right we have time derivatives
of maximum order two, but then the same thing must be true for the left!
ao
Removing
we see that:
(allall)AO- aO(oIlA ll ) can only contain time derivatives of maximum order one! sider the equation of motion for the scalar part (ollall)AO -
0° (oIlAll) = 0 ,
we observe that there is something wrong. equation.
It is not a second order
By explicit computation you can easily find that it reduces to
o
(3.53)
the scalar part even prescribe
tt
~a
and
oA+
and ~
t
l
,
because once
+
at (t l ,x)
has to obey: I'l~ =
-
0'" at (V
7
• A)
For this reason we say that equation (3.53) straint.
In fact, we cannot
at the initial moment
we have prescribed:
0
This shows us clearly that
is not a dynamical quantity.
~
~
VE =
Le.
This is not an equation of motion at all!
then
So if we con-
AO
is an equation of con-
This leaves us with the three remaining field variables
as dynamical variables.
A
We may rewrite their equations of motion as:
(0 all)Ai - ai(a All) = 0 II
Le. 2
aA = flA ot 2
(3.54)
II
v[v . A +
~]
so the Maxwell equations reduce to one equation of constraint and three equations of motion. blems.
But we still have not solved our dynamical pro-
We are still allowed to make gauge transformations and thus the
dynamical evolution is completely indetermined. must choose a specific gauge.
To get rid of this we
It is tempting to use the Lorenz gauge,
I
112
(1.22)
not only because the Lorenz condition is Lorentz
invariant, but also
because the massive vector field automatically obeys this condition. Using the Lorenz
~
condition we can now eliminate
too.
That solves
our dynamical problems:
... A
The only true dynamical quantities are the three space components They are governed by the equations of motion:
a21 at 2
(3.55)
The scalar part
~
is eliminated by the equation of constraint
and the Lorenz condition: (3.56) We should, however, be a little careful with the iteration procedure. There are no problems with the A-variables. But consider the scalar part. It should satisfy (3.56) at all times, if our method is to be consistent.
Worked exercise
3.8.1
Problem: Use the iteration procedure and the equation of motion to show that the equation of constraint and the Lorenz condition are automatically preserved at all times. From this exercise it follows that the method is in fact consistent. You might also be worried about the fact that we can still perform gauge transformations, provided
X satisfies:
(1. 23)
Thus , you might think that we could repeat the argument from before to show that there are infinitely many solutions to the equations of motion, all satisfying the boundary condition. But this is not so. The only solution to the equation (1.23), which satisfies the initial data~ x(it,O) = 0 is the
o
and
x-function which vanishes identically!
In the case of the massive vector field you can show by a similar analysis that only the space components
Ai
are true dynamical quan-
tities. This time the Lorenz condition is automatically satisfied and therefore the scalar part AO is automatically eliminated. Observe that when we constructed the Lagrangian density for the massive vector field: (3.46)
L
we directly generalized the Lagrangian density for the gauge potential.
113
I
Especially we kept the gauge invariant kinetic energy term. L
= - ~ (a~Aa
-
aSA~)
a (aa AS - aa A ) -
~ m2A~Aa
But the massive vector field is not a gauge field, so you might wonder if it would not have been much easier to start with the following Lagrangian density:
L
(3.57)
1
This would be a direct generalization of the Klein-Gordon field.
Com-
puting the equations of motion you easily find:
(3.58)
Consequently each of the components satisfies the Klein-Gordon equation. Thus, in this case, all four components are true dynamical variables. But why did we not choose that Lagrangian density? have seen, the scalar part the vector part
Ai
AO
represents a spin
represents a spin
adopted the Lagrangian density
I-particle.
Because, as we a-particle, while Hence, if we
L' , we would construct a field theory
which, when quantized, would represent a mixture of spin I-particles!
gy-momentum tensor. Exercise
0-
and spin
Furthermore, this would lead to troubles with the enerThis is the content of the following exercise:
3.8'.2
Problem: Calculate the canonical energy-momentum tensor corresponding to the Lagrangian (3.57). Show that it is symmetric but that the energy density is indefinite.
3,9
THE COMPLEX KLEIN-GORDON FIELD Now suppose that we have two real classical fields:
~1
and
~2·
We will base their dynamics upon the Lagrangian density:
(3.59) Clearly there are no interactions between the two Klein-Gordon fields as their equations of motion decouple: (dllo ll -
(3.60)
2
m )
(dlla ll - m2 )
Now we perform a trick which is very useful. fields
and
We fuse the two real
into a single compZex field:
114
I
(3.61) We can then rearrange the Lagrangian density as follows:
(3.62)
This is only a fancy way of writing the same thing, but what about the equations of motion? vary
~l
~2
and
To derive the Euler-Lagrange equations we should
independently, getting:
a
~=
oL
and
)J a (o)J 11)
HI
a
aL
n;
aL )J 0 (a)J 1 2 )
But due to the formulas:
~l + i
~
4i
~1
-
}
{
and
H2
~1 =
~ (
~2 =
~i(~-q;)
these equations of motion are equivalent to the following equations of motion: oL
aq;
(3.63)
oL = o)Jo(o)J~) ~
where we formally treat
and
and
as independent variables:
L(~,~,a)J~,o)J~)'
L
For instance, we get in the above case (3.62)
o
~
=
1 - 2
aq;
m 2 ",'I'
+ 12 a )J a)J",'I'
or
o =
~[o)Jo)J
which, if you decompose
-
~!
the real and imaginary part. Gordon equation
(3.64)
m2]~ reproduces the Klein-Gordon equation for So the complex field obeys the Klein-
too:
0,
Now, returning to the Lagrangian density: (3.62) you observe that i t is invariant under the substitution: (3.65)
~ (xl'" e
-io.-
~ (x)
115
I
Thus, we have discovered asymmetry. But when there is a symmetry, there should also be a conservation law. (This is known as Noether's theorem).
Let us examine this a little closer.
n
Let
be an arbi-
trary space-time region, and consider the corresponding action: (3.66)
J L(
5 =
n 5ince
is invariant under the substitutions
L
=
I
+
e
-ia
e-ia~, eiao)J
L(eia
n is a constant function of
a
a.
Hence, differentiating with respect to
we get:
o
r· r (
J ~L
But since
n
n
5 '(0)
+ a,j, OL ) _ (" o~ + 0 aL )] d 4 ].1"'0 (0)J
was chosen arbitrarily, this is consistent only if: .[
oL
oL
~ (<Pa;p+ d)J<Pa(a)J
dL
-
oL
(
-
]
=
0
Using the equations of motion we can rearrange this as:
o i
r
aL a].ll
-
oL
]
Consequently we conclude that because of the symmetry and the equations of motion, the following four-vector, (3.67)
=
J].I
i[ <Pa_a_L _ (a)J
_ - __0£__ ]
will obey the equation of continuity (1.29)
For this reason we will call anything to do with
J].I
a current, although it need not have
electromagnetism~
to Q =
In the same spirit we will refeI
JJ d x O 3
t=t O as a charge, although it is not necessarily an electric charge!
I
116
~(x)
A symmetry like:
+
eia~(x)
is called an internal symmetry,
in contrast to a space-time symmetry where the Lagrangian density is invariant under a coordinate transformation: yl.l = al.l
XV
V
But internal symmetries and space-time symmetries have one thing in common:
They produce conservation laws!
seen that the symmetry JIJ
~(x)
+
eia~(x)
For instance, we have just leads to a conserved current
:
Exercise 3.9.1 Notation: Let Wp = A\.l (1) + iA\.l (2) be a complex vector field. Let G\.lV denote the corresponding complex field strengths: G\.lV = dpW V - dVWp The Lagrangian for the massive vector field is immediately generalised to 2 L = - ! G GPv - ! m wWll 4 \.lV 2 \.l Problem: (a) Show that the Lagrangian is invariant under the internal symmetry: Wp + e iCl Wll W)J + e -ia w\.l (b) Show that the corresponding conserved current is given by: JV =
i(w 2
\.l
C;)Jv -
W\.l G\.lv)
3.10 THE THEORY OF ELECTRICALLY CHARGED FIELDS AS AGAUGE THEORY Let us now examine the complex Klein-Gordon field: (3.62)
In this case the conserved current
is:
J\.l
(3.68)
Can we give a physical interpretation of this current?
Suppose
you want to construct a theory of an eZectrioally charged field, then you would certainly expect that we could define a reasonable fourcurrent
, and this electric current should satisfy the equation
JI.l
of continuity
a]J J\.l =
(1.29)
0
But then the complex Klein-Gordon field is an obvious candidate. ever, the complex field cal fields
~1
and
~2
~
= ~1 + i~2
How-
actually consists of two classi-
Hence, if we quantize it, we expect that
117
I
it represents two kinds of particles! Can we understand this in a simple way?
Yes!
In a relativistic
theory we may have production of a particle-antiparticle pair.
For
instance, a photon may, split into an electron-positron pair: +e +
y-+e
and if the particle has positive charge, then the antiparticle has negative charge.
Thus , if you want charge conservation, the theory of
a charged field must represent both of these particles, one with positive charge and one with negative charge! This makes the complex Klein-Gordon field an even more obvious candidate. But if q, is going to be a charged field, it must interact with the Maxwell field. In fact the current J~ must act as a source for the Maxwell field.
The question is then:
Can we construct
a suitable interaction term for the total system consisting of the charged field and the electromagnetic field: L
=
LA, + LI + L 'I'
Aa
=-!(~) (o\lq,) I m2ijiq, + 2 ~ - 2:
LI -
!4
F
as F
To answer this we may take advantage of the gauge symmetry!
etS
We know
that the theory has to be invariant under the gauge transformation (1.3J.)
but then we can use what we have learnt studying quantum-mechanical systems. The combined theory is gauge invariant provided we exchange the usual derivatives
0a
with the gauge covariant derivatives:
(2.44) Therefore we suggest the following total Lagrangian: (3.69)
L
= ~[(o~ -
ieAjl)q,][
(o~ - ieA~)q,] - I2
2-
m $<1> -
I
4
Fas F
as
This is invariant under the combined gauge transformation: (3.70)
Using this total Lagrangian, we have, in fact, gained something. In the original Lagrangian we were only allowed to make the transformation, (a) where
a
was a constant.
to make the transformation, (b) where
a(x)
q,(x)
-+
eiaq,(x)
But in the combined theory we are allowed q, (x) ... eia(x) q, (x)
is space-time dependent.
It is customary to refer to (a) as a gauge transformation of the
I
U8
first kind and to (b) as a gauge transformation of the second kind. Hence, incorporating the
~axwell
field, we have extended the symmetry
from a gauge symmetry of the first kind to a gauge symmetry of the second kind. We have seen that there is a natural way to incorporate the interaction between the complex Klein-Gordon field and the Maxwell field. In what follows we shall refer to
= cp 1 +
cp.
iCP'2
as a charged Klein-
Gordon field and we shall base the theory upon the Lagrangian density:
(3.69)
Now, what about the current?
Since we have extended the Lagrangian,
it is not necessarily equal to (3.68) any more. involving
It may contain terms
All:
+ terms involving
All
In fact, the old expression must necessarily be wrong because it is not gauge invariant! derivatives (3.71)
all
This suggests that we simply exchange the usual
with the gauge covariant derivatives
Jll
Dll
giving
~ !t$Dllcp - ~DllcpJ 2
This is a real, gauge invariant quantity and in the absence of the Maxwell field it reduces to the old expression (3.68).
Is it conserved?
The easiest way to see this is to calculate the Maxwell equations. From
we get
and
aL aAa
=
ie [~Da~ _ ~Da$J
2
'l'
'l'
which immediately shows that:
[ II a - 3 a (dllA)J= II ie -ollF lla ;-(dlld)A 2
[~Da~ 'l'
- a 'l'-CPDcpJ
Hence, we read off the conserved electromagnetic current:
(3.72) which up to a constant factor reproduces (3.6.8).
119
Exeraise
I
3.10.1
Problem: Consider the charged Klein-Gordon field. can be rearranged as
JaAa - e 2A A~$$ ~ is the electromagnetic current (3.72). L
where
J'1
Exeraise
Show that the interaction term
I
=
3.10.2
Problem: Consider a charged particle moving in an electromagnetic field. the interaction term (2.14) can be rearranged as (3.73) s = fJaA d 4x a
I
where
Ja
Exeraise
Show that
is the electromagnetic current (1.34) 3.10.3
Notation: Consider a charged massive vector field coupling corresponds to the substitution:
w~
Here the rule of minimal
with v~wv = a~wv
Problem:
J~
3,11
-
ieA~Wv
Show that the electromagnetic current is given by: = ie [w (VUWv - VVwlJ) -
2
~
W (V~Wv ~
- vV;ju)
1
CHARGE CONSERVATION AS ACONSEQUENCE OF GAUGE SYMMETRY,
We conclude the discussion of charged fields with sane general remarks. We saw
above that it was easy to couple a aomp ~ex field well field
Aa
If
=
+ i
to the Max-
was described by the free Lagrangian density:
Lo
=
Lu(
then the total Lagrangian density was obtained simply by replacing the ordinary derivatives with the gauge co-variant derivatives and adding the usual piece for the Maxwell field: L
= Lu(
-
~
FaS FaS
This Lagrangian density is obviously gauge invariant.
We can now give
a general definition of the current associated with the charged field: Consider the action for the combined system:
S
= So
+ S
I
+ SA
a
We have decomposed it into the actions corresponding to the free fields
and
Aa , and the interaction
term.
120
r
Consider the interaction term: Sr = Sr(,A Ct ) rf we produce a small change in the Maxwell field
this will produce a small change in the interaction term: Sr Here
OSr
->-
Sr + oSr
will depend linearly on
oACt , so we can write it as
J J Ct (x)oACt (x)d 4x
(3.74)
n The coefficient
JCt
which measures the response of the interaction
will be defined to be the current! Carrpare with exercise 3.10.2 in the case of charged particles. We nrust show that this is in agreanent with our previous ideas: First, we observe that,
(3.75)
Sr =
J Lr(,a~,A~)d4X Q
where the interaction term does not depend on
a~Av
since the presence
of the Maxwell field comes from the gauge covariant derivatives. we replace
ACt
by
ACt + soACt Sr(E) =
rf
and consider the displaced action,
J Lr(,a~,A~
+
EOA~)d4x
Q
then dS oSr =
r
dsls=o
J n
aLI aA~
4
oA~ d x
from which we read off that
JCt(x)
(3.76)
aLr a ACt
This is in agreement with the Maxwell equations aL
aL aA Ct
aLr
= - a F~Ct Ct ~ since the interaction term is the only term which depends explicitly
a~ a(a~ACt)
on
i.e.
aA
ACt . We can now give another argument for the conservation of electric
charge.
We will only use the part of the action containing
:
r
121
Sq, = So + Sr This is a gauge invariant action. q,(x)
+
Performing a gauge transformation
eie[EX(x)]~(x)
Av(x)
+
Av(x) + EdVX(X)
we get: S(E)
ILo
(e ie [EX(x)]
n
n
which is constant. Hence, we conclude as usual:
o_
dS" - dE: IE:=O
Let us make some assumptions.
First we assume that
equation of motion (3.63). Next we assume that X(x) boundary n This has the following consequence: eie[EX(X)] ~(xl where
~(x)
=
$ solves the vanishes on the
q,(x) + E~(X)
is a smooth function vanishing on the boundary of
n.
(We are not interested in an explicit expression of ~(x) .) NOw, observe that since q, solves the equation of motion, it will not contribute to we get:
~I dE: £=0 o
when we actually perform the differentiation! Thus
~
aLr dvX(x)d 4 x Jr ~
de: Ie:=o
n
V
=
I n
J V (X)d v X(x)d x 4
Performing a partial integration, we finally obtain
o
=-
I
n But as
x(x)
4 (d vJ V)X(X)d x
was arbitrarily chosen, we deduce
So we caught our equation of continuity! The above argument was very general and abstract! So maybe it is good to summarize the conclusions: We are studying the interaction between a charged field ~ = ~l + i$z and the Maxwell field A~ : L = Lo(
From the total Lagrangian density we get the equations of motion: d~aCl~ = ••••
( aea ~ )A B -
( aSd B) ACl = ....
Usually we use Maxwell's equations to identify the current the right-hand side of the Maxwell equation:
JS.
It is equal to
122
(dSdU)A S - (dSdS)A
U
I
= ~[~;d~~;~l
The conservation law then follows automatically from the form of the Maxwell equation. We do not use the equations of motion of the charged field at all! The new argument shows that the dynamics of the Maxwell field is superfluous. We need only bother about the ~-field:
Lip = Lo (
LI(
From this Lagrangian density we get the equations of motion of the charged field (3.63). The current is then identified through the interaction term (3.76) and the conservation law "BJ~ = 0 follows automatically from the equations of motion of the charged field. We do not use the dynamics of the Maxwell field at alZ.! What we use, however, is the gauge symmetry, so again we see that gauge in-
variance implies a
conse~ed e~ent.
Worked exercise
3.11.1
Problem: Consider the complex Klein-Gordon field
3.12
THE EQUIVALENCE OF REAL AND COMPLEX FIELD THEORIES.
As we have seen, it is very easy to couple a complex field 4> = ~ + i4>2 to the Maxwell field Aa. All you have to do is to exchange the ordinary derivat1ves da with the gauge co-variant derivatives Da = "a - ieAa . Of course there is nothing mysterious about the fact that we use complex fields. It is nothing but a trick which makes life easier, and we could easily have avoided it. To see this, let us return to the problem of constructing a charged field. The starting point is the free Lagrangian
(3.59) (we have got to have two classical fields because the theory should incorporate both particles and anti-particles). If we collect the two components into a single vector, (3.77)
we may rearrange the Lagrangian as follows:
(3.78) But then it is obvious that it possesses the symmetry,
[~21) + [C~sa-Sina ][4>1] S1na Cos a 4>2
(3.79) where
(3.70)
~
a
is a constant. 4> + e
ia
4>
Of course, this is completely equivalent to the formula:
123
I
In the same way we can construct a gauge co-variant derivative: =
(3.80)
D~
au - eA~
=
rII0 -110 J
Operating On the combined field, we find,
D;u I
a r~l]
=
U l~2
- eA lr ~2~lJ1 U
which is completely equivalent to the formula: Du~ (all - ieAu)(~l + i~2) = aU(~.l + i~2) - eAj1(-cp.2 + i~l) Hence, we can write down a gauge invariant Lagrangian density:
(3.82)
In this case "gauge invariance" means that it is invariant under the combined transformation:
~l] [ ~2
(3.83)
-Sin(ex(x»].[~l]
[cos(ex(x» Sin(ex(x»
+
Aa(x)
Cos(ex(x»
~2
Aa(x) + "aX(x)
+
Here is a dictionary which allows you to pass from the "complex" formulation to the ureal II formulation: Gauge group
[c~sa -Sina ]
Group element
S1na
Gauge vector
;1
=
[:d
~ = ~1 +
[0 -1]0
derivative
Du = au - eA u 1
Gauge scalar
= += ~I ~ 1
Gauge phase factor
exp[ia]
Cosa
Gauge co-variant
U(l)
SO(2)
=
exp{[~
~ 2 + 1
~2
'i.
i~2
Du = au - ieA u
$~
Cl -6]eJAa dx }
= (~l
- i~i(~l + i~2)
a exp[ieJ Aadx ]
r
Concerning the gauge phase factor, you should observe that:
[~-~r -[~~] From the formula
you therefore get
124
exp{x[~ -~]}
I
[1 0. 1 Cx - x 3 + xS~ -... )[01-11OJ L x + 1 x-"')OlJ+ = [C~s x -Sin x J = Cos x [10 01 lJ + Sin x[ 01 -I} 0 x Cos x 2
= Cl -
4~
5
4
3~
S~n
Thus the gauge phase factor lives in the gauge group as any decent gauge phase factor ought to do~ So you see that there is nothing sacred about complex numbers~
SOLUTIONS eF WORKED EXERCISES: No. 3.5. , aS F ) as a callA,,)
C ay So
FasFyo) aF 0 1 aF a_S_ F So [ __ + F ~ = 11 aY 11 a callA,,) yO as a COilA,,) J a COilA,,) dF aB Thus , we must first compute a callAv) aCF
a
11
11
aF a (aaAs-asAa) aB a(aIlA,,) = a (aIlA,,) inserting this, we get: naYnSo[Co)J\) - o)J')F
o.S
Sa yo
+ F (0)J\) - 0 )J~] as yo of
(0)J\) - 0 )J')F aS + Fyo (0)J\) - 0)J')
as
yo
Sa
oy
F Il " _ F"1l + FIl " _ F VIl = 2(F Il " _ F"Il) = 4F UV (due to the anti symmetry of
FIJ "), so we have shown once and for all:
No. 3.7.' Ca)
Here the first term is not symmetric, while the last two terms are trivially symmetric. Cb) Observe first that, using the equations of motion, we may rearrange the expression for eaS eaS =-C3 FSp)Aa _ FSP a Aa P P a 2 a B SP =_m A A _ F aA p
Here the first term is born symmetric, but its pre$ence is very important for the verification of the various properties eOS ought to have:
125
I
¥
(1)
This is clearly symmetric. Observe that the first piece is identical to the energy-momentum tensor for the electromagnetic field.
(2)
dSe
(3)
J a
aBd p is symmetric in SP .
since
=-J
ap(FoPA
ai(FoiA
But the last integral can be converted to a surface integral which vanishes automatically provided the fields A
(c)
~(ii2
+ 132) +
~
m2{ (A O)2 + (AI)2 + (A2)2 +
which is clearly positive definite,
(A3)~]
0
No. 3.8.1
From the equation of constraint and the Lorenz condition you
• If you prescr1be
~ (tl,X) dt
->-(
->A t1,x)
from the equations:
->-
->-
± dA
->-
V'at (tl,x)
and you can then use iteration to determine
(t
I
+ at,x) '"
~~
at
find:
and
M(tl,x) = -
2i at
immed~ately
~
and
2i at
at
(t ,;:) + I
We must then show that the equations of constraint and the Lorenz condition are still valid at t = tl + ot
Equation of constraint:
using the equation of constraint and the Lorenz condition at time tl this is rearranged as ->->+[aA ->.7->->= - 9. dA (t ,~) + ot • ~(-9 • ->A(tl,x» = -\I :it (tl,x) + ot • 8A(t l ,X) } dt
I
using the equation of motion this is rearranged as
aA lat
->-r
\I -
->-
(t I'
d
x) + ot • -
2
A
dtZ
126
I
Lorenz condition: ~( +) = ~(+) +) at tl+ct,x at tl,x + ct· ~$ (tl,x
Using the equation of constraint and the Lorenz condition at the time t is rearranged as -+-+
-+
1,
this
a"±-+-+
=-V·A(tl,X) - ct at(V'A(tl,X)) =
-9[it(tl,*)+Ot.~~(tl'*)]
= -9'A(t 1+Ot,*)
0
No. 3.11.1
The starting point is the free Lagrangian:
LO
=-
~~)(a~$) - ~m2~~
2112
We make it gauge invariant by using the rule of minimal coupling: L$($,all$,A a )
= LO($,D =
[-
$)
~m2~$1 + ~A [$al1~_~all$l 2211
..!.rf";)(a l1 $) 2
~
From this we get the equations of motion (3.63) (a a l1 )$ = m2$ + ie[A a l1 $+a (A~$)] + e 2A A~~ 11
11
11
11
and we can immediately read off the interaction term: L
I
ie A [$al1~_$al1
=
2 11
2
11
From this we get the current: J~
Of course this is identical with the previously found current (3.72): Then we compute a J~: 1.1
a
1.1
Jll = ~[$(a all )$ - $(a a~)$] _ e 2 a [AI1 $$] 21111
1.1
From the equations of motion we now get:
-~(m2$+ie(A al.l$+a (AI.I$))_e 2A AI.I$)] 1.1 11 11 e 2 a [A~;P~] ~
Inserting this, we get:
a11 JI.I
e 2 a [A~~$] - e 2 a [A I1 $$] ~
11
=0
Consequently we have reproduced the equation of motion without using Maxwell's equations!
0
127
I
chapter 4 SOLITONS 4.1 NON-LINEAR FIELD THEORIES WITH A DEGENERATE VACUUM In this chapter we will discuss in some details a few important models in classical field theory which have recently attracked great attention due to their so-called "topological" properties. To keep the discussion as simple as possible we will only consider classical field theories in (1+1)-space-time dimensions.
(Although some of the
results have suitable generalizations to higher dimensions, these generalizations are by no means trivial.) The theories we are going to consider will be non-linear field theories, i.e. in the non-relativistic case, with which we shall be mostly conserned, they are based upon a Lagrangian density of the form, (3.28) where the potential energy density U is no longer quadratic in
~
(cf.
the discussion in section 3.3). The associated equation of motion is given by (3.30)
U ' (<1»
and to ensure a positive energy density we shall as usual assume that the potential is positive definite: U(
~ 0
To obtain non-trivial results we shall however furthermore assume that it has more than one (g~oba~) minimum. Two examples are of particular interest: Illustrative example 1:
The ~~model.
In this model the potential energy density is given by the following fourth-order polynomial in (4.1)
U(
=
~:
}(<j>2_ f)2
Notice the sign of the quadratic term which is opposite of the usual mass-term (cf. the discussion in section 3.4). The potential energy
128
I
U[cp] Klein-Gordon , potential I
I
;1 Fig. 45a
+~ A
1>-axis
it
Cp-axis
Fig. 45b
density has two distinct minima,
'It
(4 . 2 ) 1> + = and and has the characteristic shape indicated on fig.45a (known as a double well). Illustrative example 2: The sine-Gordon
mode~.
In this case the potential energy density is given by (4.3)
U(CP)
=
2
11 IT(l
COSACP)
-
which clearly is periodic in cp. The potential energy density, which has the characteristic shape shown on fig.45b, thus has an infinite series of minima:
,.'l'n =
(4.4)
ZTT
n
nT
.. ,-2,-1 ,0, 1 ,2, ..
Consider the equation of motion (4.5)
a all> )J
2
=
LSinA> A
In the weak field limit, where with
Icpl«
1, we can safely replace
SinAcp
A> whereby we recover the Klein-Gordon equation:
(3.31)
Now, because the equation of motion reduces to the Klein-Gordon equation in the weak field limit and because the exact equation involves
\
a sine-function it has become customary to refer to equation' (4.5)
as
the sine-Gordon equation. So that is the origin of the funny name for our mOdel.*)
*) According to Coleman the name was invented by Finkelstein. In Coleman [1975] there is a quotation from a letter from Finkelstein: »1 am sorry that I ever called it the sine-Gordon equation. It was a private joke between me and Julio Rubinstein, and I never used it in print. By the time he used it as the title of a paper he had earned his Ph.D. and was beyond the reach of justice."
129
1
Remark: This model has a famous analogue in cJassical mechanics which allow you to "visualize" the basic properties of the model: Imagine a fixed string on the x-axis. To this stri.ng we attach an infinite equidistant series of pendulums. These pendulums are only allowed to move perpendicular to the string, i.e. they can rotate in th~ y-z-plane. The position of a single pendulum at xl' = an is then completely characterized by its angle Ij>(xn,t) relative to the y-ax1S. (For convenience we imagine that the y-axis is pointing downwards).
Y-axis
Fig. 46
Finally we introduce a small coupling between neighbouring pendulums, e.g. by connecting them with small strings. We can then write down the energy for this system. A single pendulum contributes with a) Its kinetic energy 2[dlj>(X n ,t)]2
1
2mr
at
b) Its interaction energy coming from its coupling to the neighbouring pendulums
c) Its potential energy due to the gravitational field mgr[l - Coslj>(xn,t)] The total energy is therefore given by +00
H" n;...."
[]2 + ~k2[cjl(Xn+l't)-cjl(Xn't)] 2+ mgrll-Cos>(xn,t)]
~mr2 ~(Xn't)
Furthermore the dYnamics is controlled by the Lagrangian L" T-V
"E~mr2[~]
2
~k2[Ij>(Xn+l,t)-Ij>(xn,t)]
-
2
- mgr[l-Coslj>(xn,t)]
(Compare with the Lliscussi<Jn in section 3.1). We then pass to the continuum limit where a+O while at the same time m/a + rand ka + a . In this way we obtain the following Lagrangian L"
J::_oo(~pr2[~]2
-
~a[~]2
- pgr(l-Cos»]dx
We therefore See that for a suitable choice of parameters the infinite system of pendulums is equivalent to the classical field theory in (l+l)-dimensions characterized by the potential energy density (4.3).
D
Notice that in both of the above examples the potential U(cjl) possesses a discrete symmetry which transform one minimum into another. In the cjl4-model it is the reflection Ij> + - Ij> ,
130
I
while in the sine-Gordon model it is the translation
+
cP
+ 2rrA
The existence of such a discrete symmetry is a typical feature for the kind of models we are going to examine. We proceed to investigate various configurations in a non-linear field theory. In a classical theory we will always restrict ourselves to configurations with a finite total energy. Since the total energy is given by (4.6)
H
the assumption of finite
~gy
naturally leads to the boundary con-
ditions:
~
0 . It + 0 ; U(cp) + 0 as I xl+ 00 ' ax At infinity the field is consequently static and approach a constant value (4.7)
at
+
x
lim
-+~OQ
CP(x,t)
due to the first two conditions. The third condition states that the asynptotic value must in fact be one of the global minima for the potential, i.e.
Among the configurations with finite energy we have especially the vacuum configurations. The classical vacuum is characterized by having the lowest possible energy, i.e. it is characterized by a vanishing energy density. It follows from (4.6) that a classical vacuum must satisfy the equations: (4.8)
o
and
Thus a classical vacuum is represented by a constant field
=
CPo
CPo is one of the global minima for the potential, i.e.
Notice that the assumption about several global minima for the potential implies that there are several classical vacua. A non-trivial non-linear field theory is thus characterized by a degenerate classical vacuum! Next we consider small fluctuations around a c 1 aSSlca . 1 vacuum
+ n(x,t)
~ ~o:
131 In the weak field limit, where interaction of the field,
I
In(x,t) 1«1 , we can neglect the self-
i.e. we need only keep the quadratic terms
in the Lagrangian density. In terms of the shifted field thus get! L(n,a"n) ...
~
-
!a nalln 2 II
n(x,t)
we
!U'~. 0 )n 2
2
(3.32). In the quan-
But this is exactly a free field Lagrangian, cf.
tized version of the theory small oscillations around a classical va-
.0
cuum
consequently represent free particles with the mass-square
U"(.o). If the fluctuations are large we can of course no longer neglect the self-interaction, i.e. when the density of field quanta is high the field quanta no longer act as free particles. If the various classical vacua are all connected by a symmetry transformation then
U"(.o)
will be the same for all the classical
vacua, i.e. the field quanta associated with two different classical vacua will have the same mass. In the .'-model, e.g., the mass-square of the field quanta is given by (4.9)
while in the sine-Gordon model it is given by (4.10)
4,2
TOPOLOGICAL CHARGES As we have noticed finiteness of the energy implies that the asymp-
totic values
.+
=
x
- co
lim
~~(X)
.(x,t)
are independent of time. These asymptotic values furthermore correspond to classical vacua. Let
E
denote the space conSisting af all smooth finite energy con-
figurations. As a consequence of the degeneracy of the classical vacuum this space breaks up into different sectors characterized by the different possibilities for the asymptotic behavior. Consider, e.g. the
.4_
*
theory. Here the boundary conditions are given by
x
lim ++00
.(x,t)
= :!:
This gives four different possibilities for the asymptotic behavior: I III
_....l!..
·-00 =
If
-1!.
. - oc
= +If
; ;
.+- = .+co: =
-it
II
_Jl.
IV
If
cp-oo =
-.t
;
CP+CI> =
Jl.
;
cP+<»
• -'" = +If
=
+* +*
I
132 Thus
E
E++.
The vacuum solutions
breaks up into four sectors, which we label
E __ , E_+, E+_ and
$+ belong to the sectors E++ and E __ . These
sectors are consequently referred to as vacuum sectors. What can we say about the remaining non-trivial sectors? Consider a given smooth finite energy configuration $
• A smooth deformation of $
is a one-parameter family of finite energy configurations $A(X), such that $A (x) depends smoothly upon:\ and that $A (x) reduces to the given configuration $(x) when A=O. Clearly a smooth deformation cannot "break" the boundary conditions, i.e.
x
lim
-+±co
$A (x)
$±oo
is independent of A. A smooth deformation thus necessarily stays within a single sector. You may think of $:\ (x) of the given configuration $
as representing a pertubation
. It follows that a smooth pertubation of
a classical vacuum necessarily stays within the corresponding vacuum sector. On the contrary a configUration from a non-trivial sector can not be obtained by pertubing a classical vacuum. The non-trivial sectors are therefore called non-pertubative sectors. Notice that if the model possesses a discrete symmetry operation which relate the different classical vacua, then this symmetry operation will also relate the different sectors to each other. E.g. in the $4-model the inversion will interchange the vacuum sectors, i.e. it will map E
onto E++.
Similarly it interchanges the non-pertubative sectors E_+
and E+_.
Thus the structure of the two vacuum sectors and similarly the structure of the two non-pertubative sectors are completely equivalent. As another example we consider the sine-Gordon model. It possesses the translational symmetry p integer. The different sectors in the 'sine-Gordon model can be labelled by a pair of integers (n,m) such that a field configuration belonging to E
n,m
satisfies the boundary conditions lim
x
$(x,t) = 2;n
+-00
lim X
++00
$(x,t)
21Tm A
The translation operator then maps the sector En,m onto the sector En+p,m+p. In the sine-Gordon model there is therefore a discrete series of qualitative different types of sectors labelled by a single integer q=m-n (where
q=O
corresponds to the vacuum sectors) •
133
I
Next we look for conserved quantities in our non-linear models. Due to our boundary conditions the asymptotic values lim q,(x,t) x -+±oo are conserved, i.e. independent of time. Let us especially focus upon their difference (4.11)
Q
This can be interpreted as a conserved charge with the corresponding charge density given by p[q,]
(4.12)
Le. (4.13 )
Q
q, (+oo,t) - q, (-oo,t)
We will now compare this charge with the charge associated with the complex Klein-Gordon field, cf. the discussion in section 3.9. Exercise 4.2.1 Problem: Let q, = q,1 + iq,2 be a complex Klein-Gordon field. Show that the charge corresponding to the conserved current (3.67) is given by (4.14)
Q=
I
E
R3 ab
q,a~ld3x dt
The main difference between the stJ:ucture of (4.13) and (4.14) is that the ordinary charge (4.14) depends upon the time derivative of the field as well. Notice that the conservation of the ordinary charge presupposes that q,a obeys the equations of motion: Differentiating (4.14) with respect to time we get
the above integral can be rearranged as follows
But this can be converted into a surface integral at infinity. Provided the gradients vanish sufficiently rapid at infinity the surface integral vanishes and the the charge is conserved. The conservation of an ordinary charge is thus dependent not only upon the boundary conditions at infinity but also upon the dynamics of the field.
134
I
On the contrary the conservation of the new charge (4.13) is independent of dynamics. It only depends upon the boundary conditions.
EDr this
reason it is called a topological charge, and the corresponding conservation law is called a topological conservation law. Now, whenever we have a conserved charge we expect it to be associated a conserved current, i.e. a current which together
with the charge
density obeys the equation of continuity (1.17). In the present case the topological charge density (4.12) is associated the relativistic current (4.15) Notice that this current is trivially conserved:
a~ J~ =
E~va
~
a\! ~ =
a~av'
due to the symmetry of
0
I emphasize once again that the dynamics
of the field has not been evoked in the above derivation of the equation of continuity. Since the current (4.15)
is thus "automatically"
conserved it is called a topological current. Let us investigate the topological charge (4.13) a little closer. Notice first that it is conserved due to the requirement of finite energy, but that the structure of the potential energy density is irrelevant in this connection. Thus the topological charge (4.13) will be well-defined and conserved for any relativistic scalar field theory in (l+l)-dimensional space time. In an ordinary scalar field theory we will have a potential energy density
U[~J
with a unique minimum at
~=O.
But in that case the boun-
dary conditions (4.7) reduce to lim
x
-+±oo
~(x,t)
=0
Thus the topological charge is completely trivial in an ordinary scalar field theory, since it automa¢ically vanishes. As another example we may consider a model where the potential energy density is absent, i.e. it is based upon the Lagrangian density L
=-
~a cpa\Jcp ~
In that case the classical vacuum is characterized by
~
being an arbi-
trary constant. Thus the topological charge can take any value. Between these two extremes we have the non-linear models in which we are particularly interested. In these models we have a potential energy density with a discrete set of global minima. The topological charge can therefore only take a discrete set of values. In the
~'-model,
e.g.
135
I
the topological charge is on the form (4.16)
Q
+2J<. = O'-If'
while in the sine-Gordon model it is given by (4.17)
Q
211 = -r q,
q integer.
Since the topological charges in this kind of model can only take a discrete set of values we say that it is quantized. But notice that it is quantized already on a classical level! We speak about topological quantization. Observe that the topological charge vanishes for a classical vacuum.
Non-pertubative configurations thus correspond to the topo~ogica~ charge.
non-trivia~
va~ues
of
We end this section with a remark about the sine-Gordon model. In its mechanical analogue the classical vacuum corresponds to the configuration where all the "pendulums" are pointing downwards. In general finite energy implies that sufficiently far away the "pendulums" are static and point downwards. When we go along the x-axis from
-~
to
+~
the pendulums will rotate around the x-axis, but they will necessarily rointegra~
tate an
number of times.
Thus each finite energy configuration
Fig. 47
is characterized by a winding number. Clearly the winding number coincides, apart from a constant, with the topological charge in the sine-Gordon model.
4.3
SOLITARY WAVES We proceed to look for wave-like solutions. In a free field theory
we would naturally look for
p~ane
wave solutions, i.e. solutions on
the form ~(x,t)
=
ACos(kx - wt)
Although they have infinite energy themselves, we can construct superpositions of plane waves, which
represent solutions with finite energy.
(This corresponds simple to a Fourier analysis of such a solution). In a non-linear field theory such plane waves can never be exact solutions.
I
136
In stead we will look for solutions on the form (4.18)
~(x,t)
= f(x - vt) .
Such a solution will be called a travelling wave. Especially we will be interested in travelling waves with finite energy, i.e. if we introduce the variable, ; = x - vt then f must satisfy the boundary conditions (4.19)
lim
t; ...:± ...
f(O
As f is essentially constant sufficiently far away we say that such a wave is localized. A solution on the form (4.18) which satisfies the boundary conditions (4.19) is called a solitary wave. Let us write out the equation of motion (3.30) explicitly
a2cb - at2
+
a2 cb ax 2 =
U'[~]
•
Inserting the ansatz (4.18) this reduces to the ordinary differential equation U' [f]
If we multiply this equation with ranged as
df/dt;
on both sides it can be rear-
But that can be immediately integrated whereby we obtain
The integration constant must however vanish since both df dE; and U[f] vanish at infinity. Thus the equation of motion reduces to (4.20)
with
Y
= v'1=V2
This first order differential equation can be integrated too: (4.21)
This formula gives t; as a function of f, i.e. we have actually determined the inverse function. Notice that the boundary conditions are satisfied. Close to a zero, as follows
~o'
of the potential, we can expand the potential
U[f] ~ IU'~~ o ](f-~ 0 )2 Thereby the solution (4.21) reduces to
I
137 ~
Thus
goes to infinity when f approaches a classical vacuum. The
asymptotic behavior of f is therefore given by (4.22)
If(O-<J>ol
""
exp {
~ v1JIifQ
y(s-so)}
as
lsi'"
co
Clearly we get two types of solitary waves: The first type is in-
creasing,
(corresponding to the plus-sign in equation (4.20) ). It inter-
polates between two neighbouring classical vacua, and it is known as a
kink. The other one is decreasing and similarly interpolates between two adjacent classical vacua.It is known as an anti-kink.
:o..='~-----
X-axis
----...:-:....- - - -
- - - -
-<J>~:
anti-kink
- - - -X-a
~v
-----:-:-= - - - -
-----~----
Fig. 48 Okay, let us look at some specific examples: In the <J>'-model the kink is given by f
dt =V2 If Y( .~-~o) , '" _-{f2J ~---z-_t 2 :""Artanh[-f] ~ ~ oA i.e. (4.23) In the sine-Gordon model the kink is similarly given by f Y(s-~O)
=~f
dt 12 (1-C05>"1') TriA
1>.£
~Af
If
Iil
dt
J~(1-C052t)
!IT
=
~Jst!f ~Tr
i.e. (4.24)
We can also easily determine the energyof the kink. Inserting the ansatz (4.18) in the energy functional (4.6) this reduces to +co
J(~~2(~~)2+
~ing
5
----------~----~---=~~
U[fj)dt;
the equation of motion (4.20) this reduces to
138
<j>+co
+co
+co
I
y J /2U[f] df
(4.25)
<j> -co Thus the kink energy can be computed directly from the pctential. We need not know the explicit form of the kink at all! Once again we look at some specific examples: In the <j>"-model we
+]1/1X
get
2/1 1!...3
2
(4.26)
H[<j>kink] = yJ I~(~ - i')df -]1/ IX
Y-3- A
while in the sine-Gordon model we obtain 271 / A ,-,,.-_ __
(4.27)
H[<j>kink1
= yI
~r~(I-COSAf)
df
Y~
o Notice that the solitary waves constructed above interpclate between different vacua. Consequently they belong to the non-pertubative sectors Since they actually interpolate between neighbouring vacua, they carry the smallest possible non-trivial topological charge. The corresponding sectors are called the kink sea tors respectively the anti-kink seators.
4.4
GROUND STATES FOR THE NON-PERTUBATlVE SECTORS
Now that we have gained some familiarity with the solitary waves we will rederive them from a somewhat different point of view. As w~ have seen the space of all finite energy configurations, E, breaks ~p into disconnected sectors characterized by the different possibilities for the asymptotic behaviour of the finite energy configurations. In each sector we will now look for a ground state, i.e. a configuration <j>o(x) which has the lowest po~sible static energy in that sector. Since the ground state minimizes the static energy Hstatic[<j>] =
J~(~!)2
+ U[<j>(x)] dx
it must satisfy the associated Euler-Lagrange equation
Notice that the ground state extends to a static solution <j>(x,t)=<j>o(x) of the full field equations (3.30). Actually the following holds: The ground state for a given seator aorresponds to the solution
Of the field equations whiah has the lowest possible energy.
139
I
Proof: ~(x,t)
To see this let
be a non-static solution in the corresponding
sector. Then there is a space time point (xo,t ) such that o
~(Xo,to) f'
0
Consider then another solution specified by the initial data and It has the same asymptotic behaviour at infinity as to the same sector.
~.
Thus it belongs ~
Furthermore it has the same static energy as
but
it misses the kinetic energy! Consequently a non-static solution cannot minimize the energy in a given sector.
0
Okay, let us look for possible ground states. In a vacuum sector the ground state is simply given by the corresponding classical vacuum. But what about the non-pertubative sectors? To look for possible ground states in these sectors we will first examine the static solutions in our model (since a ground state is necessarily a static solution although the converse need not be true). The field equation for a static solution reduces to 2 d:::.....z " = U' 2 dx
(4.28 )
[~l
This can be integrated once (using the same trick as before), whereby we get
(~) 2 = 2U[~l dx
Thus we get two first order equations: ~ dx
(4.29)
•
<1>-2
v'zU[~l
•
<1>-1
.-
~- -v'zU[~l
dx -
<1>0 monotone
They have two kind of solutions:
solution
~. ~,
•
~-axis ~
~2:constant
solution
(a) On the one hand there are constant solutions, corresponding to the zeroes of U, i.e. the constant solutions reproduce the classical vacua. (b) On the other hand there are monotone solutions, which interpolate between two adjacent vacua:
(4.30)
x - Xo
=
~Jv2at
They correspond precisely to the static kink and the static anti-kink. This suggests that the kink is the groundstate in the kink sector!
140
I
Notice that there are no static solutions which pass through a classical vacuum. If the model possesses more than two classical vacua there will consequently be sectors (consisting of configurations which interpolate between non-neighbouring vacua) which has no ground state! Observe that if
~(x,t)
is any solution to the field equations, then
so is the Boosted configuration: ~(x,t)
(4.31)
=
~[y(x-vt) ;y(t+vx)]
due to the Lorentz-invariance. If we apply this to the static solution ~(x)
we thus produce a solitary wave: ~(x,t)
=
~[y(x-vt)]
We can therefore easily get back our solitary waves once we have determined the static solutions! Next we want to tackle the problem of whether the static kink really is a ground state configuration for the kink sector. This will be shown
using a beautiful trick going back to Bogomolny.
He showed that in many
interesting models the static energy can be decomposed as follows: (4.32)
P(~;~)
Here
is a first order differential operator acting on
~ while
the topological term only depends upon the asymptotic behaviour 01 the field at infinity, i.e. it is constant through out each sector. Provided the topological
term is positive we therefore get the following bound for
the static energy (4.33)
Hstatic
~
{TOPOlOgiCal term}
Furthermore this bound is only saturated provided
~
solves the first
order differential equation ; (4.34 ) Any solution to this first order differential equation is thus a ground
state. A decomposition of the type (4.32) is known as a Bogomolny decomposition. The associated first order differential equation (4.34) is known as the ground state equation. Let us try to apply this to our (l+l)-dimensional models. Guided by the kink equations (4.29) we try the following decomposition
141
I
The rest term is given by <11+ co
±I ¢-
/2U[<j>] d<j> co
Thus it is a topological term only dependent upon the asymptotic behaviour at infinity! Let us specifically look at a model with a finite or infinite number of classical vacua
where the different classical vacua are related by a symmetry transformation. In the sector En consisting of all configurations which interpolate between <j>o and <j>n' we can now apparently rearrange the above decomposition as follows: (4.35)
~f{~
Hstatic[<j>]
:;:
hU[<j>]}2 dx
Thus we have shown: (a)
(4.36)
The energy in the sector E
<j>1
Hstatic'"
is bounded below by
n
InII~
d<j>
<j>o (b)
A configuration in the sector En is a ground state i f and only if it satisfies the first order differential equation:
+: n positive { -: n negative
(4.29)
This shows especially that the kink is the ground state for the kink sector. It also shows that the energy of the static Kink is given by (4.37)
<j>1 H[<j>kink]
= IhU[<j>] d<j>
which of course is in agree~~nt with our earlier result (4.25). The Bogomolny decomposition can also be used to examine the higher charged sectors. Consider as a specific example the sine-Gordon model. We know in advance that there are no exact ground states in the higher charged sectors. We know also that the static energy is bounded below by Hstatic[<j>] ~ InIH[<j>kink] Now let X1,X2, ... ,X n be widely separated consecutive points on the xaxis and let furthermore <j>+(x) denote the static kink solution centered
142
at
X~O
(i.e. we put
v~so=O
I
in (4.24) ). It satisfies the boundary con-
dition s 27f
lim X
T
+-00
Consider now the superposition
~[ ](x) = ~+(X-Xl) + ~+(X-X2) + ... + ~+(x-xn) .n;x1""'Xn It clearly satisfies the boundary conditions lim X
~(x)
=
lim
0
x
+-00
$'(x)
=
2 n )..7f
++c:o
so that it belongs to the
Fi . 49
811/)..611/1.. 47f/)..
hi A
X-axis X2
x
x
Notice furthermore that the kink solution is strongly localized, since it approaches the classical vacuum exponentially, cf. (4.22). Outside a small region of width
l/~the
kink solution essentially reduces to the
classical vacuum. This has the following consequence for the abg.ve superposition: (a) Except when we are close to the centers configuration ~(x) cuum.
X1 ,X2, .•. ,X the n is exponentially close to a classical va-
(b) Close to a center, x k ' it behaves like a kink solution. Especially the difference
:i!
"'1 v'2U
[~]
is exponentially small. As a consequence the above configuration has a static energy given by (4.38)
Hstatic[~] = nH[~kinkl
+ {exponentiallY small error}
It is thus extremely close to the lower bound given by the Bogomolny decomposition.
furthermore we clearly have k=l, •.. ,n-l
Thus we can find approximate ground ted kinks.
sta~es
conSisting of widely separa-
143
I
Apart from a configuration consisting of widely separated kinks we may also consider strings of widely separated kinks and anti-kinks. Let us this time look at the ~4-model, and let us even specialize to one of the vacuum sectors, say E__ . As before centered at x=O, while
~_(x)
~+(x)
denotes the static kink
denotes the static anti-kink centered at
x=O. We now consider the following configuration in E __ :
~(x) = ~+(X-Xl) + ~_(X-X2) + ... + ~+(x-x2n_l) + ~_(x-x2n) Notice that this time we must let the kinks and anti-kinks alternate and furthermore, since we are in a vacuum sector, there must be an even number of centers.
]lIlA X-axis
Fig. 50
As before we see that 2U[¢J
+
{exponential small error}
When the kinks and anti-kinks are widely separated the configuration ¢(x) is thus almost a static solution. Since the static energy is thus not truly stationary under deformations we call such a configuration a ~ionary
configuration.
quasi-s~a
Although a quasi-stationary configuration need
not be an approximative ground state it will almost be a local minimum for the static energy. This is clear since any deformation of the quasistationary configuration (except for a pure displacement of the centers) will either destroy one of the claSSical vacua involved, or it will destroy one of the kinks/anti-kinks involved. Thus the static energy will increase (or at least stay constant).
3H 1
SINE-GORDON MODEL
o
vacuum vacuum
o
~~~~__~~~~~__~~__~-+~s~e~ctors
n=2
n=3
n=4
kink
sector sector
sector sector
Higher charged sectors
I
144
4. 5
SOLITOr~S
Observe that asymptotically the kink solution approach the Yukawapotential (3.34). But it is a solution of quite a different type! The Yukawa-potential (3.)4) is a singular solution to the Klein-Gordon equation and (just like the Coulomb field) it signals the presence of an external point source, i.e. a foreign particle. But the kink is a smooth solution everywhere, and consequently it is not associated with any external sources of the sine-Gordon field. Notice furthermore that they are localized solutions, where the energy is concentrated within a small region, i.e. they represent a small extended object. In accordance with this kink-solutions in non-linear field theories are generally interpreted as particles -an interpretation which in the case of the sine-Gordon model goes back to Perring and skyrme*). Since such particles are associated with solitary wave solutions they are referred to as solitons. We will now look at some of the properties of the kink-solutions which justify this interpretation: The kink-solution ~+(x-xo) represents a soliton at rest centered at x = Xo . In the same way the boosted solution (4.31) represents a soliton moving with the velocity v! We can also consider the energy and momentum of a single soliton. It is given by the total energy and momentum contained in the field, Le. by +OO plJ =
J
-00
where the energy-momentum tensor is given by TlJV
=
alJ~aV~ + nlJv[~a ~aa~ + U(~)]
a We know that plJ transforms as a Lorentz vector and in this example it is furthermore reasonable to localize it, since the energy is concentrated in a very small region! Thus we can attach the Lorentz vector to the center of the soliton.
For a soliton at rest the energy E is equal to the rest mass M M = C:TOOdX
=
C] ~(*)2
Soliton in (2+1)dimensional space-
time. Fig. 51
+ U[x] }dX
and the momentum P vanishes because the solution is statiC. Due to the Lorentz covariance of plJ, we therefore see that a soliton moving *) A model unified field equation, Nuel. Phys. 31, 1962 (550)
145
with velocity (4.39)
P
v
I
has the energy and momentum
= Myv
E
= My
(in units where
c = 1
The kink-solution therefore represents a free particle with rest mass M , while the anti-kink solution represents the anti-particle corresponding to the soliton. We now specialize to the sine-Gordon model. In the higher charged sectors, where the winding number is numerically greater than one, it is tempting to see if we can find solutions which represent several solitons. 'iliere are no static solutions. When several solitons are present, they consequently no longer behave like free particles. This means that two solitons exert forces on each other. This phenomenon is closely associated to the non-linearity of the sine-Gordon equation: The superposition of two kink-solutions is no longer a solution. We have seen, however, that if we choose the superposition carefully, then it is "almost" a solution. Worked exercise 4.5.1 Problem: (u) Consider the :;t.>'I,'t superpo"it.ion
= ~+[y(x+vt+xo)]
~(x,t)
211 + ~+[y(x-vt-xo)] - )l
Show that it satisfies T (~] - Sinh(~yx) g 4 - Cosh[~Y(vt+xo)]
(b) Show that it has the asymptotic expansion: e-~Yxo Sinh(~Yx]
(4.40)
Cosh(~Yvt]
t->«>
Based on computer simulations Perring and Skyrme guessed the following analytic expression for an exact time-dependent solution (Cf. the asymptotic expansion (4.40) :) (4.41)
T9(A~s4s]= vSinh(~yx]
Cosh[ llyvt]
Remark: It is definitely not "trivial" to verify that this is in fact a solution. In the end of this section we shall present a general method to solve the equations of motion whereby we can derive (4.41 ) "relatively" easy.
Exeraise 4.5.2 PrOblem: Show that the above solution (4.41) can be expanded asymptotically as 211 ~ss(t,x) ~+[Y(x+vt+xo)] +
(4.42a)
t+-«>
" (4. 42b)
~
ss (t,x)
t++<">
14.6
I
has a very simple According to exercise 4.5.2 the solution (4.41) interpretation: In the remote past it represents two far separated solitons moving towards each other and in the remote future it represents two far separated solitons moving away from each other. Thus it describes a scattering process between solitons!
Scattering between two solitons.
(Sketch based upon a computer simulation)
Fig. 52
Observe that the two solitons are identical, so it is not possible to say if the two solitons move through each other or if they are scattered elastically: All we can say is that there are two particles going into the collision center and two particles going out. Each of these particles carries an energy and momentum Pl~ and P2~ which can be calculated by integrating the energy momentum tensor over the central region occupied by the particle (Outside this region, the energy momentum tensors dies off expoX-axis nentially and we can safely neglect it). If you look back on Abrahams theorem (Sect. 1.6)' it is now clear that we have decomposed the total energy momentum as Fig. S3 p~ = Pl~ + P2~ where
Pl~ , P2~
themselves transform like Lorentz vectors. (Of course
this decomposition only works in the remote past and the remote future, where the solitons are far separated. During the collision it has no meaning to speak of individual solitons!). But the total energy-momentum stored in the field is conserved. Thus we get
147
I
Pl~(-~) + P2~(-~) = Pl~(+~) + P2~(+~)
(4.43)
i.e. the total energy and momentum of the solitons are conserved during the scattering. We can also compute the energy of the 2-soliton solution: This is most easily done using the asymptotiC expressions
(4.4~
Since they consist of two far separated solitons moving with velocity v
we simply get:
(4.44)
where
M[ 4>]
M
=
2~ly
is the rest mass of a single soliton.
Perring and Skyrme found two other interesting analytic solutions. They can be obtained from
(4.41)
using various "analytic continuations".
First we apply the "symmetry-transformation"
(x,t)
~
(it,ix)
and obtain
the singular complex valued solution:
Sinh[~-1-' Tg[AP(t,x)]= v 4
t]
(f:V2
COSh[~--V,I1_v 2
Then we perform the SUbstitution
)
x]
v ~1 v
whereby we obtain the regular
solution: (4.45)
-(t,x) Tg[ A4> S 5 ] = I Sinh["yvt] ,,_ 4 v Cosh[~yxl
Exercise 4.5.3
Problem: (a) Show that (4.45) represents a scattering process between a soliton and an anti-soliton by investigating the asymptotic form of the solution as t+-oo and t->+oo. (b) Show that its energy is the same as the 2-soliton solution: (4.46)
Performing the 'analytic continuation"
v
-+
(4.45)
iv
can further-
more be transformed into the periodic solution: with
(4.47)
r
I
,1l+v 2
This solution is strictly localized at all times within the region ixl < (~r)-l
since it falls off exponentially due to the denominator.
It has winding number
0, so it belongs to the vacuum sector. It re-
presents clearly a periodic oscillating configuration, called a bpeat-
148 her. The energy of the breather can also be obtained by "analytic conti-
I
Slowly moving breather.
nuation" 2M
Il+v2 Observe that it is less than the total energy of a far separated solitonanti-soliton pair. The breather is interpreted as a bound state of a soliton and an anti-soliton. For this reason the particle it represents is called a bion. (Bion is an abbreviation for bi-soliton bound state). When investigation the breather it is in fact more natural to introduce the cyclic frequency wand the "amplitude" n: (4.49)
W
= Ilrv
!
n
v
In terms of these variables the breather takes the form: Sin(wt) nCos h (nwx)
(4.50)
with
n
Notice that when t=nX (n integer) the breather momentarily degenerates to a classical vacuum. At these times the energy of the breather is thus purely kinetic. Exercise 4.5.4 Problem:
Show, by explicit calculation, that the energy of the breather at the time t~O is given by
w2
1
E = 2M(l - ~2)2
(4.51 )
where M is the soliton mass, cf. the earlier obtained result (4.48).
Now suppose that
w«
Il
•
Let us furthermore assume that Sin[wt) is positive, say get the following asymptotic expansions:
l~
eXP{Il(X + 1l-11nf2Bs:nwt)}
A<j>b Tg[-4- 1
""
2w
eXP{Il(X _ 1l_ 1ln [2US:TIWt)) }
1T
O
when x«O when x""O when X»0
149
I
On the other hand we get from (4.24) that
a
static kink, respecti-
vely anti-kink, is given by "<j>+ Tg[~]
(4.52)
When t
1T ~
is not too close to 0 or
we know by assumption that
IlSinwt w
is a large positive number. Consequently we can interprete the asymptotic behavior as fOllows: We have (a)
a soliton to the far left with the center x (t) ~ _1l-11n[2BSinwt] o w
(b)
an anti-soliton to the far rigth with the center xo(t) ~
(c)
1l-'1n[2Bs~nwt]
a classical vacuum ( <j> '" 2"Tf ) in the central region between the soliton and the anti-soliton.
This thus confirms our interpretation of the breather as a bound state of a soliton and an anti-soliton. The breather belongs to the vacuum sector. Thus it is not a nonpertubative configuration. Notice the following two extreme limits of the breather solution: (a)
When
w
~
0 the mass of the breather approaches twice the
soliton mass and we further get "<j>b t Tg [ =--!}l"T---.. 4 ] ~ Cosh[llx]
as
w
~
0
But this is the same as the limit of the soliton-anti-soliton solution
«4.45) when we let v
~
O.In this limit the soliton-anti-soliton pair
thus becomes unbound. (b)
When
w
~
11 the mass approaches 0, while <j>b
+
O. In this
limit the breather is thus just a small pertubation of the vacuum.
4.6
THE BACKLUND TRANSFORMATION The existence of exact multi-soliton solutions is a peculiar
property of the sine-Gordon modeL E.g. there are no exact multi-soliton solutions in the <j>4-model. Interestingly enough it turns out that there exists a systematic method for solving the sine-Gordon equation known as the inverse spectral transformation. Due to its complexity we shall however confine ourselves to a discussion of the Backlund transformation, which is a related powerfull technique that allows a systematic computation of
150
I
all the multi-soliton solutions. To simplify the analysis we introduce the so-called light-cone coordinates: x+ = ~(x+t)
(4.53)
x
= ~(x-t)
The derivatives with respect to the light-cone coordinates are given by
,• d-
=
1... ax - ata
In light-cone coordinates the sine-Gordon equation thus reduces to 2
X Sin[A
(4.54)
for simplicity we put A = ~ = 1 in the following. The crucial idea is to reduce the solution of this second-order differential equation to a pair of first order differential equations, cf. the philosophy behind the Bogomolny decomposition. In the last century Backlund discovered the following remarkable pair of first order differential equations: (4.55)
"[h::h] '" + 2
aSinr~] 2 L
Differentiating the first Backlund equation with respect to x- we get
tsing the second Backlund transformation the right hand side can be further reduced to (4.56)
"_"+[~]
=
COs[~]Sin[~]
Similarly we get from the second Backlund equation (4.57)
"+"- [~] 2
Consequently we get by adding and subtracting (4.56) and (4.57) (4.58)
"+"_
We have therefore shown: Any two functions
which satisfy the Backlund equations
(4.55) must necessarily solve the sine-Gordon equation too!
In other words: The sine-Gordon equation is the integrability condition for the Backlund equations. We can now reformulate the above observation in the following way: Suppose Then we respect we have
we have been given a solution
151 I given solution of the sine-Gordon equation into another solution. The operator Ba is known as the Back lund transfo1'mation (with the scale parameter a). As an example we apply the Backlund transformation to the classical vacuum. The Backlund transformed vacuum ~ solves the first order differential equation: a .. = 2aSin 2 a ~ = ~Sin2 +'"
-
2
a
2
Introducing the new variables f,;
=
ax+ + !x-
at;~
=
2Sin~
a
ax + - a1-x -
n
they reduce to
o
with the obvious solution (4.59)
Tg[2] 4
= exp[t;-f,; 0 ] =
exp[y(x-vt-s )] o
But that is precisely the kink-solution (4.24)! By applying the Backlund transformation to the classical vacuum we therefore create a single soliton. In principle we can now obtain the two-solution (4.41) by Backlund transforming the one-soliton solution, etc. Remarkable enough it turns out however that further integration of the Backlund equations can be reduced to pure algebra. This important observation is due to Bianchi, who showed that successive Backlund transformations commute and that furthermore the following non-linear superposition principle holds for ~o, ~l
Bal [~o], ~2
~3
Ba2 [~l]
=
= Ba
Ba2[~o]
and
[~2] 1
(4.60)
(this is known as Bianchi's permutability theorem). We leave the details in the proof as an exercise: worked exeraise 4.6.1 Introduction: Consider the following Backlund transforms of a given solution
~o:
~l=Bal[~o];~2=Ba2[~o];~~=Ba2[~l];~~~Bal[~2]
We are going to show that the integration constants leading from ~l to ~~ , respectively from ~2 to ~~' , can be chosen in such a way that ~l coincides with ~~' .
152
I
Problem: (a) Assume for a moment that the Backlund transformations B and B a2 mute and denote the Common value of ~l and ~l' by ~3' al Show that ~3 has to satisfy the consistency relation
COffi-
with
(4.61 )
(b) Return to the general situation where-the Backlund transformations need not apriori commute. From ~O'~I and ~2 we now construct a new function ~3 by imposing the condition (4.61). Show that ~3 is the Backlund transform of ~l with scale parameter a2, i.e. show that ~3 satisfies the Backlund eQuations:
(Similarly it follows that ~3 is the Backlund transform of ~2 with the scale parameter al' This shows precisely that the integration constants can be chosen such that ~l and ~l' coincides with the function ~3 defined by (4.61).) (c) Show that the consistency relation (4.61) is eQuivalent to the Bianchiidentity (4.60). In the Bianchi identity (4.60) we now substitute for vacuum and for
~l
and
~2
~o
the classical
two kink-solutions with different velocities.
We then obtain:
Tg[~]
(4.62)
1 +Tg [ ~ 2 ] Tg [ ~ I
]
a2+al exp[S2-S!']-exp[Sl-SO] a2-al 1+exptS2+S1-s!~s!1
For simplicity we further put a = al = - 1 (i.e. V2=-Vl) and sA = Sl' =0. a2 Then (4.62) precisely reduces to (4.41). This constitutes the promised derivation of the two-soliton solution. Remark: If one has patience enough one can work out 3-soliton solutions, 4-soliton solutions etc. In fact it is possible to write out the general N-soliton/anti-soliton solution explicitly. It is given by the following simple expression:
Cos~
(4.63)
=
1 -
2a\.la\.l~{Det[M]}
Mij
a.:a. 1
i j COSh[6 :6 ]
J
For sOliton/anti-SOli ton-solutions we put I-v. a~ 1
= __1 I+vi
6. 1
= ::y. (x-v.t-t;.) 1
1
1
Bions are included by allowing the a!s to be complex. In that case they must pairs of Hermitian conjugate numbers7
The formula for 8i
is then modified to
o~cur
in
153
4.7
I
DYNAMICAL STABILITY OF SOLITONS Let us now turn to the question of the stability of solitons. Ac-
tually this has already been solved, since we have shown that the static kink is the groundstate for the kink sector and as such it is clearly stable. However the argument was based upon the Bogomolny decomposition. It is thus essentially a topological argument, i.e. we have shown that the soliton is topologically stable. It is not always possible to give topological arguments for the stability of a static solution. It will therefore be useful to rederive the stability of the soliton from a conventional dynamical point of view. By analYSing the dynamical stability we will furthepmore be able to extract useful information about the particle spectrum in the model. Okay, we want to investigate the stability of a static solution ~o(x}. This is done by studying small fluctuations around the configuration:
~
(x,t) "" ~o (x) + n (x,t)
In(x,t) I «
1
Inserting this in the equation of motion we can expand the potential energy density to the lowest order U'[<j>(x,t») ""
U'[~o(x)l
+ U"[o(x»)n(x,t)
Thus the linearized equation of fluctuations becomes: (4.64)
This is not an ordinary Klein-Gordon equation because U "[~o (x) 1 is a function of x. Since it is not explicitly time-dependent we can,however, decompose n on normal modes: (4.65)
Inserting this we get
i.e. each mode is a solution to the eigenvalue problem: (4.66)
Notice that this is just an ordinary Schrodinger equation for a "particle moving in the potential W(x) = U"[o(x»)". It can be solved (in principle) and we should especially look for the eigenvalues. Observe especially that if there is a negative eigenvalue then the
154
I
corresponding cyclic frequency w must be purely imaginary. Thus the k "oscillation" exp[iwkt] in fact grows exponentially in the future (or in the past). Clearly this contradicts the stability, according to which a small fluctuation must stay small. Thus we have deduced the following criterium for dynamical stability A static solution ~o(x)
is dynamically stable provide1 the ei-
genvalue problem
(4.66)
{-
£..: + U'Ho(X)J}~k(X) ax2
Wk~k(X)
has no negative eigenvalues.
We can also derive this criterium by looking at the static energy functional. A stable static solution must at least be a local minimum of the static energy functional. Consider now a small pertubation: ~(x)
+ E:1jJ(x)
= ~o(x)
Expanding the potential energy density to second order we get
In the last two terms we perform a partial integration whereby we get
H[~ o J +E:Jr{_~o+ U'[~ 0 J}l/J(X)dX , dX2
+
!E:2Il/J(X){-~ +U"[~ 0 J}l/J(X)dX dX2
But here the middle term'vanishes due to the equation of motion, cf. (4.28). Thus the change in the static energy is of the second order in E: (4.67)
oH[l/J]
=
~E:2I
l/J(X){-
~
dX2
+ U'H (x) J ~l/J(X) dx 0
J
We must now demand that the quadratic form OH[l/JJ is positive (semi-) definite. If we solve the eigenvalue problem {- ::2
+
U"[~o(X)J}~k(X)
we can now decompose the fluctuation l/J(x) l/J(x)
=
~
=
wk 4>k(X)
on the eigenfunctions:
ak~k(x)
Inserting this the quadratic form (4.67) reduces to (4.68)
oH[l/J)
Thus we see again that a negative eigenvalue would be catastrophic, Since fluctuations along the corresponding mode ~k(x) would lower the static energy.
155
I
Okay, having motivated the eigenvalue problem (4.66) we now proceed to solve it. Notice that the "potential" W(x) = U"[<j>o(x)] is strongly localized and satisfies the boundary conditions: (4.69)
lim
= U"[<j>+co]
W(x)
x ~ut
here the right hand side reproduces precisely the mass-square of the field quantum, cf. the discussion in section 4.1. Making a trivial shift in the eigenvalue we can further reduce the eigenvalue problem (4.66} to the case where the "potential" vanishes at infinity.
ILLUSTRATIVE EXAMPLE: THE SPECTRUM FOR THE SCHRODINGER OPERATOR Let us look at foliowing equivalent problem: Find the eigenvalues for the second order differential operator
(4.70)
-
~
W(t)<j>
+
dt 2
=
A<j>(t)
(Notice that we have momentarily replaced the position variable x with the time-variable t.
where W(t) vanishes exponentially at infinity.
This is because we can now appeal to our intuitive knowledge of classical mechanics. This is a standard strategy in physics: When you want to discuss the qualitative behaviour of a solution to a differential equation you try to find a mechanical analogue.) If we rearrange the above differential equation as follows 2
d " ~
dt 2
= -A<j>(t)
+ W(t)<j>(t)
we recognize it as Newtons equation of motion for a particle of unit mass, which is under influence of partly a constant force, -A, and partlya time-dependent force W(t). Notice that the time-dependent force only acts for a very short time. For simplicity we assume in the following that W(t) is a negative function, i.e. it corresponds to an attractive force. When discussing the particle motion, x between two cases: (a)
Positive eigenvalues, A = w2 (i.e.
=
<j>(t), we now distuinguish the continuous spectrum).
When ), is positive the constant force is attractive, i.e. it drags the particle toward the center x = O. In the remote past it will therefore oscillate xCt) ~ a -<X'Coswt + b -ex:Sinwt t
+- ac
for a short time it comes also under the influence of the time dependent force W(t), but is then left again to the sole influence of the constant force, i.e. in the remote future it oscillates again
156
xCt)
~
a +00Coswt
t "'+0<
+
I
b +00Sinwt
Except for a change in amplitude and a phase shift it thus has the same qualitative behaviour in the past and the future.
X-axis
X-axis
t-axis
DISCRETE SPECTRUM
CONI'INUOUS SPECTRUM (b)
Fig. 55
Negative eigenvalues, A =_w 2 (i.e.
the discrete spectrum).
In this case the constant force is repulsive. We shall investigate the motion of a particle, which in the remote past is "almost" at rest at the origin, i.e. the particle motion is subject to the following asymptotic behaviour ~Ct)
exp[ wtl
~
t
-+-00
What will happen with this particle? Especially we will be interested in the following question:
lnder what circumstances does it return to
the origin in the remote future.
(This corresponds to an eigenstate) .
Okay, were it not for the attractive force W(t) the particle would just fly off to infinity. Thus for A less than the minimum of W(t) the total force is always a repulsive force and the particle flies off. When A passes the minimum of W(t) there will be a short time where the particle experiences an attractive force too. For a value Ao sufficiently high above the minimum
th~
attractive force will just be sufficient
to decelerate the particle, send it back to the origin, so'that when W(t) dies off, the particle has just the reverse of the "escape-velocity"
(i.e. it approaches the origin exponentially), cf. fig.55.
When
A is slightly higher than Ao the attractive force will become greater. hus the velocity obtained will be too great and the particle pass through the origin and escapes to infinity.
For an even higher value Al
the attractive force will now be great enough to decelerate the particle, send it back through the origin, decelerate it again and leave it with the reverse of the "escape-velocity", so that once again it approaches the origin exponentially, cf. fig. 55 . And that is how the game proceeds!
157
I
Equipped with the experiences from the mechanlcal analogue the following classical theorem
shoul~
not come as a great surprice:
Theorem 1 The spectrum of the second order differential operator
d2 dx 2
+
we x)
with x
lim+ Wex) = 0 -+
-00
consists of a continuous spectrum, A=k2, with positive eigenvalues extending from zero to infinity, and a finite discrete spectrum -
'" <
Aa
AI
<
<
A2
< ••.. <
An
< 0
The eigenfunctions of the continuous spectrum can be characterized by the following asymptotic behaviour (A=k 2 ):
exp[ikx] (4.71)
when
x«
0
{ aek)exp[ikx]+Sek)exp[-ikx] The eigenfunctions
when
x »
0
of the discrete spectrum can be cha-
racterized by the following asymptotic behaviour (AK=-K ' ):
(4.72)
_ {c_.ex)eXP[+KX]
when
x«
0
C+",ex)exp[-Kx]
when
x»
0
~
The ground state <poex) has no nodes, i.e. it vanishes nowhere. The lowest non-trivial eigenstate
This concludes our illustrative example.
a
Okay, let us return to the eigenvalue problem (4.66). To show the dynamical stability of the soliton we must show that if
~o=
dx'
U'[
(x)] 0
Differentiating this once with respect to x we get 3 d:::.....:t A. o= U"[
This we rearrange as
{ _ d' dx '
+ U"[
]}~o=
0
A comparison with (4.66) then immediately shows that function with the eigenvalue
d
A=O. Recall then that the kink, respectively
158
I
the anti-kink, are monotonous functions. Thus the assiciated zero-mode has a constant sign throughout the whole x-axis. But then the zero-mode d~+
/dx
has no nodes, i.e. it is precisely the ground state for the
eigenvalue problem (4.66): This shows that there are no negative eigenvalues. In the above argument the precense of a zero-mode ated with the finite-energy static solution
~o(x)
d~o/dx
associ-
is very central. Its
existence can be understood in the following way: The model has translational symmetry, i.e. the configurations ~o(x)
has the same static energy.
~o(x+a)
and
For a small
a
we can expand the displaced
configuration as
~o(x+a) ~ ~ (x)+a dPO (x+a) o
dala=O
Clearly dPOCx+a) da1a=O Thus we see that a pertubation along
d~o/dx
leaves the static energy
invariant. According to (4.68) it must then be a zero-mode, i.e. an eigenrnode with the eigenvalue O. To conclude: The zero-mode comes from translational symmetry.
Before we leave the stability considerations we want to mention that it is some times possible to prove the instability of a static solution by examining a particular simple deformation of the given configuration. The most important example is Derrick's scaling argument: Consider a scalar field theory in D space dimensions based upon the Lagrangian density L
= -~(d ~a) (d)J~a)
)J
_ U[~al
The corresponding static enerBY functional is given by =
~f (~) (~)dDX ddx
dX
+
fU[~aldDX
Hl[~al
+
H2[~al
d
R
R
Suppose ~a(x) is a static solution and consider the scaled configuration
a
~A (x)
= ~
a
(Ax)
The scaled configuration has the static energy
159 If ~a
I
is stable this must be stationary at A=l i.e. we must demand
o This implies the following stability condition (4.74 ) If
Hl
and
D>2 H2
the coefficients
(2-D)
and
D
has opposite signs. Since
are non-negative they must both vanish. Thus we conclude:
A non-trivial static solution in a scalar fie ld theory is unstable if the dimension of the space exceeds 2.
If
D=2
there is a small loop-hole. The condition (4.74) this
time reduces to H2[~1
=
0
and this can be fulfilled by a non-trivial configuration provided the potential energy term is absent. This is known as the exceptional case in two dimensions and we shall return to it in section 10.8. If
D=l
the condition (4.74) finally reduces to
i.e. a static configuration corresponding to a single scalar field can only be stable provided it satisfies the virial theorem; (4.75)
~J(~)2dX = JU[~(X)ldX
Interestingly enough the kink and the anti-kink in fact satisfies this identity point-wise, i.e. not only are the integrals identical but so are the integrands too! This follows immediately from (4.29). Moral: With a single exception Derrick's scaling argument shows that there are no stable solitons in higher dimensional scalar field theories. To stabilize the static solutions we must therefore extend the model somehow. This can, e.g., be done by coupling the scalar field to a gauge field. We shall return to this in section 8.7,8.8 and 10.9.
160
4.8
I
THE PARTICLE SPECTRUM IN NON-LINEAR FIELD THEORIES Finally we will give a brief (and naive) discussion of the par-
ticle spectrum in a
(l+l)-dimensional scalar field theory with a dege-
nerate vacuum. When we include quantum mechanical considerations we see that there are two basic types of elementary particles: (a) On the one hand there are the field quanta represented by small oscillations around a classical vacuum. A field quantum has the mass
m=~ o
(4.76) where
~
is a classical vacuum.
(b) On the other hand there are the solitons/anti-solitons, which c~~ssical
are present already on the
M
(4.77)
fV2U[~l
=
d~
~1
where
~l
and
~2
level. A soliton has the mass
are consecutive classical vacua.
(Strictly speaking
there are quantum mechanical corrections to this mass formula but we shall neglect them.) Apart from these elementary particles there may also be composite particles, Which can be interpreted as bound states of the elementary
particles, i.e. the field quanta and the solitons/anti-solitons. E.g. we have seen in the sine-<;ordon model that a soliton and an anti-soliton can form a bound state - the bion. Such a bion must then also be included in the particle spectrum. In the classical version its mass can be anywhere between zero and two soliton masses. Quantum mechanically the mass of the bion is quantized, i.e. it can only take a discrete set of values. We shall return to this in section 5.6. There is also the possibility of having bound states consisting of a field quantum bound to a soliton. Let us consider the interaction between field quanta and a single soliton in the weak field limit, i.e. we consider small "osci1lations" around a kink-solution ~(x,t)
=
~+(x)
+ n(x,t)
As we have seen the "oscillation" n(x,t) can be decomposed on normal modes ~(x,t)
=
~ ~k(x)exp[-iwktl
k
where
~k(x)
is an eigenstate for the eigenvalue problem {_ d_2
}
+ UII[~+(X) 1 ~k(x) = wk~k(x) 2 dx Let us introduce the "potential"
(4.66), i.e.
161
I
(4.78 ) and the "shifted" eigenvalue (4.79)
A
= w~
2
- m
Then the eigenvalue problem reduces to
~_ d L dx2 2
~ W(X)}~k(X)
A~k(x)
=
with
lim W(x) x
-+±oo
o
Thus we can take over the results from theorem 1: (a) There is a continuous spectrum consisting of non-negative eigenvalues
0<1,<00. They correspond to
frequencies w satisfying the inequak
lity Asymptotically the fluctuation is given by (4.80)
n(x,t) = ~k(x)exp[-iwkt] ~ exp{i[x/w~-m2 - wkt]} x«O
This corresponds to a free particle with momentum P and energy E given by P =
/w~
-
m2
and
Notice that E2- p2= m2 ,i.e. the rest mass of the particle is precisely m. Thus it is interpreted as a field quantum which is scattered on the soliton. When the field quantum is far away from the soliton it beha-
ves like a free particle, but when it is close to the soliton they interact in a complicated way and the approximation (4.80) correspondingly breaks down. (b)
There is also a finite discrete spectrum AO < Al < .••• < An < 0
As we have seen the groundstate AO corresponds to a zero-mode, i.e. AO
=
_m 2
•
The groumd state arises from the translational symmetry of the model. It represents no increase in the energy relative to the soliton, but simply reflects the fact that the
kink-solution,~+(x),is
is just one particular member of a continuous
degenerate, i.e. it
family,~+(x-Xo),of
kink-
solutions. The other discrete eigenstates, if they are actually present in the model, are more interesting. In that case the fluctuation dies off exponentially when we go away from the soliton. They are interpreted as bound states of a field quantum and a soliton. Notice that the total mass of the bound state lies between M and M+m. The question then arises if there are models with non-trivial discrete eigenstates.
162 Exer>cise 4.8. 1
Problem: (a)
I
Consider the ~!model. Show that the potential (4.78) is given by
(4.81 )
with
W(x) = - Cosh 2 [axl
(b)
a =
7z
Consider the sine-Gordon model. Show that the potential (4.78) given by
(4.82)
W(x)
with
Cosh2[axl
a
1S
II
Fran exercise 4.8.1 we see that in our most popular models, the ~' model and the sine-Gordon model, we must determine the discrete spectrum for the Eokhardt potential: w(
x
)
-
-
2 a s (s+l) - Cosh 2 [axl
ILLUSTRATIVE EXAMPLE: THE SPECTRUM FOR THE ECKHARDT POTENTIAL. The eigenvalue equation is given by (4.83)
-
~ dx
2
2
- a s(s+1) Cosh2 [axl
~
=
A~
By a suitable transformation we can eliminate the transcendental function in front of
~.
Consider first the substitution:
~(x) = Cosh-s[axlw(x) By a straigth forward computation we get the following equivalent equation for w (x) : d 2w
dw
- --- + 2asTanh[axldx - a 2 s 2 w(x)
(4.84 )
dx 2
=
AW(X)
Since the Eckhardt potential is an even function, i.e. it posseses the symmetry, x ~(x).
~
-x, we need only look for even and odd eigen functions
Furthermore Cosh[axl
is an even function, so we need also only
look for even and odd solutions to (4.84). Especially we need only solve equation (4.84) on the positiye semi-axis, i.e. for
O~x
The odd ei-
genfunctions then correspond to the boundary condition w(O) = 0 while the even eigenfunctions correspond to the boundary condition dw (0) = 0 dx We can now make the further variable substitution ~
=
Sinh2[axl
The eigenvalue equation (4.84) then reduces to (4.85)
163
I
The even solutions can be expanded in a power-series like
~a I;n n=O n while the odd solutions can be expanded in a power-series like (4.86a)
w+(I;) =
(4.86b)
w_ (I;)
=
1
00
I; 2 l: an <:n n=O
substituting (4.86a) into (4.85) we get the following recursion relation: 1 A n(n-s)+ij"(az+s 2 ) a n +l = (n+\) (n+l) A discrete eigenvalue occurs when this power-series terminates after a finite number of terms, i.e. when
4a 2 (n
Since
~+(x)
=
~
(l+I;)->;sw+(I;)
(l+I;)->;sl;n
I; -+ 00 we must furthermore demand 2n < s in order to get a norrnalisable wave function. In the odd case we get the recursion relation A (n+>;) (n+>;-s) + 1ij"(az+ s2)
Ie : (n
25 3 n
+ :2
+ :2)
This generates a polynomial (corresponding to an eigenfunction) when
Since
~
-
(x)
=
(1+1;)
-\sw
-
(1;)
we must furthermore demand 2 (n+l.;) < s The even and odd case may be comprised in the following way: The eigenvalues are given by (4.87) Ak = Even
K
-
a 2 (s_k)2
k integer, k<s
then correspond to even eigenfunctions, while odd k correspond
to odd eigengunctions. The associated eigenfunction is given by when k is even (4.88)
~k(X)
when k is odd
164
I
where
P denotes a polynomial of degree n, which is obtained from n the above recurrence relations. Notice that ~k(x) has precisely k nodes
0
in accordance with theorem 1:
Okay, we can then apply the results obtained in this example to the ~'-model and the sine-Gordon model. (a) In the ~4-model we get from (4.81) that s=2. Consequently there are precisely two discrete eigenstates. The corresponding eigenvalues are given by Al = - a' =
and
-
~ll'
.
According to (4.9) and (4.79) they correspond to the frequencies
o
i.e.
and
wf =
The first one is of course just the zero-mode, while the second one represents a genuine bound state! We can also determine the corresponding eigenfuction where
~1
(x). It is odd and corresponds to the case
w_(~) reduces to ~~. Thus ~, (x) is proportional to Sgn[x] (1+0 -l~~ =
Sinh[~x] Cosh' [~xl
(b) In the sine-Gordon model we get from (4.82) that s
=
1. Con-
sequently there is preciely one discrete eigenstate (corresponding of course to the zero-mode). In the sine-Gordon model there is consequently no bound state of a field quantum and the soliton.
SOLUTIONS OF WORKED EXERICSES: No •. 4.5.1(a)
Using the trigonometric identity
Tg(a+s-i) ; Tg(a)Tg(S)-l Tg(a)+TgtSJ and the expression for the static kink solution A~+
Tg(~)
exp(llx)
exp(2\lYx) - 1 exp[lly(x+vt+Xo )l+exp[lly(x-vt-x0 )1 (b)
To find the asymptotic behaviour as
lim Cosh(x+a) Coshx X ~ a:
exp(a)
i.e.
t~~
we use that
Cosh(x+a)
RI
x
~
ea.cosh(x) '"
o
165
I
No.4 .6.1 (a)
If we introduce the notation
a
exp(,,)
=
we can write down the Backl¥Ud equations in the following condenced form:
(4.89a)
d+(,/ll+<jlo)
2exp[::"11Sin[¥1 + 2exp[::"21Sin[~1
(4.89b) d+(
+
(4.89cJ
2exp[::"21Sin[P2; PO l + 2exp[::"11Sin[
d+(
(4.89d) d+(
By adding and subtracting the first two we get +
d+(
=
+
::Zexp[::"11Sin[Pl;P01+ Zexp[::"21Sin[~1
Similarly we obtain from the last two +
- ) d+ (
=
+
+ + P1.:.P.1 -Zexp [+-~2 1Sin [P.d.! Z 1+ Zexp[-"llSin[ 2 1
AS the left hand sides coincide we therefore deduce tpe consistency relation
CXP[::"11{Sin[~1~Sin[Pl;P01}
eXP[::"21{Sin[Pi;Pll:;Sin[~1}
==
Using the well-known trigonometric identity +
Sinx :: Siny
=
-
ZSin[~lCos[~l
the consistency relation reduces to
Using the condenced notation introduced in the exercise this can be written as
(4.61)
exp[::"llSin[A+l == exp[::"21Sin[A_l +
(b)
Differentiating the consistency relation (4.61) we get
exp[::"llCos[A+ld+A+
=
exp[::"21Cos[A_ld_A_ +
Writing out
+ +
A+ this leads to
{eXP[::"llCoS[A::l- eXP[::"21COS[A+l}d::
eXP[::"21COS[A~1}d;:
:; {eXP[::"llCOS[A::l- eXP[::"21COS[A:;1}d::
166
{exp[!AllCOS[A!l-
(*)
eXP[!A21COS[A;1}a!~3==
!{eXP[!AllCOS[A!1-eXP[!A21COS[A;1}a!~1 •
+
+ -4exp [+-AI 1exp [+-AzlCos [A_1Sin[ ~ 2 1 +
;2{eXP[~AllCOS[A!1+ eXP[~A21COS[A;1}eXP[~AzlSin[PZ;POl The first term on the right hand side is just what we want, but the remaining two terms need a little massage! Introducing. the abbreviation
B ; (p3+PO)~(pZ+Pl) 4
+
we have
+
!B ; A + +
~; ~B+ +A 2 +
+
P3-~1
2
; B
+
A
+-
+
If we introduce these abbreviations the last terns in (*) reduces to
exp[:AlleXP[~A21{2COS[A;lSin[B!lCOS[A~1+ 2COs[A+1Sin[A;lCos[B+l ;
-4COS[A_1Sill[A+1COS[B+l}
'- - - - - - - - - - - - - - - ~- - j
+
-
-
exp[~A21eXP[~A21{2COS[A_1Sin[A_1COS[B+1-2COS[A_1Sin[B+1COS[A_l}
+ + + + Using the consistency relation (4.61) at the combination indicated by the broken line this is further reduced to:
2exp[~Allexp[~A21rCoslA 1. . -+ lSin[B+- lCos[A+- 1+ CostA+- lSin[A+- lCos[B+- l} - 2eXP[~AzleXP[~A21{COS[A_1Sin[A_1CoS[B+l+ COS[A_1Sin[B+1COS[A_l} +
+
-
+
+
-
2eXP[~A21{exp[!AllCOS[A!1- eXP[~A21COS[A+l}{Sin[B!lCOS[A+l+Sin[4;lCoS[B~1==
2eXP[!Azl{eXP[~AllCOS[A!1- eXP[~AzlCOS[A+l}Sin[P3;~11 Inserting this back into (*) this fin~lY reduces to
a~~3 ; !a~~l + 2exp[~A21Sin[P3;~11 (c)
Writing out the consistency refation (4.61) explicitly as
aISin[(p3-PO)~(pZ-PI)1
azSin[(P3-PO)~(Pz-PI)1
it can be rearranged as
aISin[~lCos[~l + aICos[~lSin[PZ~Pll
== a Sin[~lCos[PZ-~ll - a Cos[~lSin[~l 4 4 4 4 This immediately leads to
(az+al)Sin[~lCos[~l == (az-al)Sin[~lCoS[~l from which the Bianchi identity (4.60) follows trivially.
0
167
I
ehapler 5 PATSINT.BGRAI,S AND INSTANrONS 5,1
THE FEYNMAN-PROPAGATOR REVISITED In the remaining chapter we would like to include a few aspects
of the quantum theory of fields and particles.
To simplify the dis-
cussion we begin our considerations with quantum mechanics of a single particle in one space dimension. In ordinary quantum mechanics the central concept is the Feynmanpropagator
K(X lt lx ;ta ). As we have seen it denotes the probb b a ability amplitude for a particle to move, i.e. to propagate, from the space-time pOint (xa,t a ) to the space-time point (xb,t ). b The propagator governs the dynamical evolution of the Schr6dinger
wave function according to the rule (2.18)
W(xb;t b )
=
JK(Xb;t b I xa;ta)W(xaita)dxa
and as a consequence all information about the quantum behavior of the particle is stored in the propagator.
Furthermore it satisfies
the group property (2.25)
K(xc;tclxa;t a )
= JK(Xc;tclxb;tb)K(Xb;tbIXa;ta)dXb
It will be useful to recall what we have learned so far about the propagator.
For simplicity we consider a physical system character-
ized by the Lagrangian dx 2
tm«(lt)
- V(x)
Then we have previously shown that
1)
The propagator can be
~ep~esented
as a
x(tt)=2b =
(2.21)
J
path-integra~
t
dx 2-v ( x) 1dt } D[x(t) exp{fiifbm [Z(iIT) t
a
where we sum over all paths oonneoting the spaoe-time points (xa;t a )
2)
and
(xbitb).
The propagator is the unique
so~ution
to the Sohrodinger equation,
1
I
satisfying the initial oondition o(x b - xa)
K(xb;talxa;ta)
(Cf. the disoussion in seotion 2.4.
In the mathematioal terminology the
propagator is thus a Greens function for the Schrodinger equation). ;;)
The propagator has the following Moomposition on a complete set of eigenfUnotions for the Hamiltonian operator:
(2.70)
1< (xb;t b Ixa;t a ) =
I
n=o
~
En (t b - tal 1
cf. the worked exercise 2.11.;;. Of these characterizations only the last two have an unambigious meaning.
The first one, involving the path-integral, is only a for-
mal description since we have not yet indicated how one should actually perform such a summation over paths! Before we discuss the properties of the propagator further we will indicate yet another characterization of the propagator. that in quantum mechanics a physical quantity
T
a Hermitian operator eigenvalues
An
T
Recall
is represented by
and that this operator can only have real
which represent the possible values of the quantity
T, when we try to measure it in an actual experiment.
In general a
wave function can be decomposed on a complete set of eigenfunctions for the operator
The coefficient
T: an
is then interpreted as the probability An' i.e. in an actual experlinent we will
amplitude for measuring the value measure the value
An with the
E.g., the position
Xo
~ility
Pn ~ 1~12. is represented by the multiplication
operator (5.1)
The corresponding eigenfunctions are given by (5.2)
Wxo (x)
=
o(x-x o )
Notice that these eigenfunctions are not normalizable.
This is re-
lated to the fact that the position operator has a continuous spectrum, i.e. a continuum of eigenvalues.
The eigenfunctions, however,
169
I
still satisfy the completeness relation (2.69):
I
WXo (x 1 ) WXo (x 2 )dx o = JO(XO-X1)O(XO-X2)dXo = o(x 1 -x 2 ) Similar the momentum
is represented by the differential
Po
operator (5.3)
PoW (x)
= - Hi.
1:£. ax
The corresponding eigenfunctions are given by the plane waves Wk (x) = __1__ exp[ik x]
(5.4)
nrr
o
with
0
p
o = ilk 0
(Notice that we parametrize the momentum eigenfunctions by number
t~e
wave
ko' rather than the momentum
Po = ilk )' The momentum o eigenfunctions are not normalizable, but, like the position eigenfunctions, they satisfy the completeness relation (2.69):
Here we have used that the Fourier transform of Dirac's delta function is simply 1, i.e.
Jo(X)eXP[iYX]dX
= exp[iyx] lx=o
By Fourier's inversion formula we thus get
°
(5.5)
(xl
= 2111
f
exp[ -iyx] dy
Exercise 5.1.1 Problem: Compute the free-particle propagator by means of the formula (2.70), which in this case reduces to -+2
+00
K(xb,tblxa;ta)
where
~k
(x)
=
f ~k\Xa/Wk(xb) :;---r::--"T"
exp[-
i
E
E-.
2m (tb-ta']dk
is given by (5.4).
Consider now the time-evoZution operator
i exp [ - E HT].
It has
the matrix element
Ix > is the position eigenfunction (5.2). Using a complete a set of eigenfunctions for the Hamiltonian H we can now expand the
where
position eigenstate Ix a > =
If
I
n
<wnlxa>lw n >
we furthermore use that
170
I
then the matrix element reduces to i
-
L -Wn(x a )
<xblexp[- fi HTl IXa> =
i
<xb1wn> exp[- fi En T ]
n
I
Wn(xalWn(xb) exp[-
n
%En T ]
According to the third characterization we have thus shown:
4)
The propagator can be represented as a matrix element of the time evolution operator exp[-
K(~;Tlxa;O)
(5.6)
%
BTl , i.e.
<~Iexp[- ~
=
Hrllxa>
As an especially important example of how one can extract infor-
mation from the propagator, we will now show how one can in principle calculate the energy spectrum of a particle.
For this purpose
we consider the trace of the time-evolution operator (5.7)
G(T)
Tr{exp[-
~
HT]}
= I<xo1exp[-
i
HT]lx o > dx o
jK(Xo;T1Xo;0) dx o Using the third characterzation of the propagator this reduces to (5.8)
where we have used the normalization of energy eigenfunctions
~n(x),
Thus the trace of the propagator decomposes very simple in terms of
harmonic phase factors on ly depending upon the energy leve 1-s! Introducing a "Fourier transform" (5.9)
G(E) =
%J
G(T)exp[i ET]dT
o we in fact immediately get: (5.10)
G(E)
i Y J n n=o
exp[-
~(En-E)TldT
o
Consequently the energy levels show up as poles in the transformed
trace of the propagator.
171
I
In the above discussion we have been focusing upon propagation in the "position space". We could equally as well use other physical properties, such as the momentum, as our starting point. Thus we define K(Fb;tblp;t) as the probability amplitude for a particle, which is released with moment~ a p at the time t . . l ) denotes a to be observed w1th momentum Pb at the t1me t . Thus K( Pb;t ap;t b the probability amplitude for a particle to "proBagate in momentum ~pa~e" from the point (p;t) to the point (pt,;t ). According to the basic rules of quantum b mechanicsath~ momentum propagator is now related to the position propagator as follows K(pt,;tbl Pa;t a ) = H
Here the various terms can be
a) <x Ip) is the probability amplitude for a particle with momentum Pa to be at theapo~ition x. b) K(~;tblxa;ta)a is the p,obability amplitude for a particle to propagate from xa to x • b c)
The momentum propagator is thus obtained from the position propagator by a "double" Fourier transformation. Consider e.g. the free-particle propagator 2 I m im (x b -x a ) (2.28) VZTI1b(t b -t a ) exp[zt t b -t a ] In momentum space this is converted to K(ptiTIPaiO)
r-m- 1 II i m (~-Xa) = V2:ilir 2rr exp{K['2 --r-
2
-
This is just a Gaussian integral, cf. (2.27), which can be calculated in the usual way by "completing the square". Using the identity, m
(~-Xa)2
'2 - r
- ptxb
m
+ Paxa =
'2'I' [xa
the x -integration cancels the factor formu1a: 2 K(Pb;T IPa;O) = exp[-
i Pa K 2m
1
T] 2rr
T nor.. 2 + iii(Pa - ""r')] -
;f2TI~~T
J exp~(Pa i
P~
2iii T + (P a -Pt)~ ,
in front and we are left with the - Pb)xb]d~
But according to (5.5) the last integration just produces a a-function! fore finally get: 2 Pb - P a i Pa (5.12) a(--~-) exp[- fi' 2m T]
We there-
This simple result has an intuitive explanation: Since the momentum of a free particle is conserved, it follows that once we release a free particle with
172 I momentum p it will necessarily still have momentum p once we decide to measure it. a This explains the a-factor. Furthermore a a free particle with momentum p will have the definite energy E = p2/2m. Thus the propagator has the charact~ristic phase factor indicated by the d~ Broglie rule, cf. (2.32), exp[-
i 1i
ETl
= exp[- ifl
2
Pa Tl 2m
Evidently the free-particle propagator is much simpler in momentum space. This is a general feature of quantum theory and therefore it is the momentum propagator which is most often displayed in the literature.
5.2
ILLUSTRATIVE EXAMPLE: THE HARMONIC OSCILLATOR As an example of how to use the preceding techniques we now take
a closer look at the harmonic oscillator, which is characterized by the Lagrangian: (5.13 )
The corresponding equation of motion is given by 2 d x 2 dt Notice that
w
2 w x
is the cyclic frequency of the oscillator.
First we must calculate the propagator.
Sinde the Lagrangian is
quadratic, we can use the same trick as for the free-particle propagator, cf. the worked exercise 2.4.1.
The propagator consequently
reduces to the form,
A(T)exp[~
(5.14 )
t
f L(xcll
dx
d~l)dtl
o where
x
= xcI
(t) is the classical path connecting
(xa;O)
and
(xb;T) • To determine the classical action we notice that the general solution of the equation of motion is given by xcl(t)
=a
Coswt + b Sinwt
we shall then adjust the parameters
a
and
b
so that the boundary
conditions
= xa are satisfied.
If
wT
~
nIT
and this is easily obtained and we get the
unique solution (5.15)
a
and
b
Xb - xa CoswT Sinwt
173 If on the contrary
wT = nrr
I
(corresponding to either a half period
or a full period) we are in trouble.
This is because a classical
particle, after a half-period, necessarily is in the opposite pOint. After a full period it is similarly necessarily in the same point. Classical trajectories for the harmonic oscillator.
X-axis
wT-axis
Fig. 56
For
wT = nrr
we can therefore only find a classical path if
xb = (-l)n Xa , and, if that is the case, any classical path passing through (xa;O) will in fact also pass through (Xb;T)! In analogy with optics we say that the classical path has a caustic nrr.
when
wT =
At the caustics the propagator thus becomes singular! Okay, neglecting the caustics for a moment we can complete the
calculation. The classical action is given by S[X cl ]
mw =:r
2 2. 2 [(b -a )CoswT S~nwT - 2ab Sin wT]
and substiting the values (5.15) for
a
and
b
we now get after a
lengthy but trivial calculation: (5.16 ) As expected this is highly singular at the caustics, i.e. when
wT
nrr. We proceed to calculate the amplitude
A(T).
It can be deter-
mined by the same trick as the one we used for the free-particle propagator, i.e. by using the group property (2.25). invariance in time the group property reduces to
Substituting the expression (5.14) we now get
By translational
I
174
This integral looks pretty messy, but after all it is just a harmless Gaussian integral. Using the identity
3Z
imw (X +X )C05WT Z - ZX3xZ ~[ SinwT + Z imw Sinw(T Z+T,) zh SinwTZSinwT, [xZ we can immediately integrate out the x -dependence whereby the right 2 hand side reduces to
Most reassuring the x 3 ,x 1 -dependence is the same as on the left hand side and we can therefore factor out the phase factor completely. We then obtain the following functional equation for the amplitude
A(TZ+T,) A(TZ)A(T,)
/
ZninSinwTZSinwT, mwSinw(TZ+T,)
A(T):
mw
It has the general solution
but as in the case of the free-particle propagator we shall neglect the additional phase
a=O.
exp[iaT], i.e. put
(This corresponds to
a normalization of the energy,and has no physical consequences). Notice that, in the limit where
w
+
0, the function
A(T)
reduces
to the free-particle amplitude A(T)
= J2nlhT
as we expect it'to do! Consequently we have now determined completely the propagator for
the harmonic oscillator. (5.17)
K(XbIT!Xa,O)
In its full glory it is given by
=.; 2TIif:WSinwT e:xp[
2fi ~l~wT{
(X; +
x!)QOSWT - 2X b Xa }]
I
175
Notice too that it also reduces to the free-particle propagator in the limit of small
T.
Expanding the phase to the next lowest order
we in fact get im til 2 2 } exp[2fl SinwT {(X b + xa)COStilT - 2xb x a ]
...
2 2 2 T im 2 imtil exp[ 2fiT (x b - xa) ]exp[- ~ {Xb + xa + xbx a } 3] Here the first phase factor is the free-particle phase factor, while the second phase factor reproduces the interaction phase factor, T
t JV(x)dt]
exp[ -
o integrated along the free-particle path, i.e. the straight line x
= xa
+
(xb-Xa)t/T.
Thus the "infinitesimal propagator" is in
accordance with our previous ansatz (2.53), which we used in our derivation of the Schrodinger equation. included the phase factor
exp[iaT]
Observe also that if we had
this would no longer be correct.
To be honest, the above formula,
(5.17), for the propagator of
the harmonic oscillator is not quite correct.
This is due to the
fact that we have been rather careless when performing Gaussian integrals of the type +""
(5.18)
J e- iax2
dx =
~
The above formula is obtained from the rigorous formula, (2.27)
[
A
by performing an analytic continuation.
>
0
But this analytic continua-
tion is actually double-valued, since it involves the square root of a complex number.
This gives rise to a phase ambiguity in the Gauss-
ian integral (5.18).
As a consequence the propagator has a phase
ambiguity (corresponding to a phase
~1
or
~i).
Below the first caust-
ic, i.e. when T < ~ , the formula (5.17) is correct simply because til it is in correspondance with the free-particle propagator in either of the limits
til
+
a
or
T
+
O.
But once we have passed the first
caustic, we no longer control the phase ambiguity. to this problem in the end of this section.
We shall return
I
- 176
Now that we have the Feynman-propagator at our disposal, we can easily calculate the energy spectrum.
The trace of the propagator
is given by G(T)
V2(C05WT - 1)
2iSin!wT
But this is easily rearranged as (5.19)
G(T)
- e -iwT
~ e-i(n+!)wT n=Q
A comparison with (5.8) then immediately permits us to read off the energy spectrum (5.20) In the case of the harmonic oscillator the possible values of the energy are thus evenly spaced and, perhaps a little surprising, the ground state energy is not zero as in the classical case, but is instead given by (5.21) The zero-point energy is often explained by referring to Heisenberg's uncertainty principZe, according to which the indeterminacies of the momentum and the position are bounded below by (5.22)
<x>
~
fi
Unlike the classical case, the particle therefoce cannot be at rest in the bottom of the potential well, since this would cause both <x> and
E
2
=~ 2m
+ tmw2<x>2
We want to minimize this subject to the constraint (5.22). Replacing
~/<x>
by <x>
Except for the factor !, which we cannot account for by such a primitive argument, this is the same as the zero-point energy (5.21).
Once we have the propagator at our disposal we also control the dynamical evolution of the Schrodinger wave function.
As in the
free-particle case we can now find an exact and especially simple solution to the Schrodinger equation. the normalized wave function
At the time
t=O
we consider
177
(5.23)
1jJ(x,O)
=
'rmw
y'~
I
mw 2 exp[ - 2fl(x-a) ]
This corresponds to a Gaussian probability distribution centered at the point
x=a.
At a later time
t
the wave function will, accord-
ing to (2.18) evolve into IK(y;tIX;O)1jJ(X,O)dx
1jJty,t)
4/iii;; 2/ lItJJ V ~ V21TlflSinwt
=
I
illtJJ 2 2 exp[2li.Sinwt[(y +X )Coswt - 2yx} -
2 2li.(x-a)] dx
IlliJ
As usual this is just a Gaussian integral. Using the identity
illtJJ 2 2 2fISinwt [(y +x )Coswt - 2yxJ -
lItJJ 2 m(x-a)
lItJJie iwt . t 2 2 1W (aSinwt - iy)] - ~[y 2li.Sinwt [x - iethe integration can be immediately performed, and the wave function at the time t reduces to
(5.24)
~(y,t)
· · WK exp[- ~(y-aCoswt) 2 ]exp[- l~t]exp{_ ~[2aySinwt -
~
=
V
2
~ Sin2wt]}
Apart from a complicated phase factor this is of the same form as (5.23).
Consequently the corresponding probability distribution,
P(x,t)
11jJ(x,t) 12 =
/§I
exp[ -
~w
(x - a Coswt)2]
is still a Gaussian distribution, but this time it is centered at x = a Coswt This is highly interesting because this means that the wave packet
oscillates forth and back following exactly the same path as the classical particle! If
a=O
the probability distribution reduces to the stationary
distribution
corresponding to a classical particle sitting in the bottom of the potential well. state.
function for (5.25)
Thus it is closely related to the classical ground
It should then not come as a great surprise that the wave a=O, i.e.
178
I
represents the quantum mechanical ground state. the simple time dependence
exp[- ~ Eotl. with
equal to the ground state energy
!hw.
Notice that it has Eo
precisely being
It must therefore in fact
be the eigenfunction of the Hamiltonian with the eigenvalue
!fiw.
Exermse 5.2.1 Introduction: Consider a free particle, where the corresponding wave fUnction at the time t=O is given by ~ x2 1jJ(x,O) = ~2 exp[-1j"(?"l Problem: a) Show that its wave fUnction at a later time t is given by
(Unlike a particle which is at »rest« in the bottom of a potential well, this wave packet thus spreads out in time!) b) Show that the corresponding probability amplitude in 'momenttun space« is given by
i2 an
~(k,t) =
2(t)
HIt 2 2 exp[ - k a (t)l exp[8ma2cr2(t)lexp[
and as a consequence
'!\ "2
Also the rema~nlng eigenfUnctions can easily be extracted from the propagator. With the third characterization of the propagator in mind we decompose it as follows K(y;Tlx;O) = e- iwT / 2
mw 2' T exp{ mw 2'wT [~(y2+x2)(1+e-2iwT)_ 2xye- iwT l} 11fl(1-e- lW ) 11(1-e- 1 )
vi
If we introduce the variable K(y;Tlx;O)
z=e-
iwT
this can be rewritten as follows
r-mw= z-~ exp[- ~(X2+y2)l~ exp{ J..
(2
2) 2
_ ~[ x +Y1_~2-
This should be compared with
,'"'
E ~~ (y) exp[_i E Tl = z-· E ~ (x)~ (y)zn
n=O n
n
fJ. n
n=O jn
n
If we put
~n(x)
= ~ exp[-
~x2J(2nn! )-!Hn(~)
and introduce the rescaled variables
u=~ it follows that (5.26)
and
Hn(u)Hn(v)
v=~, has the generating formula ( 1 -z 2
,
)-.
exp
2 [uvz - (u2 +v 2) Z 2 1 1_Z2
2
xyzl}
179
I
Since the left hand side is a Tay~or series in z we can actually find 2-n H (u)H (v) by differentiating the right hand side n times and thereafter put z o. I~ fol~ows trivially that H (u) is a polynomial of degree n in u. In fact it is a Hermite polynomial and thg above formula is the so-called Mehler's formula. We leave the details as an exercise:
Worked exercise 5.2.2 Introduction: mula
The Hermite polynomials Hn(x)
z
are defined by the generating for-
n
n=o Problem:
a) Deduce Rodriques' formula 2 d n _x2 Hn(x) = (_On eX e n dx b) Show that e
-x
2
1
liT f
_t,2
e
+ 2it,x dt,
and consequently H (x) = (_2i)n e x2 n
liT
Jt,n e-t,2 + 2iE,x dE,
c) Prove Mehler's formula
(5.28)
L
n=o
=
[l-z2j
-! exp [ 2xyz -
(2
2) z2
x 2+ Y - z
Using Mehler's formula we can in fact recalculate the propagator. must then determine the eigenvalues En and the eigenfunctions ~n(x) Hamiltonian: 2 2 h d 2 2 (- + jlllL1l X )$ (x) = E ~ (x) 2m dx2 n n n
First we for the
(This eigenvalue problem is treated in any standard text on quantum mechanics!). Once we have the eigenvalues and the eigenfunctions at our disposal we can finally explicitly perform the summation in (2.70) by means of Mehler's formula. Notice that the phase ambiguity is still present. The Taylor series in Mehler's formula has convergence radius 1, due t~.&¥e poles at z = ±1. We are especially interested in the behaviour at z = e ~ • We are thus actually working directly on the boundary of the convergence domain! This boundary, i.e. the circle Izl= 1, is decomposed into two disjoint arcs by the poles z = ± lJ Furthermore the poles correspond precisely to the caustics wT = nTI. At two times t and t , h separated by a caustic, we have thus no direct relation between the pftases.
180
I
Let us finally tackle the problem of the phase ambiguity in the propagator for the harmonic oscillator. Below the first caustic we know that the exact propagator is given by K( ~,·TI Xa'·0) -- e-i1T/4 v'ZnMInwT rrrr;;- exp {imu 2 2) COSW:r - Z~Xa]} mSinwT [( ~+Xa
(5.17)
When
T
=~
1T T <w
this reduces to the particular simple expression 1T.
K(~;Zwlxa'O)
=e
-i1T/4 !iii;
vin% exp[-
.1Illl
lh~xal
From the group property (2.25) and the translational invariance in time we know that
K(Xb;~IXa;O)
= =
!K(Xb;!wIX;O)K(X;z:IXa;O)dX i1T Z e- / i1TJexp[- i~(Xb+Xa)xl~dx
which by (5.5) reduces to
This takes care of the first caustic. We now proceed in the same way with the second caustic 1T K(xbl : lx a ;O) = J K(x b ; ~IX;O)K(X: ~IXaIO)dX
Continuing in this way we finally obtain the propagator corresponding to an arbitrary caustic: (5.29 ) But the propagator on a caustic serves as the initial condition for the propagator on the subsequent segment, i.e. we must demand lim T ...
!!!
+
K(xb;T!xa;O)
=e
-in.~
n o[x b - (-1) Xa1
w
Consider the expression e
- i 1T /4 / mw { imw [( 2 2) T 1} V21T~ISinwTlexp ZnSinwT Xb+Xa Cosw - ZXbXa
I
181
(Notice that it differs from Feyruman's expression (5.17) by an inclusion of the absolut value of SinwT under the square root). If wT
=
nrr
0<£«1
+ WE:
we can approximate the above expression by
e-i1T/41z~£exp{(-1)n e
-i1T/4
rm--
b
lR-£[(X +X!)(_1)n - 2x x a ]} b
rim [
n
v'~exP2n£xb-(-1)Xa
]2}
Compairing this with the "infinitesimal" free particle propagator we see that it has the limit O(x
b
-
(-1)n Xa )
All we need to get the correct boundary condition is therefore just an inclusion of the additional phase factor -i ~ Ent [ WT 1 e
1T
where Ent[xl is an abbreviation of the entire part of X, i.e. the "greatest integer below x. The correct propagator for the harmonic oscillator is consequently given by the following formula:
(5.30)
This is the famous Feynman-Soriau formura which takes into account the proper behaviour of the propagator at the caustics. It was discovered by Soriau in 1975 and has an interesting geometrical-topological interpretation, which we unfortunately cannot explain within this simple framework. It has also been derived by adding a small anharmonic term, say £x4, to the potential, and then studying the limit of the anharmonic propagator as £+0.
5,3
THE PATH-INTEGRAL REVISITED
Up to this point we have been treating the path integrals in a handwaving way. Although we are definitely not going to turn it into a rigorous concept, we shall now try to make it a bit more precise. The basic idea is to construct a limiting procedure, which allows us to compute the path integral as a limit of ordinary multiple inte-
182
I
grals (in much the same way as an ordinary integral can be obtained as the limit of Riemann sums). There are several different ways in which this procedure can be deduced. Let us first concentrate on the same reasoning as we used when we derived the Schr5dinger equation (section 2.9).
we slice up the time interval from
ta
to
tb
in
N
To compute
equal pieces:
with By repeated use of the group property (2.25) we then get x(\)=~
J exp{~S[x(t)l} D[x(t)l x(ta)=Xa (TO simplify the notation we have put xa = Xo and xb = X ). At N this point the formula is exact, but it presupposes the knowledge of the propagator we want to calculate!
In the limit where
N -+-
co
we
can however use that the propagators involved become infinitesimal propagators.
We can therefore replace them by the approximate
expression, cf. K(y;t+£lx;t) ~
(2.53),
r-rn-
i m (x-y) 2 {i x-+-v} 2 £ 1 exp - fi V(~)£
V~ exp[fi
Here the first two terms comes from the free-particle propagator while the last term comes from the interaction amplitude.
We thus
arrive at the following limiting procedure:
J f
N
. [Vz=Ik:"l rm N •.• expLr;:i E 2m = 11m N -+-
with £
CD
1T
£
"1<.=1
(_)2 + 4< 4<-1 4< ~-1 - V(-2-)£ldx1··~_1 £
183
I
In fact this is the original formula given by Feynman. We can rederive the same formula from a slightly different point of view: Consider an arbitrary (continuous) path leading from (xa;t ) to a (xb;t b ). Using the time-slicing we can approximate this path by the piece-wise straight line connecting the intermediary pOints (xo;t )' o •.. , (xN;t N). Rather than summing over all paths we now restrict ourselves to a summation over piece-wise linear paths. The action of a piece-wise linear path is given by
Piece-wise li-
t-axis
t-axis
Fourier
near path
path
Fig. 57b
Fig. 57a
V(x)dt
+ •••
The
~entiated
. exp Cfi:
action is therefore approximately given by {i
s [x (t)]} '" exp ¥i
N
L
k=l
~
(xi-x _,)
2
i -=---=-...!--
x.
-V(~
+
x } 2 i 1 )£
£
But since the piece-wise linear path is completely characterized by its vertex-points (x 1 , ••. ,x _,), we can now sum up the contributions N from the "N'th order" piece-wise linear paths simply by integrating over the vertex coordinates. In the limit where N + 00 the piecewise linear paths "fill out" the whole space of paths and we therefore propose the following limiting procedure:
I
184
x(1)=~
(5.32)
J exp{iS[xCt)l} D[x(t)l
~
x(ta)=xa
Comparing this with (5.30), we see that it is given by almost the same expression, except that we have "forgotten" the integration measure:
(/21T~fi£) N For various reasons, we will however prefer to neglect the integration measure. Rather than calculate a single path-integral, we will therefore calculate the ratio of two path-integrals (both involving a particle of mass m). Then the measure drops out and we end up with the formula
x(tb)=~
J
exp{k S[x(t)l} D[x(t)] xCta)=xa (5.33)
!M.
xCtb)=~
J exp{k So[x(t)l} D[x(t)l x(ta)=xa
def.
lim N+ex>
Usually So is the free-particle action so that we are actually calculating a propagator relative to a free-particle propagator. When we calculate a path integral we need not necessarily sum over piece-wise straight lines. Other "complete" families of paths can be used as well. Suppose e.g. we want to calculate a propagator of the form
185
I
K(D;TIO;O)
By slicing up the time-interval we again break up an arbitrary path, x = x(t ); 1 1 ... ;x _ 1 X(t _ ). This time we will approximate the given path N N 1 x(t) by a Fourier-path, i.e. a path of the form x(t), in segments connecting the intermediary pOints
N-1 k=1
kt k Sinn T
a
so that
I
x(t) = If we choose the coefficients
a k
N-1
I
(5.34)
k=1
a
k
Sin [ n.]!] N
j=1, ••• ,N-1
we evidently obtain that the Fourier path passes through the same intermediary points as the given path, cf. figure 57b. Again the approximate path is completely parametrized by the vertex coordinates
(x 1 ' ••• ,x _ ) and we can therefore sum up the N 1 contributions from such paths by integrating over the vertex coordi-
nates.
In practice it is however more convenient to integrate over
the Fourier components (x , ..• ,x _ ) 1 N 1 we evidently have: between
(a1, ••. ,a _ ). Since the relationship (5.34) N 1 (a , •.. ,a _ ) is one-to-one and smooth 1 N 1
and
J. .. I
J...
I
But the transformation (5.34) is in fact linear, so that the Jacohiant (a , ... a _ )! We can therefore forget about it N 1 1 and we finally arrive at the formula:
is independent of
x(T)=O
Jexp{~[x(t)l}D[x(t)l
x(O)=O
(5.35)
x(T)-O
Jexp{~o[x(t)l}D[X(t)] x(O)=O
To see how it works, we shall now return to our favorite example:
TEST CASE: THE HARMONIC OSCILLATOR According to exercise 2.4.1, the path-integral can be expanded around the classical path, whereby we get
186
x(T)=O
I
T
K(X ;Tlx ;0) = exp{ks[x ]}Jexp{if[1(¥t)2 -1 ix2]dt} D[x(t)] Z 1 C1 x(O)=O 0
According to (5.35) this can be further rearranged as i
. 1
K(x Z;Tlx 1;0) = exp{w>[xclnKo(O;TIO;O) 11m
N + '"
.
where
Ko
N-1
. kTIt
.- exp{~[ l: ,\Sm-r-l }da1 ••• ~1 J----.-.-...:""N'"'--1-------J J.- Jexp{1.KSO[k~,aksm-Tl . kITt } da1·-~-1 ~1
.
is the free-particle propagator, which is given by
=.j2rr~'tJ.T
Ko(O;TIO;O)
The remaining problem is therefore the computation of the multiple integral. Using the orthogonality relations for the sine-function it is easy to calculate the action which reduces to S{
N-1
I
k=1
a k Sin
+
k t}
=
mT N-1
T
I
k=1
2 k 2 TI2 2 a k (-:2 - w ) T
But then it is trivial to perform the integration over Fourier components, since the exponent is not only quadratic but even diagonal in (a , ••• , a _ ) ! 1
N 1
.
wf4TIin)N-1 N-1 k 2 rr2 2-l n (-2 -w ) mT k=1 T
N-l . kTIt .•• exp{fi S[ l: akSln~l}da ... da N_ 1 k=1
J J
To calculate the same multiple integral for the free particle action we simply put w=O i~ the above formula. Thus we finally get:
But using Euler's famous product formula for the sine-function, (5.36)
sin(TIx)= rrx[
n
k=1
this precisely reduces to
(1
-
2 x )]
~
187
j 27Tifi
I
m SinwT
exp{~ 11
s[xcll}
in accordance with (5.17)! What about the phase ambiguity? This can also be resolved if we are a little more careful! Consider the analytic extension of the Gaussian integral. If we carefully separate the phase from the modulus we get 7T i· 4 when A > a e +'" 2 iAx dx Tn . 1T Je -~'4 e when A < a
r
jU 1m
=/¥ =
Now, going back to the action of the Fourier path, (5.37)
i
exp lii S 1
we observe that the analogue of
2 2}
2 (k 2 7T an --:2 - w ) k=1 T
N-1 = exp {imT 4n I A
is negative when
k < wT/1T, and
positive when k > WT/1T. Consequently there are Ent[wT/1Tl ti ve terms" (where Ent [x 1 denotes the entire part of x). multiple integral is therefore actually given by
"negaThe
For the free-particle propagator this ambiguity does not occur. corrected formula for the propagator thus comes out as follows:
The
(5.38)
But that is precisely the Feynman-Soriau formula! Now that we have seen how the limiting procedure based upon Fourier paths work, let us return for a moment to the problem of the integration measure. If you want to calculate a path-integral directly, i.e. you do not calculate a ratio any longer, you must necessarily incorporate an integration measure in the limiting procedure. Feynman suggested that one could use the same integration measure as for the piece-wise linear path. This assumption leads to the formula:
I
188
x(T)=O
exp{i}[~(dX)2_ TIO 2 dt
(5.39)
J x(O)=O
r;;- N J
lim N ...
V(x)]dt} D[x(t)]
(V21iThi) a>
J . N-1 k t ... exp{i' S[ L akSirr.f-]}dx ... dx N- 1 k=1
lim N ...
a>
But here the right hand side diverges as you can easily see, when you try to calculate it in the case of a free particle, cf. exercise 5.3.1 below:
Exercise 5.3.1 Introduction: Consider the matrix A involved in the linear transformation (5.34), i.e. 'k Sin [,T I ] j,k 1, ... ,N-l Ajk N
a) Show that
N-1
AXT
N
I
=2
Det LA.]=
and
[~]-Z
b) Show that the series involved in the limiting procedure (5.39) is given by lim N-""
1
[N]
r (N)
TI
N-1~ 21TihT
and use Stirlings' formula for the r-function to verify that it diverges. In fact, by considering the case of the free particle propagator, it is not difficult to construct the correct integration measure, which (when you integrate over the Fourier components) turns out to be given by:
XjT) =0 i exp{[ S[x(t)]}
D[x(t)]
=
1T N-1
Nl!~ (vrz)
rm
r(N)[V~]
NJ
r
. N-1
1
.
k1Tt]} N-1
_jexp{~[k:1akSln.y- d
x(O)=O (As a consistency check you can u~ this formula to rederive the propagator of the harmonic oscillator). So now you 'see why we prefer to neglect the integration measure: Every time we introduce a new limiting procedure, i.e .• a new denumerable complete set of paths, we would have to introduce a new integration measure!
5,4
ILLUSTRATIVE EXAMPLE: THE TIME-DEPENDENT OSCILLATOR
As another very important example of an exact calculation of a propagator we now look at the quadratic Lagrangian: L
dx 2 = tm(at)
- tmW(t)x
2
a
189
I
Since it is quadratic we can as usual expand around a classical solution, so that we need only bother about the calculation of the following path-integral x(tb)~O
J
t imfb dx 2
exp{211 [(CIt)
x(t }~O
ta
a
Let us first rewrite the action in a more suitable form.
Performing
a partial integration we get tb S [x (t)]
=- ~ J [ ta
Here we have neglected the x(t a )
tions
= x(t b ) =
O.
boundar~
terms due to the boundary condi-
Inserting this we see immediately that
the above path-integral is actually an infinite dimensional generalization of the usual Gaussian integral since it now takes the form:
x(tb)=O
J
t
im b dZ exp{- 2fJ. fx (t) [dt Z
+
W(t)]x(t)dt}D[X(t)]
ta
x(ta)=O
To'compute it we ought therefore to diagonalize the Hermitian operator -
d
2
dt
+ WIt)
2
At the moment we shall however proceed a little different.
By a
beautiful transformation of variables we can change the action to the free particle action!
Let
f
be a solution the second order
differential equation
d2 {-2 + W(t)} fIt) dt
(5.40)
i.e.
f
=
a
belongs to the kernel of the above differential operator.
The solution can be chosen almost completely arbitrarily within the two-dimensional solution space. that
f
The only thing we will assume is
does not vanish at the initial endpoint: f (t a ) " O.
(Notice that
fIt)
is not an admissible path, since it breaks the
boundary conditions!).
Using
f
we then construct the following
linear transformation, where x(t) is replaced by tha path y(t):
r
190 t
(5.41 )
x(t)
f (t)
J
~ f (s)
ds
ta Here we assume that the transformed function y(t)also satisfies the boundary condition y(t ) = 0 a Differentiating (5.41) we obtain t
x' (t) = f' (t)
J Yf'(~sl
ds + y' (t)
f' (t)
"""f'(t) x(t) + y' (t)
ta so that the inverse transformation is given by t
(5.42)
y(t) = x(t) -
I
f' (s) x (s) ds
fTs)
ta We can now show that the above transformation has the desired effect. Using that t
x"
f"
(t)
J Lill f(s)
(t)
d
s
+ ft (t)f (t) fIt
+ y" (t)
we obtain:
{;i$
t
+ W(t)}x(t) ={f"(t) + W(t)f(t)}
I[~'(~S)]dS
+
f'n~r(t)
+y"(t)
ta But here the first term vanishes on account of (5.40).
Consequently
the action reduces to:
~
S[x(t)] =
-1 J
t
[F(t)f' (t)y' (t) + F(t)f(t)y"(t)] dt
with
F(t)
= J L...!& f(s) ds
ta Performing a partial
integrat~on
on the second term this can be
further rearranged as: tb
s[x(t)
- 1 [X(t)Y'(t) Jt
1
a
But here the boundary terms vanish due to the boundary conditions satisfied by
x(t).
Thus we precisely end up with the free-particle
action in terms of the transformed path
y(t)!
191
I
There is only one complication associated with the above transformation, and that is the boundary condition associated with the final end-point
tb'
The boundary conditions satisfied by
x(t)
are transformed into the following conditions on the transformed path
y(t): tb
J
~dS
o
ta The second boundary condition is non-local and therefore not easy to handle directly.
We shall therefore introduce another trick!
the identity, Ii (x(t )) = -,}; b cf.
J exp
Using
{-ia x(t ) }da
b
(5.5), we can formally introduce an integration over the final
end-point: x( t ) arbitrary b
x(1)=O
2~
J exp{iS[X(t)]}D[X(t)]
Jf~eXP[-iaX(~)]eXP{kS[x(t)]}da D[x(t)]
x(ta)~O
x(ta)=O
This is because the integration over
a
now produces a Ii-function
which picks up the correct boundary condition!
(Notice that if we
attempt to calculate the path integral by a limiting procedure we must now also integrate over the final end-point
x ). N
Changing
variables we then get
Here the infinite dimensional generalization of the Jacobi-determinant is independent of linear.
y(t)
because the transformation (5.41) is
The remaining integral is Gaussian.
"Completing the square"
the whole formula therefore reduces to ~
1
OX = #ed liy ]
J
exp{ -
-'" with
y(tb ) arbitrary b dt im bdy 2 f (\)f -z-}da exp{crifCat) dt}D[y(t)] t f (t) t a yet )=0 a a t
ill Z 2
2iif
t
yet)
~
yet) -
~f(~)ff~) ta
•
J
I
192 At this point we get a pleasant surprise: the a-integration!
We can actually carry out
Furthermore the remaining path-integral is with-
in our reach, since it only involves the free particle propagator. In fact we get:
y(1)
arbitraIb
Jexp{~ {(~~)2dt}D[y(t)] y(ta)=O
== f -0>
Ko (x;1b IO ;ta )dx
a
==
lz C1b 7Tih
0>.
-t )
a
LeXP[~
2
(1, ~tidx ==
This should hardly come as a surprise since, by construction, the above path-integral represents the probability amplitude for finding the particle anywhere at the time
tb.
The total path integral thus
collapses into the simple expression:
- W(t)x 2]dt}D[X(t)]
(5.43)
It remains to calculate the Jacobiant! using a very naive approach.
We shall calculate it
(The following argument is only includ-
ed for illustrative purposes, it is certainly not a rigorous procedure!).
As in the approximation procedure for path-integrals we
discretize the linear transformation by introducing a time-slicing. The paths
x(t)
and
y(t)
are then replaced by the multidimensional
points
The linear transformation (5.42) can then be approximated by T
- N
n
f' (t k )
(x k " x k - 1 )
k~ 1 ~ ---'2"---:":"--:""
(This is actually the delicate pOint, since the discrete approximation of the integral is by no means unique, and the Jacobi-determinant turns out to be very sensitive to the choice of the approximation procedure).
Okay, so the Jacobi matrix has now been replaced
by a lower diagonal matrix. from the diagonal, i.e.
The determinant thus comes exclusively
I
Taking the limit
N
we then find:
-T
N
lim N +
exp[log n (1 k= 1
a>
£'(t k ) T
N
lim N +
exp[ E log(1 - ! £( t )
N) 1
k
k= 1
a>
tb
f'J11 exp[-! T(t)dtl
fta
Consequently
Inserting this into
(5.43)
our formula for the path-integral fi-
nally boils down to the following remarkable simple result:
The path-integraZ corresponding to the quadratic action (5.44)
t t m b dx 2 2 m b d2 S[x(t)l == 2 f[(dt) - W(t)x ldt == - 2 fx(t) [dt2 ta ta
+
W(t)lx(t)dt
is given by x(~)=O
(5.45)
f
exp{~[x(t)l}D[x(t)l
x(ta)=O where
f(t) is an almost azobitrary soZution to the differentiaZ equation 2
{~2
+
W(t)}£(t)
the onZy constraint being that
=
0
f(t a )
~
O.
Notice that the above formula in fact includes the free-particle propagator (with W(t)=O) and the harmonic oscillator (with W(t)~2). You can easily check that
(5.45)
in these two cases by putting
reproduces our previous findings
f(t):1, respectively
f(t)=Cosw(t-t ). a
194
5.5
I
PATH-INTEGRALS AND DETERMINANTS Now that we have the formula for the propagator corresponding to
a quadratic Lagrangian at our disposal, we will look at it from a somewhat different point of view. tb Sex (t)
;:J
1
x(t)
d
Since the action is quadratic,
2
{~+
W(t)} x(t)dt
dt
we can "diagonalize" it.
For this purpose we consider the Hermitian
operator d
2
- -:-2 - WIt) dt which acts upon the space of paths,
x(t), all satisfying the bound-
x(t ) = x(t ) = O. b a normalized eigenfunctions ¢n(t):
It possesses a complete set of
ary conditions:
A given path,
x (t),
can now be approximated by a linear combination tb
N xN(t)
I
with
an¢n (t)
n=1
an
f
¢n(t)x(t)dt
ta
Notice that the corresponding action of the approximative path reduces to
The summation over approximative path's can therefore immediately be carried out since it is just a product of ordinary Gaussian integrals:
I. ·f exp{~s In the limit where
[XN (t)
N
+
=
1 }d3., ... da N the approximative paths fill out the whole
space of paths and the path-integral is therefore essentially given by
By analogy with the finite-dimensional case we define the determinant of the Hermitian operator eigenvalues.
_a t2
- WIt)
to be the infinite product of
Of course the determinant will in general be highly
195
I
divergent but we can "regularize" it in the usual way by calculating the ratio of two determinants.
From the above calculation we then
learn the following important lesson:
The path integral corresponding to a quadratic Lagrangian is essentially given by the determinant of the associated differential Operator, i.e·X(s,)=o ~
f
(5.46)
exp{~fX(t)[-:tr2-
x(t )=0 a
MDet[-~2- W(t)J}-~
W(t)]x(t)dt}D[x(t)] =
ta
hlhere the right hand side should actually be interpreted as a limiting procedure (i.e. it is a short hand version of the follOhling expression: N
(5.47)
lim
_1
L',(N)[ IT;')
N ->- co
2
fexp[~ l: Ana~l(vj~~/dNa N
lim L',(N)f.-
=
n=1
N ->- co
n"'1
Notice that we have included a factor
(12 ~ft f in
11
the integration
over the generalized Fourier components, i.e. the proper integration
measure for evaluating the determinant is given by
~
(5.48)
k=1
[/2:Hi da k ]
Apart from that we still need a further integration measure
L',(N)
since the determinant is only proportional to the path-integral.
The
above characterization of the path-integral is the one used by e.g. Coleman [1977]
.
Since we already know how to compute the path integral we can now extract a relation for the determinant.
To avoid divergence problems
we calculate the ratio of two determinants.
According to (5.45) it
is given by
Det[-a~- W(t)]
(5.49)
Det[-a~- V(t)]
In this formula it is presupposed that at
tao
fW
and
fV
does not vanish
It can however be simplified considerably by going to the
singular limit, where
fW
and
fV
does vanish at
tal
Let us de-
I
196 note by
{- a~ -
fO
the unique solution to the differential equation,
W
W(t)}f(t)
=
0,
which satisfies the boundary conditions
2.. fO (t ) dt w a Similarly we denote by
f~
= 1
the solution which satisfies the boundary
conditions:
o In the above identity (5.48), we can then put fO
f = fO + d 1
f1
W+ E W
V
V
V
It follows that and Finally the limit of the integral tb
f
dt ta [f (t)]2
w
fO at tao But since almost all W the contribution to the integral comes from an "infinitesimal neigh-
diverges, due to the vanishing of
ta (in the limit where
bourhood" of
s+O), it follows that i t di-
verges like
Consequently lim
s
+
0
tr, f -w
dt
{
t
f2 (t) }
a Thus the identity
I{
(5.49)
ftr, e
dt (t)}
,V a collapses into the extremely simple detert
minantal relation: Det[ -a~-W(t) ]
(5.50)
DetH~-V(t) ]
If, eg., we put V(t)=Q
(and consequently
Det[-a~-w(t)l
DetH~l
f~(t)=t-ta) it follows that
197
I
This can be used to calculate a propagator (corresponding to a quadratic Lagrangian) relative to the free-particle prop~gator Ko: . (5.51)
K(~;tblxa;ta)
K(O;tbIO;t a ) exp{~[xcll}
==
(
1
Det{-d~-W(t)}]-2 KO(O;tbIO;t ) Det{-d~} a
-I
m
2TIiflf~ ( tb )
exp{~s[xCl]}
E.g. in the case of the harmonic oscillator we put W(t)=w 2 and consequently 1 fW(t); ;;Finw( t-t ). In this way we recover the by now well-known result (5.11). a /
Rema:r>k: Incidenti.lly the
investigation of determinants corresponding
to linear operators has a long tradition in mathematics.
E.g. the
basic determinantal relation (5.50) has been known at least since the twenties*).
This is very fortituous, since in our derivation some
dirt has been swept under the rug.
The passage from quotients of
path-integrals to quotients of determinants only works if the integration measure
n(N)
introduced in (5.47)
of the potential function
WIt).
is actually independent
Since the result we have deduced,
i.e. the determinantal relation (5.50) is known to be correct, we have thus now justified this assumption. In fact the determinantal relation (5.50) beautiful reasoning which emphasizes its basic who are acquainted with more advanced analysis argument: Consider the expressions 2 Det(- a g(A)
can be proven by a very general and position. For the benefit of those we include the main line of the
WIt) - A) t 2 Det(- a - V(t) - A) t
and h (A)
is important for the following that A is treated as a complex variable. can be proven quite generally that a differential operator of the form,
It
- a2t
- WIt)
has a discrete spectrum of real eigenvalues 1.. 1 ,1.. 2 ' ••• , An' •••
*) J.H. Van Vleck,
Proc.Nat.Akad.Sci.
li,
178 (1928).
It
198
I
which are bounded below and which tend to infinity as n~. Each of these eigenvalues has multiplicity one, i.e. the associate eigenspace is one-dimensional. When A coincides with one of these eigenvalues, say A = An' it follows that the "shifted" operator 2
{-at - wit) - An}
has the eigenvalue zero, so that its determinant vanishes. As a consequence the function g(A) becomes a meromorphic function with simple zerOs at the eigenvalues A~ and simple poles at the eigenvalues A~. Consider next the function to the differential equation
f~+A(t).
{- at2
-
By definition it is the unique solution
w(t)
-
A}f (t) = 0
which satisfies the boundary conditions
flta) It fo llows that
=0
is an eigenvalue of the operator
WIt)} if and only if
o
f IW+A) It b ) = 0
and when that happens f(W+A)(t) is in fact the associated eigenfunction (except for a normalization factor). As a consequence the function heAl V becomes a meromorphic function with simple zeros at AW and simple poles at A k The quotient function n' g (A) /h IA) is thus an analytic function without zeroes and poles! (Thus it has a behaviour similar to e.g. exp[A]). Furthermore, from general properties of determinants (respectively solutions to differential equations) it can be shown that g(A) (respectively heAl) tends to 1 as A tends to infinity except along the real axis. The same then holds for the quotient, which as a consequence must be a bounded analytic function. But then a famous criterium of Liouville guarantees that it is constant, i.e.
Specializing to
5,6
A=O
**=
1
we precisely recover the determinantal relation (5.50).
THE BOHR-SOMMERFELD QUANTIZATION RULE
we shall now encounter the important problem of computing quantum corrections to classical quantities. Especially we shall consider quantum corrections to the classical energy. In the case of a onedimensional particle we know that it can be found by studying the trace of the propagator. Inserting the path-integral expression for the propagator, this trace can be reexpressed as
I
199
x(T)=x
(5.52)
j J eX~t~S[X(t)]}
G(T)
-00
x(O)=x
D[X(t)]dx o
o
== J exp{~S[X(t)]} D[X(t)] x(O)=x(T) where we consequently sum over all path's which return to the same pOint.
This expression is known as the path-aum-traae integral.
There are now in principle two different approximation procedures available for the evaluation of this path-cum-trace integral. first method is the hleak-aoupling approximation.
The
Let us assume that
the potential has the general shape indicated on fig. 58.
To calcu-
late the low-lying energy states we notice that near the bottom of the potential well, we can approximate the potential by V(x)
RJ
W"
(0)x
2
I
iV"(O)x 2
E
:
E"
E3
E2 El
Eo
X-axis Fig. 58
Thus the problem is essentially reduced to the calculation of the path-cum-trace integral for the harmonic oscillator! already solved, cf.
This we have
(5.19-20) and the low-lying energy states are
consequently given by (5.53)
with
v"(O)
mw
2
The second method is the so-called WKB-approximation which can be used to find the high-lying energy states. non-perturbative method.
It is thus essentially a
It leads to the so-called Bohr-Sommerfeld
quantization rule, which in its original form states that
200
f pdq
(5.54) where
q
= n
I
• h
is the position coordinate,
we integrate along a periodic orbit.
p
the conjugate momentum and
Notice that the left hand side
is the area bounded by the closed orbit, cf. fig. 59, so that the quantization rule states that the area enclosed in phase-space has to be an integral multiple of
h.
The original WKB method which was
based upon the construction of approximative solutions to the schrodinger equation, will not be discussed here.
But before we jump out
in the path integral version of the WKB method it will be useful to take a closer look on classical mechanics and the original derivation of the quantization rule (5.54).
p-axis Phase-space diagram.
--+--"'------""'t'---f-"'--..... q-axis
ILLUSTRATIVE EXAMPLE:
Fig. 59
THE BOHR-SOMMERFELD QUANTIZATION RULE
Let us first collect a few useful results concerning the action. Suppose
(xa;t a )
and
(xb;t b )
are given space-time points.
Then
S(Xb;tb[xa;t a ) will denote the action along the classical path connecting the two space-time points.
The
res,ult~ng
function of the initial and final
space-time point is known as Hamilton's principal funation.
(If
there is more than one classical path connecting the space-time points (xa;t a ) and (xb;t b ) it will be a multivalued function). All the partial derivatives of Hamilton's principal function have direct physical significance:
as
(5.55) Here
~ Pa
as
Pa
ata
E
as
at b
is the conjugate momentum at the initial point
the conjugate momentum at
xb
and
served along the classical path).
E
E
xa ' P b is the energy (which is con-
201
I
Proof: A change, 6x b , in the Epatlal position of the final space-time point will cause a change, 6x(t), in the classical path. The corresponding change in the action is given by tb
J [~~
6S =
6x(t) + aL Qx(t) J dt ax
ta
~
tb CaL
J
x
t
-~~J dt "aX
6x(t)dt + [aL "aX
6x(t) ]
a But here the integrand vanishes due to the equations of motion and we end up with: 6S
= a~(tb)QX(tb) ax
It
-
aL ) 6X b = -;-(t b
aL(t )6x(t ) • a a ax
Pb 6X b
ax
follows that 6S 6X b
Pb Similarly we may consider a change,
6t , in the temporal position of the b final space-time pOint. Notice first that
The corresponding change in the action is also slightly more complicated,
[~ 6x(t) ax
+
~ Qx(t)Jdt ax
ta where the first term is caused by the change in the upper limit of
As before we can now make a partial inte-
the integration domain.
gration leaving us with the formula 6S
=
aL L(t b )6t b + -;-(t )6x(t b ) b dX
=
L(t b )6t b -
aL' ) J 6tb = [ L(t ) - -;(tb)x(t b
ax
b
But the energy is precisely given by
aL
•
~(tb)x(tb)6tb
dX
202 E =
p~
I
aL
- L
a~
X- L
so that we end up with the identity
-
6S =
as
Le.
Eot b
3t
-
~
E
b
Notice that the above considerations were in fact anticipated in the discussion of the Einstein-de Broglie rules, cf. the discussion in section 2.5. After these general considerations we now return to the discussion of a one-dimensional particle in a potential well like the one sketched on fig.
59.
In the general case there will exist a one
parameter family of periodic orbits whi9h we can label by the fundamental period
T: x = xT(t)
Each periodic orbit will furthermore be characterized by its energy E, which thus becomes a function of the period
T, i.e.
E
= E(T).
From (5.55) we learn that dS = _ E
(5.56 ) where
OT S(T)
is the action of the periodic orbit
xT(t).
We can now
introduce the Legendre transformation of the action function
S(T).
It is defined by the relation: dS W(E) = S(T) - dTT=S(T) + E'T where
T
should be considered a function of
E.
(Notice the simi-
larity with the passage from the Lagrangian to the Hamiltonian).
In
analogy with (5.56) we then get: dW _ dS dT
dT
dE - OT OE + T + E • dE
(5.57)
= -
dT dT E dE + T + E CiE = T
Finally we can get back the actfon function by a Legendre trans formation of
W(E),
(5.58)
where
S(T) E
= W(E)
- T'E
= W(E)
dW - dE • E
should now be considered a function of
T.
Notice that
the Legendre transformation is in fact given by the phase integral in (5.54) T
W(E)
i.e.
S(T) + E'T
f [m(£E) o
T
2
- V(x)]dt +
2
f [m(~~} o
+ V(x}]dt
203
T
(5.59) since
I o
WIE) dx = m at
p
quantity
W(E)
I
2 m(dx) dt
dt
for this simple type of theory.
It is thus the
which we are going to quantizel
From (5.57) we now get,
where
dE
= -T1
w
is the cyclic frequence of the periodic orbit.
dw = ~ dw 2~
w
with
2~ =T
Following
Bohr we then assume that only a discrete subset of the periodic orbits are actually allowed and that the transition from one such periodic orbit to the next results in the emission of a quantum with energy
nw, where
w
is the frequency of the classical orbit.
The
last part of the assumption is Bohr's famous correspondence principle.
It follows that
W can only take a discrete set of values and
that the difference between two neighbouring values is given by
= n.
2~
Thus the quantization rule for W(E ) = n
where
c
2~(n+c)ft
=
2~·n
W can be stated as +
2~ftc
is a constant which we cannot determine from the corre-
spondence principle.
You can look upon this formula as the first
two terms in a perturbation expansion for powers of
1/n.
W(En)
where we expand in
Bohr and Sommerfeld simply assumed it was zero, but
using the WKB-method it was actually found to be ,.
Thus the correct
quantization rule for the above case is given by (5.60)
As an example of its application we shall as usual consider the harmonic oscillator. features:
This example turns out to have two remarkable
1) The assumption of a one-parameter family of periodic
solutions labelled by the period consequence the action function monic oscillator.
T
breaks down completely.
SIT)
As a
cannot be defined for a har-
2) The quantization rule (5.60) is exact.
Since the action function cannot be defined, we shall define W(E)
through the phase integral (5.59).
A general periodic orbit
is given by x(t)= ACoswt +
BSinwt
The associated energy is consequently
E
=
mW 2 (A2 + B2)
while the phase integral turns out to be
I
Thus W(E) = 2rr E w
and the Bohr-Sommerfeld quantization rule reduces to the well-known result (5.20)
o
Now that we understand the Bohr-Sommerfeld quantization rule we return to the path-cum-trace integral (5.52). This time we are going to expand around a classical solution x(t) = xcI (t) + nIt) The potential energy is expanded to second order: V[x(t)]
2
V[Xcl(t)] + V'[xcl(t)]n(t) + Pl"[x c1 (t)]n (t) Inserting this, and using the classical equation of motion, the action then decomposes as follows Rl
T
S[X(t)]
Rl
S[Xcl(t)] +
b{I(~)2-!V"[XCl(t)]n2(t)}dt d2
T
1
+ 1bn(t){-~- ffiV"[Xcl(t)]}n(t)dt
Consequently the path-cum-trace integral reduces to (5.61)
G(T)
Rl
7
l:
-00
x
n(T)=O
exp{~[xcl(t)]} cl
J expHKn{-~- ~Vff[xc1(t)]}ndt}D[n(t)]dxo
n(O)=O
where xcl(O) = xcl(T) =xo' But the remaining path integral we know precisely how to handle. According to (5.45) it is given by
j __m_-J""-Td-t 2rrifJ.f(O)£(T)
with
£2 (t) j
o In fact we can relate f to the classical path: solves the Newtonian equations of motion: 2
d XcI m~
- V'[x Cl (t)]
Differenting once mOre, we thus find 3 d x
m
~ dt
= -
V"[x lIt)] C
The classical path
205
I
Consequently we can put
Thereby the path-cum-trace integral reduces to (5.62 )
T dt
dx
IlxdP o
0
where we sum over all classical paths satisfying the constraint xcl(O) = xcl(T) = xo.
Despite its complicated structure it is
essentially a one-dimensional integral of the type +00
I
exp{if(x)} g(x)dx
Such an integral can be calculated approximatively by using the Thus we look for a point
stationary phase approximation.
the phase is stationary, i.e.
f' (x ) = O. o
Xo
where
We can then expand
around this point: fIx) "'" f (x o ) + ,f" (x o ) (x-x o ) g(x) Rl g(x ) + g' (x o ) (x-x o ) o
2
In this approximation the integral then reduces to a Gaussian integral: +co
(5.63)
f
exp{if(x)}g(x)dx
Rl
g(xo)eXp{if(Xo)}vI£,,(:~)
Using this on the path-cum-trace integral (5.62) we see that we need only include contributions from the classical paths which produce a stationary phase, i.e.
o = xa ao
But according to (5.55) this gives
o i.e. the momenta at
t=O
and
P b - Pa t=T
are also identical.
Thus the
stationary phase approximation selects for us only the purely periodic solutions! od
T.
They can be parametrized by their fundamental peri-
Corresponding to a given
T
we should therefore only in-
clude the contributions from the periodic orbits:
206
I
x T/ 3 (t) r •.. rXT/n (t) I •••
XT / 2 (t)
Before we actually apply the stationary phase approximation we can therefore restrict ourselves to a summation over these periodic orbits:
N::>tice that the parametrization of the periodic orbit is only determined up t, i.e. we can choose the starting point quite arbitrarily on the closed path. p-axis Each of the x-values between the turning points x 1 and x 2 occurs furthermore twice on the path, cf. fig. 60. We are now ready to perform the x -integration. The x x 0 ~~____________-+____+-__-+~2~.. action of the periodic orbits does X-axis not depend upon the choice of the starting point, and neither does the expression to a translation in
T
Fig. 60
J
o
dt (~) 2
Thus the only contribution to the integral comes from x2 +00 dx o 2 dt = ~ I~T(O) I x 1 Cit
J
f
J
i
n
Having taking care of the xo-integration we must then compute the integral over the reciprocal square of the velocity. Here we get ~2
T
Jo
•
dt
(x T / n )
2
2n
J
3
(2[E(T/n) - V(xB-~) dx m
The same integral can however be obtained by a different reasoning! Consider the phase integral W(E), which is given by x 2
WeE)
= 2n
J /2[E
- vex) 1 dx
x1 for the periodic orbit in question. respect to E we obtain:
Differentiating twice with
207
I
-3/2
dT dE
(2[E - V(x)])
dx
consequently dt
T
Jo
•
(x T / n )
_ m3/2 dT dE
2
Putting all this together, the path-cum-trace integral thus finally reduces to: G(T)
(5.64)
Rl
1.
I
_1_
m I2nifl n=1
-[T]n nT~dEI dT
exp{i nS - } ft
T T~
n
This was the hard part of the calculation! energy levels.
Now we can extract the
As usual this is done by gOing to the transformed
path-cum-trace integral and looking for the poles. path-cum-trace integral is given by, cf. i
G(E)
(Notice that
E
G(T) exp{fi·ET} dE
now both represent an integration variable and the
energy of the periodic orbit. latter as
The transformed
(5.9)
Ecl I).
To avoid confusion we shall write the
Inserting the above expression for
G(T)
we
obtain
G(E)
i
Rl
1
iiiKv'21Til'i
""
n~,Jo
Here it will be preferable making an exchange of variables, i.e.
~
~
= Tin,
is the fundamental period of the periodic orbit in question:
G(E)
Rl
. 1 ~v'21Ti~
""""f· ~ dT E exp{1f(ET+S[T])}T/I~Iv'n
n=1
T
o
The T-integration is then performed by use of the stationary phase approximation.
A stationary phase requires
a o = 1T(EI
+ S[T])
Thus for a given value of
E
=E
as
- ~
=E
- Ecl
we get the main contribution from the
periodic orbit which precisely has the energy (5.63) the stationary phase approximation now
E! give~:
According to
208
G(E)
Rl
~ jZTTift
2- _,_
mI'lv'ZTTifl n='
n
I
_,_
fd2S cr:rr
Using that dE
and
dT
W(E)
+ ET (E)
S (T)
the formula finally reduces to
(5.65)
G(E)
Rl
~
T(E)
~
exp{%nW(E)}
n='
i mt
T
(E)
This clearly has poles when (5.66) and thus we have precisely recovered the old Bohr-Sommerfeld quantization condition (without the half-integral correction term).
The
reason we missed the half-integral correction term is, however, very simply:
It is due to the fact that we have not taken into account
the phase ambiguity of the path integral.
precisely as for the
harmonic oscillator we should pick up phase corrections. pression (5.62) is completely analogous to (5.38).
The ex-
The additional
phases then come fran the singularities in the integrand
T
-,
Uo f2 ~:) ] These singularities correspond to the zero's of
f(t)
=
x(t), i.e.
In the present calculation the x and 1 are thus analogous to the caustics. Each time we pass
to the turning pOints turning
~ts
a turning pOint we therefore phase-factor
e
iTT/2
points in the orbit
, cf.
e~pect
.
(5.29).
xT/n
that we pick up an additional
Since there are
2n
turning
we consequently gain an additional phase
e- iTTn = (_1)n which should be included in the formula for the propagator.
The
transformed path-cum-trace integral (5.65) is thereby changed into
(5.67)
G(E)
Rl
~
T(E)
~
n=1
(-1)nexp {%nW(E)}
209
I
This causes a shift in the poles, whichare now given by (5.68)
and that is precisely the correct quantization rule according to the original WKB-calculation. Notice too that near a pole we get the expansion
Using the approximation 1 + exp4i W(E)}
Rl
it follows that
1 + exp{k W(E
ll
G(E)
)
}exp{~ T(E) (E-En )} = 1 - {1
+
~ T(E) (E-En )}
behaves like G(E)
Rl
1 E--=-E n
when
E
Rl
En'
This should be compared with (5.10) and it clearly
shows that the approximative path-cum-trace integral has the correct asymptotic behaviour close to a pole. we have been working hard to derive a result which, as we have seen, can in fact almost be deduced directly from the correspondance principle!
When we are going to use the path-integral technique in
quantizing field theories we will actually have to work still harder Before we enter into these dreadful technicalities I want to give a few examples of almost trivial applications of the preceding machinery. The first example is concerned with the free particle.
To find
the energy levels we enclose the free particle in a box of length L
and use periodic boundary conditions.
Thus we have effectively
L.
replaced the line by a closed curve of length effect that the energy spectrum is discretized. uous free-particle spectrum back by letting
L
This has the We get the contin-
go to infinity.
The free particle can now execute periodic motions by going around and around the closed circumference:
The periodic orbit
given by xT(t) =
Lt T
It has the following energy and action: m dx 2 _ mL 2 E(T) = Z(dE) SIT)
-;;Z
ET
x
T
is
210
I
Consequently W(E) ; SIT) + ET ; LI2mE Since there are no turning pOints in this problem the quantization , rule is given by (5.66), i.e. Ll2mEn
21Tnfl
i.e. (5.69 )
These are the correct energy levels in the discrete version.
To see
this we notice that the free particle Hamiltonian is given by
~
fi2
d2
H=----
2m dx2
Since the SchI1.5dinger wave function must be periodic, the eigenfunctions are given by 21Tnx] o/n(x) = exp [ ± 1. -Land the corresponding eigenvalues precisely reproduce the above result (5.69)
The second example is concerned with a field theory.
In the
discussion of the sine-Gordon model we found an interesting family of periodic solutions:
The bions (or breathers).
In a slightly
changed notation, where we emphasize the cyclic frequency are given by (cf. ~(x,t)
(5.70)
= ~4
w, they
(4.50): n Sinwt ] with Arctan [ Cosh(nw~
n
/].l_w 2
= ---w--
and
O<w<
~
•
The corresponding energy is given by (cf. 4.51): E = 2M!j1 _
(5.71)
w~ \j
where
M is the mass of a single soliton.
As we have seen it re-
presents a classical bound state of a soliton and an anti-soliton. We therefore expect that it will generate quantum bound states which should be recognizable in the energy spectrum.
Since (5.70)
is a one-parameter family of periodic solutions (which can easily be labelled by the period
T), we can in fact use the naive Bohr-
Sommerfeld quantization rule. the form
For a classical field theory it takes
211
21Tnh
WeE)
(5.72)
(where
1T(X,t)
calculate
with
I
T
+00
o
-00
JJ
WeE)
1T(X,t)
~
is the conjugate momentum density).
WeE)
dt dx
Rather than
directly from the phase integral we shall use the
basic relation dW dE =T =
(5.57 ) From (5.71) we see that
W
dw dE
(5.73 )
21TW
-1
can be replaced by 21T)1-
1
(1
-
E2
--.,-)
E
and we get
-'
4r.1"
This can trivially be integrated WeE)
E=O, i.e.
But using that state, where
= 41T~-1M ArCsin(~) W=)1,
+ C
corresponds to the classical ground
W obviously must vanish too, we see that
C=O.
A
trivial application of the naive Bohr-Sommerfeld quantization rule then gives: (5.74)
E
o<
n
Notice the upper limit of crete set of energy states. When
n
n
< :~
n, i.e. there is only a finite disThis comes about in the following way:
grows so does the fundamental period
T, which by (5.57)
and (5.73) is given by T(n) But when
T
unbound and for
dW dE
21T
the corresponding soliton-antisoliton pair becomes n
above the treshold, the classical state thus
becomes unstable and breaks up into a separate soliton and antisoliton. In the above derivation we have been using the naive Bohr-Sommerfeld quantization rule.
As in the case of the single particle we
can improve this uSing the path-integral version of WKB.
In the
multidimensional case the WKB-formula, however, becomes more complicated than in the onedimensional case, where we just had to include a "half-integral" correction term.
For the special case of the
sine-Gordon model it can nevertheless be shown that the energy spectrum of the bion-states is still given by (5.74), with a suitable redefinition of the parameters
M
and
)1.
212
I
INSTANTONS AND EUCLIDEAN FIELD THEORY
5,7
At this point we introduce a very important trick, which will enable us to extract non-trivial information about the quantum mechanical ground state of a system.
Let us consider the Minkowski
space with its indefinite metric ds 2
=-
dt
2
+ dx
2
+ dy2 + dz
2
Formally we can turn it into a positive definite metric by using Weyl's unitary trick, i.e. by replacing the Minkowski time the imaginary time
1
t
with
= it:
Obviously we can think of this as a kind of analytic continuation if we regard the 4-dimensional Minkowski space as being a 4-dimensional real subspace of an 8-dimensional complex space. The above trick has been widely used in special relativity, where it allows you to avoid the distinction between upper and lower indices.
But it is also widely used in quantum field theory, where it
has had profound implications for our understanding of a variety of phenomena.
In quantum field theory, it has further been given a
special name:
The analytic continuation to imaginary time is by
quantum field theoreticians known as performing a Wick rotation. To illustrate the applications of the Wick rotation we consider for simplicity ordinary quantum mechanics in one space dimension.
As we have seen, the most important ingredient in ordinary quantum mechanics is the Feynman propagator this to be an analytic function of
K(Xb:tblxalta)' tb
and
t
If we assume
, we can perform a
a wick rotation whereby we produce the so-called Euclidean propagator: KE(XbITbIXa;Ta)'
E.g. the free particle propagator is Wick rotated
into the following Euclidean propagator: 2
m (X b - Xa) } exp { - 2fi T - T b a
(5.75)
cf.
(2.28).
Similarly the propagator for the harmonic oscillator
is Wick rotated into (cf.
(5.17»
where the trigonometric functions have been replaced by hyperbolic
f,·"..,t-; ons.
Notice that the caustics have disappeared and so have
213
I
the phase ambiguities - the Euclicean propagator is only singular when Tb = Ta! In the Minkowski space the Feynman propagator can be reexpressed as a path integral. The same is true in the Euclidean case. To Wick rotate the path integral we notice that the substitution, t ~ - iT, generates the following change in the action Tb
sex (t)]
- V (x)} dt
~
i
J
+
V(x)}dT
1a
consequently the Wick-rotated path integral is given by
x~Tb)=~
J XCTa)=xa
Tb exp[iJjc*)2 + VCX)}dT] D[X(T)] Ta
where we sum over paths in the Euclidean space. Let us introduce the abbreviation Tb (5.77)
SE[X(T)]
J{jC*) 2
+ Vex) }dT
Ta The quantity SE[X(T)] is known as the Euclidean action. Notice that the Euclidean action is positive definite. The Euclidean path-integral is now given by X(Tb)=~
J
exp{-iSE[X(T)]} D[X(T)]
X(Ta)=Xa This is one of the main achievements of the Wick rotation. Rather than being an oscillatory integral, which can never be absolutely convergent, we have now turned the path integral into a direct generalization of a nice decent Gaussian integral where the exponent is negative definite and furthermore it is at least quadratic. This puts the path integral on a much more sound footing, where we avoid completely the divergencies and phase ambiguities which plague the Minkowski version. The Euclidean path integral can be defined through limiting procedures like (5.31) or (5.33). If they are quadratic, i.e. of the Gaussian type they can also be diagonalized, so that we can use relations like (5.45) (suitable modified). In the Euclidean case it is in fact even possible to define it in terms of a rigorous integration theory on an infinite dimensional measure space using the so-called Wiener measures. Returning to the Euclidean propagator we now get from (2.21) that it can be reexpressed in terms of an Euclidean path integral:
214
I
X(Tb)=~
f
(5.78)
exp{-k SE[x(T)l} D[x(T)l
X(Ta)=Xa The other characterizations of the Feynman propagator can also be carried over to the Euclidean version. Notice that the Wick rotation turns the Schrodinger equation into the Heat equation: t\.2
(5.79)
= 2m
2 ~
a ax
$(X,T)
- V(X)$(X,T)
As a consequence the Eualidean propagator is the unique solution to
the Heat equation:
satisfying the boundary aondition KE(xbiTa!XaiTa) = 6(x b - xa) Next we observe that the Hamiltonian operator is unaffected by a Wick rotation (since it does not contain the time). This evidently leads to the characterization: The Eualidean propagator has the following deaomposition on a aomplete set of eigenfunations for the Hamiltonian operator: (5.80) cf.
(2.70). Finally it can be reexpressed as a
ean time evolution operator (5.81)
1
matri~
element of the Eualid-
~
exp [ - fi HTl , i.e. 1
~
KE(XbiT!Xa;O) = <xblexp[ - fi HTl !xa>
As we have seen, the Euclidean formalism offers a chance of constructing in quantum mechanics a rigorous path integral which is not plagued by singularities and phase ambiguities. To some extent this is also true in quantum field theory. Consider the Lagrangian density
(For technical reasons U[~] should be at most a fourth order polynomial in
215
I
i.e. that one can get rid of various divergent expressions in quantum field theory using a so-called renormalization procedure). In the Euclidean formalism this Lagrangian density leads to the following Euclidean action (5.82)
SE[~l
=
J{!(~)
2 +
!(:~)
2 +
U[~l}d~dT
=
J{!6~Va~~av~
+
U[~l}d~dT
Notice that apart from a change of sign, the Minkowski metric has been replaced by the Euclidean metric. As in quantum mechanics, the Euclidean action is positive definite and the Euclidean path integrals
are of the nice infinite dimensional Gaussian type with a negative definite quadratic exponent. As a consequence it has been possible to construct a rigorous integration theory for such path integrals in one and two space dimensions. In three space dimensions (corresponding to our actual world) there are, however, still unsolved technical problems. In quantum mechanics the Euclidean formalism with its associated Euclidean path integrals is a funny trick, which is not really necessary (after all, it is easy to solve the original Schr5dinger equation). In quantum field theory it is on the contrary the only way one knows to construct a rigorous theory: First one must construct the Euclidean version and thereafter one wick rotates back to the Minkowski space. we now turn to a very interesting aspect of the Euclidean field theory. In a Minko.wskian field theory we have seen that the path integral is dominated by classical solutions with finite energy, and this we have been using to calculate a semi-classical approximation of the path integral. In the Euclidean version we can do the same, i.e. we expect the Euclidean path integral to be dominated by classical solutions to the Euclidean equations of motion. To investigate Euclidean configurations a little closer we look at the space of smooth configurations with a finite Euclidean action, Le. In analogy with the Minkowski case a Eualidean vaauum is defined as a configuration with vanishing action. Since the Euclidean action is given by
216
(5.82)
SE[$] =
f{i(~~)
2
I
2 +
i(l!)
ar
+
U[$] }drd,
this implies that
11 = 11 = U[$] aT ar
0
i.e. a Euclidean vacuum (like the Minkowski vacuum) corresponds to a zero point of the potential energy density:
with Next we observe that finiteness of the Euclidean action implies that the field asymptotically approaches a Euclidean vacuum. We then look for solutions of the Euclidean equations of motion. Obviously a Euclidean vacuum is such a solution, but there may also be non-trivial solutions with the following two properties a) It has a finite Euclidean action; b) It is a local minimu m of the Euclidean action (i.e. it is stable) • Since such a non-trivial solution is localized not only in space, but also in (Euclidean) time, 't Hooft has proposed to call it an instanton. The basic role played by the instantons is then that we expect it to dominate the Euclidean path integral and thereby help us with the calculation of the Euclidean propagator. At this pOint you have probably recognized a similarity with our discussion of solitons in chapter 4. This is no accident since there is an extremely close connection between instantons and solitons which we can formalize in the following way: Consider a Minkowskian field theory in D space dimensions. It is characterized by an action of the form
When studying solitons we look for a static configuration $(r,t) = $(r) with the following two properties: a) It has a finite static energy; b) It is a local minimum of the static energy (i.e. it is stable). Here the static energy is given by the expression: 2
E = f{!(li) static ar Next we consider a field theory in D spacetime dimensions, based upon the same potential energy density U[$]. The Euclidean version (5.83)
217
I
of this theory is based upon the Euclidean action
which is identical to (5.83)! As a consequence we therefore derive the following extremely important principle: A static soliton in
D
to an instanton in
D
space dimensions is completely equivalent
spacetime dimensions.
We shall apply this principle to solitons in one space dimension. They correspond to instantons in one spacetime dimension, i.e. in zero space dimensions! This example is therefore somewhat degener~. Now a field in zero space dimensions can only depend upon time, i.e. it is on theforrn $a(t). Consequently a field theory in zero space dimensions corresponds to a system with a finite number of degrees of freedom, i.e. to ordinary mechanics. If there is only one field component $(t), we can identify it with the position of a particle moving in a single space dimension, i.e. put x = $(t). Let us contretize this: A particle moving in a one-dimensional potential V(x) is characterized by the Lagrangian L
= !m(~)
2
- V(x)
In the Euclidean version this corresponds to the Euclidean action 2
+oe
SE[$ h)] =
f
Om (~~)
+ V ($) ]dT
where we have denoted the Euclidean path by x = $(,). Since , is now a space like variable, we can finally denote it by x whereby the Euclidean action becomes + V[ $ (x))} dx
which is precisely the same as the expression for the static energy in a (1+1)-dimensional field theory. Notice too a very pleasant surprise: Although we have started out with a non-relativistic theory for a particle moving in a potential V(x), we see that the corresponding Euclidean formalism corresponds to the static version of a relativistic theory in (1+1)-spacetime dimension! We can now take over the results obtained in the sections 4.3 4.4 concerning the kinks in (1+1)-dimensional field theories.
218
I
First of all a kink in a (1+1)-dimensional field theory corresponds to an instanton in ordinary quantum mechanics. E.g., a particle moving in a double well 4
~
V(x)
(x
2
2 2 - a )
is associated an instanton given by (5.84)
$(T) = a Tanh {a::f (T - 'oj}
cf. (4.23).
Similarly a particle moving in the periodic potential V(x)
)l4 f1 T l - Cos (IX)] 11 $
is associated an instanton given by 4
$(T) = ~Arctg
(5.85)
cf.
{ )lIT-To)}
e
(4.24).
Furthermore we have seen that a theory with a degenerate vacuum is characterized by a topological charge Q and that the static energy is bounded below by the topological charge due to the Bogomolny decomposition. In the Euclidean version this corresponds to a decomposition of the Euclidean action (5.86)
SE[$(-r)]
+00 m
2
J
{~ +" PV(*) d, m
P
$+ + [Q]
with
Q =
J$-
~)d$
where $(T) is a configuration interpolating between the classical and $+. This leads to the following simple picture: vacua L The space of all smooth configurations with finite Euclidean action breaks up into disconnected sectors characterized by different values of the topological charge Q. In each of these sectors the Euclidean action is bounded below by the topological charge
we expect the contribution from a given sector to the Euclidean path integral to be dominated by a configuration which saturates the lower bound, i.e. from a configuration satisfying the differential equation (5.87)
d$ = + flV($) err - m
In the lowest non-trivial sectors this leads precisely to the oneinstantons or anti-instantons (like (5.84) and (5.85). In the
219
I
higher sectors wekocw that we cannot saturate the bound exactly, but we can still corne arbitrarily close by considering multiinstanton configurations consisting of widely separated instantons and anti-instantons, cf. the discussion in section 4.4. Due to the close relation between solitons and instantons, Polyakov has suggested to refer to instantons as pseudo-particles, a label which is also currently used in the literature.
5.8
INSTANTONS AND THE TUNNEL EFFECT
Having understood some of the formal properties of instantons we now turn to the question of their physical significance. We know that solitons correspond to a particle like excitation of the field. In a similar way we will now show that instantons are associated with the so-called tunnel effect in quantum mechanics. Let us first recall that instantons in ordinary mechanics are associated with a particle moving in a potential well which has a degenerate minimum (like the double well or a periodic potential). As a consequence the classical vacuum is degenerate and formally the instanton·s represent Euclidean solutions which interpolate between two different classical vacua. We now proceed to examine the quantum mechanical ground state of such a system. To be specific, let us concentrate for a moment upon the motion in a double well (cf. fig. 63 page 231). Classically there are two ground states represented by the configurations and
x_It) := -a
corresponding to a particle sitting in the bottom of either of the wells. Quantum mechanically we must solve the Schrodinger equation and thereby determine the energy spectrum to see which eigenfunction has the lowest eigenvalue: _ fl2 d 2 ,,, + ;;;....x. V(x) ljI(x) '" E ljAx) 2m dx2
Based upon the classical analysis we can give the following qualitative description of the low lying energy states: Each of the two wells correspond to a harmonic oscillator potential. Thus the two lowest eigenstates will be related to the ground states of these two oscillators. Furthermore the potential V(x) is invariant under the inversion, x ~ -x, and consequently the eigenfunctions of the Hamiltonian can be chosen to be eigenfunctions of the parity operator pljI (x) = ljI (-x)
220
I
i.e. they can be chosen as either even or odd functions. In the case of the double well, the two lowest eigenstates are therefore of the form 1 { $ (x-a) - $ $ (x) = --
1jJ+(x)
-
12
0
0
1 (x+a~
J
where $0 is the ground state wave function for a harmonic oscillator (i.e. it is given by (5.25) where we put t=O). There are now two possibilities: Either the quantum mechanical vacuum is degenerate (like the classical vacuum), i.e. $+ and $_ has the same energy, or the Classical degeneracy has been lifted, i.e. $+ and have different energies. To find out which possibility is actually realized, we must compute the associated eigenvalues, i.e. solve the Schrodinger equation. We will not do this for a general potential, but to get a feeling for what is going on we look at a particular simple example:
w_
ILLUSTRATIVE EXAMPLE:
THE DOUBLE SQUARE WELL
~
Consider the double square well ~ shown on fig. 61. Because the Vex) potential is constant "piece by piece" it is trivial to solve the V Schrodinger equation. Due to the 0 symmetry of the wave functions we need furthermore only solve it in the regions I and II. Since the II I IV / III I potential at x=b is infinitely high, the wave function must :/ / / / / / / / / I vanish at x=b. At x=a we must -b -a a b Fig. 6-1 furthermore demand that the wave function and its derivative is continuous. This can be simplified by demanding that the logarithmic derivatives match,i.e.
~ ~ ~ ~/;
~
~ ~
~ ~
(5.88)
Okay, now for the solutions. In region I, the solution to the Schrodinger equation is given by $I(x) = A
sin{vI~
(X-b)}
~ ~ ~
:;~
x
I
221
(where A is a normalization constant which will soon drop out of the calculation). In region II, the solution depends on whether the total wave function is even or odd. For an even wave function we get the solution
{!
2m (V0 - E)
$II(x)
=
B
Cosh
2
fi.
x}
while in the odd case we get
x} In the even case the boundary condition (5.88) now reduces to (5.89a)
/2niE
Tan {v~ (b-a)
}
= -
E)
a}
f j2_m_(V....::o~_E_)
a}
~ Coth {/2m(vfi.2 v~ o
while in the odd case it reduces to (5.89b)
Tan{~~
-
(b-a)}
;--p;o
V~ Tanh 1 / L
2
fl
In the limit where Vo + 00, i.e. where the potential barrier is infinitely high, the two conditions collapse into the single cond~ tion
Tan{vI~
(5.90 )
(b-a)}
=
0
(Notice that both Tanh x and Coth x tend to 1 at infinity!). In this case the spectrum is therefore degenerate, i.e. both the even and odd wave functions have the same energy which is given by (5.91)
E -
fl2
n
2 2 11
n = 1,2,3, •••
- 2m (b-a) 2
(cf. the free particle spectrum (5.69». For finite Vo the spectra are however different in the two cases due to the difference between the hyperbolic tangens and cotangens functions. We can furthermore estimate the two lowest lying energies. If Vo is high, we expect the correct energy to be close to the "degenerate" value 2m (b-a)
i.e. we can safely put E = Eo + oE
2
222
I
Using the approximation
j
2mE
?
We then get
Tan{j:~E
(b-a)}
oE Tan { 2Eo
I~ ~
(b-a)
}
Tan{~2m~0 (b~a)}
(Notice that = 0 and use the addition theorem for the tangens TI function!). As loEI« Eo' we can furthermore forget about the tangens function itself so that the boundary conditions (5.89a) and (5.89b) finally reduce to:
!iiiiV
2EoTI
7(b~--a~)~v~2~m~V-o Coth{V'[2°.a} (5.92)
2E o11
- (b-a)V2mV
(even case) (odd case)
°
rhus the effect of including an additional potential well is one the one hand to produce a general shift in the ground state energy ;riven by (b-a) I2mV0 In the other hand it causes a small split between the energies of the even and odd wave functions. Using the asymptotic relations Coth x
1 + e
'"
-2x
and
Tanh x
1 _ e- 2x
x+oo
He in fact get that the even wave function has the lowest energy and that the energy split between the even and odd case is given by (5.93)
E_ -
E+ "'"
4Eoll (b-a) 12mvo
exp
[1 -!i
(IZmvo ) 2a
Returning to the case of a general double well we thus see that the classical degeneracy has actually been lifted. The symmetric ground state wave function suggests that there is an equal probability of finding it somewhere between the wells. This is related to the tunnel effect. Classically a particle with energy E+ cannot penetrate through the potential barrier, so it is totally unaffected by the presence of the second well. Quantum mechanically there is,
223
I
however, a small chance for the particle to be in the region between the wells. Semiclassically we can thus describe the ground state in the following way: Most of the time the particle is vibrating close to the bottom of either of the wells, but from time to time it leaks through the barrier and moves from one well to another, i.e. it tunnels forth and back. Notice that the instanton solution (5.84) precisely interpolates between the two classical vacua and we can therefore think of it as representing the tunneling between the two vacua. Similarly we may think of the multiinstanton solutions as describing the tunneling forth and back between the two vacua. Let us further consider the Euclidean equations of motion and the associated Euclidean energy: Minkowski version (real time)
(5.94)
Equation of motion Energy
m
E
;
dZx dt 2
Euclidean version (imaginary time) 2
;
dx im(dt)
d x m -2 d.
- V' (x) 2 +
V(x)
E
;
-
;
V' (x)
!m(~~)
2 +
V(x)
The Euclidean energy (which is the Wick rotated version of the Minkowskian energy) is indefinite, but it is still a constant of motion. In the Minkowski case, the demand of vanishing energy leads to the classical vacua, i.e. x(t) ; + a But in the Euclidean case it leads to the instanton equation (5.87), i.e.
dx d1:
;
+
-
I/V(x) • 21m
Thus the instantons are zero energy solutions which interpolate between the classical vacua. (The same is "almost" true for the multiinstantons) . This is now a general aspect of the tunneling phenomena: The existence of a Euclidean zero-energy solution which interpolates between two different classical vacua signals the presence of tunneling between these vacua, and this tunneling lifts the degeneracy of the classical vacuum.
I
224
Let us in the light of the above interpretation of the instanton return for a moment to the double square well. Here we have shown that the energy split is given by 4Eo11 {1 ~ (b-a)~ exp - ~ 2a /2mV0 o
}
This can be related to instantons in the following way: to (5.86) the action of a single instanton is given by
According
+a
f
~d$
Za/ 2mvo
-a Thus the energy split is precisely proportional to exp { This strongly suggests that the energy split is related to some Euclidean path integral of the form
t
So} .
x(oo)=+a
f
exp{- iSE[x(T)]} D[X(T)]
x(-oo)=-a
where we have evaluated the path integral by expanding around the instanton solution in the usual fashion. We shall make this precise in a moment. As a final remark in this section, we consider a scalar field theory in (1+1)-spacetirne dimensions based upon a potential density U[$] , which has a degenerate minimum. This includes our favourite examples: The $4-model and the Sine-Gordon model. what happens when we quantize such theories - does the vacuum degeneracy survive the quantization or is it lifted by instanton effects? The vacuum degeneracy persists! An instanton in this case would correspond to a soliton in two space dimensions, but the existence of such a soliton would violate Derrick's theorem (see sec. 4.7). Thus a scalar field theory in (1+1)-dimensiohs cannot support instantons. It can also be argued that two different classical vacua are separated by an infinite potential barrier. To be specific, let us consider the $ 4-mode l. To be able to "visualize" the argument let us furthermore use the string-analogy of the field theory, cf. section 3.1. The potential energy U[$] corresponds to a double valley separated by a ridge of height ~4/1X, see fig. 62. A classical vacuum corresponds to a string which lies in the bottom of one of the valleys. Now if we want to convert one vacuum into another, we would have to lift the string a=ss the ridge. It requires the energy
225
I
U-axis U[] -axis
X-axis Fig. 62
4 LL A
to lift the string across a ridge of length infinite amount of energy to lift the string long ridge. The same type of argument shows quantum mechanically stable, i.e. that it is into one of the vacua.
5,9
INSTANTON CALCULATION OF THE LOW LYIMl
Thus it requires an across an infinitely that the soliton is not allowed to decay
L.
ENE~Y
LEVELS
Now, that we know the general aspects of the instantons, we shall try to make an explicit calculation of the energy-split in the double well. As a byproduct of this calculation we will also be able to verify the previous given description of the ground state. Since we are not interested in the high-lying energy states we need not evaluate the total path-cum-trace integral. It suffices to calculate the asymptotic behavior of the propagator as T goes to infinity. This follows from the decomposition (2.70), which in the Euclidean formalism becomes: (5.80)
KE(xb;TblxalTa) = nL
~
En T }
In the large T-limit the right-hand-side will clearly be dominated by the ground state. Also the "amplitude" will be related to the wave function.
226
I
Consider e.g. a single potential well as the one sketched on figure 58. The classical ground state is clearly given by
which in fact is the absolute minimum of Euclidean action, so this path will dominate the path-integral. Consider a small fluctuation around the classical ground state n (T)
x(T)
Expanding the potential to second order, we get 2 2 2 V[X(T)] R1 W"(0)n (T) = tmw n (T) with The Euclidean propagator from given by
0
to
0
mw 2 = V"(O)
is therefore approximately
But the a~tion corresponding to the classical ground state is zero and the remaining path integral is just the Euclidean propagator for the harmonic oscillator. Consequently we get (cf. (5.76)):
we_then go to the large T-limit where we can replace the hyperbolic Sine by an exponential function, so that we end up with (5.95)
TI
T
KE(O;"! 0; - 2) '"
,tmW - ~'T VTiK e
T+oo
On the other hand we get from (5.80) the following asymptotic behavior: (5.96) A comparison therefore reveals that (5.97)
$0 (0) ""
4!IDw VTiK
and
This is in precise agreement with the previous obtained results concerning the harmonic oscillator (see especially (5.25)) respectively the weak coupling limit of the low-lying energy states in a single potential well (cf. (5.53)).
227
I
Having "warmed up" we now· turn our attention to the double well, cf. fig. 63. In this case the classical ground state is degenerate X.± (,) '= :!: a
and we shall correspondingly take a closer look at the following four propagators: K(-a;il- a ; -~)
K(+a;il+a;-~)
K(+a;il- a ; -~)
K(-a;il +a; -~)
The first two connect a classical ground state with itself, while the last two connect two different classical ground states. The calculation of their asymptotic behavior will be performed in two steps: In the first step we will deduce the general structure of the asymptotic behavior leaving a single parameter uncalculated. This will especially put us in a position where we can deduce the qualitative behavior of the quantum mechanical ground state. In the second step, which will be performed in an illustrative example, we will then calculate in all details the missing parameter. Then this will finally give us an eXplicit formula for the energy split. Okay, to be specific, we concentrate upon the propagator (5.98)
In this case the path integral is in particular dominated by the instanton solution (in the large T limit). Besides that we will also get contributions from strings of instantons and antiinstantons. Let us denote such a path by X
where
[n;'1"""n]
(t)
n is the total number of instantons and antiinstantons, while
'1"""n are the consecutive positions of the consecutive centers. The centers consequently obey the constraint (5.99)
-
i < '1
< '2 < .•• <
'n <
i
In order to get a string of widely separated instantons and antiinstantons the difference between two consecutive centers should actually be at least of the order of ~ (the width of the instanton). In the large T limit the subset of configurations where some of the instantons and antiinstantons overlap, will, however, be very small compared to the total set of configurations satisfying the constraint (5.99) and therefore we will neglect them. Notice too
228
I
that since we propagate from -a to +a the integer n must necessarily be odd. Okay, according to our general strategy for semiclassical approximations we must now expand around each such quasi-stationary path
and approximate the potential by 2 V[X(T)] '" V[X(T)] + W" [X(T)] n (T) The corresponding contribution to the path integral then decomposes as
where SE is the Euclidean action of a particle moving in the time dependent potential U(n)
=
!V"[x
(T) 1n
2
[nn 1'···.· ,Tn1
The remainingpath integral corresponds to the Euclidean propagator of a particle moving in the above potential. Let us for notational implicity denote this propagator by -
T
K (0;2 iO ;
n
-
T
2)
where we have suppressed the dependence upon the centers (T1' .•• ' Tn). Notice that the Euclidean action of such a string is approximately given by S
E
[~[ n;T1' ••. "n1 1 "" ns0
where So is the Euclidean action of a single instanton, cf. (4.38). Finally we must sum up the cpntributions from all the quasistationary paths, i.e. we must sum over n and integrate over T , ... ,T . The total approximation of the propagator (5.98) is thus n 1 given by
(5.100)
229
I
We then turn our attention to the funny propagator first the case of a single instanton centered at T=O. the potential -
U[n]
=
-
!V"[x(T)]n
Kn. Consider Notice that
2
degenerates to a harmonic oscillator potential except when we pass right through the instanton, i.e. outside the small interval [-~,~] the potential is given by
This suggests that we should compare the distorted propagator K with the propagator of the harmonic oscillator, which we will denote by Kw. We therefore put (5.101) K (0·!10· w '2 '
where ~ is supposed to be small. Using the known expression for the Euclidean propagator of the harmonic oscillator (5.76) we then get -
T
T
K(01-2 I01 - .".) ,t;
if mw = ~2lT'fi Sinh[wT]
Since we are going to investigate the asymptotic behavior in the large T limit, we can safely approximate the hyperbolic Sine by an exponential function and we end up with K(01j101 -
i)
~ ~e-WT/2 ~
Before we proceed to the case of multiinstantons, we want to make two important remarks about the behavior of the correction term ~ in the large T limit: In the large T (1)
~
(2)
~
limit
is independent of T, and is independent of the center of the instanton.
Proof: Suppose ~ values of T:
depended upon
T
and consider two different large
IZlFZ7l\ -(1
+(1
o
230 According to the group property (2.25) we then get ~ T2 T2 K(O;TIO; - T)
T2 T1 ~ T1 = ff~K(O;TIY;T)K(Y;Tlx;
[- T1/2;T1/2]
Outside
tor propagator.
n
= 0,
x
T1 ~ - T)K(x;
the curly propagator reduces to the oscilla-
Furthermore
small, except when
I
and
K(y;T1/2Ix; - T1/2)
is exponentially
yare close to the "classical" vacuum
cf. the behavior of the ground state wave function (5.25).
In the above integral we can therefore safely replace K(y;T1/2Ix; - T1/2) by ~(T1)Kw(Y;T1/2Ix; - T1/2). property once more we therefore get
Using the group
T2 T2 T2 T2 K(O;TIO; - T) ~ ~(T1)Kw(0;T!0; - T) ~(T2)
i.e.
argument.
= ~(T1)'
The second statement is proven by a similar
0
Notice too that the correction term corresponding to an antiinstanton is the same as for an instanton, since they generate the same potential
Urn].
We now turn our attention to the multi-instanton case.
Here the
"particle" moves in the potential
For large
T
and widely separated instantons and instantons:
T/2
- T/2
we then get exactly as before that the potential reduces to the oscillator potential except when we pass through one'of the instantons or anti-instantons.
As in the above proof we
From the group property (2.25) we get
c~n
safely replace
K(xk;Tk!xk_1;Tk_1) by
Using the group property once more we therefore deduce TI TnT T) ,n -wT/2 .~ Kn (0;-,: 0; - Z) "" ~ Kw (0;'11 0; - Z """ e Vifi
231
I
Inserting this our approximate formula for the full propagator (5.100) now reduces to
But neither So nor ~ depend upon the intermediary times '1""'n' Consequently the time integration can be trivially executed. Using a symmetry argument we quickly get
Inserting this we can finish our calculation: (5.102) n
=
rmw e -wT/2 v"iTh
Had we considered the propagator from -a to -a itself everything would have been the same except that this time n should be even. We would therefore end up with the analogous formula (5.103)
KE(-a;il-a; -
i)
~ ~
e-
wT 2 / Cosh
{~T
exp { -
~
T...oo
Finally the propagators are invariant under the sUbstitution a ... -a
-a ... a
putting everything together we have thus shown:
(5.104a)
(5.104b)
So}}
232
I
Hx)
-a
+a
Fig. 63a
Fig. 63b
Comparing this with the general formula (5.80), we see that the ground state has split up into two low-lying energy states $0 and $1' The energy split is given by (5.105)
E1 - Eo = 2fi.tI exp [ -
and the wave functions $0
and
~1
~
So]
satisfy:
Thus the ground state wave function $0 together with its close companion ~1 peaks at the classical vacua ~a, but the peaks are only half as great as the corresponding peak for the oscillator ground state, cf. (5.25)
~o (+a) = $0 (-a) = 12. ~ = $1 (+a) = -
$, (-a)
Clearly the true quantum mechanical ground state $0 corresponds to an even combination of the corresponding oscillator ground states $_ and ~+, while $, corresponds to an odd combination. Notice that in accordance with the node theorem in section 4.7 the even combination has no node while the odd combination has precisely one node. We have thus in all details verified the qualitative description outlined in the preceding section. Let us make some general comments about the method we have used: If we imagine that the instantons are some kind of particles (which they are not!), then we can think of an arbitrary distribution of instantons and anti-instantons as a kind of "instanton gas". To calculate the Euclidean propagator we are then summing up various types of configurations in this "instanton gas", but the crucial
233
I
point is that the calculation is dominated by configurations consisting of widely separated instantons and anti-instantons. For this reason the above method is often referred to as the dilute gas approximation.
We can now check if the method is self-consistent, i.e. if the main contribution to the approximation (5.104) really does corne from widely separated instantons and anti-instantons. Consider the intermediary result (5.102):
KE(a;il- a ; -
i)
~~
e-
wT/2
L
1 n!
{LIT exp [ - ~ Sol}
n
n odd
T~oo
Now, in a series like
L
n n!
xn
the main contribution comes from the terms where n
~
x
(As long as n < x, the factorial is beaten by the power and vice versa). In our case the main contribution thus comes from terms where
i.e.
But niT is the mean density of the instantons and anti-instantons, and since the right hand side is supposed to be small, this mean density is very small. Thus the main contribution really comes from "dilute gas" configurations. As another application of the dilute gas approximation we now consider a particle moving in a periodic potential, sayan electron moving in a one-dimensional crystal. Thus the potential is very analogous to the potential in the Sine-Gordon model, and you should keep the Sine-Gordon model in the back of your mind when reading the following discussion. In the case of a periodic potential the classical ground state is infinitely degenerate (cf. fig. 63b) Xj (T)
=
ja
j
= ...
,-2,-1,0,1,2, ...
We shall correspondingly try to estimate the propagator which connects two such classical ground states which we label by the integers Land J' + ••
234
I
The corresponding path integral is dominated by multi-instanton configurations, i.e. by strings of widely separated instantons and/or anti-instantons.
But notice that this time there is no restriction
on the positions Qfthe anti-instantons relative to the instantons. If we label such a multi-instanton configuration by
(where
n
is the number of instantons,
m
the number of anti-in-
stantons etc.) then the only restrictions to be placed upon
n,
m
and the centerpositions of the instantons, respectively the antiinstantons, are accordingly given by n - m
=
j+ -
- 2T
L
< ~1 <"'<~m < 1: 2
Following the same chain of arguments as in the case of the double well we therefore obtain the following asymptotic estimate:
Using that 211
~ vab = 211
f
e i6 (a-b) de
o we can in fact get rid of the restriction on the summation over and
m
n
and we thereby get
But then we can explicitly perform the double summation which, in fact, decomposes into the product of two separate summations:
_ ,/mW -V'rt
e
-wT/2
1
2rr
f
211
-26T
e
exp
{
-
~ u
S }cose 0
e
-ie(j+ -
j_)
de
o So the calculation has come to an end, and after a little rearrangement we get the final result:
235
I
(5.107)
Now, how are we going to interpret this asymptotic behavior? If we compare it with the general expression for the asymptotic behavior of an Euclidean propagator (5.80), we see that this time the classical degenerate vacuum has in fact been split up into a contin-
uous band of low-lying energy states e.
~e(x)
labelled by the angle ~e
Furthermore the energy of the eigenstate
ground state
(relative to the
e=O) is given by
(5.108)
E(EI) "" 26exp {-
i
so} (Cose - 1)
Notice also that '" '!'
(.
EI
Ja ) -_
4~w -
-1- e ije 11ft I21i
This suggests that the eigenstates
~e
is a superposition of the
form
~(x-ja)
(5.109) where
~
(x)
e
ije
is the oscillator ground state wave function, cf.
(5.25).
Notice that the periodic potential possesses a discrete symmetry, since it is invariant under the translation
x ... x + a If we neglect tunneling effects, and correspondingly try to represent the quantum mechanical ground state by the discrete series of wave functions
The effect of the in-
stantons is thus, precisely as for the double well, to restore this symmetry:
The correct eigenstates
are now eigenfunctions of the translation operator:
is
~e(x)
All the preceding results are in fact well-known from solid state physics, where wave functions, which like
the translation operator, are known as Bloch
are eigenfunctions of ~aves.
236
5.10,
I
ILLUSTRATIVE EXAMPLE: CALCULATION OF THE PARAMETER
6
So far our discussion of the double well has been dominated by qualitative arguments.
Now we will finish the calculation, so that
we in principle can compute the actual figure representing the difference in energy between the two lowest lying states. We have already deduced the formula: (5.105) In this formula
So
is the action of a single instanton, which
according to (5.86) is given by +a
(5.110)
=J
So
12mV(x) dx
-a Thus it is a computable number once we have specified the potential. The other parameter
6
is, according to (5.101), given by -
T
T
K(O;~I 0; - ~)
Here
K
is the Euclidean propagator corresponding to the motion of
a particle in the potential U[x]
=
2 lV"[x (1:)]x o
Xo (1) is the one-instanton solution (5.84), and Kw is the propagator of a harmonic oscillator with the frequency w. In
where
T-+oo
x(+I)=o
f
2
T
T
_f +J2 dx 2 1 eAll-~ f(d1) + w2x 2}d1]D[X(1)]
x(-z)=O
-2
One way of calculating this ratio is to use the relationship between path integrals and determinants. thus get
lim
T"''''
According to (5.46) we
237 T~
But in the limit where
I
we know in fact that the differential
operator, (5.111)
m
V" [~(-rll
has a zero eigenmode given by (5.112 ) i.e. by construction
~o
is an eigenfunction of the differential
operator (5.111) with the eigenvalue O.
This is the old story which
we encounter every time we have a symmetry in the Lagrangian which is broken by the particular solution in question, cf. the discussion of the translational mode in connection with the solitons (sec. 4.7). But the other differential operator
,2
(5.113)
- w2
- 0
has no zero eigenmode.
It's lowest eigenstate is given by: 7[2 Le. ~1 (-r) = Sin[;: 1 +-::2 T
As a consequence only the first determinant goes to zero in the large
T
limit.
Evidently this implies that as
But this is a disaster!
T ...
00
The whole discussion preceding this illu-
strative example was based upon the assumption that
6
is a very
small quantity, which is furthermore independent of
T
for large TI
So where did we go wrong? Fortunately it turns out that all the conclusions in the preceding discussion are in fact correct, but that we have been a little
6, where we have in fact committed
sloppy in the introduction of subtle but serious error.
To understand that, we must reconsider the whole philosophy behind our approach: To calculate a path integral, we proceed in two steps:
First we find the quasistationary paths, i.e. in the pre-
sent case the multiinstantons
x[
ni
-r 1'···
,'f n
1(-r)
Next we consider the fluctuations around each such configuration:
x(-rl = x[ T~e
n;'1''''''n
1 (-r)
+ n (,)
path integral is then evaluated as a sum of contributions coming
from each quasistationary paths, i.e.
238
I
(where the sum over the instanton centers integration!).
'1' ... "n
is actually an
Each of the "shifted" path integrals is thereafter
evaluated in the Gaussian approximation.
The problem now arises
because the set of quasistationary paths is labelled not only by a discrete parameter, but also by continuous parameters.
To study
this in detail consider the one-instanton contribution. from a curve in path space, the instanton.
x, o (,)
where
'0
It comes
labels the center of
we perform the Gaussian approximation corresponding to fluctuations around counting".
x
'0
we in fact make a "double-
The fluctuations along
the instanton curve in path space include contributions from the nearby one-instanton paths
but they have already been included
'0 Thus we see that when performing the Gaussian approximation we ought only include fluctuations
when we "sum" over the positions
which are perpendicular to the instanton curve in the path space! Notice that (5.114 ) Consequently the fluctuations along the instanton curve are precisely generated by the zero eigenmode, i.e. these fluctuations are of the form
This indicates a connection with our determinantal problem:
On the
one hand it is the zero mode which generates fluctuations along the instant on curve (and thereby causes a "double counting" in our naive approach).
On the other hand it produces a zero-eigenvalue in the
determinant (as
T + "') •
We can now resolve the problem in the following way: When calculating the shifted path integral in the Gaussian approximation, we
239
I
must only integrate over fluctuations which are perpendicular to the instanton curve.
Let us denote the resulting path integral by:
Let us furthermore consider the associated differential operator (5.111).
It possesses a complete set of normalized eigenfunctions
{ -a T2
-
!V "[ X0 (T) m
where we know that where
T
l}
n (T)
is proportional to
goes to infinity).
dXo/dT
(in the limit
Normally we would integrate over.all
the "fourier components" in the shifted path, nIT)
=
ao
I
an
n"'1 but in the above path integral we have excluded integration over a ' o since this is to be replaced by an integration over the instanton center
To'
This causes, however, a slight normalization problem.
In the standard evaluation of the path integral (leading to the determinant) we use the integration measure
cf.
(5.48).
We must now reexpress this in terms of
dTo'
According
to (5. 114), we get
and consequently
The correct integration measure associated with the integration over the instanton center is therefore given by (5.115 )
j~dTo
When we originally performed the integration over the instanton center, we missed the factor in front of ed in the corrected definition of 6.
dTo'
This will now be includ-
To sum up the correct defini-
I
240
n
tion of
is therefore given by
(5.11b)
=~ T+oo lim
lim T
+
00
[
!'Jet{-a~
- lVII[XO(T))}]-l
Det{-a~ -
w2
}
(where the stroke indicates that we shall omit the lowest eigenvalue in the evaluation of the determinant!). computed using the following strategy. T
T
[- 2; + 7 1
This expression can now be On a finite interval
the differential operator (5.111) will not have an exact
zero eigenmode (i.e. the smallest eigenvalue vanishing in the limit where
T
+
Ao(T)
will only be
Using the determinantal rela-
00).
tion (5.50) we therefore get:
(5.117)
AO(T)Sinh[WT1]lz Wfo(!;T)
/5
Ao(T)eXP[WT1]!; lim
T
+
00
fo(~)!
It remains to calculate a)
Calculation of
By definition
[ 2wf (!;T) o
f
fO(T)
and
Ao(T):
T
o
(7): is the unique solution to the differential
equation (5.118)
{- a2T
-
1m
V"[ 2 (T)l} fO(T) 0
o
which satisfies the boundary condition
o
o
241
cf. the discussion in section 5.5.
I
One particular solution to the
above differential equation is
dx0
1
S
In the large
T
-,
-
o
d1:
limit this is properly normalized and must there-
fore have the asymptotic behavior:
since (5.111) reduces to (5.113) for large
1:.
Here
Co
is a con-
stant, which by construction, is extracted from the asymptotic behavior of the one-instanton solution deL (5.119 )
lim 1:-+0>
Here.the parameter So
is given by
(5.110) .Thus Co can be computed once
we have specified the potential. follows that
$0
Since
Xo
is an even function.
is an odd function, it
Consequently we may summa-
rize its asymptotic behavior as follows: (5.120)
I
CO e
$0(1:)
Co
-W1:
e W1:
as
T
as
1"..... T ...,..
~
+00
The second order differential equation (5.118) has two independent solutions. $0
The second one can be constructed explicitly from
as follows:
Put 1: dS
Jo $~
(5.121 )
(Notice that $0
has no nodes and that
$0
since the integral is 1:-dependent).
(S) ~o
identity we now get 1: I
~o (1:)
$~(1:)
~~ (1:)
~~ (1:)
dS
J0 $~ (S)
1
+~
and 1: dS
J $~ (S) 0
They imply the following two identities: (5.122)
W[$o;~o]
d1/J
= $0
erro -
~o
is not proportional to
Differentiating the above
d$o
I
242 and (5.123) (The second one suffices to show that ~o
Since
~o
is another solution). ~o
is an even solution to (5.118), it follows that
is
an odd solution. We are also gOing to need the asymptotic behavior of
~o'
From
(5.120) we get T
C e -WT o
JC2 d~2WT e o
The asymptotic behavior may thus be summarized as follows
(5.124)
~O(T)
R;
{-
1 WT 2C w e o
as T ...
-WT 1 e 2C w o
as T
+00
-+
We proceed to investigate the particular solution the one we are really interested in. combination of
~o
and
fo
which is
We know that it is a linear
~o:
fo(T) = A~O(T) + B~o(T) Clearly
A
and
B
can be expressed in terms of Wronskians:
and
But these Wronskians may also be computed directly using the boundary f . o·
conditions satisfied by W[~o;fol
=[<11
W[~o;fol
=[~ o ~dT
and
o
d~
dfo_ f dT 0
df]~ =_ f
df
d~
f
0
dTO]IT
T
~o(- 1)
T
=-
T
~o( - 1)
"2
Consequently (5.125) From this expression, which is exact, we can extract the asymptotic behavior of
fo:
243
1
2w
Especially we get: in the limit of large
(5.126 )
T.
b) Calculation of Ao(T): To compute the lowest eigenvalue we must first determine the solution
fAIt)
to the differential equation
(5.127 ) which satisfies the boundary conditions
T T f A(- 2)=3,f A (- 2) The eigenvalues
0
A are then determined from the subsidiary condi-
tion
For small values of
A we can now use the approximation:
(5.128)
fA (,) ~ f o (') + A
dfA (B\I
A=O If we introduce the function dfA
= cn:-Ih)
A=O we get from (5.127) that it is the unique solution to the differential equation: (5.129) which satisfies the boundary conditions T d T
,
by: atP o (') + 131/10(') +
J T
- 2
[1/Io(s)tP o (') - tP o (s)1/Io(')] fo(s)ds
244
I
But from the above boundary conditions it follows that both S vanishes, i.e. ~ is given by
~
and
,
~(,)
(5.130)
J
=
[1/Io(s)tP o (') - tP o (s)1/Io(')] fo(S)ds
T
- 2" Inserting this in (5.128) we thus end up with the following approximation: fAt,)
~
,
J
f o (') + l
[1/Io(s)tP o (') - tP o (s)1/Io(')] fo(s)dS
T
- 2" But
fo
is given by (5.125).
properties of
tPo
and
1/1
T
Inserting this and using the symmetry
we now get in the large
0
2"
T
f l ('2") ... T.,.."
fo(i) +l
J {tP~(f)1/I~(s)
-
1/I~(i)~(s)}
T
limit:
ds
T
- 2" But in the same limit T
2"
+00
J !(S)dS ... J
.~(S)ds
T
2" while T
2"
J 1/I!(S)dS ...
A •
e
wT
T
- '2 Consequently the first term is bounded (since tP o2 (.'!'.) goes like 2 e- WT ) while the second term grows exponentially. Inserting the asymptotic behavior of fo and' 1/1 we thus end up with: 0
It fOllows that the lowest eigenvalue is given by (5.131 )
lo(T)
'"
4C~
we-
wT
245
I
T f o (2)
Now that we have calculated both return to the formula for
~.
and
Ao(T)
we can
Inserting (5.126) and (5.131) into
(5.117) we obtain:
rn ~ SW
C
(5.132 )
__ 0_
o
Thus the energy split (5.105) is given by E1 - Eo ::: 2C
(5.133)
o
VmS~flW
~
exp { -
So}
This ends our discussion of the double well. As an example of the application of the above machinery you can now work out the detailed formula for the energy split in the double well specified by the fourth order potential *)
,4
T
v(x)
(x
2
2 2
-
JT)
*) Details can be found in: E. Gildener and A. Patrascioiu, Phys. Rev. D16, 423 (1977) .
SOLUTIONS OF WORKED EXERCISES: No. 5.2.2
(a) From Taylor's formula we get Hn(x) =.a::.... az n
exp[2xz_z 2 ]=e
Iz:O
X2 n -"- exp[_(x_z)2] "Zn
Iz-O
2 n a {exp[_(x_z)2]} = (-1) ne x ---
axn
Iz=O
(b) By completing the square _u 2+2iux = _(u_ix)2_x 2
we can immediately perform the Gaussian integral:
'"-u2+2iux
Je
du
246
I
Using Rodriques' formul~ we now get 2 n (_2i)n x 2J n -u 2+2iux i=lln x a J -u +2iux r= e - - e du du fiT e u e .11 axn (c)
I
n=O
r
. L 2....n (-2i )2n e X2+y2J JUnv n e-U 2+2·lUX-V 2+ 2lvy dudv n=O 2n n! 11
We can now easily perform the Gaussian integrals whereby we obtain
PART II
THE MOBIUS STRIP
PARITY VIOLATION!
BASIC PRINCIPLES AND APPLICATIONS OF DIFFERENTIAL GEOMETRY
II
248
GENERAL REFERENCES TO PART II: M. Spivak, "A Comprehensive Introduction to Differential Geometry", Publish or Perish (1970). S.1. Go IdberC], "Curvature and Homology", Academic Press, New York (1962). G. deRham, "Varietes Differentiables.
~'ormes,
Courants, Formes
Harmoniques", Hermann, Paris 1960 V. Guillemin and A. Pollack, "Differential Topology", PrenticeHall, Englewood Cliffs (1974). 1.M. Gelfand and G.E. mic Press, New York L. I. Schiff, "Quantum
~,
"Generalized Functions", Acade-
(1964). ~Iechanics",
McGraw-Hill Kogakusha ltd (1968)
A. Jaffe and C.H. Taubes, "Vortices and Monopoles", Birckhauser (1981) •
249
II
ehapter 6 DIFFERENTIABLE MANIFOLDS TENSOR ANALYSIS 6.1
COORDINATE SYSTEMS
To simplify our discussion we shall work entirely with subsets of Euclidian spaces! l-le will assume the reader to be familiar with Euclidian spaces, but to fix notation some useful properties and definitions are collected in the following table:
II
n The Euclidian spaceR consist of all n -tuples (xl, ... ,xn) n where xl, ... x are arbitrary real numbers.
i Rnis a vector space with an inner product ['inea:t' s truature :
Addition:
I
x+y
Multiplication with a scalar:
Ax=
, [Xx"'nl ]
A
[~Xx"'nl] A
I
Algebraic structure
Inner product: <xiY> = xlyt+ ... +xnyn The inner product has the following characteristic properties: 1. Symmetry: <xlp =
r
The inner product generates the Euclidian norm:
250
II
Metric: The distance between two points x and y is defined as
II
d(x;y) =
x-yll = .; (xl_yl)2+ ... +(xn_yn)2
Balls: The open ball with center Xo and radius <: is given by
B(x ;E) = { x€ Rnl d(X;\) <<: } o . "", ... , ,
--
,
~,
E
,
,\
r---:, 0,
' ..... _-,,,
Open sets: A subset A c Rn is called open if each point in A can be surrounded by a ball lying entirely in A.
ITopological I structure
Neighbola'hood: A is a neighbourhood of Xo if it contains a ball centered at Xo •
,....... -...,
• + • ' ...... _~I
A
Convergence : A sequence xn converges towards Xoo
if:
A~B,is called continuous if it has one of the following two , equivalent propert ies :
Continuity: A map, f :
1. It transforms 'every convergent sequence into a convergent sequence, i. e. x ..... xCI) implies n that f(X ) -+ f (xc), n 2. The preimage f- 1 (U) of every open set U in B is itself an open set in A.
251
II
n Observe, that once we have equipped the Euclidian space R with a metric, the rest of the definitions are standard definitions that are common for all metric spaces. In what follows we shall especially be interested in subsets of Rn which might be referred to as "smooth surfaces". Consider, for instance, the unit sphere 52 in the three-dimensional Euclidian space R': 5 2 = { xE R'I (XI)2+(X2)2+(X')Z= 1 } If you prefer you may think of 52 as a model of the surface of the Earth. "ow we can clearly characterize a point on the surface through its Cartesian coordinates (X I ,X 2 ,X'). This on the other hand is rather clumsy because one of the Cartesian coordinates is superfluous: We can use the equation (X I )2+(X 2 )2+(X')2= 1 to eliminate e.g. x'. Because it suffices to use two coordinates when we want to characterize points on S2 we say that 52 is a smooth two-dimensinal surface in R'. In geography it is customary to use lattitude and longitude as coordinates. In our mathematical model they are referred to as polar coordinates (6,lp) 6 being the polar angle while
~
,
is
the azimutal angle.
Let us try to sharpen these ideas. Let M be a subset of the Euclidian space RN . We want to give an exact meaning to the statement:" M is a n-dimensional smooth surface". The key concept is that of a aool'dinate system: Let U be an open subset of Rn and let ~: U ~ M be an injective map. AS ~ is injective, i t establishes a one-to-one correspondence between pOints in the domain U and points in the image ~(U). Thus
each point P in ~(U) is repren sented by exactly one set of coordinates (xl, ... ,x ) (see fig·65 ),
We may therefore characterize the points in ~(U) through their coordinates. But that is not enough! To be of any interest ~ must respect the topology on M. Being a subset of RN we know that M is equipped with a metria. Obviously we must demand that points in M lie close
252
II
to each other if and only if their coordinates lie close to each other! We can formalize this in the following way: Let P be an arbitrary pOint in
~(U).
The coordinates are expected to
give information about the properties of the surface around P. But then it is necessary that the coordinates covers a neighbourhooe of P. We can achieve that by demanding that
~(U)
should be an open subset
of M.
u
(x
,~, I'
(x +e:
,x)
I'
IX
"" +e; ) Fig. 65
Hence if Q is sufficiently close to P, then Q lies in represent it by a set of coordinates:
(XI+EI
I ' ••
xn+e:
TI
~(U)
and we can
).
Now, let(P i )i=1, ... be a sequence in ~(U) convergirg to anoint P in ~(U) and look at the corresponding coordinates: (Xi""'X~) and
(X!"",X~). We want the following to hold: Pi converges to P oo if and IX~) converges to (x!, •.. ,X~) I but this v!e can achieve by demanding that 4> is a homeomorphism, i. e. that 4> and the inverse map
only if (xi,'" ~-I
are continous maps! n By demanding that 4> is a homeomorphism from an open subset UC R to
an open subset 4>(U) c M we have thus guaranteed that the coordinates respect the topology of M. But it would be also be nice if we could express t:1e smoothness of M in terms of the coordinate map 4>. First we observe that ~ can be regarded as a map, U ~ RN, Y 1 = ~ 1 (x 1,
••• ,
xn)
But then we must obviously demand that if> is a smooth map, i.e. that the components
~l, ••• ,~
.2.£j ax
have partial derivatives
~ '
J"
ax ax
k'
a3~i
J" k I' etc. ax ax ax
of arbitrarily high order! If
~
is not smooth we could easily be in
trouble as you can convince yourself by considering the example:
253
(x,y,1 xl)
II
(See figure 66 )
~
(x,y, Ix!)
y-axisj
~aXiSx-axis FIG. 66
But even that is not enough! Given a smooth map
(6.1)
We say that trix
lr1£1 I ax
and that
j
is regular at a point xo~(x~, ... ,x~) if the Jacobi ma-
has maximal rank at that pOint. The maximal rank is n x
o
is regular simply means that the n-colum vectors
~ ax
, ... ,
~ ax
~ ax
~ ax
are linearily independent. We will demand in U. This is
a
to be regular everywhere
rather technical assumption, but we will give a sim-
ple geoQetrical interpretation in a moment. Let us first look at an example which shows that the assumption is necessary:
~iS \ x-aus
x-axis FIG. Ip.
this case
67
is a smooth map, but the Jacobi-matrix:
254
is not of maximal rank when x = O.
Il
Thus
~
is singular along the
y-axis. The corresponding points in M form the sharp edge! Let us sillMilarize the preceding discussion:
Definition 1 n Let M be a subset of~. A coordinate system on M is a pair (
~ (~
(b)
~
is a homeomorphism trom the open set U onto the open set <prU) respects the topology of M)
is a smooth regular map.
(~respects
You may consider a coordinate system
the smoothness of M).
(~,V)
as a parametrization of
a "smooth surface", but this parametrization need not cover the whole surface. Actually there need not exist a single coordinate system covering a qiven "smooth surface". The sphere,e .g., cannot be covered by a single coordinate system (this will be demonstrated in section 6.2). Let us look again at the Euclidian subset MCR N . Suppose that we have a family of coordinate systems
(~i,Ui)i€I
all of the same dimen-
sion n, and suppose furthermore that this family covers M, i.e.
Mc
In this case some of the coordinate systems may of course overlap, which means that the same point P is represented by different coordin nates, say (x1, ••• ,x ) with respect to the coordinate system (~ ,V ) 1
1
and (yl, ... ,yn) with respect to the coordinate system (
coordinates!
255
R~ To investigate
t~is
U12~
FIG.
69
closer we now intrOduce the sets (See fi0. 69) -1
-1
U21 ~
(
Observe, that the overlap region,CPl (U I )n
and CP2 are continuous maps we con€lude that Ul' and D21
are open sets. We now introcuce the transition functions
(See fig. 70)
_1
CP21 cP
12
2 0
,Pt
-1
'1>2
~
(6.2) ~
Suppose P is a pOint in the overlap region. Then P is represented by n (Xl, •.. ,x ) and (y 1, .•. ,yn) ,where
two sets of coordinates,
(x 1 , ... ,x n ) ~ CP-:(P) arrl
(yl, ... ,yn) =
gjJ (Y+,",Y
(xl ~,xn)
n
n
R
R
FIG. 70
But then it is evident from the diagram, fig. 70,that: (yl, ... ,yn) = CP21 (xl, ... ,x n ) Thus the transition function CPu has a simple meaning: It
new coordinates (yl, ... yn) in temzs of the oZd ones n (x 1 , ... ,x ) = CP12(yl, ••• ,yn)
(x1, ...
:xr).
e~resses
the
In a similar way:
256
II
i.e. the second transition function $12 expresses the old coordinates in terms of t:1e new ones! As
and
it follows
t~at
-I
I
0
the transition functions are homeomorphisms. It will,
however, be important later on that
~~e new coordinates (yl, ... ,yn)
not only depend continuously but even smoothly upon the old coordinan tes (xl, ... ,x ). Therefore we demand that the transition functions are smooth functions. vfuen the transition functions are smooth we will also say that the two coordinate systems are smoothZy re Uzt:ed. The preceding discussion motivates the fOllowing definition:
Definition 2 Let M be a subset of RN. An atZas on M is a fcurrUy of coordinate systems ($i'Ui ) i EI with the following properties: (a)
The famiZy (
Any two coordinate systems in the fami Zy are smooUy rdated.
(b)
/
6.2
DIFFERENTIABLE MANIFOLDS
Let us look at subset MCR pose f
:
N
equipped with an atlas (
M... R is a map from M to R. We know what it means to say that
"f is continuous" because M is a metric space. Let
Po
be an arbitrary
pOint in M. Now we also want to assign a meaning to the statement: "f is differentiable at the,point PO" Let us choose a coordinate system (
=
$tixL •.. ,x~).
Then we can represent f by an ordinary Euclidian function
1 defi-
ned on the domain of the coordinate system 0
I
,
n
y=1 (x', ••• ,x ) with n
(Xl, ••• ,X ) eOI
where
I =
fO
)~
'Q~~ ~o) I";) i _____ f_=_f_.
Fig. 71
/
f(P
I
But as f itself is defined on a neighbourhood of Po we see that the Euclidian representative 1 is defined on a neighbourhood of (x~, ••. ,x~
-
II
257
Consequently we define Definition 3 f is differentiab le at the point Po if and only it. the Euclidian representative is differentiable in the usual sense at (xL ... ,xo). You might be afraid
t~at
f
the definition is meaningless, because you
could choose another coordinate system
(~2,U2)
covering PO. Then
f
would be represented by another Euclidian function i'and what are we going to do, if it turns out that
i
is differentiable at (x5, ... ,x~),
while i'is not differentiable at (y!, ... ,y~)? But do not be afraid! i'
is gi-
ven by the expression
"f' = fO~2t which you can rearrange in the following way: -1
i'= fO~2= fO(~10~ I)O~2 -1
(fO
0
( 1 O~2) = iO
•
As the two coordinate systems are smoothly related, we know that the transition function ever
i
~lZ
is a smooth function. Thus
we see that when-
is differentiable, so is i ' !
Because an atlas makes it possible for us to introduce the concept of a differentiable map, f: V
+
R, we say that an atlas generates a
differentiable structure on M. Now suppose we have been (~j,Vj)j€J
and
~~ven
two different atlasses
(i,Ui)i€I
which cover the same subset M. If all coordinate sy-
stems in the first atlas are smoothly related to all t,1e coordinatesystems in the second atlas, then they will obviously generate the same differentiable structure, i.e. a continuous map, f:M tiable relative to
t~e
~
R,
will be differen-
first atlas exactly when it is differentiable
relative to the second atlas. This motivates the following concept Definition 4 An atlas (~i,Vi)il:;[ on a subset M is called maximal if it has the following property: Whenever (~,V) is a coordinate system which is smoothly related to all the coordinates systems (~.,V.) in our atlas, then (~, V) itself belongs to our atlas, i.e. N, V) = ( .,V.) for ~om~ jQ;.X
J
Thus
J
.
a maximal atlas comprises all possible smooth exchanges of the
coordinates on M.
258
II
Given an atlas, say ($i,Di)iEI' you may easily generate a maximal atlas. You simply supply the given atlas with all possible coordinate systems that are smoothly related to the given atlas. The use of maximal atlasses will be important for us when we later on are going to study the principle of general covariance. We can now formalize the definition of a
differ~iable
structure
Definition 5 A differentiable struetu:ro on an Euelidian subset M is a maximal atlas on M. Finally we can give a precise definition of what we understand by a "smooth surface", which we from now on will refer to as a differentiable
manifold: Definition 6 A differentiable manifold M is an Euelidian subset M equipped with a differentiab le struetu:ro. Let M be a differentiable manifold. If we can cover it with a single coordinate system, we call it a simple manifoZd. In that case a map, f:
M~R,can
be investigated through a single coordinate representative:
f
(Xl, •••
,x n )
In general, however, we will need several coordinate systems to cover
M. A map,f: M--R,must then be represented by several coordinate representatives: -
I
n
fi (x , ... ,x )
corresponding to the various coordinate systems
(~i,Di).
In the oVer-
lapping regions Q
ij = $i (Di)1) $j (D j )
the coordinate representatiJes are tpan patched together through the relations -
I
n
fi (x , •.. ,x )
-
I
n
fj (y , •.• ,y )
where (See fig. 72 ) Finally we mention that a n-dimensional differentiable manifold M n is frequently denoted M , where the dimension of M is incorporated explicitly.
259
II
l'i g. 72
The sphere.
Illustrative example:
We are now armed with heavy artillery! Let us return to the sphere S2 and see how the machinery works. We want to construct an atlas on the sphere, and thus
we must con-
struct some coordinate systems. If U denotes the unit ball in R2, we may define a map, ~: U~S2,in the following way:
~(XI,X2) = (xl,x', ';1-(X ' )'_(X 2 )2) Clearly this map has a simple geometric interpretation: If we identify R' with the x ' -x 2-plane in R 3 , then ~ is nothing but the projection along the x 3 -axis! (See fig.73). Let us check that (q"U) is acI 2 3 (X ,X ,X )
tually a coordinate system: U is obviously open and q,(U) is the northern hemisphere,
x
2
(U) ={xES 2 [x 3 > O}
,
which is open too (when regarded as a subset of the topolo-
I 1<1>
gical space S2!) As (Xl ,X2) "',; 1- (Xl) 2_ (x2) 2 x
l
2
(x ,x )
2
FIG. 73
is a continuous map it follows that
is continuous!
-1
is
nothing but the projection on the first two coordinates
II
260
-1
Hence
is continuous. T.hus we have shown that
of S2. Then we must
We know that the square-root function 'it is smooth when t>O. (1\t t=O it is not smooth due to the fact that:
d
2~
(It)
-> '"
when
t
->
0+ ).
But we have restricted U to values of (Xl,X2) where (Xl)2+ (x2)2<1. Thus there are no troubles and
[~] ax6
~ ax d
2
axr ~ ax
a
0
~ ax ~ ax
J
0 -Xl 1-(xl) 2_(X2) 2
_x 2
J
1- (Xl)
2_
(x 2 ) 2
~ and we see that ax] always have rank two due to the upper square unit matrix. Consequently we have also shown that
To see that they form an atlas we must check that they are smoothly related, (s ee figure 75).
261
II
x
2
FIG. 75
From the diagram you read off:
(XI,X2)~(XI,X2,
A_(XI)2_
Hence the transition function yl=
~21
(X 2 )2)
is given by,
Xl
and the corresponding domain
UJ2 = [
-r
~(XI,A_(XI)2_
(X 2 )2)
(X I ,X 2 )
I x2
> 0,
U1 2 by
(XI)2+
(X 2 )2 <
1}
As the square root never obtains the value 0, the transition function ~12
is obviously smooth: In a similar way you can check that all the
other transition functions are smooth and thus we have shown in all details how to construct an atlas on S2. Consequently S2 is a differentiable manifold and we shall refer to the coordinates constructed above as standard coordinates. What about the polar coordinates correspond to a coordinate system ~(8,~)
=
(Sin8Cos~,
(8,~)?
(~,U)
Sin8Sin~,
But what can we choose as our
In our new context they
where:
Cos8) ~-axis
domian U? There are two kinds of troubles: The first trouble is not so serious. Due to the peri-
8-axis
odicity of the trigonometric functions several values of the coordinates actually describe the same point: ~(8,
~)
~(8,
~
~)
~(-8,~
(8,
~
+ 2IT) + IT)
262
II
But the second trouble is more serious. At the northpole and the southpole the coordinate
~
is completely indeterminate! Let us com-
pute the Jacobi matrix:
[a~~] ax]
2Jtl
~ll
as
a~
~2
£.t.2
ae
a~
~3
£.t.3
ae
a~
CosS·Cos~
-SinS·Sin~
CosS·Sin~
SinS·Cos~
- SinS
0
At the northpole (S=O) or the southpole (S=n) we see that sinS =0, and therefore we find
[~ which shows that
~
o o
o
is singular when S=O or S=n.
Thus
the coordinate
system breaks down at the northpole and the southpole and we will have to exclude these from the range of $ lar pOints.
because they are singu-
(Let me emphasize once more: The sphere itself is per-
fect and smooth at the poles. The singularities
onZy arise because
we have chosen a "bad" coordinate system!). The best thing we can do is therefore to put U equal to something like
u
= {
(S,~)
[0 <S
O<~<
2n }
You see that we have not only missed the poles but also the arc r joining the poles. The arc "singularity" is due to the periodicity in
~.
Clearly we may construct an atlas using two coordinate sys-
tems of this kind
FIG. 77
(see figure 77) .
11
263
II
Now we have constructed two different atlasses on the sphere. Does it mean that we have constructed two different types of differentiable structures on the sphere? No, because if we try to express the polar coordinates in terms of the standard coordinates, we find that the polar coordinates depend smoothly upon the standard coordinates and vice versa: 8
x 2= Sin8' Simp
Arcsin[
I
xr
Arctg [
+ (x')' 1
(Xl) 2
x2
1
Thus all the polar coordinate systems are smoothly related to the standard coordinate systems! So they are obviously equivalent, i.e. if a function,f : S2~ R,is a differentiable function when expressed in polar coordinates it will also be differentiable when expressed in standard coordinates. Thus it is irrelevant whether you use standard coordinates or polar coordinates on the sphere, since they will generate the same maximal atlas and consequently the same differentiable structure. ~fuen discussing the sphere S2 you might have wondered why it was
necessary to use several coordinate systems to cover the sphere. It required 6 standard coordinate systems, respectively 2 polar coordinate systems, to cover the sphere! If we were smart enough, might we. then hope to find a single coordinate system covering the whole of the sphere? However, it is impossible to do that:
Any atlas on the sphere con-
sists of at least tuJo coordinate systems. The reason for this peculiar fact is purely topological. It is intimately connected with the fact that the sphere
~s
logical space, since it: is a i:Jour:¥ied closed subset of R single coordinate system (CP,U) covering then
cp
would be an homeomorphism of U cR '" '"
U homeomortpic onto
~he
a
compact
topo-
If we had a
whole of the sphere,
onto the sphere 52:
• 52
But this forces U to be compact itself. Therefore U is both
compact and
open. But the only subset of an Euclidian space which is both open and compact is the empty set. This is a contradiction! Hence the sphere 52 is not a simple manifold. However, if you cut out just one point from the sphere, then the rest is not compact and nothing can prevent us from covering the sphere minus one pOint with a single coordinate system! For simplicity we cut out the northpole. Then stereographic projection defines a nice
II
264
coordinate system (see fig. 78)
2
R
....
onto
s
6.3
PRODUCTMANIFOLDS AND MANIFOLDS DEFINED BY CONSTRAINTS
Now consider two Euclidean manifolds ~e RP , Nne Rq where ~ is an m-dimensional and Nn is an n-dimensional manifold. We can form the product set M X N consisting of all pairs (xiY) with
x€M and yEN. Clearly M x N is a subset of the Euclidean space RPx Rq
=
FIG.
struct an atlas covering M x N. Let
(~,U)
79
RP+q.We want to con-
"
be a coordinate system
on M and let
(~,V)
M
be a coordi-
nate system on N:
N
~III\I!(UXV)
1 - - -.......- - - . -xI M
FIG. 80
~(U)
Then we define a coordinate system on the product set M x N as follows: (6.3)
~19~:UxV"'MxN
~
&
~ (Xl, ••• ,xmiyl, ... ,yn)
Clearly U x V is an open set in Rn x Rm
=
Rn+m and it can be
265 ~
checked in all details that a regular map.
~
is a homeomorphism and that it is
Thus it is a nice respectable coordinate system. If
(~i,Ui)iEI is an atlas for (~i8 ~j,Ui8Vj)
8
II
M and
(~j'Vj)jEJ is an atlas for
N, then
is an atlas for M x N. "Ie call M x N the proJu.ctmani-
foZd of M and N. By construction M x N becomes a
m+n
dimensional ma-
nifold. Some examples may throw light on this rather abstract procedure:
, "..---- .... ....
The eyZinder: SlxR, fig. 81. We have used polar coordinates on Sl. The cylinder is supposed to be extended to infinity in both directions.
The. torus: SlxSI In figure 82 the two circles :.ave been drawn with very different radii
a > b.
(Observe that strictly speaking SlxSl should be constructed as a subset of 2
2
,
R x R = R , but that is not so easy to visualize. Observe also that if a
~
FIG.
b then
82
the torus constructed on figure 82 becomes a smooth surface with self-intersections. such a surface is not a Euclidean manifold.) We will now sketch a very general method to construct Euclidean manifolds.
First we remind you about the following problem from
classical calculus: Let f: Rm+ n ~ Rn be a smooth function and consider the equation:
::. • 1 (x 1,
•••
m ,x ,y1, ••• ,yn) =0
f(x,y)=o i.e. {
fn (Xl, ••• ,xm ,y1, ••• ,yn)=O
Suppose (xo,Yo) is a solution of the equation f(x,Y)=O. Under what circumstances can we find a
neig~UrhOOd
can solve the equation f(x,y)
U of (xo,Yo) in which we
=0 with respect to y and obtain y as
a smooth function g of x.i.e.y=g (x)?
266
II
n R
Rmt-n
y
';;o,Yo)
f(;;:,y)
a
u
x
FIG. 83
This problem is wellknown. To motivate the solution we study the f(xo+~x;yo+~Y)
linearized problem, where
is replaced by
f(xo;yo) + Dx f·~~I+ Dy f·~YI Thus the equation f (x,y) = 0
is replaced by the equation
Dxf· ~~I + Dyf. ~YI = 0
,
where we have taken into account the assumption that (xo,yo)
is a
solution, i.e. f(xo,yo)=O. Now the linearized equation can be sol= provided = af i ved with respect to ~YI Dyf=(-a--.) is a regular square matrix. In that case we :furt:heJ::nore get: yJ
= = - (Dyf) = -1 . (Dxf)· = ~xl =
~YI
.
With this in our mind we now state the sOlution to our problem:
Theorem 1 (Theorem of impHdt functions) mrn ~ Rn be a smooth function. If U f is a regylar squaro matrix, when Let f: R it is evaluated at a_solution (XO,Yo) to th~ equation f(x,y)=O, there exists a neighbourhood U of (xo,yo) and a smooth function g:Ff' ~ ~ such that: 1)
2}
f(x,y)=
=
D~
=-
°iff y = g(x) =_1
(DI)
for aU (x,y) in the neighbourhood U.
{ljxf}
(i.e. the partial derivatives of y with respect to x are found by imp licit differentiation:
ai +:d.. a)
~
al a)
o)
Using this cornerstone of classical analysis, we can now show: Theorem 2 Let f: Rq .... rf be a smooth function (q > p). Let M be the set of solutions to the equation f (iii) = o. If Mis nonempty and if f has maximal rank throughout M, then M is differentiable manifold of c1:imension (q - p).
267
II
We will not go through the proof in all details, but let us Rketch how to construc t coordina te systems on M. Suppose M is not empty and suppose Zo belongs to M, i.e. f (;,) = O. We know that:
l~
Z
~j
dZ I
Df--
Z=Zo
dZ
~fP
q
af P q
dZ I
dZ
has rank p,Which is the highest possible rank. Hence we can find p columns which taken togethe r constitu te a regular square matrix. To Simplify let us assume that the last p columns are independ ent and let us introduc e the notation : = G,y) i.e. (zl , ... ,zq) = (xl, ... ,Xq-P,yl , ... Then D f is a regular square matrix. Accordin g to the theorem of implicit functions y We can now find a neighbou rhood U of Zo = (xo ,Yo) of the form U = { (x,y) I x~-£< xi< x~ + e , y~ -e< yj< y~ + e } and a smooth map g:R q- P ~ RP such that
z
?)
(x,y) EM n U iff TI(U)
= { x
I x~
y
- e< x<
x~
g(x) when x belongs to TI(U), where + e }
i.e. TI(U) is the pr~jection of U into Rq - P , cf. fig. 84.
Fig. 84
x:
. x 1-ax~s
2
.) -ax~s
Rq-P
But this means that restrict ing ourselve s to U we can define a coordina te system ($,TI(U)) , $:TI(U) ~ M,in the followin g way:
$ (Il) = (iC ,g(iC) ) Such a coordina te system will be called a standard coordina te system. We must now check that $ really defines a coordian te system: a) Since U is open, M n U automat ically becomes an open subset of M. As the projection map IT = f1is triviall y continuo us, we see that $ is a homeomorphism of the open subset TI(U) onto the open subset MnU. b)
$
is a smooth function since g is. Furtherm ore
Dx $ =
' [
, '. 0
o
.,
~"'l
$ is triviall y regular since
dyl
where the left part of the Jacobi matrix is the unit (q-p) x (q-p) matrix which is triviall y regular.
268
II
We can clearly cover M by standard coordinate systems and it is not too difficult to check that they are all smoothly related. Thus we have defined an atlas, and this shows that M is a differentiable manifold of dimension (q-p), see fig. 85.
If M is generated as the set of solutions of an equation f(z)=O, we say that M is defined by an equation of constraint. Clearly the sphere S4 is defined by the following equation of constraint: (Xl)2+
(X 4 )2+
(X 3 )4=
I
If you review the illustrative example of the sphere, you will in fact find that most of it was an exemplification of the above abstract discussion. The assumption that f has maximal rank is essential. Consider for instance the following trivial example: Let f: R4 ~ R be the smooth map gif (x,y) = X 4_ y 2 • af Of Then Df= (ax; 3y) = (2x,-2y) is
ven by
singular at the point (x,y)
R
(0,0)
But the pOint (0,0) clearly be-
u
longs to
M
= {(x,y)
Ix 2 -y2= a}. In accor-
Fig. 86
dance with this we easily find that M is not a I-dimensional differentiable manifold.
It is impossible to
find a smooth map ~ which maps an open interval bijectively onto an open neighbourhood of (0,0) in M. (See figure 86). As another illuminative example we consider the smooth map, given by
f:R3~R,
269
II
which is singular at all points on
z- axis
M={(x,y,z) Ir<x,y,z)= 0 }= {(x,y,z) [x2+y2=1, z
y-axis
= O}. The subset
M is,nevertheless, clearly a
; 0
I-dimensional differentiable manifold (it is isomorphic to the circle S 1). IVhat goes wrong is the dimensionality. If f had maximal rank, then M would have been a
2-dimensionaZ manifold! As a final example of differentiable manifolds defined by con-
straints,we look at the configuration space for a system consisting of a finite number of particles in classical mechanics. Consider for instance a double penduZwn: It consists of two particles where the positions are constrained through the equations (ql)2+(q2)2+(q3)2= I (q'_ql)2+(q5_ q 2)2+(q6_ q 3) 2= I
Since the smooth map,f: R6~ R2, given by (ql)2+
(q3)2_ I
(q2)2+
[ (q'_ql)2+
(q5_ q 2)2+
(q6_ q 3) 2_ I
is regular on the configuration space M, we conclude that M is a 4dimensional differentiable manifOld. In fact M is isomorphic to S2xS~
6.4
TANGENT
VECTORS
\'le are now in a position to introduce tangent vectors. Consider an Euclidian manifold M c RN (figure 89) Let P be a point in M. Then we have a lot of vectors at P in the surrounding space RN. Observe that a vector is defined as a directed line segment PQ • Thus
a vector is characterized
not only by its direction and length but also by its base point P!
N
R
2.70
II
Hence if two vectors have different base pOints P and p', then they are different vectors, even if they have the same length and direction. Let us return to the vectors with the base point P. Not all of them are tangent vectors to the smooth surface M! Let A:
R
~M
be a
smooth curve passing through P. Then A generates a tangent vector at p, the velocity vector, (6.4)
.... dA vP=dt
Ji
t
=a
The set of vectors generated by
aU possible smooth curves in M passing
through P forms a set of vectors called the tangent space and it is denoted by: Tp(M)
I~
To investigate the structure of this tangent space we introduce (~,U)
a coordinate system
which
covers a neighbourhood of P. To simplify we choose the coordinate system so that P has the coordinates
(0, ... ,0)
(see figure 91). Consider the curves: (6.5)
~(t,O,O,
AJlt) =
... ,O)
; •.. ; An(t) = et>(O,O, ... ,O,t)
They are called the coordinate lines. By differentiation we now obtain the tangent vectors
..
(6.6)
e
n
They are called the canonical frame vectOl's. To investigate 1
dA dt
-n
x
,Jt
a
2
x
them a little closer we will
1
write out the parametrization of the coordinate lines in
U
their full glory:
Adt)
FIG. 91
~l(t'O'.'.'O)l = . , .•. ,
[
·N et> (t,O, ... ,a)
By differentiation we get
[
~~ (0, ••• ,a ,t) et> (0, ...
,a ,t)
l.
271
...e
II
a
1
ai{T
dl'l
= dt I t=O
, ...
/
a$N
axr
...e
dAn n
_
dt1t=0
axn a
"3XTI.
I (0, .•. ,0)
I (0, ••. ,0)
Here the coordinates are the Cartesian coordinates in RN and we will refer to them as the extrinsic coordinates of the vector. Now take a look at the Jacobi matrix 3i{T
~ ax
a$N
~n
a
i
(-ax-r) ai{T
ax
It is obvious that the n collDTlYl vectors in the Jacobi matrix are nothing but
the extrinsic coordinates of the canonical frame vectors. But we have previously demanded (
~ .)
ax] to be of maximal rank n. Therefore the column vect·ors are linearly independent! Thus, we have shown that the canonical frame vectors are
linearly independent. H"ext we want to show that the canonical frame vectors span the whole tangent space T (M). Let ~ be an arbitrary tangent vector. ... p p Then v is generated by a smooth curve A. Let us write out the parap
metrization of A: First we observe that A has the coordinate representation: I(t)
with respect to a coordinate system (cp,U) , cf. ::ig. 92.
Fig. 92 We can then write down the parametrization of A with respect to the Cartesian coordinates in the surrounding space
RN.
272
[q,l (~l (t) , •.. ,x
A(t)
n (t»
II
1
q,N (Xl (t) , ... ,x n (t» By differentiation we get
...vp
(6.7)
dA dt! t=O
n l ~ dx + •.• + ~ dx ax ~ ax dt aq,N dx n ~dXI· ax ~ + ••• + axrr dt
dx l
...
dt el
+ ••• +
qxn ... at en
conseque:
t ly we see that the vectors ;;1' ... ';;n really span the tangent vector vp. On the other hand, if ~ is a vector spanned by ~l' ... '~n i.e.
u=a1el+ ..• + an~
n
then ~ is a tangent vector, since it is obvious that it is generated by the curve A(t) with the coordinate representation A (t). -+
-+
As e l , • • • , en are linearly independent we therefore conclude that the tangent spaae Tp (M) is a n-dimensional veatorspaae. The vectors -+ -+ e l , ..• , en constitute a frame for this vectorspace. Formlla (6.7) now motivates the following definition DEFINITION 7
Let
tp be
sentation:
a tangent veator generated by A and let A have the ooordinate repre"A(t)
(x1(t) •••. ,:rr(t)) •
. di a1- = at
Then the numbers are aaUed the intrinsia speat to the aoordinate system
aoordinates of vp with re-
(q,.U).
Observe that the intrinsic coordinates a i are simply the components of;j with respect to the canonical frame (e 1 , · · · ,en) . p . The numbers al. are also referred to as the aontravariant components of v P • We shall explain this mysterious name in a moment! Now, let us see what happens if we exchange coordinates, say from (x1, .•. ,x n ) coordinates to (yl, ... ,yn)_ coordinates. Then the same tangent vector vp is described by two set of numbers, (al, •.. ,a n ) with respect to the x-coordinates
...
...
III
(1)
n { (al, ... ,a ) with respect to the y-coordinates 121
(2)
We want to find the transformation rule which allows us to pass from x-coordinates to y-coordinates.
II
It will be convenient to introduce the notation 0 21 for the Jacobi matrix of the transition function, i.e. i
021
=
(~)
ax J Okay, we are ready: Let +vp be generated by the smooth curve A. Then A has the coordinate representation, xi
= xi (t)
,
in x-coordinates, and it has the coordinate representation, yi = yi (x j (t) ) in y-coordinates. Consequently we obtain
di
i
(6 .8)
~I
= di
~
a i dxt fu (it =
ai ~
a1
(II
In a matrix formulation we may collect the intrinsic coordinates into a column vector
Then we can rearrange the transformation rule in the following way: (6.9)
This is the most impontant of all the formulas involving the tangentvectors. Let us also investigate the connection between the old canonical frame vectors (Il ]. and the new, canonical frame vectors \£I ~ ].,.NOte that , ~p- has the old coordinates &)J = = (0, ... ,1, ...• ,0), where the number 1 is in the i' th posi tion. But OCJW we may carplte the new coordinates from the above formula:
e,
aj = (21
oI
~ • ak axk
(J)
-
~ axk
• ok
+
i
~ axi
ConsequenUy the vector ~Ii has the following decanposition along the nfM eanonical
·f,,-~me
vectors:
(6.10)
-+
~Ii
e"a j (21J (21
-+
e,
(21J
274
II
Let us fix the notations: ~21 and ~12 are the transition functions with the corresponding Jacobi matrices 0 21 and 0 12 , y = ~21{X)
(6.11)
5 21 =
X=¢12(Y)
[5]
As ~21 and ~12 are inverse maps we conclude that 0 21 and 012 are reciai . a ~ procal matrices! Thus ~ and ~ are components of reciprocal maJ ayJ ax trices and we may therefore rearrange formula (6.1U in the following way (6.12) A quantity which transforms in the same way as the canonical frame vectors is said to transform ao-v~antly (co-varians = transform in the same way). Hence a
covariant quantity is a quantity Wj which
transforms according to the rule axi (6.13) w. = w. --. {2jJ 11I~ ayJ when we exchange the coordinates. Observe that we have in the lower position: wi' It is referred to as covariant that a covariant index is transformed with the matrix right.
put the index i index. Observe axi from the ayj i
. dx The canonical frame vectors transformed with the matr~x -.. A quan~ ayJ tity which transforms with the reciprocal matrix axJ is said to transform aontra-variantly (Contra-variantly = transforms in the opposite way). Hence a contravariant quantity is a quantity a i which transforms according to the rule
(6.14 )
a (21
i
~ axJ
a
j
(I)
when we exchange 'the coordfnates. This time we have put the index i ~n the upper position: a i • It is referred to as a contravariant index. Obs~!'r a"i ve that a contravariant index is transforI'led ",ith the matrix .::...L...ax j from the left. As we have already seen the components of a tangent vector transforms contravariantly! In the preceeding sections we have only discussed the local aspects of functions and tangent vectors, i.e. we have focused upon the i"l\'!rediate neighbour-hood of a specific IXlint P. We will row briefly discuss how we can introduce scalar fields and vector fields gn a manifold M. Let us look at a scalar field first. In our geometrical interpretation, a scalar field is simply a smooth function ~:M ~ R.
275
II
If we introduce two coordinate systems, say x-coordinates and y-coordinates, we may now repr.:sent q, by the ordinary Euclidean functions and and these are related through the
~ (xi)
rule
~ (yi)
=
(I)
transfo~ation
(2)
where xi and yi are coordinates of the same space time point p. When we investigate such a scalar field p .... q,(P), we will always introduce coordinates and use the Euclidean representative. :')e then "forget" the bar, and write the corresponding Euclidean function as
...
Next we have vector fields. To construct a vector field A on an arbitrary manifold M, we attach a tangent vector, Ap,to each pOint P in our manifold, (see figure 94 ).
Fig. 94 n
If we introduce coordinates (x1, .•. ,x ) on M, then the tangent vector P(xl, ... ,x n ) can be decomposed along the canonical fra-
at the point
....
....
me vectors el, •.. ,e n as follows: i
n
1
-+
a (x , ... ,x lei We say that the vector field is smooth, if its components ai(xl, ... ,x n ) are smooth functions of the coordinates. In general we want to investigate a vector field A through its components, ai(x), but one should be careful. If the manifold M is a non-trivial manifold, then several coordinate systems will be needed to cover it! (Compare the d~sion of 8 2 in section 6.2) x2
[T" i
(~
--..
xl
Fig. 95
y
A
2
EJ
U2
l
276
If we have two such coordinate regions S"h and the region
Q 12,
ai(x) and ai(y) to describe the vector field (II
(2)
gion
Q 12
Q2
which overlap in
then you will have to use t\qO coordinate expressions
A.
In the overlapping re-
these two coordinate expressions are row connected through the
tr ansf orma tion formula (6.8): ai, n ~ a.f(xl, ... ,x )
ax]
(I)
METRICS
6.5
n In the ordinary Euclidian space R we have an inner product , <xly>
=
xlyl+ ... + xnyn
which is symmetric, bilinear and positive definite. We want to extend this concept to the tangent vectors on a manifold M. To each tangent space on the manifold M we therefore associate an inner product. This family of inner products is called a metPia and it is denoted with the letter g. If vp and ~p are tangent vectors belonqing to the same
tan-
gent space Tp(M) we denote the value of their inner product by:g(;p:~p)
R i1Q
FIG. 96
-+
Observe that we do not define the inner product of tangent vectors vp and
UQ
with different base pOints.
As in the Euclidian case we shall assume that g is symmetPia,
(6.15) and bi linear , (6.16)
-
277
II
But we shall not demand that the lCletriCS is positive definite. We only demand that it is non-degener>ate, Le. the only vector which is orthogonal to all other vectors in Tp(M) is the zero vector: (6.17)
g(VpiUp)~O= for all ~p in Tp(M) implies Vp=
e , ... ,en
Let us introduce coordinates and let canonical frame vectors. The nunbers
1
a
be the corresponding
(6.18)
are referred to as the metric coeficients. Let vp and up be two tangent vectors with the coordinates aiand b i . T'le can then express the inner product of vp and,u p ent~rely in terms of the metric coeficients and L the coordinates a and b L : (6.19)
At each pOint P the metric is therefore completely characterized by its components gij(P), Thus we may regard the metric coefficients as functions defined on the range of the coordinate system. If ~Ie identify P with its coordinates (x!, ..• ,xn) l.,re may furthermore regard gij as ordinary Euclidean functions In gij= gij (x , ... ,x ) t'le say that the metric is smooth if g ij depends smoothly upon the coordinates. In ""hat follows we will only work with smooth metrics! Let'us summarize the discussion:
Definition 8 Let M be a differentiable manifold. A smooth metric g is a family Of inner products, defined on each of the tangent spaces, with the following properties: d) g is a bilinear map, which is symmetric and nonderenerate.
b) The metric coefficients gij depend smoothly upon the coordinates
(Xl,."
n ,x ).
The metric coefficients are often collected in a matrix G = (gij)' which by construction is symnetric. Let us denote 17.f ~I and I), the coordinates of two tangent vectors vp and up' Then we may rearrange the formula for the inner product in the following way: j g(vp;u p )= gijaib = aigijbj= Next. we want to show that the matrix is a column vector with the property
a7 GE,
G is
regula:!>, Let us assume that
E,
278
II
We have finished the proof as soon as we can show that
h,
is necessari-
ly the zero vector! Let E, be the components of a fixed tangentvector up and denote by a,the components of an arbitrary tangentvector vp' From GE,= cr,we conclude: -+ -(GO I )
a,
i.e.
=
g(vp;~)
...up
0
for all
=0
for all
Consequently vanishes because g is non-degenerate. It follows that 0 as we wanted it to be.
0
E,=
AS G = (gij) is a regular matrix we knolll that it has a reciprocal matrix! We will denote the elements of the reciprocal matrix with gij. By construction we get
(6.20)
Then we want to see how the metric coefficients transform .,hen we pass from one coordinate system to another. But that is easily done. Let there be given two coordinate systems, x-coordinates and y-coordinates.lf ~afl are the old metric coefficients corresponding to x-coordinates and 8~fl are the new metric coefficients corresponding to y-coordinates, we now get
9:nD '4~ "
=
gee jeD) (2\a (2l"
= gee
lIlY
-+
-+
ax Y axO
(1)
(I)
aya
g(eYieo)--
~
ax Y • aya '
e ax O)
(11 0
axY
(~yo aya
aye
axo ayfl
where we have used (6.12). So the metric coefficients transform covariantly: (6.21)
Worked exercise: 5.5.1 Problem:
L:~
gij be the covariant components of the metric. Show that the components
glJ of the reciprocal matrix transform contra-variantly.
Exercise: 6.5.2 Problem: Consider the Euclidean space R3 with the usual metric. In Cartesian coordinates the metric coefficients reduce to the unit matrix
GCartesian '"
[ ~ ~ ~] 001
279
II
Let us introduce spherical coordinates
through the relations:
C~stp
X] y [z
[r sin e ] r sin e Slntp r cos e
and cylindrical coordiantes
X] [:
(r,e,~)
=
(p,~,z)
through the relations:
[P costp 1 P
s~ntp
Show that the metric coefficients in spherical respec:ively cylindrical coordinates are given by (6.22)
[~
ITspherlcal .
o 00 ] r2 o r 2 sin 2 e
ITcyllndrlcal . .
(Hint: It is preferable to work in the matrix langmage. Let D12 the Jacobi matrix corresponding to a coordinate transformation. Then the transformation rule(6.21) reduces to (6.23)
=
i
~j be ay
'* =
(~) = Dl~?P12
We will need some elementary ?roperties of the metric. F~rm the regularity of ~ we conclude that the determinant, g = Det[~l,is non-zero. From the transformation rule (6.23) we furthermore get (6.24)
It follows trivially that the sign of g is independent of the coordinate system we work in. Clearly the determinant depends continously upon the coordinates. Now let us assume that our manifold M is connected (i.e. any two pOints in M may be joined with a snooth curve within M.) If g was positive at a pOint P and negative at a point Q, then g would have to be zero somewhere between P and Q. Since this is excluded we obtain the following lemma:
Lemma g has a constant sign throughout M
Being symmetric we know that ~ has n real eigenvalues AI"",An. We also know that ~ is regular, therefore none of the eigenvalues can be zero. (Remember that Det[G 1 = A1···A ). It can be shown that the numn ber of positive eigenvalues is independent of the coordinate system.
II
280
(The actual value of AI, .•. ,A depends strongly on the coordinate n system in question!) How let us assume again that our manifold is connected. Then we can-extend the previous argument to show that the number of positive eigenvalues is constant throughout our manifold. ·We will especially be interested in two cases:
Definition 9 If a~l the eigenvalues are strictly positive we say that the metric is an a) Euclidian metric. In that case the metric is positive definite. A maniford with an euclidian metric is called a Riemannian manifold. (If the metric is not Euclidian we speak of a pseudo-Riemannian manifold). b)
If one of the eigenvalues is negative and the rest of them are positive we say that the metric is a Minkowski metric. In this case the metric is nondefinite. (Hence a manifold with a Minkowski metric is a special example of a pseudoRiemannian manifold).
Let us investigate the Minkovlski metric a little closer:
Lemma 2 Let M be a manifold with a Minkowski metric. Let T (M) be the tangentspace at the p ·point P. Then we can decompose T (M) in an orthogonal sum p
T (M) = V_ P
~
U+
where V_ is a one-dimensional subspace, such that the restriction of g to V_is negative definite, and U+is a (n-1}-dimensional subspace, such that the restriction of g to U+is positive definite. Proof: Choose a coordinate system covering P and let gi" be the metric coefficients with respect tolthis coordinate system. Then
G=
(gij)
has one negative eigenvalue AO and (n-l) positive eigenvalues AI, ..• ,A _ . Let V_be the eigenspace corresponding to AO and let U+be n l the subspace spanned by the eigenspaces corresponding to AI, ... ,A _ . n 1 The restriction of g to V_is negative definite. To see this let ~ be an eigenvector with eigenvalue AO and coordinates a i . Then ~ ~
g(v,v) Hence
v has
=
i j i i i i a gija = a (Aoa ) = AO (a a ) < 0
a negative square. ~
~
Similarly the (n-l) eigenvectors UI, ... ,U _ with eigenvalues n 1
281
II
Al, ••• ,A n _ l have positive squares. But they form a frame for l4and therefore the restriction of q to U+is positive definite. Finally we must show that V_ is orthogonal to U.. It suffices to show ~
~~
~
that v is orthogonal to each of the frame vectors U1, . . . ,U _ . But n l i j i i g(V;U ) a gijb a (Akbi) Ak (aib ) k and i j g'(Uki V) b gija As
AO*A
vanish.
+ ...
k
(A k is positive and AD is negative) it follows that g(v;l1c) must
0
This lemma motivates the following classification:
Defpiition 10 iet M be a differentiable manifold with a Minkowski metric: (1) A vector with negative square is called Time-like. (2) A vector with zero square is called a null-vector. (3) A vector with positive square is called space-like. The null-vectors forms a cone, the light-cone, which separates the time-like Vectors from the space-like vectors. (See fig. 97 ) x
x
o
1
Fig. 97
!f.lustrat1ve example:
The sphere II.
Let us look at a sphere with an arbitrary radius r and let us introduce spherical coordinates:
X] = [ y
[r Sine r Sine
z
r cose
COS~]
21f
Sin~
y
1f
a
x
282
intrinsic
Here (S,tp) are the
II
coordinates on the sphere and (x,y,z) are
extrinsic coordinates. At the pOint (S 0 ,tpo) we have the canonical
the
eSam
frame vectors
.. aXl as
-+
e
[, co.a,
as az
ax atp
Costpo
r CosS o Sintpo
lY
s
...
e tp with the extrinsic coordinates:
-r SinSo
-+
e
lY atp
[-'
Sin6 0 Sin., r SinSo Costpo
az alp
as The
=
1
0
intrinsic coordinates are by definition -+
e tp
and The frame vectors
e
e(?
s and are obviously orthogonal to the radial vec~r
r
and they are or-
thogonal to each other. We also want to construct a metric on the sphere. But here we may use the usual inner product
1
+ V 2 U 2 + V 3 U 3 associated with R 3. Thu;, for any =
V1U
two tangent vectors
vp
and ~
we put: (6.25)
This is called the
induced metric. Let us compute the metric coeffici-
ents:
[: ~:::: ~:::] -r SinSo
g(jl(jl= [-r SinSoSintpoir SinSoCostpo; 0]·
[
-r SinS OSinlAJ] r Sins~costpo
283
II
Therefore we find (6.26)
1 g;J' (So ,tpo) = r2 [0
•
0 2 ] Sin So
Observe that the metric becomes singular at the north pole and the south pole,
01
gi'l
oj
J pole
which again tells US that the coordinate system breaks down at the poles! Clearly this example can be generalized to an arbitrary Euclidian manifold, which then becomes a Riemannian manifold, when we equip it with the induced metric. Let us emhasize however that the induced metric is not the only :oossible JT!etric on an Euclidian manifold. ~'l'e may choose the family of inner products completely arbitrary.
Exercise 6.5.3 Problem: Consider a torus with the outer radius a and the inner radius b. Use the angles S and tp as shown on the figure to parametrize the surface. Compute the metric coefficients of the induced metric in the (S,tp)-coordinates. y-axis FIG.
100
Exercise 5.5.4 Problem: Consider the unit sphere Sll-l in the Euclidian space ~. We introduce spherical coordinates as indicated below. Show that the metric components, are given as shown below Xl = r SinSISinS2 ... SinSn-2CosSn-1 x2 x3
r SinSISinS2 ... SinSn-2SinSn-1 2
r SinSISine 2 .•. cosen-
xn- l r SinelCosS 2 xn = r cose l
r 2Sin 2S 1
(6.27)
r 2Sin 2e1SinS2
g, ,(r ,SI, •.. ,Sn-l)= IJ
o
o
284
6.6
THE MINKOWSKI SPACE
Consider a four-dimensional the four-dimensional Euclidian sic coordinates as they are of use this manifold M as a model manifold M represents an event. a photon by a particle. It all specific time: REAL WORLD AT t=t o (PHOTOGRAPHY)
PARTICLE A ABSORBS A PHOTON
II
!~
manifold M which we may identify with space R4 , but forget about the extrinno importance in our example. We shall of our space-time. So each pOint P in our This event could be the absorption of happened at a specific position at a MATHEMATICAL MODEL;
\•
THE PHOTOGRAPHY CORRESPONDS TO THIS SLICE
PARTICLE B MOVES UNDER INFLUENCE OF FORCES
FIG.
101
A particle moving in ordinary space is represented by a world line in our model. The world line is a smooth curve comprising all the space-time pOints occupied by the particle during its existence. Next we want to introduce coordinates in our manifold. To this purpose we shall use an inertial frame of reference, i.e. we imagine an observer equipped with a standard rod for the measurement of length and a standard clock for the measurement of time. Equipped with these instruments he can characterize each point in space by means of three Cartesian coordinates (X 1 ,X 2 ,X 3 ) and he can characterize any event by its position in space (X 1 ,X 2 ,X 3 ) and the time xo.
Hence any event is characterized by a set of four coordinates (xo ,Xl ,x 2 ,x 3 ) . In our mathematical model such an
o
1
2
3
+(x ,x ,x ,x )
p
•
Observer corresponds to a specific coordinate system, ~: R4~M,covering the Whole space-time manifold. What do we mean by an inertial frame of reference? Physically an inertial frame of reference is characterized by the following property:
285
II
Definition 11 An inertial frame of reference is a frame of reference where any free particle moves with constant velocity. The coordinate system corresponding to an inertial' frame of reference will be referred to as an inertial frame and we speak about inertial coordinates. Consider the worldline of a free particle. In inertial coordinates it is parametrized in the following way, xo+
[~:l
xI'" alxo+bl
Le.
where al is the constant velocity of the particle. In the coordinatespace this is
the parametrization of a straight line, cf. fig.103
[!l] [~ljl
[:;J -
x2 - a2 >'of b 2
4 R
FIG.
NOW,
x
3
a3
b3
103
suppose we have been given two inertial frames of reference.
Then we have two coordinate systems, (2, R4), covering the space time manifold. The transition function
which exchanges the
old coordinates (xO,x',X 2 ,X 3 ) with the new coordinates (yO,y',y2,y3) .is referred to as a
Poincare transformation. What can we say about the
'::ransi tion function? Observe that the world line of a free particle Ls represented by a straight line in both coordinate systems.
cx
FIG
.
~
104
Thus the function
y=
/
/ I
1
I
AUSXS
+ bU.
;1\
maps straight lines onto straight lines
and we conclude that it is a linear function
/J.
'{i
286
II
To proceed we must invoke the following assumptions:
BASIC ASSUMPTIONS OF SPECIAL RELATIVITY:
I
The speed of light has the same value in aU directions in aU inertial frames of reference, i.e. the speed of light is independent of the observer. II AU inertial frames of refenence are equivalent.
Let us introduce matrix notation. Put
o I o o
n
(6.28 )
0 0 I
0
~l
We can then show
Theorem
;3
A Poincare transformation connecting two inertial frames is given through the formula
Yl=A~1 + where
51
A satisfies
/11.29)
===+ AnA
A+nA
n
Proof: We have normalized our units so that the speed of light is one, c=l. Suppose P and Q are two pOints lying on the worldline of a free particle. Then the speed of this particle is given by v
=
j
(XI_X I )'+ {x'-x' )2 + (X 3_X 3 )' Q P Q P Q P
If we introduce the notatipn ~XI=XIQ-XIP we therefore conclude that the Q lies on the worldline of a photon if and only if
points P and
-(xQ-X~)'+(xQ-X;)'+(X~-~)'+(X~_X;)2
=
0
i.e.
But the conditiOn for P and Q to lie on a worldline of a photon must be valid in any inertial frame. Therefore we conclude (6.30) But here
~y 1=
YI Q - Yip
'!'hus (6.30) can be rearranged as
287
II
iff Hence the two symmetric matrices nand A+nA generates the same lightcone K, where
But then they must be proportional. Therefore we conclude that (6.31) In this way we have attached a proportionality factor A to the POinca~21
re transformation
YI =
which was given by
~21 (XI)
= Ax/.+
51
The inverse POincare transformation ~12 is gonsequently given by =_1 XI = ~12(YI) = A A-1bl
y, -
Thus ~12 is characterized by the matrix A-I lation of eq.
A simple algebraic manipu-
(6.31) gives us
i-~ = (A-I) +n (A-I) so the proportionality factor corresponding to
~12
is }. As the two i-
nertial frames of references are equivalent we conclude
1
A=I
Le.
A= ± 1
But from (6.31) we also get (Det[A])2= A Consequently ).=+1. We have therefore shown and The last formula easily be converted to give =-1 An- I A+ = n But n
-I
n and therefore A also satisfies A n A+ = =n
0
This theorem motivates the definition
Definition 12 A matrix A with the proporty (6.32) A+n A = A A+ n is called a Lorentz-matrix. The Lorentz matrices constitute
n
a group called the
Lorentz group, which we denote 0(3,1). A homogenous transformation YI = AXI
where
A is
a Lorentz matrix is called a Lorentz transformation.
288
II
Exercise; 6.6.1 Problem: (1) Show that the matrix n itself is a Lorentz matrix. What is the physical interpretation of the associated Lorentz transformation? (2) A Lorentz matrix which preserves the direction of time is called orthochronous. A Lorentz matrix which preserves the orientation of the spatial axes is called proper. Show that a Lorentz matrix is proper and orthochronous if and only if it has the properties, AOo > 0 ; Det
A=
1
and that the matrices with both properties constitutes a subgroup of the Lorentz group (which is denoted SO (3,1) ). (3) Show that the spatial rotation group SO(3) can be considered a subgroup of the Lorentz group.
Exercise: 6.6.2 Problem: Consider the matrix:
Ae;p[i il!l Show that it is a proper orthochronou~ Lor~ntz matrix. Consider the corresponding Lorentz transformation,y = AX,and show that it corresponds to a transformation between two observers, where the second observer moves with the velocity v along the x-axis of the first observer. This transformation is called the standard Lorentz transfo~ation.
We will now construct a metric on our s?ace-time manifold M. Let and be two tangentvectors at the same pOint P. Let a, and 5, p their coordinates relative to an inertial frame and consider the number ~;nb,
u
This is actually independent of the inertial frame. To see this we introduce another inertial frame. Let the corresponding Poincar~-transformation be given by Then the Jacobi matrix reduces to • = av a D21 = (~) = 'A S
ax
According to (6.l4)the new coordinates of ~p and up are given by A~, and AE, . Hence in the new inertial frame we comp·.lte the number
(Aa,) +n (AE,)
a; cJ~+nA) b number a;nEI =
l
is really independent of the and this shows that the inertial frame. Thus we may define: The inner product between two -+ ... tangent vectors up and vp is given by -+-+ =+== g(up'v p ) = al nbl
289
II
where at and 5, are the components of ~p and ~p relative to an arbitrary inertial j'rame.
In this way we have d~fined a metric! We can easily read of the metric coefficients:
(Compare with
(6.2~.)
dinates. Furthermore
The~
n is
are clearly smooth functions
o~
the coor-
scnnmetric and regular and it has the eigen-
values -1, 1, 1, 1 so it is a nice Minkowski metric. The four dimensional manifold R' equipped with the above metric is called Minkowski space. Illustrative example: Spherical symmetry in the Minkowski space. Very often we study physical systems having some kind of symmetry, and it is then advantageous to choose coordinates which reflect this symmetry. If the physical problem in question has spherical symmetry, you would naturally introduce spherical coordinates. The transition functions from inertial coordinates (xo,x,y,z) to spherical coordinates (t,r,8,t?) are given by: x
rSin8Cosq>
y
r
Ix 2 +y2+z2
8
rSin8Sinq> ; z = rCos8
The corresponding Jacobi matrices are given by: (1) From inertial to spherical coordinates:
(6.34 )
aya D21=(~)=
=
0
o
o
~
!L
r
r
-.Jf.._z__
rZ..;:;-r::::zz
-
v - X2:ty2
0
z
r
X~
0
o
r2Yr~2
2_z2
- r i y;t:z2
x
o
x 2+y2
(2) Frcm spherical to inertial coordinates: 0
(6.35) ])12=
dX
a
~
=
0
0
0
Sin8Cosq>
rCos8Cosq>
-rSin8Sinq>
0
Sin8Sinq>
rCos8Sihq>
rSinSCosq>
0
CosS
-rSinS
0
We can then find the metric coefficients in spherical coordinates. Using the transformation rule.(6.23) we easily get
290 -1
(6.36)
II
" ,,:m,']
o
Observe that the part of the metric which is associated with the polar angles is nothing but the Euclidian metric associated with a sphere of radius r. (See (6.26) ). Observe also that the metric breaks down on the z-axis«: 8 = O,n. That is just the old coordinate singularity we found on the sphere. So strictly speaking spherical coordinates do not cover the whole space time manifold! Now lpt us also look at a tange~t vector A (i.e. a four vector). Let us for a moment uSe inertial coordinates (A ,Ax,AY,Az ). Furthermore we assume that the vector part ,(AX ,AY,Az) ,is a radial vector, i.e. on the form A(Sin8Costp, SinSSintp, CosS) , where A is the spatial length of the space vector. This is the typical form of a four vector in problems with spherical symmetry. Let uS try to find the spherical coordinates of this four vector. Here we should use the transformation rule (6.8). Thus we get FIG. 105
0
=,
0
A,
0 0
0
AO
A"
CosS Sin S r
A SinScostp
A
A SinSSintp
0
A CosS
0
0
SinSCostp
SinSSintp
Cos S Costp - Cos 8 Sin (jl r r Costp Sintp - rSin 8 rSin8
0
So, in spherical coordinates, we get the simple expression: At = A°;Ar = A; AS = Atp = 0
Exercise: 6.6.3 Problem: Introduce cylindrical coordinates Write down the transition functions. Show that the Jacobi matrices are given by
(6.37)
[~
D21
(6.38)
D12
[!
a
(t,p,~,z)cf.figure
0 Sintp
Costp _ Sintp
~
p
p
0
0
0 0 costp -pSintp Sintp pCostp 0 0
and that the metric is given by
(6.39)
gall
[-'0
0
p2
,]
--------
106:
~J !J
P(p,
y-axiS x-axis
FIG.
106
291
6.7
II
THE ACTION PRINCIPLE FOR ARELATIVISTIC PARTICLE
Now, consider the world line of a particle, which need not be free. Let it have the parametrization, XCX = XCX(A)
t a
FIG. 107
Then the speed is given by the formula ds dt which
~lies
Thus
ds dA dt dA
l,dX') 2+
(dxl) 2+ (dx3) 2 dA dA
VidA
<
1
dxO
dA
that
->
the tangent vector vp is always timel ike. ,'Ie therefore call the
world line of. an ordinary particle a timeUke curve. Now let us look a little closer at the .,orld line: Let P, and P 2 be two pOints on the "iOrld line corresponding to the parameter values A, and A2' Then .le can define the arc-length of the arc P,P2 by the follm·ling formula r----------::-
(6.40)
A2 / dcxdxS ArcJ.ength=L v'-gas(X(A)) d~ dA dA
Observe that the arc-length is independent of the coordinate system chosen because the integrand is coordinate independent:
Observe also that the arc-length is independent of the
param~trization:
If we exchange the parameter A with a new parameter s, such that A = A(S) with A, = A(sll
and A2 = A(S2)
we get
292
II
S
dA ds ds
dx dA dA
Hence it is a reasonable definition. We "lill try to <Jive a physical interpretation of the arc-length. Let us look at a free particle first. As the particle is free we may choose an inertial
fra~e
of reference which follows the particle, i.e. the particle is at rest in this inertial frame. Consequently we can parametrize its
X
1.
world line as
Observe that the parameter A represents the time measured on a standard clock which is at rest relative to the particle! In this particular simple coordinate system you get
Thus we have shown:
The arc-length corresponding to a free particle is the time interval measured on a standard clock which is at rest relative to the particle! Then we look at an arbitrary particle moving in space. h'hen the world line is no longer strai<Jht the above argument does not apply. Let us however assume that the rate of a standard clock is unaffected by ac-
celerations. Then we may approximate the curved world line with a world line which is broken straight: The broken straight world line corresponds to a standard clock which is piece by piece
lO~
free. Therefore the above argmnent applies and the arc-length of the broken straight world line is the time interval measured on this standard clock! Going to the limit
we conclude:
293
II
The arc-length of a worldline is equal to the time measured on a standard clock following the particle. This will be referred to as the proper time T of the particle. Returning to the world line of a 9article mowing in an arbitrary way we may now use the proper time T of the particle as a parameter. It is connected to the old parameter A via the formula (6.41)
Differentiating this formula ,Ie get
i.e. in a symbolic notation we have
This formula is often "sC!Uared" whereby we get (6.42)
dT
2
a B =-<JaBdx dx
and dT 2 is then referred to as the square of the line element or just the
line element. Using the proper time T as a parameter we get the tangent vector cx (6.43) UCX = dx dT \V'hich is referred to as the four Velocity.
The four velocity has a very
simple property. If we square it we get -+ -+ dxcxdx S dxcxdx S [d A] 2 g(up1llp) = "bSdT dT = gaSdA dA [dT =-1 Thus the four velocity is a unit Vector pointing in the direction of time.
FIG.
110
294
II
Even in special relativity there is nothing sacred about inertial frames. We may introduce any coordinate system we like! We should, follo~]ing
however, be aware of the
fact: If (q"U) is not an inertial
frame, then a) The metric coefficients no longer reduce to the trivial ones, i.e
'*' nCtS
gCtS
b) The world line of a jreeparticle is no longer parametrized as a straight line, i.e. Ct
d 2x (f[2
0
'*'
u
/
/ FIG.
11 1
1~ I
XCt=XCt(T
As a free particle has a non-trivial acceleration in curvilinear coordinates we say that it experiences fictitious forces. I!e call the forces fictitious because they owe their existence to the choise of the coordinate system. If we return to the inertial frame the fictitious forces disappear. Let us try to find the equations of motion of a free particle expressed in arbItrary coordinates. It would be nice if we could derive these equations of motion from a Lagrangian principle, i.e. the worldline should extremize an action of the form
S
A2
=f
Ct dx Ct L(x idA) dA
Al
Observe that the Lagrangian depends on all four spacetime coordinates and their derivatives. In a relativistic formulation time and space coordinates must be on the same footing. In an Euclidian space we know that a straight line extremizes the arc-length. This suggest the following theorem:
Theorem 4 The relativistic lagrangian Ct
(6.44)
azct
L(x 'dA)-
or a
ree particle is given by azS -getS d)'" dA
azct
so the corresponding action is pl'OpomonaZ to the proper time
-
295
II
(6.45 )
The equations of motion can be written on the form d 2 xv v dza dzS (6.46)
JT2
r
aadT
dT
where rV aa ' t he so-called Christoffe l fie ld, is given by dg dg dO = ~g~v[ ~ + _~ _ ~ 1 (6.47) B 'Q( ... dX
dX
dX
Proof: The Lagrangian principle leads to the following Euler-Langrage ec:uations
!~~
=
;:~] j~
:" [
=
0,1,2,3
3(dT)J
The evaluation of these equations using (6.44) is strai<,!ht fon-lard but tedious so we leave it as an exercise:
Worked exercise 6. 7. 1 Problem: Show that the Euler-Lagrange equation corresponding to the action (6.45) reduces to
To conclude the proof
~le
must show that the ec:uations of motion redu-
ces to
in an inertial frame. But in an inertial frame we know that the metric coefficients are constant, gaB
=
naB' so their devivatives vanishes.
a
If we compare this with the expression for the Lorentz force in
inertial coordinates, d 2x
v
m --a;r-2 = you see that we may identify -mrv 0.6 with the field strengths of the fictitious
forces. Let us look at the electromagnetic force once more. lIe knm·r that in terms of inertial coordinates it may be rewritten as (1. 29)
= g~V [~_~] dX~
ax
II
296
it is expressed as a combination of derivatives of the r~axwell
Thus
field Aa . For this reason ~le say that the Maxwell field act as poten-
tiara for the electromagnetic forces. Returning to the curvilinear coordinates we have seen that the Christoffel field may be expressed as a combination of derivatives of the metric coefficients: ag ago ag~o La~V[ ~a + ~~ ~~l (6.47) rv -W dX a - ~
as
"-
For this reason
~le
say that the metric coefficients act as potentials for the
fictitious forces. Finally we notice that if we parametrize the world line by the frame time, i.e. put A = XO
(=t)
arranged as:
, then the relativistic action can be re-
t2
s
(6.48)
-m
J!I-~2
dt
tl EOr a non-relativistic particle, i.e. when
v « l , this reduces in the
lowest order approximation to the non-relativistic action tz t2 S ~ -m J (1-~~2)dt J~m~2 dt + m(t2- t l) tl
tl
(except for an irrelevant constant) • Consider an arbitrary manifold Ylith a metric, Ylhich can be a
~ie
mannian metric or a 11inkowski metric. A curve connecting two points which extremizes the arc-length is called a
geodEsic. In the above ana-
lysis we have made no special use of the fact that ,'Ie Ylere Ylorking in Binkowski space. Consequently geodesics on an arbitrary manifold are characterized by the geodEsic equation (6.49)
d2x
a
d?
= _ ra
~V
dx~ ds
where s is the arc-length and
v dx ds
ra
~v
are the Christoffel fields. The
Christoffel fields are defined through the formula (6.47) • But unfortunately this formula is very inconvenient for computational purposes. ~Jhen
you want to compute the geodesic equations it is therefore preferable to
extract them directly fran a variational principle. Of course you could use the
,
Euler-Lagrange equations coming from extremizing the arc-length itself
But this leads to very complicated computations due to the square root. So in practice one uses the following useful lemma
297
II
Lemma 3 Let y be a geodEsic: (1)
Then y extremizes the fUnctional A2
I =
(6.50)
f
d:/).
ax S
l.; gOf',(X(A)) dX F
d)"
Al Fza>theY'l1lore, when XCt=XCt(A) extremizes the funational I, then Ct ax ax S is constant along the geodEsic, so that we can idEntify gCtS(x(A)) 7i."A F the parameter A with the arc-length s.
Proof:
Consider the Lagrangian Ct dxCt L(x iax-) = l.;g~V(X(A»
dx~ dx v ax- ~
It leads to the Euler-Lagran0e equations
Le.
but they are identical to the sreodesic e0uations if we can show pro·· perty (2). This is done in the following way: Ct dxo Let XCt=XCt(A) extremize the functional I. Observe that L(x iax-) Ct dx is homogenous in the variable ax- of degree 2. Thus it satisfies the Euler relation 2L
dxCt We must show that L(XCt(A);ax-)iS constant alonCT XCt=XCt(A). Consequently we consider the total derivative: dL dA
USing the Euler-Lagrage equations this is rearranged as dL 2 d)..
where we have taken advantage of the Euler relation
• But this im-
plies immediately that ~~ = O,i.e. A must be proportional to the arclength s.
D
II
298
Observe that when you have found the geodesic equations you can directly extract the Christoffel fields as the coefficients on the rigth hand side. By an abuse of notation the above functional is often denoted as
Exercise: 8. 7 • 2 Problem: Consider the two-sphere 8 2 with the line element ( cf. (6.26) ), ds 2 = dS 2 + 8in2Sd~ (1) Determine the geodesic equations and show that the meridian ~ is a geodesic. (2) Compute the Christoffel field.
= ~o
Worked exercise: 6.7.3 In Minkowski space the line element is given by dT2 = _ dt 2 + dr 2 + r2(de 2 + 8in28d~2)
Probl~m:
when it is expressed in terms of spherical coordinates, cf. (6.39). Compute the aSBociated ChriBtoffel field.
Okay, you might say: This is very nice. We can work in arbitrary coordinates if we are willing to pay the price of fictitious forces and non-trivial metric coefficients and all that, but why should we do it? After all, we have at our disposal a nice family of coordiante systems, the inertial frames, which gives a simple description of physiscs! There is however a subtle reason for working in arbitrary coordinates which was pointed out by Einstein. Fictitious and gravitational forces have a strange property in
co~on:
If you release a test particle in a gravitational field its acceleration is independent of its nass. Two test particles with different masses will follow each other in the gravitational field. This is known as Galilei's principle and it has been tested in modern tines with an extremely high preCision! ROTATING FRAME OF REFERENCE
INERTIAL FRAME OF REFERENCE
x
i
(THE PARTICLES ARE
BO'l'Ii AT REST)
FIG.
112
(THE PARTICLES APPEAR TO BE ACCELERATED)
299
II
But the fictitious forces have the same property: The acceleration of the test particle depends only on the field strength -
rVas '
and
the field strength itself only depends on the metric coefficients. Hence if we release free particles, then their accelerations relative to an arbitrary observer, say a rotating frame of reference, .rill be independent of their masses. Esc:>ecially if
~le
release two free partic-'
les close to each other, then they will stay close to each other! Okay, you might say, I accept that they have this property in common, but have they anything else in cOl'!1I!lon? E'or instance I can transform the fictitious forces completely away by choosing a suitable coordinate system. But you cannot transform away the gravitational field! It is true that I cannot transform away the gravitational field -completely. But I can transform it away locally! Consider a gravitational field, say that of the earth. Now let us release a small box so that it is freely falling in the gravitational field. Inside the box we put an observer with a standard rod and a standard clock.
EINSTEIN BOX
Thus he can assign
coordinates to all the events inside his box (See fig.113) But now the miracle happens. If he releases partic-
OBSERVER
les in his box, they will be freely falling too. Therefore they will move with a constant velocity relative to his coordinate system, and
lLJ
they will act as free particles relative to his coordinate system! Consequently we have succeeded in trans-
SURFACE OF THE EARTH
FIG. 113
COORDINATE SYSTEM IN BOX
forming the gravitational field away inside the box by choosing a suitable coordinate system! When we say that we have transformed away the gravitational field inside the box, this is not the whole truth. Actually if we were very carefull we would find that if we released two test particles as shown on figure 114 they would slowly approach each other. Consequently there is still an attractive force between the two test particles. This force is extremely weak and is known as the Tidal forl;J2. The Tidal force owes its existence to the inhomogenities of the gravitational field, and it is a second order effect, so we may loosely say, that we can
300
II
transform away the gravitational field to first order inside the box! EINSTEIN BOX
So there is a good reason why we might be interested in studying fictitious forces:
It might teach us something about gr(]1)i tationa l forces!
SURFACE OF
Fig. 114
THE EARTH
+ 6.8
CevICTORS The next concept we want to introduce on our manifold M is a co-
vector:
Definition 13 A c ovector
~p
at the point P is a Unear map
~p: Tp(MJ ....
R
Here we will use a notation introduced by Dirac. Given a tangent vector it by
!vp>'
vp
we denote
The symbol
I>
<w I~ > +P P
is
called a keto Similarly a covector ~p is denoted <~pl, where the symbol
vp
~p
R
at Fig. 115
is written as follows
(6.52)
where the symbol < I > is called a bracket. Using this bracket the linearity of
~p
may be expressed in the follo1fling way
To justify the name a vector space
If
covector, we must shm"l that the ~
and
4
are
covectors form
covectors, i.e. linear maps from the
tangent space, we define their sum
w + 4p as follows
+p
301
II
(~+ 4p) (irp) = ~p (vp ) + 4P (vp ) ~p
(you can easily verify that ~
the product of a co-vector
(A~p)
(vp )
+ 40 is a linear map) am. s:imilarly we define
and a scalar A in the following way:
= A~ (Vp )
Using the bracket notation you may rewrite these rules as:
<~p + 4plvp>
a) Sum:
=
<~plvp>
b) Product with a scalar :
+ <4p!V P >
Iv....p > =
A<~p
I....v p >
Consequently the bracket is bilinear in both of its arguments. So the
covectors at the point P form a vector space, ,1hich mathe-
maticians call the dual Vector space, and which they denote by
Let us investigate the structure of this vector spa· ce a little closer. Pick up a
covector
~p.
Introduce a
R
coordinate system around P. Ne can then define
dinates of the
Fig. 116
the COOI-
co vector in the following way:
(6.53) Using the coordinates of
~p
we can express the value of
....
....
~p ~t
a tan-
L
gent vector vp in a simple way. If vp has the coordinates a ,
\ole
find
(6.54) Observe, that an expression like w.a ....
i
is coordinate independent, be-
L
cause <~plvp> is defined in a purely ~eometrical way without any reference to the coordinate system used on M. This is a general featza>e of
..
expressions where We sum OVer a repeated index. ~p
Now let
Xi. Then
~p
have the coordinates wi' and 4P have the coordinates
+ 4p gets the coordinates
<w + 4P y Ie.> .... L In a similar way
Thus
A~p
=
<wple.> +
x•••
gets the coordinates
we see that once ,'Ie have introduced coordinates we may identify
* the dual space Tp(M) with the vector space Rn in the sense that the
302 covector
II
wn is represented by its coordinates (wl""'w
~~
*
n
~
Especially
we see that Tp (M) is a n-dinensional vector space. 'lb each point p on our manifold
....
'.,;e
have therefore now introduced two n-dimensional vectorspaces
When discussing tangent vectors we noticed that a tangent vector
vp had a simple geometrical interpretation. It could be regarded as a velocity vector corresponding to a smooth curve A passing through P. Now we want to give a similar, geometrical interpretation of a
R
covector. Consider a dif-
ferentiable
function,f:M~R,
defined in a neighbourhood of
f(A(t»
P and let A be a smooth curve
passing through P (see fig.ll7). Then t
~
f(A(t»
ry function fran
---=+-----+~
is an ordina-
t-axis
Fig. 117
R into R merely describing the variation of f along A.
Being an ordinary function we can differentiate it at t df (A (t» dt
=0
It = 0
Let us investigate this number a little closer. It represents the rate of change of f in the direction of A. If we introduce coordinates (xl, ... x n ) around P we can represent f by ap ordinary Euclidian funcn tion, y = f (Xl, ... x ), and \\Ie can furthermore parametrize the curve A as xi
~P has the intrinsic coordinates
xi(t). Then the tangent vector
ai
dx
i
It
(it
=
0
USing these parametrizations we now obtain (6.55)
df(A (t» dt
d
It=O
-
i
= dt f(x (t»
()f
-.a
It=o
d
This formula shows that the number dt f(A(t» tangent vector
vp
i
()X~
depends only on the
and not on the particular curve generating it. It
also shows that this number depends linearly on the tangent vector
vp.
Ne therefore define a linear map on the tan<Jent space Tp(M) by -+
vp
df(A(t» dt
where A is any smooth curve generating tangent space Tp (M)
is nothing but a
vp.
But a linear map from the
covector: This
covector .Till
be denoted df , and we refer to this expression as the exterior deriva-
tive of f.
303
II
We have thus shown (6.56)
d I....v p > = dt
f(
A(t)) It
=
0
=
af a i axi
If you compare this with (6.54) you can immediately read off the coordinates of df . af ., -af , ... ,af df.( -) ax~ ax' ax n
(6.57)
is nothing but a generalization of the gradient vectors
But then df
you know from elementary vector analysis in Euclidean spaces! Using gradient vectors, we may analyse the structure of the dual
* space Tp(M)a little closer. Let ?obe a point on M, and let us intron duce coordinates (xl, ... ,x ) around Po' Then we have constructed a frame for the ordinary tangent space T Po (M)
in the fOllowing 'lay:
First we introduced the coordinate lines on M,
see fig. 118,
AI(t) = cp(t,O, •.. ,O); ... ;An(t) = cp(O, .•. ,O,t) and then the coordinate lines generated the canonical frame vectorsdAI
~
It=o
• . .... e = ~ , ... , n dt It=o
Now we want to do something similar. First we observe that (O,t)
we can define a smooth function on M in the following ''lay: Q(xl, ••• ,xn)~ Xl This function is referred to as
the first coordinate function, and for simplicity we ''Jill denote it by Xl. Thus Xl (0) = X6 .lhere 0 (xg, ... ,x~). In a similar Iolay we can define the
has the coordinates
other coordinate functions
xi, i=2, .•• ,n "lhere xi (Q) = • xi0 .
Being smooth functions these coordinate functions x~ vlill generate
dx i at the point Po' These
covectors
covectors have particularly
simple coordinates, e.g. dXI gets the coordinates axl axl axl dXI: (--, - - , •.. ,--) = (1,0, ... ,0) n axl ax 2 ax and similarly
dx': Therefore the
(
°,
n flx : '(
1, ••• , 0) ; •• :;
covectors
°,°,... ,
1)
dzl; ... ; dzn serve as the canonical frame vectors in
the dual space T;rM) . If an arbitrary (6.58)
~p
=
covector-
~P
has the coordinates Wi' ."Ie noW get
WI dxl+"'+Wn dxn
304
II
This formula is in particular valid for the gradient vector df Therefore
\ole
(6.59)
conclude
df
Here you see something strange! This formula looks exactly like the well-known formula for the differential: df ... df di i dx But in modern geometry the idea of infinitesimal increments have been burried becaUSe of the logical difficulties involved in their use. The symbol a is now interpreted as an exterior derivative and the miracle happens: Many of the formulas involving the exterior derivative look exactly as the «old« formulas inVOlving differentials. So in modern geometry we work with the same formulas, they have just been given a new interpretation.
Finally we may investigate how the coordinates of a covector ~p are changed if we exchange the coordinate system. Let the old coordinates be \~Ii and the ne'lT coordinate:, be (6.12)we then get
I~
I~
dX
L
I~j = <~p, &Ij > = <~p, tfd. dyj>
axi
w, --, (ilL dV J
So the coordinates of a co-vector are transformed according to the rule (6.60)
w. (2)J
Thus the coordinates transform their name: covectors!
covariantly, and this justifies
We can also easily find out how the canonical frame vectors transform. If we introduce two coord~nate systems, then the old coordinates (xl, ..• ,x n ) are connected 'to the nelol coordinates (yl, •.. ,yn) via the transition functions
yl
yl (Xl, ••• ,x n )
yn
yn(x1, ... ,xn)
But here yi
=
yi(x1, •.• ,X n ) is nothing but the new coordinate func-
tions expressed in terms of the old coordinates! A simple application of (6.59) then gives: (6.61)
305
II
Up to this point we have been working with a manifold without a metric. At eact point of the manifold "le have learned how to construct two n-di-
* mensional vector spaces Tp(M) and Tp(M). Let us now assume that we have also been given a metric g on the manifold M. Then we can do something strange: We can fuse the two vector spaces Tp(M)
* and Tp(M) to-
gether so that they become indistinguishable! I.e. we may construct an identification procedure where each tangent vector \-lith an unique
covector
~p
Vp
is identified
in a purely geometrical nanner.
To see how this can be done, let us choose a tangent vector
Vp.
Then ,'Ie construct a linear map, Tp (M)~ R ,in the following way
But a linear map, Tp (M)
~R,
is nothing but a
tangent vector vp has generated a
covector, and therefore our
covector ~/hich will be denoted as
g(vp ; .)
Vp
Let us introduce a coordinate system around P. Let have the coori dinates a . What are the coordinates of the corresponding co-vector g(Vpi')? They are easily found. Denoting them by a i . k . (6.62) a i = g(vp,e i ) = gjkaJoi = gjiaJ
we get
i Thus the tangent vector with coordinates a generates the covector j with coordinates a. = g .. a . 1. 1.J. The map, I:a1.~a. = g .. a J , is a vector space isomorphism between * 1. 1.J Tp(M) and Tp(M) because g.. is a regular matrix. Therefore we may use 1.J * this map to identify Tp (M) with Tp (M). This is called the canonical iden-
tification! Exercise: 8. 8. 1 Problem: Let wp be covector characterized by the coordinates w.• Show that the corresp~nding tangent vector is characterized by the comp6nents wi where ( 6 • 6 3)
k
w =g
ki
wi
'
It should be emphasized that tangent vectors and covectors were born in different vector spaces with different geometrical properties. If there is a metric g on the manifold, we may, however, identify them. Hence you will often see that when there is a metric, one simply «forgets« the dual space, i.e. the covectors. But then you have only tangent vectors at your disposal. Given a tangent vector, you can now characterize it by two sets of numbers: a)
You can
decomp~se
....
;p along the canonical
frame vectors e , 1.e.
i
Fig. 119
II
306 i .... a e.
1
where the numbers a b)
i
are referred to as the
contp~-variant
components.
You can use the numbers a
........
i
= g(vp,e ) .
i
They are referred to as the covariant c~onents. If~. is a unit vector, then a i is simply the orthogonal projection of vp onto ~i! 1
Due to the rules and
(6.64)
a
i
vIe often say that the metric ct:Jefficients are used to "paise and lowep indices"!
Illustrative example:
The gauge potential outside a solenoid.
Let uS consider the Minkowski space M. Here we have a metric g, so we do not hacovectors. A vector field A could for instance represent the gauge potential in electromagnetism. We have previously computed the components of the gauge potential outside a solenoid in an inertial frame (~ection 1.4). ve to distinquish between tangentvectors and
2
A
Boa [0, _ --y---2-
The spatial part was circulating around the z-axis. This makes it natural to introduce cylindrical coordinates (t,P.~,z). After all there is nothing sacred about inertial coordinates in special relativity! In exercise 6.6.3 we have computed the Jacobi matrices for the transition functions and the metric coefficients. We are therefore ready to find the contravariant components of the potential in cylindrical coordinates. Using formula (6.8) we get 0
Ao.=B oa 121
Cos~
2 •
2
0 0
_
0
0
Sin
0
Sin~ Cos~
P
P
0
0
0
_
Sin~
P Cos~
0
0
Boa 2
~
0
1
P
;J2
0
0
but as we know the metric coeffi~ients (6.39) we can easily find the covariant components in cylindrical coordinates. Using that A =g 0.
0.8
pf>
i.e.
GAcontrav.
Acov.
we then obtain At= We can also
o·, A p= 0 ; A ~
2
Boa 2 - '• A z = 0
A using
cylindrical coordinates
307
II
Thus A is a space-liNe vector, and its magnitude falls off like ~. If
H=l
dX~
then the contravatiant components d~
d~
= gUV a
g~V
l
V
Thus
in spherical coordinates, for instance, you get (compare section 6.6)
Returning to the potential A we can now write down a gauge transformation in a coordinate free manner as follows
(6.65) where X is an arbitrary scalar field. (If you write it out in coordinates, you get
which coincides with a gauge transformation in inertial coordinates. Therefore it is valid in an arbitrary coordinate system). To see how (6.65) works, let us consider the pot~ntial outside a solenoid, and let us work in cylindrical coordinates: 2
(At;>Ap>A(?>Az)
Boa = -2-
(0,0,1,0)
As the scalar function we choose the coordinate function -Boa 2 -2i.e.
~
multiplied by the constant
-Boa 2
X(t,p,~,z) = -2- ~
Then the "differentiel« dx gets the coordinates
£X (£X dt ' dP and A+
dx.
~
£X)
, dip , dZ
2
Boa -2- (0,0,1,0)
the coordinates Boa 2
2
2 Boa ~ (0 0-1 0) + -2(0,0,1,0) = 0 ",
This does not mean that we have managed to gauge away A completely. The coordinate function ~ is not defined throughout the whole of spacetime. We must exclude a halfplane bounded by the x-axis, where ~ makes a jump. Hence all we have shown, is that if we cut away a halfplane bounded by the z-axis, then we can gauge away A in the remaening part of spacetime. So we have arrived at the same conClusion as we did the first time (Section 1.4). At the same time we have lerned an important lesson: Every time you use a cyclic coordinate (i.e. a periodic coordinate) there is a singularity associated with this coordinate, and the corresponding coordinate jUnc-
tion does not dEfine a smooth scalar field on the whole of our manifold.
II
308
6.9
TENSORS
By now yOU should begin to have a feeling of how to construct the geometrical concepts associated with a manifold. We consider linear maps to be the natural geometric concept associated with the tangentspace. Metrics were constructed as bilinear maps on the tangentspace, and
covectors were constructed as linear
ma~s
on the tangentspace.
r'!e can now generalize the concepts of a metric and a
covector. Consider
a map Fp ( ' ; " ' ; ' ) with k arguments. The arguments are all tangent vectors
vJl~
.. 'Vp(k),
and Fp
is supposed to .... (1)
.... (k)
) map such a k-tuple (v p ; •. ;v p • -+ (1) .... (k) Lnto a real number Fp(vp"";v p ). The set of all k-tuples (V~l); .• ;vp (k) ) is denoted ,Tp(M)X. "xTp(M),
Fig. 120
v
k factors
and thus Fp
is defined as a map
Fp : Tp(M)x •..• xTp(M)~ R To be of any interest .re will, of course, assume that Fp is linear in each of its arguments
....
(k)
AF p ( .... ....... • .... (k) ) + v (.1)., •• ,up, .•. ,v p
~'p
-+
(11
Fp(v~
....
....
, ••• ;AU p + ~wp; •.. ;vp
I:
=
)
. .... (k) ) vp(0., ... ...... ,wp, ... ,v p
( ....
l'1e express this property of Fp by saying that it is a multilinear map. This is the kind of map we will be interested in and we ,o[ill give it a special name:
De fini tion 1 4 A cotensor
Fp
is a multilinear map
Fp : TiM)x ••• xTp(M) ~ R
If Fp has k arguments we say that Fp is a co tensor of rank aU c
Clearly a
co vector
wp
is nothing but a
metric gp is a symmetric, non-degenerate
k. and the set of
cotensor of rank 1 and a cotensor of rank 2.
309
II
The next thing we must investigate is the possibility of introducing "coordinates" to facilitate computations Il7ith show' you that these
cotensors (and to
cotensors actually correspond to the tensor con-
cept you are used to in ordinary physics.) For simplicity we will illustrate the machinery with co-tensors of rank 3: Let us start by introducing coordinates (x1"
n .. ,x ) around the
point P in M. If Sp is a co-tensor associated with the tangent space T?(M) we will refer to the numbers (6.66)
Sijk
= Sp(ei:ej:ek ) n
as the components of Sp with respect to the coordinates (x1, .. "x ) Let
~ ~ ~ up,vp,w
j
i
a ,b,c
k
b e tree h ar b'~trary tangent vectors with the coordinates
n
• respectively. We can then easily express the value of
Sp(~p;Vp;;p) in terms of the components of Sp and the coordinates of
....
~
~
up,v p and wp ~ ~ ~ _ i~ . j~ . k~ Sp(upiVpiWp) - Sp(a ei,b ej,c e k )
_ i j k ~.~.~ _ i j k - abc Sp(ei,ej,e k ) - Sijka b c
Observe, that allthough the specific components S'J'k and coordinates i j k • a ,b,c depend strongly on the coordinate system, the number i j k Sijka b c is independent of the coordinate system!This is an example of the iIrp::>rtant rule: Whenever we contract all upper indices and lower indices, is the resut-
ting number an invariant, i.e. it is independent of the coordinate system employed. Let us check how the components of Sp transform under an exchange from x-coordinate"
~o
~
~
~
~
~
~
§"bc (
=
y-coordinates,
Sp(e ;~h;e ) \2,a 12,... l21==
= dx i
~ dX i ~ dx j , ~ Clx k Sp(e,--; (1)1, Clya l~b dyb' (~ClyC Clx j dX k
Sp(e'ie,;e~)--:-E --UI~ CllJ Ill"" Clya dy dyc
k i j S dx dx Clx U1ijkdya dyb Clye
where we have used (6.12) and the multilinearity of Sp' Thus you see that the components transform as the name
(6.67)
~-tensor)
(~abc
:
covariant
co~ponents(which
justifies
II
310 l'le have various possibilities for
mani~ulating
tensors:
1. the sum of two co-tensors: If Sp and Tp are cotensors of the same rank k, then their sum, Sp + Tp ,is defined as the
follo~1ing
-
cotensor of the same rank k
-+ -+ ~ (Sp + T p) (UI;"';u k ) = Sp(ill;"';u k ) + Tp(u1;"';U k ) -+
-t
and the sum is characterized through the components S.
1.1
•••.
1.k
+ T .•••. 1.1
1.k
2. Product of a cotensor and a scalar: If Sp is a cotensor of rank k, and A is a scalar, then we can define a new cotensor of rank k, ASp' through the formula (ASp)
(ri I;' .. ;rik )
= ASp (ri 1 i· .. ;ri )
k
The product of a cotensor and a scalar is characterized by the CCIIIpOnents AS· ... , 1.1 1.k 3. Tensor product: If Sp and Tp are
cotensors of rank sand t, then their tensor procotensor of rank s + t
duct,SpSTp,is defined as the following (Spit T,,)(ri ; ... ;ri -
1
S
iV 1 i ... ;Vt
)
The tensor product has the components
4. Contraction:
v , ... ,v,
If Tp is a cotensor of rank k, and are tangent vectors, III l;cl then their contraction is defined as the follo~ling cotensor of rank (k-t) -+
T" (v(1) ;
-+
~
••• iV t';;";;')
\9.
and the contraction is characterized through the components i
l
i
Til •.. i~, i H 1 . . . . . i all) .•. ~ k It is important to observe that the
f cotensor space Tp (O,k) (M)
actually carries a linear structure, i.e. it is a vectorspace. Once we have the tensor product at our disposal, we can discuss how to build up
cotensors of arbitrary rank systematically using the co-
vectors as building bricks •
311
II
(0,1) l'Je have previously studied the space of covectors T*(M)=T p p and especially we found the canonical frame vectors dx 1 ; ••• idx n • These are the co~ctors which we shall use as the basic units!
(M)
Consider first an arbitrary covector Wp. If it has the components wi we can decompose it in the following way
.
(6.58)
W"
=
w.dx ~
i
Next we consider a cotensor Fp of rank 2! Let it have the components Fkt . Then we can construct another cotensor of rank 2 i
Fijdx fldx ~'Jhat
j
are the components of this F ij d x
i
~
cotensor?
dj-o ... ) x (ek;e t
has the components Fkt too, and we conclude that are identical cotensors, i.e. (6.68)
This is an important result: Any
cotensor of rank 2 can be expres-
sed as a linear combination of the cotensors: dx 1 fldx 1 ; dx 2 fldx 2 ; ••. ; dx n 8dx n j Therefore the co tensors dxifldx generate the whole tensor spa ce Tp (0,2) (M), so you see that they act as canonical frame tensors. Now you can guess the rest of the story! Let Tp be a cotensor of rank 3 with the components T ijk , we can then decompose it in the following way (6.69)
I
etc!
Consider the metric g. It is a cotensor of rank 2. If we choose a coordinate Gystem, the~all cotensors of rank 2 can be expanded in terms of the basic cot ensors dxafldx Especially we may expand the metric tensor a S (A ) g = gaS dx fldx This formula is interesting, because formally it looks very much like the «square of the line element« (B)
But (A) has an exact meaning: It states that two well-defined cotensors are identical, while (B) only has a symbolic meaning. It is a mnemotecnic rule of how to
312
II
write down the arC length of a curve
dA (or even worse: dt is an infinitesimal arc length, and dt 2 is the square of an infinitesimal quantity, whatever that is). Hence you see again that the formulas of modern differential geometry very often resemble formulas from «old days« involving «infinitesimal quantities« and other mysterious things. Using the notation g = gusdXUaaxS we can write the metric on a 2-sphere as
(6.26) and the metric on the Minkowski space expressed in spherical coordinates as
(6.36)
g
= _ dtadt
+ dr8dr + r" ( deade + Sin 2 edq:sd q»
Let us look at a point P on a manifold Mn. To this point P we have attached the tangents pace Tp(M), and the tanqent space has been used to generate all the attached the
cotensor spaces. But to the point P we have also
* cotanqent space Tp(M), and that is a II-dimensional vec-
torspace with a structure very similar to that of the tangent space. T'le can therefore generalize the 1iscussion of
cotensors in the fol-
lowing way:
De fini tion 1 5 A mixed tensor Fp of type (k,~) is a multilinear map
* * Fp : Tp(M)x ••• )(Tp(M) x TiMJx •.• ){Tp(M)'" R !.'~
k-factors
i-factors
A mixed tensor of type (k,l) has rank k+l, and the space of all mixed tensors of type (k,l) is denoted T~k,2) (M).
A mixed tensor of type (k,O) is siI!\ply called a tensor of rank k • Let us focus the attention on the tansors of rank 1. A tensor Ap of rank 1 is characterized by its components Ai Ap( ax i ), and these components transform contra variantly. Thus there seeI!\S to be an intimate connection between tensors of rank 1 and tangent vectors. In fact we can identify theI!\: Let Vp be a tangentvector. Then the map ~p
<w
"'p
Ivp >
313
II
* is a linear map Tp(M) ~ R, since the bracket is bilinear. Therefore vp
...
generates a unique tensor of rank 1, which we also denote vp'
Exercise 6.9.1
.
...
...
Problem: Conslder a tangent vector vp ....Show that the components of vp as a tangent vector and the components of vp as a tensor of rank 1 are iaentical.
Ylhen we have tensors of an arbitrary t:'pe at our disposal w" can ge-
neralize the contractions. Let, for instance, Tp be a tensor of type
(1,2). It is then characterized by its components
n with respect to some coordinates (x1, ... ,x ). Here the index a transforms contravariantly and the indices band c transform
covariantly.
But then we may contract the indices a and b obtalltingthe quantity:
As we sum over the index a, this quantity is characterized by only one index
The important point is now that the quantity Sc transforms covariantly! This follows from the computation j a S Ta av Ti ax axk (Toc (21 ac axi C!l jk aya a~/
a j where we have used that (~) and (~) are reciprocal matrices. But axl. aya if Sc transforms covariantly, we may regard it as the components of a cotencor Sp' We say that Sp is generated from Tp by contraction in the first two variables. This is obviously a general rule: Whenever you
contract an upper index with a Zower index, the resuZting quantity transforms as the components of a tensor. (where the degree of course is lowered by two!). The only trouble with contractions is that they are almost impossible to write in a coordinate free manner! As contractions play an important role in the applications, we will therefore write down many equations involving tensors using component notation.
314
II
In the following table l'/e have summarized the most important properties of mixed tensors:
A mixed tensor Tp of type termined by its components i
Components
T
l • .. i
k
.
. = T (dx
JI'''~
i
l
(k,~)
is completely de'
; ... ; dX1k;~. ; ... ;~. )
P
JI
J~
with respect to a coordinate system. The indices i l , ... ,i transform contr8:variantly , and the indices k jl, •.. ,j~ transform covariantly.
If Sp andTp are mixed tensors of the same type and A 1S a real scalar, then you can form the mixed tensors Sp+Tp and ATp of the same type (k,~). Furthermore these mixed tensors are characterized by the components (k,~),
Linear structure
(6.71)
Tensor product
If Sp and Tp are mixed tensors Of type (kl'~I) and (k 2 ,P-2), you can form the mixed tensor S1' ~ Tp of type (k l + k2 '~l +P-2). The tensor product is characterized by the components
(6.72)
S
il ..• i k
il ... i 1. T JI"·Jt
1
k
2
jl·"j~2
If Sp is a mixed tensor of type (k,~), you can form a mixed tensor Tp of type (k-l ,~-1) by contracting a contravariant and a covariant index. The contraction is characterized by the components
Contraction
(6.73)
In what follows we are going to deal a lot with tensor fieZds.
To con-
struct a tensor field T of rank 3 we attach to each point P in our man nifold a tensor Tp of rank 3! Let us introduce coordinates (xl, ... ,x ) n on M. Then the tensor at the point P(xl, •.. ,x ) is characterized by its components T
abc
(x 1 , ... ,x n )
315
II
\·7e say that the tensor field T is a smooth field if the components Tabc(X', ... ,xn) are smooth functions of the coordinates. If nothing else is stated we will always assume the tensor fields to be smooth.
The unit-tensor field.
Illustrative example:
Let M be an arbitrary manifold. Then we construct a mixed tensor of type (1,1) in the following way. At each point P we consider the bilinear map
The corresponding components are i .... To( dx ie.) • J
'" < ax
i
" l e.> J
i.e. the Kronecker-delta! It is a remarkable fact that the components of this tensor have the same values inall coordinate systems. As the components are constant throughout the manifold, they obviously depend smoothly upon the underlying coordinates! We therefore conclude:
The Kronecker-delta oi. are the components of a smooth tensor field on M, called J
the unit tensor field of type (1,1).
Exercise 6.9.2 Problem: Show that the Christoffel fields r
V
as
=
l.
..,g
v].! [a/>:].l(:t
a/>:S].!
g a aS ]
ax
ax].!
--S+-a --ax
are not the components of a mixed tensor field. (Hint: Show that they do not transform homogenously under a coordinate transformation. )
Exercise 6.9.3 n Problem: Consider a point P on our manifold. To each coordinate system (x', ... ,x ) around P we attach a quantity T·lJk with two upper indices and one lower index. Apriori the upper indices need not transform contravariantly, and the lower index need not transform covariantly. Show the following:
If the quantity U~ given.by Q, " RU k - T1-J S
-
k
ij
are the components of a mixed tensor of type (1,1) .Ulhenever S1.ij are the eomponents of a mixed tensor of type (l j 2), then p1-J k are the components of a mixed tensor of type (2,1)
316
II
The method outlined in exercise 6.9.3 is very useful when you want to show
that a given quantity transforms like a tensor. Clearly it
can be generalized to mixed tensors of arbitrary type. Suppose now that we have attached a
metria g to our manifold M. Then
we have previously shown (Section 6.8) how to identify tangent vectors and covectors using the bijective linear map
I:T;(M) - Tp(M)
,
generated by the metric. If we write it out in components it is given by
Exercise 6.9.4 Problem: Show that I(
axk )
is characterized by the components gki
11/e can now in a similar way identify all tensor spaces of the same rank. For simplicity we sketch the idea using tensors of rank 2. Let T be a
cotensor of rank 2 characterized by the
covariant com-
ponents T ij . Then we identify T with the following tensor of rank 2: II (T)
(~;4)d~f.T(I(~) ;I<';~»
The tensor I 1 (T) is characterized by a set of contra variant components which we denote T ij • They are given by
Tk~ = II
(1)
(
In this way we have clearly constructed a bijective linear map II: Tp (0,2) (M)
_
Ip (2,0)
(M) •
tie can also identify T with the following mixed tensor of type (1,1)
Here I 2 (T)is characterized by the components
In this way we
hav~
generated a bijective linear map
12: T(O,2) (N)
....
T(l,l)
(M)
3"1:7
II
On a manifold M with a metric g we have therefore a aa:noniaaZ identifiaa-
tion of the tensor spaces 0f rank 2:
g
Tp (0,2) (M)
g \
Tp (1,1) (M)
(2, 0) (M) •
Obviously these results generalize immediately to all tensors of higher ranks. If there is a metric g on our manifold M is it customary not to distinguish
bet,'leen the various types of tensors of the same rank.
one simply speaks of tensors and a given tensor T is then represented by various types of components. For instance a tensor T of rank 2 is represented by the following four types of components T ij ,T
ij
,T i
j
,T
ij
and these components are related to each other through the formulas: Ti. J
=
g
j
=
g
Tij
=
k .. gik T j T 1J
=
gkjT i
ik
T
kj
T.
1
kj
:iJc
~j
ij
=
g
Tij
=
gikg~jT
Tik T
g
Tk~
(6.75)
k
k~
This is known as the art of raising and lowering indices!
The metria aotensor.
Illustrative example:
Let us apply this to the metric
cotensor itself.
find the mixed components of g and the contravariant Raising one of the indices i
g j
=g
ik
gkj
\"le
Thus we want to components of g.
find
i
=a .
Consequently the mixed ca:nponents are nothing but the Kronecker-delta! As we have already used the notation gij to denote the reciprocal matrix we will for a moment use gij to denote the contravariant
compo-
nents of g. Raising both of the indices we find: gij
= gikg2jgk~ =
gikojk
=
gij
So we may relax: gij actually denotes the
contravariant components of
g too. Hence the metric g is characterized by each of the following components (6.76)
and
where gij and gij are reciprocal matrices and Oij is the unit matrix.
II
318
TENSOR FIELDS IN PHYSICS
6.10
To see how a physical quantity can be represented by a tensor, we will consider the electromagnetic field. It is characterized by the
E and B
field strengths The components
~,
measured in an inertial frame of reference.
... ,Bz are collected into a skew symmetric square
matrix 0
~
Fas
~
E
-~ 0
-~
-Bz
0
By
z
-E
z
-By
Bz
Ex
-Ex
0
If we exchange the inertial frame of reference
5 1 ,characterized by iner-
tial coordinates xa,With the inertial frame of reference 5 z ,characterized by inertial coordinates yS, then it is an experimental fact that the electromagnetic field strengths transform according to the rule (6.77)
VJa S
wherelA
is the reciprocal Lorentz matrix, i.e.
This allows us to define a
cotensorfield F of rank 2 on Hinkowski spa-
ce. Let P be a space-time point, and
ITp,vp two tangent vectors. Then
we define (6.78 ) where F S,UU,v S are the comP9nents relative to an arbitrary inertial a . frame. That the number on the rigth hand side is independent of the inertial frame follows immediately from the transformation rule (6.77). Thus in our mathematical model the electromagnetic field is represented by a skew symmetric
cotensor field F of rank 2.
Of course, we do not have to restrict ourselves to inertial frames. As a specific example let us consider a pure monopole field. Let us assume that the monopole is at rest at the origin of the inertial frame of reference
SI. We want to find the components of the monopole
field in terms of spherical coordinates
(t,r,e,~).
me the electric field strength E vanishes and
In the inertial fra-
II
319
....
B
....
r r3
....2...... 4rr
where g is the strength of the monopole. Therefore
=--.!L
F (1)06
4rrr2
0
0
0
0
0
0
~
x..
0 0
r
z r y r
r
x
0
r
x
0
r
Using the transformation rule (6.67) we find
i.e.
The Jacobi matrices were computed in section 6.6.Inserting (6.35) we get
F
(6.79)
J.4rr
121
0
0
0
0
0
0
0
0
0
0
0
0
0 -Sine
Sine 0
Thus only the F e(? -term is non-vanishing in this particularly simple coordinate system. The corresponding contravariant components become
i.e.
T'Te have already determined the metric coefficients in spherical coor-
dinates, see (6.36):
G
[~
0
0
0
1
0
0
0
0
.
=-1
J..e.G
=
-1
0
0
1
r2
0
0
0
0
r 2 Sin 2 e
0
0
and from this we obtain the contravariant
components of
0
0
0
0
1 '1:
0
F:
2
0
1 r 2 Sin 2 a
320
paS
( 6.80)
(2)
--.:l.. 411
II
0
0
0
0
0
0
0
0
0
0
0
0
0
~
r"Sin8
~
r"Sin8 0
Observe, that we cannot find the "strength" of the field just by observing the
covariant (or the contra variant) components of F.
To do that you must form an invariant quantity. If you have a tensor T of rank 2 you can form the following two invariants: a a
A) The trace of T·
T
B) The "square" of T: ~TasTaS But the electromagnetic field tensor F is skew symmetric, i.e.
F(~p,rtp)=-F(Up'~p),so the trace vanishes automatically. Concerning the square, you get:
Therefore the "strength" of the field depends only on the radial coordinate r, and it varies as !2! r
Illustrative example: The four-momentW7l and the energy-momentum tensor'. As another example of a tensor in Minkowski space we consider the energy-momentum tensor T . This is a s~etric tensor of rank 2. In an inertial frame the contra variant components T(l have the wellknown interpretation:
Energy:
( 1.37 )
ae T
Energy]
;:!e~si:tz _I:,. ___c;:r.::eEt_
~ Mome~tum: [ densl.ty ,
,
Momentum currents
For a sourceless electromagnetic field F the energy-momentum tensor can only depend on F i.e T = T(F).In an inertial frame we have found the following relation (1.41 )
As naB are the contra variant components of the metric g we conclude that the general relation is
II
(Since the left and the right hand sides are components of tensors which coincide in an inertial frame they must be identical). By definition the tensors are constructed as multilinear maps on the tangent space
Tp(M) and the cotangent space ~(M). You may find it strange to represent a phys1cal quantity by a multilinear map, so let us look at some examples:
A) Let us lssume that we are observing a particle moving in spacetime with a fourmomentum P. The four--momentum was introduced as a tangent vector along the worldline a
pa
= m dx dT
but we can identifY it with a co vector P i.e. a terize it by its covariant components Pa
cotensor of rank 1, and charac-
R p
Fig. 121
As observers we are ourselves moving through space-time along a certain world line ~ith a certain fo~r-velocity As P is a linear map, Tp(M)~ R, it will map into a real number
U.
U
the energy which the observers will measure, because the observers are at rest relative to this inertial frame. On the other hand, the four-velocity will have the coordinates ~ = (1,0,0,0) with respect to this inertial frame. We therefore get:
U
a
So we conclude that-
B) Consider the energy-momentum tensor T of a certain field configuration. Let us assume that we are observing this field configuration, and that we are moving with a certain four-velocity As T is a bilinear map,
U.
Tp(M) x Tp(M)
~
R
it will map the pair of tangent vectors (U,U) into the real number T(U,U). To find the physical interpretation of this number, we introduce again an inertial frame, where we ar.e momentarily at rest. With respect to this frame T is characterized by the covariant components ~B where Too is the energy density of the field which we will measure! Our four-velocity, on the other hand, is characterized by the components un (1,0,0,0). Therefore we find,
=
T(U,U) = TaSU~S = Too and we conclude that T(U,U) 1:S the energy density of a field with energy-momentwn T measu:t'ed by an observer with four-velocity U.
n
322
SOLUTIONS OF WORKED EXERCISES, No.
6.5.1
Let gij and gij be the components with respect to two coordinate systems If)
I!1J
(~2'
U2 )
Then, by definition: lj,fuk =
0\ (and If)i 1iik 0\) j
=
But we know that gij transforms covariantly! Consequently dxil. axm (~k = ayJ ayK 1~9.m Substituting this we get m ij axil. ax oi l~) ayj a/ (?~m = k
ak
Multiplying with the reciprocal matrix -I- we get axP i .. e.
Multiplying with the reciprocal matrix .,.Pq we further get ('i')
,
Le.
g
ij ax q
(2)
ayi
-, = -
ayJ
g
pq
al III
ak
Multiplying with the reciprocal matrix L we finally obtain q ax
i.e.
g
ik
(21
So finally we get the formula we wanted! After a lot of index gymnastique you. see with your own eyes that gij transforms contravariantly!
No.
0
6.7.1 Taking the Lagrangian adx dA
a
~----,-
/
L(xj-) =-mV-g
dxa dx S
'as dA dA
r J
, d ~a aL as a starting point the Euler-Lagrange equatlon -aL = -X axa d a(a:xagaS dx a dxll
~dX~
2V
reduces to
can
323
II
:J~l agalldxdx+ S a
~ t---;,:s ax ax
2 aJ dx
gall
~
(V That is obviously a mess, but you should not be surprised of that because even a straight line may be characterized by a complicated parametrization: xa
= aaSin(etgA)
+ ba
Hence even a simple curve like a straight line may solve a complicated differential equation, if we choose a crazy parametrization! Let us take advanta~e of the fact that we can choose our parameter A as we like it, so let us use the proper time T as the parameter! Hence we put A = • in the equation of motion and use the simple relations:
.'dA !..- V-= a
V-=1,
Then the equation of motion is simplified considerably!
2
g
a
d x -all d T2
o
(Here we have split the term:
Performing the substitutions a+a,a~ in the last term, this is rearranged as
I f we multiply this equation with gllV we get: 2
a
d x oVa -2 d.
g
1 llv ya [a + - g + 2 axS
2 V d x 1 llV =>--+-g
6.. 2
and we are through!
2
D
ag
Sll
ax
a
a ] dx di = 0 dT d. axll
_ agas
324
II
No. 6.7.3 If we extremize
=
s
we get the Euler-Lagrange equations
which we may rearrange as 2
d t
d? = 0
~ = r (~~/ (6.81)
d 2e
+ r Sin2e
(~~) 2
dr
2 dr de dIP 2 = - ;:: de dT + sin e Cos e (de)
g dT
= _
~ dr dIP _ 2 Cot e de
dT dT
These equations then comprise the equations of motion for a free particle. But we can also Use them to read off the Christoffel fields. Using the formula d 2x
a
d?" '" - r
a
dx].l dx
].lv
V
dT dT
you immediately get the following nonvanishing components of the Christoffel fields: rr - r Sin 2 e - r
r~e
(6.82)
a
e re rU)
er rIP
rq>
q>r
r
r
r
r
a
(/XI)
- sin
(!XI> 1 r\P = rIP r Sq> tP9
e Cos Cot e
e
~
325
II
ehapter 7
DIFFERENTIAL FORMS AND THE EXTERIOR CALCULUS 7.1
INTRODUCTION
We want to extend the differential calculus, to include tensors. Consider an arbitrary Euclidean manifold M and a cotensor field T of rank 2 on this manifold. If we introduce coordinates (xl, .•. ,xn ), we can characterize T by its co-variant components TetS(x I , ... ,x n) Here you might try to differentiate the components of T ,'lith respect to one of the coordinates x~, i.e. you form the components n a~ TetS(XI, ... ,X ). The question is: Is this new quantity a co-tensor field of rank 3, i.e. do the components a T Q(xl, .•. ,x n ) coincide with ~ a" the components of some co-tensor of rank 3? To investigate this we must try to show that the quantity a~TaS transforms we introduce new coordinates (yl, ..• ,yn).
covariantly. Therefore
Wi th respect to the old coordinates (x I, ... ,xn ) the quantity has the components din
--T
aX~(11 et
B(x
I ••• ,X
)
and with respect to the new coordinates (yl, ..• ,yn), the quantity has the components din
--T
aY~(21
etS(y, ..• ,y ).
As I
n
Tets(y , •.. ,y ) (21
=
I n ax Y T .s(X , ••• ,x ) - ayet
(II Y
II
326
we get
The first term is exactly the term we are after, but the second term spoils everythin0. So the quantity with components
not a
a~Ta8
is certainly
cotensor:
We can even understand intuitively what went wrong! We are looking at a co-tensor field T and trying to form something like the partial derivative of it. Intuitively you would then try to make an expression like 1
[Tp(x l
£"
•
,_
.x~+~,
•• ,
c"
'x n ) -
T P(x l
••• ,
.x~,,
I• • • • ,
'xn) 1
••• ,
But this has no meaning at all because we are trying to subtract two tensors with
different base points! But they lie in different tensor spa-
ces and have absolutely nothing to do with each other. Of course, you could still try to form the component expression 1
[ T a8 (xI ; ..• ;x ~ +
£"
E; ..• ;Xn) -
n) Ta8 (Xl., ••. ,x .~., .•. 'x ,
1
This is legitimate, but observe that this expression has no geometrical meaning. You cannot compare two tensors with different base point P and Q just
by compairing their components. Original coordinate system
M
Changed coordinate system
M
Fig. l23a
Fig. l23b
Introducing a coordinate system which covers P and Q we may characterize the tensors Tp and TO by their conponents
327
II
and You might hope that i t I.ould make sence to say that the b/o tensors Tp and TQ are almost equal if their components
T~S(P)
and TaS(Q) are
almost equal. But watch out! Let us change the coordinate system in such a way that the coordinates in the neighbourhood of P are unaffected while the coordinates in the
nei~hbourhood
of Q are changed dras-
tically (see figure 123). Then Tp is still characterized by the components T'aS(P)
=
T~B(P)
while TQ will be characterized by the new components
T'NQ(Q) ~" Thus,
=T
Y
Yo
0
(0) ~ ~ ay~ ayS
although we have fixed_the components Tp, we can change the
components of TQ into almost everything! This shows clearly that it makes no sense to compare
cotensors
with different base points, just by comparing their components. In these notes we will not try to attack the general problem of constructing derivatives of co-tensor fields but will only show that it is possible to overcome some of the difficulties if we are willing to vlOrk only with a special kind of
7,2
cotensor.
K-FORMS - THE WEDGE PRODUCT Now let F be a
cotensor of arbitrary rank. He say that F is skew-
symmetric if it changes sign whenever we exchange two of its arguments:
If we characterize F by its components,
we see that F is skewsymmetric, whenever the components F aJ···· a k are skewsymmetric in the indices aJ, ... ,a ! The skewsymmetric cot enk sors are so important that mathematicians have given them a special name:
Definition 1 A k-form is a skewsymmetric
ootensor of rank k.
328
II
Consider a specific po.int P o.n Mn. We have previo.usly attached a who.le family o.f
cotenso.r spaces Tp(o.,k) (M) to. this po.int. No.w we
want to. extract the skewsymmetric co.tenso.rs. The set o.f k-fo.rms at k the po.int P will be deno.ted Ap(M). Let us start with so.me elementary remarks: If F and G are skewsymmetric
co.tenso.rs o.f rank k, then so. are
F + G and XF. Therefo.re the set o.f k-fo.rms
A~(M) fo.rms a vectorspace. If k>n then A~ (M) degenerates and will o.nly co.ntain the zero.-fo.rm:
A~(M)
{O}, k>n •
To. see this, let F be a skewsymmetric
co.tenso.r o.f rank k. Then F is
characterized by its co.mpo.nents F
aj ••••• a k '
but as k>n two. o.f the indices must always co.inside (since an index a can i o.nly take the values l, •.• ,n), whence F aj ••• a vanishes! So. in the k case o.f k-fo.rms we do. not have an infinite family o.f tenso.r spaces! We will use the convention that co.tensors of rank 1 are co.unted as l-fo.rms. Of ~o.urse, it has no. meaning to. say that a co.vecto.r is skewsymmetric, but it is useful to. include them amo.ng the fo.rms. In a similar way it is useful to. treat scalars as O-fo.rms. Co.nsequently the who.le family o.f fo.rms lo.o.ks as fo.llo.ws:
*
n
n+l
A~ (M) =R;A~ (M) =Tp (M) lJ\.~ (M) l ••• lAp (M) lAp
n+2
(M) ={O} lAp
(M) ={O}; •••
Wo.rking with o.rdinary co.tenso.rs we have previo.usly intro.duced the tensor pro.duct: If F and G are arbitrary co.tensors then the tenso.r pro.duct F €II G def ined by
is a tria
co.tenso.r o.f rank k+m. But if we restrict o.urselves to. skewsymmeco.tenso.rs this co.mpo.sitio.n is no. lo.nger relevant because F ~ G
is no.t necessarily skewsymmetric.(If yo.u interchange V. and 1.
say no.thing about what happens to. F(Vji ••• ;vk)G(Uji •.. therefo.re try to. mOdify this compo.sitio.n:
;rtm».
rt.J
yo.u can
We will
If Waj •••• a is a quantity with indices aj, ••• ,a k then we can co.nk struct a skewsymmetric quantity in the fo.llo.wing way (7.1)
w[ aj ••••• a 1 k
1 = k!
!(_l)TI w TI TI
(all ••• TI (a ) k
329 where we sum over all permutations
TI
II
of the indices al, ... ,ak,and
(-If
is the sign of the permutation, i.e. (-1)
TI
(-1) TI
= =
+1
if
11
is an even permutation
-1
if
11
is an odd permutation
For instance we get W[ab]
1 = 2![Wab
- wba]
and W[abc]
=
1 3! [w abc + wbca + wcab - wacb - wcba - wbac ]
The quantity w[al .... ~] is called the skewsymmetrization of wal ••• ak . ] is completely skewsymmetric, and that if ~ wal ....• ak is born skewsymmetric, then
Observe that w[a
I····
1 (This is, of course, the reason why we have included the factor k!') If we introduce the abbreviation
+l if (b l ... bk) is an even permutation of (al"
(7.2)
sgn[bl .•. b k al ... ak
]= { -1
.ak)
if (bl ... bk) is an odd permutation of(al ... ak)
o otherwise
we can write down the skewsymmetrization as an explicit summation W = 1 sgn[bl ... b k ] W [al" .ak ] k! al'"
...
coten-
, and G a
al a k m-form with components Gb " ' b ' then the tensor product is charateril m zed by the components:
Next we form the skewsymmetrization (k+m)! F G k!m! [al ..• a bl .•. b ] k m where the factor The
o
0
stat~ct~cal
(kk;m~! has been included to make life easier . • m.
factor
(k+m!) k~m!
removes »double
0
count~ng«:
II
330
Observe that sgn[Cj",Ckbdj"'bdm] and F are both skewsymmetric in (Cj",ck)' aj ... ak j •.. m Cj .. ,ck When we perform a permutation of (Cj",ck) we pick up two factors (_1)n. All permutations of (Cj ... Ck ) thus give the same contribution to·the sum. We need therefore only to consider a single representative, for instance the ordered set (Cj ... c ) k c~araterized by the property cj< ...
L
L
Cj<",
dj< .••
sgn[Cj"'Ckdj"'~]F aj ... ak bj ... bm
G Cj .•. ck dj ... ~
where we have obviously avoided «doublecounting«.
Observe, that
Each of the terms Frr(aj) ..• G"'rr(b ) transform
covariantly. Therefore
m (k+m)! F G k!m! [aj ... a k bj ... bml coins ide with the components of a skewsymmetric
cotensor of rank k+m.
This motivates the following definition:
Definition 2 Let F be a k-form with components F and Gam-form with components Gb a r.ak 1·· .b m then the wedge product FAG is the (k+m)-form with components: (k+m) !
kIm!
Let us give a Simple example: If A and Bare
covectors,
i.e~
I-forms, then their wedge produc AAB
has the components
From (7.3)
this you see that AAB
=
[A €I B - B €I Al
Exercise 7.2.1 Problem: (a) Consider a 2-form F characterized by the components Fab and a one-form B characterized by the components Bc' Show that FAB are characterized by the components
331
II
FabBc + FbcBa + FcaBb
(7.3)
(b) Let A,B,C be 1-forms. Show that (7.4)
The wedge product A has some
s~ple
algebraic properties,
Theorem 1 The wedge product is associative and distributive: (7.5)
(FAG)AH
=
FA(GAH)
(7.6)
FA(AG + llH)
(7.7)
If F is a k-form and G is a m-form, then FAG = (-lJ
=
AF A G + llFAH
Observe expeciaZZy that
coVectors anticommute!
Proof: Associativity follows from the simple observation that (FAG)AH has the components, (k+~+m) ! (kH) !m!
Distributivity also follows directly when you compare the components of both sides of the equation. We omit the proof which simply consist of keeping track of a lot of indices! Finally we consider the wedge products FAG and GAF: (FAG) (GAF)
(k+m) ! b b = -k-'-'- [F G b + •.. al ... a k I'" m .m. al ... a k b I··· m b l ... bmal .•. a
=
(k+m)! k!m!
=
(GAF)
k
[G
F b I ' " b m al ••• a k,+ ...
Consequently (FAG)
al.· .akb l ... b m
b l .•. bmal" .ak
Observe that it will require k'm transpositions to obtain al ... akbl ... bm from bl ••. bm~I ... ak (It will require k transpositions to move b
through al .•. a , it will then require another k transposim k tion to move b _ through al .•• a ect.) Using that GAF is skewsyrnmem l k tric we get (GAF)
b l .· .bmal" .ak
(-l)km(GAF)
al" .akb l ..• b m
332
II
Combining these two formulas we finally obtain
(-1) kIn (GAF)
aj ... akb j ... bm
showing that
FAG
=
U
(-l)kInGAF
The most surprising of these rules is no doubt the special "commutation" rule (7.7). There is an easy way to remember this rule. If F is a form of even degree, we call it an even form and similarly a form G of odd degree is called an odd form. The rule (7.7) can now be summarized in the following form: EVen forms aZways commute, odd forms anti-commute. So if you consider an expression like
and you want to interchange the order of the forms, then you only have to worry about the odd forms: Every time you interchange two odd forms it costs a sign. Now let us look at a 2-form F. Introducing a coordinate system we may decompose F along the basic cotensors dx i 0 dx j , i.e. F = Fijdx
i
®.dx
j
where Fij are the components of F with respect to this coordinate system cf. (6.68) . But F ij is skewsymmetric iIi i and j. Consequently we can rewrite Fij as Fij = ~(Fij-Fji). Thus we can rearrange the decomposition of F in the following way F
=
\(Fijdx
i
0 dx
j
j - Fjidxi 0 dx )
Interchanging the dummy indices i and in the last term we further get i j j i j F = \(Fijdx 0 dx - Fijdx 0 dx ) = \FijdxiAdx This may obviously be generalized to arbitrary k-forms:
Theorem 2 If F has the components P.
. we can expand F in the followirJ{f way
~j. "~k
(7.8)
F
=
1
P.
. d::;i IA ••• A d::;ik
k! ~j • • '~k
Exercise 7. 2. 2
MP.
Problem: Let us introduce coordinates on a manifold The coordinate system generates canonical frames (tl •... ,tnl and (dx1 ••.• ,dxll)
a, (dx
(a) Show that
333 ~ A ..
·N1x)b
b 1'"
k
II
[
al"'~l = Sgn b, •.• b k
is an even permutation of (l ..• k) 1S an odd permutation of (, ... k)
...
...
(b) Let (Ul ••. Un) be a n-tuple of ~angentve4to~s,.where Uj is characterized by the contravariant components A1 j , i.e. u·=eiA1 .• Show that
ExerlYise 7.2.3 Problem: Let T be a form of maximal rank n. Observe that T is characterized by the single component T j • • • n since all other components can be obtained from this using a permutation of the indices. Show that T can be decomposed as (7.10)
In the rest of this section we will try to develope a gearet.rical interpretation of differential forms, to help you to understand more intuitively the concept of a k-form. We start by considering the ordinary 3-dimensional Euclidian space R3. Let us investigate the k-forms associated with the tangentspace at
the origin: Canonical frame: 1
O-forms I-forms 2-forms 3-forms
dx ; dy dz ; dXAdy; dYAdz; dZAdx dXAdYAdz
Consider the basic two-form dXAdy. Remembering that dXAdy is defined as a linear map, Tp(M)xTp(M)~R,we want to compute its value on a pair of tangentvectors dXAdy(~,v). Observe that dXAdy has the components
(cf. exercise 7.2.2). Consequently we get
But this is easy to interprete. We use that ~,v span a parallellogram, which we project onto the x,y plane (See figure 124). But then we conclude: dzAdy(7., is the area of this projection. For this reason we say that dXAdy defines a unit of area in the x-y plane. Observe also that dXAdy generates an orientation in the x-y plane. The projected area
v)
dXAdy(~,v) is positive if and only if the projection of (U,v) defines
334
II
z
~::-+---+-----,. y
x Fig. 124 Exercise 7.2.4 Problem: Consider the basic one-form dx. Show that dx defines a unit of length along the x-axis in the sence that dx(~) is the length of the projection of ~ onto the x-axis.
Exercise 7.2.5 Problem: (a) Show that dxAdYlldz defines a volume-form in R3 in the senc:e that dxA~lIdz(Ul;U2;U3) is the volume of the parallelepiped spanned by Ul,U2 and U3·
dxAdYlldz defines an orientation in R3 in the sense that (Ul,U2,Ull is positively oriented if and only if dxAdYlldz(UI·,U2;Ul) is
(b) Show that
positive. (Hint: Decompose the triple (Ul,U2,U3) as
U. = ~.Ai. J
1
J
and show that i Det(A . ) J
Compare with the discussion in section 1.1) .
Obvio~sly
the above analysis depends on very special properties of
the Euclidian space R3. It id however possible to convert it into a purely geometrical form suitable for generalizations to arbitrary Euclidian manifolds. Consider once more the basic two-form dXlldy. The coordinate-functions x and y defines two stratifications of the Euclidian space: ••. ,X
... ,y
-2, x -2, Y
-1, x -1, Y
0, x 0, Y
1, x 1, Y
2, ... 2, •.•
Together these two stratifications form what is generally known as a "honey-comb structure" (see figure 125) .
II
335
y
x
Fig. 125
If we return to a parallellograrn spanned by two tangentvectors ~
and ~, then it is clear that the projected area is simply equal to the number of tubes intersected by the parallellogram. We therefore conclude: d::;Ady (~;
v)
is the nwnber of x-y-tubes, which is intersected by the paraUeUogram
spanned by the ~ and
Exercise
v.
7.2.6
Problem: (al Consider the basic one-form dX. The coordinate function x defines a stratification: ... ,x '" -1, x = 0, x = 1,.. .Show that dX(;;) is equal ~o the number of hyperplanes x = k that are intersected by the vector u·
(bl Consider the basic three-form dXAdYAdz. The and Z generate three strali!i~ations, which tu:e. Show that dXAdYAdz~u;v;w) is equal-+t~ ta~nQd in the parallelep~ped spanned by u,v
coordinate fUnctions, x,y together form a cell-structhe Q:umber of cells conand w. _
We can nOw transfer the above machinery to an arbitrary manifold. Consider a two-dimensional manifold M with local coordinates (X I ,X 2 ) . Let Po be a point on M and consider the basic two-form dX l Adx 2 at Po' The two coordinate functions Xl and X2 generate two stratifications on M: xl= ••• ,-l,O,l, •.• and x 2 = ••. ,-1,O,1, ••• These two stratifications divide the surface into a great number of cells (see figure 12b).
x2 (b l ,b2 \
1\ ",.
Fig. 126
(a l
II
336
Consider two tangentvectors
u and v at
Po. This time
u and v span
a parallellogram which lies outside the manifold M. We must therefore use a trick. In coordinate space (a l ,a 2 ) and (b l ,b 2 ) span an ordinary parallellogram. The number
dx l "dx 2 (i!;v) is therefore equal to the number of cells contained in this parallellogram. But since this parallellogram is a subset of the coordinate domain U, we may transfer it to M. The image will be referred to as the
"parallellogram" swept out by
u and v.
Therefore we conclude in
the usual way: dXIAdX2(~;V) denotes the number of cells contained in the paraZZeZlogram swept
out by
7.3
u and v.
THE EXTERIOR DERIVATIVE That was a long algebraic digression. Let us return to our main pro-
blem. We want to construct a differential operator which converts a cotensor field of degree k into a
cotensor field of degree k+l, i.e.
something like T
al ... a k
(xl; ... ;xn) .... a T (XI; ... ;X k ) 11 al· .. a k
We now restrict ournelves to skewsymmetric
/
cotensor fields. They
playa key role in differential geometry and are called differential
fonns.
If it is clear from the context that we are working with a skew-
symmetric
cotensor field, then a differential form of rank k will be
referred to simply as a k-form. The set of all smoth differential forms of rank k will be denoted
A~(M). Observe that Ak(M) is an infinite-di-
mensional vectorspace for D
~.
Then we have previously introduced the differential ope-
rator d which converts the scalar field degree one, i.e. a
covector field
d~.
~
into a differential form of
If we introduce coordinates then
~ is represented by an ordinary Euclidean function ~(XI; .•. iXn) and d~ is characterized by the components all~(xl; ... ;xn). It is this differential operator d we want to extend! Consider a differential form of degree one, i.e. a co-vector field A characterized by the components
337
II
A (Xl, ... ,xn) o. Differentiation this we get the quantity
__ d_ A (xl, ... ,xn) -_ d~Ao.(X I , ... ,xn) dX~ o. but this is of little interest because it is not skewsYmmetric in and o.. Therefore we antisyrnmetrize it and get
~
This is skewsyrnmetric and if we can show that it transforms covariantly,then we have succeeded. Introducing another coordinate system we get the new components 23 A (2)[ Pl2jCt 1
But A (y) (2)o.
2Ig)[\l{~o.l
:)
a~"
:\
'1
(2l a
\vt.
dY
(V! / ~ ,
o.
. ,,,p)
ax S = AS(X)-. Inserting this we get (1) ayo. axS ax S a [A (x)-l_a_[A (x) 1 ayo. ayo. (1) s ayP Il)S ayP aA ax Y axS a&s ax Y axS a"x S ....!ll6. ----- + A - lllS ay~ ayo. ax Y ayll ayo. ax Y ay a ayP
a"x S A -----
lllSayaay~
Using the fact that partial derivatives commute, we observe that the spoiling terms cancel each other! Finally we get ax Y axS ax Y axS =aA ---aA-u)YU)Sax p ayo. lllYl1tl ayo. ayP Here we interchange the dummy indices Sand Y in the last term and get 2a
A
l2j[ \l{zjlll
Y
S
= [a A - a A 1 ~ ~ mYtnS tuS(l)Y axP ayo.
Y S 2a A ~~ (u[Y(llSl ayP ayo.
So everything works! The expression 2a[~ASl coins ides with the components of a differential form of degree 2 which we will denote dA. This may immediately be generalized:
Definition J If F is a differential form of degree k characterized by the components F-40 •• ik(:x;J. then dF is the differential form of degree k+l characterized by the components (k+l)! k!
a
F
[p i
l •••
ikl
The differential operator d: i<-(M)-+N<+l (M) is called the e:cterior derivative.
338
You may think of d as a kind of 1-form characterized by the components fact a~ transforms covariantly according to the chain rule:
a
(2)~
V
II a~.
In
\}
_a_=_a_ ~=a ~ ay~ ax\} ayW (l\\}ayll
The construction of the exterior derivative is then formally equivalent to the «wedge-product« : dAF. This also explains the statistical factor. (Compare with definition 2).
Exercise 7.3.1 Problem: Let F be a 2-form characterized by the components FaS' Show that dF is characterized by the components (T .11)
The ",o,r-.;; ·'5<"'" .:, .. :" .. h,-,-V8 has several important properties. It is a linear operator, i.e.
d(F+G) = dF + dG
(7.12) as you can ing one:
~sily
check. But the most important property is the follow-
Theorem 3
(Poincares lemma)
(7.13)
dd = 0
i.e. it automatically vanishes if you apply it twice. Proof: Consider first a scalar field ~ represented by the Euclidian function ~ (Xl, ••• ,x n ). We then get (d~)ll = a~~ , which :iInplies that (d 2~)
~v
=a \} a
~
~- a
~
a\} ~ = 0
This shows the mechanism in the cancellations. We then consider an arbetrary differential form F of rank k. Let F have the components F. . We then get for dF 11"
.1
k
and similarly for d 2 F (k+2) : (k+l)
!
339
II
But this sum vanishes, because if we write out all the terms they occur in pairs d d of. a " J
I · ••
and
. Jk
As Sajl ... jk is generated from aSjl ... jk by applying one transposition they will have opposite signs, i.e. we may collect the two terms into
and this vanishes automatically.
0
Before we derive more rules we introduce some more notations. Exercise 7.3.2 Problem: Let F be a k-form characterized by the components F
(7.14)
... an and Gam-form al characterized by the components Gbl .•. b ' Show that we can decompose FAG as m 1 a ~ b b b dx I A... A da Adx I A... Adx m k' , F G b .m. al"'~ I .. • m (Hint: Use the distributivity of A)
Exercise 7.3. J which is not necessarily skew-symmetric. Show Problem: Consider a quantity w al'''~ that a ~ a ~ w[ _ 4x I A... Adx =w dx I A... Adx (7.15 ) al"'''j(
al ... ~
If F is a differential form of degree k we can decompose it along the basic k-forms (7.8). If we form the exterior derivative dF, we know that it is characterized by the components
Therefore we can decompose dF in the following way i
F
1
=
i
(k+l)!
1 -k.'
(k+l) ! . j dxll"dx 1" ... Adx ~ d[ 11 F.l.1·· ·l.k
i i d[ F. . jdxllAdx I A ••• Adx \l l.1·· ·l.k
i
k
k
But according to exercise 7.3.3 we are allowed to forget the skewsymmetrization! In this way we obtain the formula
340
(7.16)
II
dF
We are now in a position to prove a generalizatbon of the familiar rule for differentiating a product (Leibniz' rule):
~ (f.g) dx
=
df .g + f. ~ dx dx
Theorem 4 (Leibniz' rule for differential forms) If F is a k-form and G is an J,-form, then d(FAGj = dFAG + (-l)k FAdG
(7.17)
Proof: If you canbine (7.14') and (7.16) you .i.nrnediately obtain 1 II i ik j j J, d(FAG)-k' n, 3 (F. . G. .)dx Adx I A... Adx 'idx I A...Adx • N • II 11 . . . 1k J I ..• J J,
According to Leibniz' rule for ordinary functions this can be rearranged as
II i ik j I J, . (3 G. .)dx Adx l A... Adx Adx A..• Adx II J I· •. J J,
1
+ -k",F. • N.
11·· .1
k
i
In the last term we ~d like to nove dx ll through dx I A•. . Ad$. anticamn.rte, and therefore we IIUlSt pay with a factor (_l)k:
i
k
•
But covectors
We m~ think of d as an odd form,(cf. the discUssion after def. 3). In Leibniz' rule we have two terms and
In the last term we have interchanged d and F. This costs a sign if F is an odd form.
341
II
Observe that if ~ is a differential form of degree 0 i.e. a scalar field, then ~AG is just a fancy way of writing ~(xl, ... ,xn)G. In that case Leibniz' rule degenerates to (7.18) We can now recapture (7.16) in the following way. Consider a k-form 1
F
i
-k! F. i (x)dx lA ... Adx l.l ... k
ik
If we keep the coordinate system fixed, we may treat F. . (x) as a l.l·· .l.k scalar field. Applying the exterior derivative we therefore obtain l. i,~'.· ,
:d',
q.-.
i
k
i i (x)]Adx lA ... Adx k
1 i i k + k! F. . (x)Ad[dx lA ... Adx ] l.l·· .l.k i
i
Using Leibniz' rule once more we see that d[dx lA ... Adx k ] automatically vanishes since it only involves double derivatives (theorem 3). We are therefore left with the first term. If we use that
we finally recapture (7.16)~ We conclude this section with a definition of two important types of differential forms:
Definition 4 (aJ A differential form F is alosed if its exterior derivative vanishes, i.e. dF
=0
.
(bJ A differential form G is exact if it is the exterior derivative of another differential form F, i.e.
G = dF • Observe that an exact form G is automatically closed, since G :implies that
2
dG .. d F
=
0
but the converse need not be true! See
dF
(Examples will be given later on.
,e.g., 5eC"tWn 7.8 TNhcre the case of the IlDnopole field is diocussc1.)
Exercise 7.3.4 Problem: Let F be a closed form and G an exact form. Show that FAG is exact.
342 A
II
closely related type of differential form is given by;
Definition 5 A k-form W is (JaZZed sirrrpZe if there exist k smooth scaZarfieZciB
(7.19)
w = d
The basic forms generated by a coordinate system and the simple forms are intimately connected. If
yk
n
n
X
is an admissible exchange of coordinates and in the new coordinates, (yl, .•. ,yn), we can decompose Was W = dylA ... Adyk • So W is a basic form generated by the new coordinates. Consider a simple form d
A simpZe form is cZosed and exact.
Proof: 2
As d =
0 it follows trivially that dtPIA ... AdtPk is closed.
The exactness follows from the fOlllUlla
dcPIA ... Ad
0
Worked exercise 7.3.5 Problem: Let flo •.•
,fit: R.... R be smooth real functions. Show that
is a simple form.
343
7.4
II
THE VOLUME FORM n Suppose the manifold M is also equipped with a metric ~. Then we
can speak of areas, volumes etc. in the tangent-space Tp(M). Let us make this more precise. Consider an n-tuple (Ul"'.'~n) which generates a "parallelepiped". We want to find out how to compute the n-dimensional volume of this "parallelepiped". If we introduce coordinates n (xl, ... ,x ) it would be tempting to put ~
-+?
1
Vol [Uli ••• jU n ] = dx 1I ••• lIdx
n
-+
:-1'
[uli •• .)'un ]
by analogy to the lower dimensional Euclidian cases. (See exercise 7.2.5) But here you $hou:d be careful since the canonical frame generated by (xl, .•. ;xn ) needs not be orthonormal. The tangent space Tp(M) is simply isomorphic to the standard Euclidian
n
~pace
-
R . Recall the standard definitions of a volume in the n-
dimensional Euclidian space. Let (e:, ..• ,e~) be an orthonormal frame and consider an arbitrary n-tuple of vectors (~l""'~n)' They span a "parallelepiped" with the volume
(7.21)
........
V01(VI; ... ivn)
=
-
Det[B]
i where the matrix elements B , are the coordinates of ~, J J
~,
J
= e~Bi, ].
J
-+ -+ • ""'0 -+0) Furthermore (VI""'V n ) generates the same orientat].on as (el, ••• ,en
if and only if Det[B] is positive. Now let us introduce coordinates around the point P. Then the canonical frame (el, ... ,e ) generated from the coordinates needs not to n be an orthonormal frame, but we can decompose it on the orthonormal frame (e~, •.. ,e~) : (7.22)
....
ej
=
....0
i
eiE j
Let furthermore (u""',;n) be an n-tuple of tangent vectors, where
~j is characterized by the
contravariant components Ai,. From exer)
cise 7.2.2 we know that (7.9)
DetCA)
We now get -+
uj
i _""'0 k i eiA j - ekE i A j
-+
From (7.21) it then follows that (7.23)
It remains to interprete Det[E]. Consider the coordinate exchange
344
II
Observe that the orthonormal frame (elo, ... ,e~)
is the cononical fra-
me generated from the (yl, ... ,yn)-coordinates.
(This follows directe-
ly from (7.22) and
(6.10». The metric coefficients gij corresponding
to the (yl, •.. ,yn)-coordinates therefore reduce to the Kronecker-delte 0ij' The transformation formula (6.21) for the metric coefficients may be rearranged as
i.e.
[Detd~)12
g = DetLGl
i.e.
Inserting this into (7.23) we have thus determined the volume form in the tangent space Tp(M):
(7.24) later on we shall recover this formula from another point of view. The volumeform also
contro~the
orientation of the tangentspace. This
follows immediately from (7.9) which shows US that (~I, ••• ,Un) generates the same orientation as the canonical frame (~l, ... ,en) if and only if dx 1 A ••• Adx n (ir 1 i • •• ; irn ) is positive. In the preceding discussion we have been focusing upon a single tangent space. We will now try to extend the discussion of the local aspects of the volume form to its global aspects. Our final aim is to construct a global differential form
t
duces to the local volume form given by for this we must first discuss Let M be a manifold and Po
~he
which at every point
p
re-
(7.24), but as a preparation
orientability of manifolds.
a fixed point in M. Consider the set of
all coordinate systems surrounding Po. Each coordinate system generates an orientation on the tangent space Tpo(M). Now consider two sets of cOordinates: From
(x 1 , ••• , xn)
and (y 1 , ••• , yn) •
(6.12) we know that
....e.
(2jJ
Therefore we conclude that (x1, .•. ,x n ) and (yl, ... ,yn) generate the same orientation on the tangent space Tpo (M) provided the Jacobiant
345
[dX~l
Det
II
is positive.
dyJ
Okay! Let us choose a positive orientation in our tangent space Tpo(M). This is done by picking up a specific coordinate system (
is an arbitrary coordinate system, then it generates a positive
orientation provided
is positive and it generates a negative orientation provided its Jaco-
biant is negative! In what follows we will always assume that we have chosen a positive
orientatio~
in our tangent space Tpo(M).
This was the local aspect,
(i.e. we have been focusing at what hap-
pens in a neighbourhood of a point Po). Now we turn to the global aspect of orientation. First we consider a simple manifold, i.e. a manifold which can be covered by a single coordinate system
(~o,Uo).
It
generates an orientation in each of the tangent spaces on M, and we ~eclare
these orientations to be the positive orientations!
Then we consider an arbitrary manifold. Suppose we have two coordinate systems
(~I,UI)
and
(~2,U2)
which overlap. If we are going to let
them generate the same orientation in the tangent spaces we must be careful in the overlapping region
nI2=~1 (UI)n~2(U2)
.We must demand that
they generate the same orientation in the overlapping region n 12 , i.e. that
Det[dx
is positive throughout the overlapping region. This
motivates the following definition:
Definition 6 A manifold M is said to be orientable if we aan ahoose an atla~ (~i>~i)i£I.suah that any two aoordinate systems in our atlas generate the same o~entat~on> ~.e.
det(~) dys
is positive throughout all overlapping regions.
When a manifold is orientable we can choose an atlas with the above property. This atlas will then generate a specific orientation in each of the tangent spaces on M, which we arbitrarily declare to be positive! Let us end this digression on orientability with an example of a manifold which is
not
orientable: The MObius strip (see fig. 127 ).
346
Fig. 127a
CONSTRUCTION OF THE MOBIUSSTRIP
Fig. 127b
NON-ORIENTABILITY OF THE MOBIUS STRIP
II
Exercise 7.4.1 Problem: Let MTI be a differentiable manifold. Let us assume the existence of an nform 0 which is nowhere vanishing. If we introduce coordinates (xl, ... ,xn) we can decompose 0 as 0 = n1 ... n(x)dx1A ... Adxn . That 0 is nowhere vanishing implies that I"lr ... n(x) is nowhere zero. (a) Let (xl, ... ,~) denote an arbitrary set of coordinates. Show that the sign of O(~l; ... ;~n) is constant throughout the range of the coordinate system. (b) Show that 0 generates an orientation of M by defining a coordinate system (xl, ... ,xn) to be positively oriented if and only if O(el; ... ;en ) is positive. (c) Show that an n-tuple (~l, ... '~n) is positively oriented with respect to the' orientation generated by Q if and only if O(~l;' ... ;~n) is positive.
Exercise 7.4.2 Problem: Let Mn be an n-dimensional manifold in Rn+1. i.e. M is a smooth hY.Persurface. Let us assume the existence of a smooth normal vectorfield rt(x) to M. i.e. to each point x in M we associate a normal vector ti(x) to the tangentspace. Show that M is orientable. n 1 (Hint: Let (x 1 ••.. ,x + ) denote the extrinsic coordinates in ~+1. Let (~l •.•.• ~n) be an arbitrary n-tuple tangent to M. Define a smooth n-form o on M in the following way ...... ... o(...Vl; ..• ;V...n ) = dx 1A..• Adxn+1 (n;vl; ..• ;V ) n and apply exercise 7.4.1) Consider the sphere and the Moebiusstrip in R3 (see fig. 128.129)! You can see with your own eyes that the sphere allows a normal field. while the MOebiusstrip does not!
347 The sphere
II The Moebiusstrip
Fig. 128 Fig. 129 As a preparation for the construction of a global volume form we are going to investigate two special quantities which are not true geometrical objects, Le. they can only be defined in connection with a coordinate system on the manifold. The first quantity is intimately connected with the metriC g on our manifold. To be specific let us assume that g is a Minkowski metric. Consider a point Po in M and a coordinate system covering Po. We can then characterize g in terms of its components ga~
=
1 gaS(xo, ••• ,xn) o
and we may compute the determinant g = Det[ga~] This determinant is necessarily
negative, because g is a Minkowski
metric. It is this determinant we are going to examine. Obviously g(xo) is not a scalar, i.e. it does not transform according to the rule q(yo) (~)
=
q(xo) when we exchange the coordinate system. To find the
trl
transformation rule we use that
=g
(6.21)
ga~
transforms
ax Y axe
(l)ye aya ay~
We can rearrange this as a matrix equation: (6.23) If we evaluate the determinant, we find
Le. (7.25)
g (Yo) (21
covariantly:
348
II
This transformation property makes it natural to look at V-g(x), and we conclude: Lemma 2 On
a manifold with a Minkowski metric y.:g7XJ transforms according to the pule
(7.26 J
v-g(yJ (21
(On
a ().
= I Det(-7,J I y.:g7XJ ay (1) Fg by
a manifold with an Euclidian metric you just replace
Wi) .
Keeping this in mind we now introduce the next quantity, the LeviCivita symbol: +l ~f (al, ... ,a n ) ~s an even permuta~ion of (l, ... ,n) ~f
-1
(7.27)
{
o
(al, ... ,ad
~s
an odd
permutat~on
of (l, .•. ,n)
otherwise
Again we select a point Po and an arbitrary coordinate system to this pOint. Then we attach the Levi-Civita-symbol to this point Po. This does not corresponds to the components of a
cotensor, since by
definition it transforms according to the rule: (7.28)
S
al •.. a n
(Yo)
=
S
al ..• a n
(xo)
However we may use these two quantities,~ and sal ... an,to construct a differential form. Let us choose specific coordinates (xl, •.. ,x n ), and consider the n-form T characterized by the components
with respect to this particular coordinate system. Let us try to compute the components of T with respect to some other coordinates (yl, ... ,yn). oAs T get (*)
~b
••• b 1
n
(Yo)
T
ilIa ••• a 1
n
a
S
ax 1 a ... a - bn I ay 1
Here we should try to study
a
l ••
ax ay
.an
a
transforms a
1
ax n
(xo)---". ••• - b Dl
ay
a
ax n -bay n
n
covariantly, we
II
349
a ax n
-bay n
a little closer. If (b 1 ,
•••
,b n )
(1, ... ,n)
then
by the very definition of a determinant! In the general case we conclude:
a ax 1 sal .•• a --b-'" nay 1
a ax n
--s-
Det
ay n
CI.
Deti~lif (bl, •.• ,b n ) is an even permutation of (l, ... ,n) ayS
CI.
-Deti~lif (bl, ... ,b n ) is an odd permutation of (l, ... ,n) aye
o
otherwise
Consequently we obtain the following formula a
(7.29)
ax
I
ay
1
-b-
a ax n
CI.
--s-
Det (ax 0) •• b n ay"
S
b1
ay n
•
Inserting this in (*) we get CI.
b (Yo) = I-g(xo)
Th
n
(2),",1'"
S
(!)
b
l ...
bn
Det(ax a ) ay"
But using (7.26) we finally obtain
axCI.
Det(¥) l1;b l
•••
(yo)
b n
IDet(~)1 ay
Sgn (Det[ axCl.]) J-q(yo) aye (2')
I-g (Yo) (21
S
bl .•• b n
E:b
1'"
b
n
350
Consequently we see that up to a sign form as lTal ,I)
•••
II ~bl
... bn has exactly the same
an! This motivates the introduction of the Levi-Civita
form:
Definition 7 Let M be an orientable manifold of dimension n. Then the Levi-Civita form associated with the chosen orientation is the n-foTIn £ characterized by the components V-g(x)
[c]
a1 .. ·a n
with respect to a positively oriented coordinate system
E al ..• a
={
n -V-g(x)
E al· .. a
with respect to a negatively oriented coordinate system n
(If the metric is Euclidian we replace V-g(x) by Vg(x) ).
Observe that the Levi-Civita form £ is only defined after we have fixed the orientation of our manifold. If we change this orientation then the Levi-Civita tensor is replaced by the opposite tensor, i.e. £
~
(:
1
'=
-t
pseudo-tensor.
For this reason physicists refer to it as a
Exeroise
7.4.3
Problem: Consider the Levi-Civita symbol E al· .. ~
We define
( 1) Show that
(7.30)
a1···an E E al ... a n
n!
(2) Consider the cases n=2 and n=3. Deduce the rules
(3) Show that the contra-variant components of the Levi-Civita form
£
are given by
(positively oriented coordinate system)
+
1
~
e:
aj ••• a n
(negatively oriented coordinate system)
351
II
(4) Let (xl, ... ,xn) be positively oriented coordinates. Show that the Levi-Civita form can be decomposed as
(7.34) n Let (x1, ••• ,x ) denote positively oriented coordinates on a Riemannian manifold Mn. From (7.34) we learn that the Levi-Civita form £ can be decomposed as £
= Vg dx11l •• • lIdx
n
If we compare this with (7.24) we see that £ is nothing but the volwne
form.
Observe that £ is globally well defined, so that we,on an orien-
table manifold,can piece together all the volume forms on the different tangent spaces to a globally well defined n-form. On a non-orientable manifold this is no longer possible. Observe also that £ not only generates the volume, but also the
orientation of the tangent spaces. If (~l"" '~n) is an arbitrary n-tupIe then £(~l;"';~ ) is the volume of the parallelepiped spanned by n
(~l""'~n)' and it is positive if and only if (~l, ... ,Vn) is positively oriented. The Levi-Civita form can also be used to generalize the cross product from the ordinary Euclidian space R3. Let Mn be a Riemannian ma-+
-+
nifold, and consider a set of (n-l) tangent vectors, Ul,""U _ . n l If we ~ontractthis set with the Lev1-Civ1ta form £ we obtain a I-form (7.35) Since we work on a manifold with a metric, this I-form is equivalent to a tangent vector
fi, and it is this tangent vector fi which genera-
lizes the usual cross product (Compare the discussion in section 1.1):
Lemma :> The vector ; is characterized by the foZlowing three properties: -+ ... -+ (1) n is orthogonal to each of the tangent vectors u 1 , .•. ,un - 1 . (2) The length ~f (3)
n
is-+equal to the "area" of the "paralleLZogram" spanned by u , ... ,u _ , n 1-+ 1-+ -+ The n-tuple (n,u 1 , ... ,u _ 1 ) is positively oriented. n
Proof: ... First we deduce a useful formula. Let v be an arbitrary tangent vector. Then bl b r: un-Iva -+ -+ ...) (7.36) !I(n;~) = nava .gE b bU... = £(ViUl;···;un_l a 1'" n-l
II
352
The rest now follows easily:
= £(~.;~ i •.• ;~ ) which vanishes automatically since t (1) g(;i~.l 1. 1. 1 n-l is skew symmetric. (2) Consider the normalized vector ti/litiU . It is a unit vector orthogonal to ~ , •.. ,~ . Therefore the volume of the parallelepiped 1 -+ n:'l -+ -+ spanned by (n IUnll),u , ••• ,u reduces to the area of the paral4.1;- 1 lellogram spanned by u1, .•. ,U _ ' Consequently we get that n 1 -+- ..• iUn_l] -+-+- ... ;u -+- - ) = g(n; -+- -+Area[uli .:.( -+nl 11-+-11 n ;u1; nl II-+-I nl) n 1
r +. IInll
-+ -+
g(ni n )
=
-+
IInll
(3) We immediately get that £(ti;~ i ••• ;~ 1
n-l
)
g(ti,ti)
-+
Lemma 3 motivates that we define n -+
-+
Ul, ..• ,U _ n l -+
> 0
0
to be the cross product of
and we write it in the usual manner -+
-+
n = Ul x • ,.xu _ n l
Exercise
7.4.4
Problem: Consider the Euclidian space ~+1. (a) Let (~'~l'" .,~ ) be an arbitrary (n+1)-tuple. Show that the volume of the parallelepiped gpanned by (~';l""'; ) is given by the familiar formula n
(bl Let Nfl be an orientable n-diemensional manifold in Rn+1. At each point in M we select a positive oriented orthonormal frame (til""'~ ) in the corresponding tangent space. Show that the unit normal vector ¥ield -+
-+
-+
n=u1x ...x U n
is a globally smooth normal vectorfield on M (Compare with exercise 7.4.2)
7.5
THE DUAL MAP As another important application of the Levi-Civita form we will
use it to construct the dual map *, which allows us to identify forms of different rank. This gives a greater flexibility in the manipulations of various physical quantities.
353
II
Let F be a k-form with the components considering the
We start by components
Then we contract this tensor with the Levi-Civita form and obtain !,/g(x) e: b b Fbl ••. bk [PositivelY orientated] k. al···an-k 1'" k coordinate system! 1 is included to make life easier! where k! As e:al"'~ is skewsymmetric we have thus produced a form of deThis form is called the dual form and it is denoted *F gree n-k * or sometimes F
Definition 8 Let
*)
be a k-form on an orientable manifold
F
Then the dual form nents
al'"
a
with a metric
g.
is the (n-k)-form characterized by the compo-
*F
~!/g(X) e:al"'~_k (7.38)(*F)
Mn
bl •.. b k Fbl •.• bk
(~~~~i metric)
n-k
~!/-g(x) e:
al···an _ k b1···b k F
bl' .. bk (Minkowski metric)
with respect to a positively oriented coordinate system. Exercise 7.5.1 Problem: (a)
Let ~ be a scalarfield on a Riemannian manifold and let be positive oriented coordinates. Show that
(xl, ...• xn)
*~ = ~;gdxlA ... Adxn
(7.39) (b )
Show that E
(Euclidean metric) (Minkowski metric)
= *1
Exercise 7.5.2 Problem: Show that the contra variant components of *F are obtained by contracting the components of F with the contra· variant components of 1:. i.e. _1_ e: a l · , • ~-k b l ,·· bk F
(7.41)
(*F)
al···a -k
k!/g
n
{
- 1
k!/-g
*)
(Euclidean metric)
b l .. ,bk
e:
a 1 ·, .an - k b l ·· ,bk
F
b " .b k l
(Minkowski metri c )
This definition differs slightly from the one generally adopted by mathematicians. See for instance Goldberg [1962] or de Rham [1955].
354
II
In this way we have constructed a map from
hk(M)
to
An-k(M)
It is called the dual map or the Hodge-duality. By construction it is obviously linear
*(AF) = A(*F)
*(F+G) = *F + *G
(7.42)
But a more interesting property is the following one: Theorem 5
**T **T
(7.43)
*
So up to a sign
* :
=
(_I)k(n-k)T
(Euclidean metric)
T
(Minkowski metric)
= -(-l)k (n-k)
is it's own inverse. Especially the dual map,
Ak(M) ~ An-k(M), is an isomorphism.
Proof: For simplicity we only consider the case of a Minkowski metric. Let F be an arbitrary k-form. Then *F has the components
1. ,I-g k!
E:
a l ·· '~-k b l ·· .bk
Fbl' .. b k
We want to compute the components of **F The first thing we must do is raising the indices of *F, i.e. determine the contravariant components of *F. According to exercise 7.5.2 they are given by the formula
Then we shall contract the Levi-Civita form with factor! )
(6)
(**F)
c l ·· .c k
= -(n-k)! -I ; : g E: c
*F (remembering the statistical (*F)al ·· .an- k
l ·· .ck a l ·· .an- k
1
k!(n-k)!
E:
cl,,·c k al·"a n _ k
E:
a l ·• .an - k b l ·· .bk
F
bl".b k
The rest of the proof is a combinatorial argument! First we observe that
But
is skew-symmetric and therefore
Sgn [bcl ... bc k ] F = k! F 1,,· k bl" .bk c l " .c k and if we insert that into (0) we are through.
0
From theorem 3 and 5 you learn the following important rule whenever you deal with the operators in the exterior algebra:
355
II
"Be wise - appZy them twice" We shall evaluate the sign associated with the double map
**
so
often that it pays to summarize them in the following table
(7.44)
Euclidean metric k even
**F
n even
k odd -F
F
n odd
(7.45)
F
F
Observe that the dual form,
**F
Minkowski metric k even
k odd
n even
-F
F
n odd
-F
-F
*F ,involves the Levi-Civita form. Thus
it depends on the orientation of our manifold. If we exchange the orientation with the opposite one, then the dual form is replaced by the opposite form
F -F
*F ~ '= For this reason physicists also call the dual form
*F
a pseudo-tensor.
&eT'Cise 7.5.3
Introduction: Let MP be an orientable Riemannian manifold. A differential form F is called seZf dual i f it satisfies *F = F and anti self dual if Problem:
Show that self dual and anti -self dual forms can only exist in Riemannian spaces of dimension 4, 8, 12, 16, .,.
Exercise 7.5.4
Let MP be an orientable Riemannian manifol~ and A a+l-form+on M. Then A is equivalent to a tangent vector a. Let (ul""'~-l) be an arbitrary in-l)-tuPle. Show that the volume of the parallelepiped spanned by (iIl , ... '~n-l,a) is given by -+ -+ -+ -+ -+ -+ -+ -+ Vol[ul;···;un_l;al = *A(ul;···;un _ l ) = ai u l x ..• x un - l ) This gives a geometric characterization of the dual form *A'
Problem:
Exercise 7.5.5 Problem:
a)
Show that al
*(dx b)
A ••• Adx
ak
_ Itg ) -
(n-k)!
E..
ll" .In-k
h .. ·jk
g
jlal jkak i l i n- k ... g dx A . . • Adx
Consider spherical coordinates in Euclidean space dual map is given by *dr
= r 2 SinedeA
*de = Sine
R3.
*
Show that the
356
II
The scalar product between two I-forms is given by
gijA.B. 1.
J
Clearly this is a special case of a contraction between a k-form and am-form
s
where S
jl·· ·jm
The result obviously is a differential form of degree would be nice to be able to express such
(k-m)
•
It
contractions in a coordinate
free manner, and as we shall see this is possible using the dual map. Thus the following pattern emerges: When you have a metric on an orientable mani-
fold, you can use this metric to construct a dual map
*
but once
you have constructed the dual map, you can forget about the metric: It has been "coded" into the dual map. The first thing we will construct is the scalar product between 1forms. Let therefore
A
and
*AAB
B
be two I-forms. Then
*A
is an (n-l)-form, and
is an n-form. But this can be dualized to give a sca-
lar * (*AABl • This scalar is a coordinate invariant formed out of Ai Bi
and therefore it must be proportional to
,
other coordinate invariants you can make out of
;g
Ai
and
Bi '
(since there are no gij'
Ai'
Bi
and
El.'
. !) Okay, that was a rather abstract argument. Let us work 1·· ·l.n out the coordinate expression for *(*AABl in, e.g., the Euclidean case C
=
*(*AABl
i.e.
*C
*AAB
Here the right hand side is characterized by the component 1 (n-ll![(*Al 12 ... (n-ll(Bl n + ... - .•• 1
(*AAB) 12 •.. n
(*Al12 ... n_l(B)n+(-lln-l(*Al2 ... n(Bll + .•.
whereas the left hand side is 'characterized by the component (*Cl1. .. n From this
= ;g
E1. .. n C
= ;g
C
we conclude
*(*AAB) which justifies our claim. The expression symmetric in
A
and
the dualization of has been raised,
B, A
Ai
but that is
*(*AABl
AiB.1.
is not manifestly
neither!
Observe that
is reflected in the fact that the index of ~ Ai We have thus shown
A
357
II
Theorem 6 L«t
A and
( 7. 47) ( 7.48)
B be l-forms, then * (*AI\B) * (*BI\Al * (*AI\B) * (*BI\Al
AiB.
(Euclidean metria)
l.
-AIlB
Il
(Minkowski metria)
This theorem has an important consequence. It tells you how to reconstruct the metria from the dual map! Sometimes it is preferable to omit the last dualization, which in both cases produces the formula
*AI\B = *BI\A = AiB.£ l.
(7.49)
The above formula suggests a generalization of scalar products to arbitrary differential forms. Before we proceed we need a technical result: Worked exeraise 7. 5. 6 Problem: Let T,U be k-forms. Show that 1 il *TI\U=,T k.
(7.50)
From exercise
7.5.6
· .. i k U..
l.l ... l.k
we see that
£
*TI\U
is symmetric in
T and
U
and that it represents the coordinate invariant
The statistical factor is very reasonable, since it removes "double counting" . Let us introduce the following abbreviation for the scalar-product of two k-forms: (7.51) Then we have succeeded in generalizing theorem 6 to arbitrary k-forms: Theorem 7 Let
T,U
be k-forms, then
(T I u)
(7.52 )
1
-- T
i l " .ik
k!
1 Ill" ' T k.
(7. 53)
U
. Ilk
il .•. i k
(Eualidean metria)
U (Minkowski metria) )Jl,,·)Jk
Exeraise 7.5.7 Problem: Show that a)
(dxalm(»
= gab
(dxa,wPldxcNlxd ) = gacJX1 - gad,fc Remark: It can be shown in general that [] al ~ bl ~ 1 iljl ~jk !-".~ (7.54) (dx 1\ ... Nlx Idx 1\."l\dx ) = k!'1 ".g Sgn l.l'''~ b)
358
E:z:eT'Cise 7.5.8 Problem: Let T rank. Show that
p,f.
be a n-form on
Consider a single point
P
i.e.
T
is a differential form of maximal
on our manifold. To this point we have at-
A~(M)
tached the finite dimensional vector space
consisting of all
k-forms. By construction the map
(T,U)
(TI u) A~(M)
is a symmetric bilinear map on the vector space
is a good chance that it actually defines a metric on
Thus there k
Ap(M) . To check this we--Ulust show that it is a non-degenerate map. Actually the
following holds:
Theorem 8 (aJ Let
M be a Riemannian orientabZe manifold. Then the scalar
product (T.UJ (TluJ defines an Euclidean metric. (bJ Let M be a pseudo-Riemannian orientable manifoZd. Then the scalar product
(T.UJ
(TIU) defines an indefinite metric.
Proof: (a)
We can choose a coordinate system at
Po
with metric coeffi-
cients gij (Po) Then the scalar product
=
°ij .
(Tlu) reduces to (TI u) = 1: i < ...
which is obviously positive definite. (b) Let
T
We must show that the inner product
(I)
is nondegenerate.
be a k-form that is perpendicular to all other k-forms, i.e.
o = (TIU) Choose
U with the component
components I2 k T •• = 0
for all
U12 ••• k = I
u.
and such that all other
Ui ... l where il< •..
variant components of
T
vanish. But then
Observe that the dual map tively onto the vector space fore no great surprise:
T
=
0 •
0
* maps the vector space A~ (M) bijecAn-k(M) The following the8rem is therePo
359
II
Theorem 9 (a) Let M be a Riemannian orientable manifold. The dual map an isometry, i.e. (7.55)
(*T I *U)
T
for all k-forms (b)
*
is
(Tlu)
U.
and
Let M be an orientable manifold with a Minkowski metric. Then the dual map * is an anti-isometry, i.e.
(7.56)
(*TI*u)
for all k-forms Proof:
and
T
-(Tlu)
U
(The Riemannian case) *(**TA*Ul
(*T I *Ul
=
(-l)k(n-kl*(TA*Ul
*(*UATl = (UITl = (Tlu) Observe that we have two signs involved. One coming from the double dualization, and the other coming from the exchange of
T
and
*U·
0
So much for the scalar product. Now we proceed to investigate the more general contractions: Let k~m.
T
be a k-form and
U an m-form where
Then we can extend the above machinery and we immediately see
that becomes a
(k-m)-form.
But the only general coordinate invariant (k-m)-form you can make and
out of
U jl " ' jm 1 m.
=r T . . Thus
*(*TAUl
1
l
···1
is the contraction k _ m jl .•. jm
must be proportional to this contraction.
following we will use the abbreviation i.e.
T'U
jl···jm U
T'U
In the
for the contraction
is the (k-m)-form characterized by the components
(7.57).
Observe that dualization itself is a contraction between the Levi-Civita form
£
and the differential form
U
you want to dualize, i.e.
*U = £·U
Ea:ereise 7.5.9 Problem: Show that (Euclidean metric) (MinkowSky metric)
Wor'ked ea:ereise 7.5.10(For combinatorial fans!) PrOblem:
(7.58)
Show that (Euclidean metric) (Minkowsky metric)
II
360
7.6 THE CO-DIFFERENTIAL AND THE LAPLACIAN Once we have the dual map at our disposal, we can construct another "natural" differential operator:
Definition 9 For an arbitrary k-form T the &T is the (k-1)-form defined by ~T ('1.5~)
=
co-differentia~
of
T,
(_l)k(n-k+l)*d*T
(EucZidean metric)
~T =-(_l)k(n-k+l)*d*T
(Minkowski metric)
denoted
In this way we have constructed a differential operator &: Ak(M) ~ Ak-l(M) It is obviously linear, since it is composed of the linear The sign of Ii is a convention which makes forthmaps * and d coming formulas easier to work with. Observe that Ii has an important property in common with the exterior derivative d
Theorem 10
o
(7.60)
Proof: Ii~
=
±(*d*) (*d*)
where we have used theorem S.
±*dd*
o
a
So once again you see that it pays to apply the operators twice! Note that there is no formula for Ii analogous to the Leibniz rule. This is due to the fact that
*(TAS)
cannot be reexpressed in a simple way
(See however exercise 7.6.1 for a possible generalization).
Eft{trcise '1. 6. 1 Problem: Prove the following generalization of Leibniz rule: Ii(T'U) = (_l)n-m liT'U +(_l)n-k T'du (7.61 ) (Yes, it is an ordinary exterior derivative in the last term!) Let T be a differential form. We have previously Seen that we can symbolically consider d as a I-form and the exterior derivative of T as a kind of wedge product: "ctr = dAT". In the same spirit we can consider the co-differential of T as a kind of contraction ~ = (-l)kd-T This follows from (7.58-59). E:urei3IJ ? 6 • 2
Problem:
(7.62) (7.63)
Prove the following identities *~T = (_l)n-k+l d*T *ctr
= (_l)n-k &*T
361
II
Observe that the above formulas hold both for the case of an Euclidean metric and a Minkowski metric!
If you think of
d
as generalizing the gradient, then you may
think of
~
as generalizing the divergenae. To see this we will work
out what
S
is in a few cases. For simplicity we work with an Eucli-
dean metric: Let cp be a a-form, then 5cp = gree of a form. (b) Let A be a I-form. If We put B = (-l)n*d*A i.e. has the If A components A.1 then ponents (a)
Thus is obvious because
a
SA , then B *B = (_l)n d*A
B
*A
S lowers the de-
is the scalar given by
is an (n-l)-form characterized by the com-
(*A)12 ... (n-l) = .Tg E: 12 ... (n-l)i Ai = .Tg An (where the other components are obtained by cyclic permutation). Taking the exterior derivative, we get an n-form characterized by the component (d*A)
12 ... n
= _ 1 _ [d (*A)
(n-l)!
1
2 ..• n
-
....
+
. ...
dl (*A)2 ... n + (-1)n-l d2 (*A)3 •.• nl + d3 (*A)4 ... n12 + (-l)n-lal(lgAl) + (-1)n-l a2 (lgA2) + (_l)n-la.(/i Ai) 1
Consequently
(-l)nd*A is characterized by the component [(-l)nd*AJ = _d.(lgAi) 12 ... n 1 On the other hand *B is characterized by the component [*BJ 12 ... n = .TgB
We thus get:
5A
= _..l..
d.(/~i) •
Ii 1 Observe the sign which is conventional: 5 really is minus the divergence! (c) Let F be a two-form. If we put G SF, then G is a one-form: G Here
= *d*F
i.e.
*G = (-l)n-ld*F
*F
is an (n-2)-form characterized by the components: 1 r F ij 1 '[F(n-l)n Fn(n-l)J (*F)12 •.. n-2 = 2! vg E 12 ... (n-2)ij = 2! vg = Ii F(n-l)n
In the rest of the discussion we will put n=4. This makes it easy to see what is going on, and the calculations can easily be generalized. d*F is then a 3-form characterized by the components: (d*F)123
=~! [d l (*F)23 - dl (*F 32 )
+ ... -
... J
dl(*F)23 + d2(*F)31 + d3 (*F)12
dl(/i~4)
+
d (/gr24) + d (/iF34 ) 3 2
362
II
But here we can artificially introduce t~~ term d4(;gF44) since it vanishes automati cally due to the skew symmetry of F2J Thus we have obtained
·4
(d*F)123 = di(fgFl ) and this formula is immediately generalized to If
(d*F)12 ... (n-l) = (-l)ndi(;gFin) G is a one-form characterized by components Gl ... n . (*G)
l2 ... (n-l)
=
,;g E l2 ... (n-l)1•
1
G
Consequently we have
Let us collect the results obtained in the following scheme:
( 7.64)
Euclidean metric
Minkowski metric
o-rorm: Il<j>
<j> l-form: IlA = -
A 2-form:
(5F)j F
=
1
;q
.
d
=
5<j>
0 SA
i (Ig Al.)
= - .l:.. d. (Ig _Lci l.
Fij)
0
= - __1__ d
(IlF)v =
r-a
-
1
r-a
~
(i=g A~)
d (i=g
F~v)
~
A general coordinate invariant expression for the components of
liT
is hard to derive. The details are left as an exercise for those
who are especially interested in combinatorics:
Worked exercise 7.6.3 Problem: given by
Let
T
be a k-form. Show that the contravariant components of
IlT
are
(7. 65 )
II
Once we control the co-differential
I
we can now generalize the
Laplacian to arbitrary Riemannian manifolds. For a scalar field this is particularly easy. Consider a
<j>
Euclidean space with the usual
Cartesian coordinates. Then the Laplacian is defined as: L\.<j>
= ;7;v·
(V<j»
=
~
dZ,j,
<.. ~ i=l d (xi) Z
In the language of differential forms this is immediately generalized to: (7.66)
-Ild<j>
363
II
Consequently (7.66) is the covariant generalization of the expression
aiai~, which is valid only for Cartesian coordinates. This suggests that we could generalize the Laplacian for arbitrary differential forms in the following way: ?
-/; ;, Sd but unfortunately this is not correct. To see this we consider a one-form , A.
Then
SdA
is a I-form
characterized by the components I
/g
a. [/g(aiAj_ajAi )] l.
~ a.[/gaiA j ] + ~ a.[/gajA i ]
;g
/g
l.
l.
It is the last term that spoils the game! On the other hand we may con-
dS
sider the differential operator know that
SA
When i t operates on
A,
we
is the Q-form - ~ a. (/gAi) /g l.
and therefore (dSA)j
= -aj(~a.
/gl.
In a Cartesian coordinate system
(/gAi»
/g
=
I,
and the two expressions
reduce to: -a.aiAj + aja.A i l.
l.
-
aja.A i l.
Adding the two expressions we therefore get that
(Sd + dS)A
is
characterized by the aontra-variant components -a.aiA j l. Observe that
dS
automatically vanishes for scalar fields! The correct
generalization is therefore apparently given by symmetric expression: Definition 10 *)
Sd + dS
-A
(7.67)
Worked exeraise 7.6.4 (For combinatorical fans) Problem: Show that in a Euclidean Space with Cartesian coordinates, the coordinate expression for the Laplace operator reduces to (AT)'
.
ll· •• l.k
= (a.J aj iT.l.l···l.k .
*) This definition differs in sign from the one generalZy adopted by mathematiaians. See for instanae Goldberg [1962] or de Rham [1955].
364
II
Exercise 7. 6. 5 Problem: Show that the Laplacian commutes with the dual map, the exterior derivative and the co-differential, i.e.
= *t.
t.*
dt.=M
If we proceed to consider a pseudo-Riemannian manifold with a Minkowski metric, then the above considerations are still valid, except that
-(&d+d5)
here represemts the d'Alembertian
o.
In an inertial
frame it is given by the expression o
= oj.lo
j.l
-
02
= atT
+
~
and its co-variant generalization to arbitrary differential forms is (7.68) -0 = dS + Sd Okay, let us return to manifolds with arbitrary metrics. In analogy with the exterior derivative we can now introduce two important types of differential forms
Definition 11 (aJ A differentiaZ form F is co-closed if its co-differential vanishes, i.e. SF = O. (bJ A differential form G is co-exact if it is the co-differential of another differential form F, i.e. G 5F Observe that a co-exact form is automatically co-closed, since G implies that
SG
=
S2F
=
SF
0
Exercise 7.6.6 Problem:
(a) (b)
Show that Show that
F is co-closed if and only if *F is closed. G is co-exact if and only if *G is exact.
In the rest of this section we consider only Riemannian manifolds. The generalization of the Laplacian to a Riemannian manifold suggests the following definition:
Definition 12 A differential form
T
is called harmonic if
t.T
O.
Exercise 7.6.7 Problem:
(a) (b)
Show that the constant function 1 and the Levi-Civita form £ are harmonic forms. Let T be a harmonic form. Show that *T, ctr and liT are harmonic forms too.
365
II
A very important class of harmonic forms are those where both the exterior derivative and the co-differential vanishes:
Definition 13 A differentia~ form T is called primitively harmonic if it is both closed and co-closed, i.e. dT = ~T = 0 • A primitively harmonic form is obviously harmonic, but the converse need not be true. This is wellknown already for scalar functions. Consider
-+
R3 ,{O}
M
d~
therefore
f
+
~(x)
and put
0
= rI ~
Consequently
Then i t is not constant and is not primitively harmonic. On
the other hand, i t is harmonic as you can trivially verify. In the above example it is important to notice
that
M
is not compact. For a
compact Riemannian manifold we shall show later on that all harmonic forms are automatically primitively hanronic (see chapter 8, theorem 8).
Exercise ?
(j.
8
Problem: Let A be an exact I-form, A = nic i f and only if ~ is harmonic.
d~.
Show that
A is primitively harmo-
Worked exercise 7. (j. 9 (n Complex calculus on manifolds") Notation: Let Me RZ be a two-dimensional manifold, i.e. an open subset of RZ. We identity
R2
with the complex plane (x,y)
C through the usual identification
z = x+iy
+>-
We introduce complex-valued differential forms by admitting the components to be complex-valued functions, i.e. a complex-valued differential form is a skew-symmetric multilinear map
F : Tp(M) x ••• x Tp(M)
+
c .
Consider the complex-valued I-form
w = [f1(x,y)+ig1(x,y)ldx + [f2(x,y)+ig z (x,y)ldY If we introduce the basic complex differentials
dz = dx+idY, then we can expand
w
d;:
= dx-idY
as
where hl(Z) If and
h~)
ay
= ~[fl+g2l
+
~[gl-fZl
and hz(z)
= ~[fl-gzl
+
~[gl+fzl
is a complex valued function we can introduce the partial derivatives through the expansion: db
= ~z z a
+
~z ;: i.e. ,d = 1(...£... -i...£...) a aZ 2 aX ay
dh
dZ
and ...£... = (...£... + i...£...)
az
ax
ay
We take it as well-known that an analytic (holomorphic) function in M is characterized by the following two equivalent characterizations (1) h can be expanded as a power-series around each point Zo in M. i.e. h(z) = ~ au(z-zo)n for z sufficiently close to zO. (2) ~~ "'0 n(Cauchy-Riemann's equations)
366 Consider the l-form w
Problem:
II
= h(z)dz:
(a) Show that the real part is the dual of the imaginary part. (b) Show that h(z) is analytic (holomorphic) if and only if w is closed. (c) Let us assume that h(z) is analytic (holomorphic). Show that the real
part and the imaginary part of h(z) are harmonic functions. h(z) be holomorphic. Show that ~ is holomorphic too. (e) Show that h( z) is holomorphic if and olhy i f dh is anti -self dual in the sense that (d) Let
*dh = -idh (Compare with exercise 7.5.3).
7.7
EXTERIOR CALCULUS IN 3 AND 4 DIMENSIONS We start by looking at the ordinary Euclidean space
clard veator anaZysis: The only admissible coordinate systems a~e the Cartesian coordinate systems. A vector a is characterized by its Cartesian components
R3.
Here we have the stan-
z
(ax,ay,a z ) • We do not distinguish between co-variant and contra-variant components as the metric coefficients are given by [g. ,] =
lJ
[:
...e
::]
x
p
o
e
y
Fig. 130
...
y y
a
~-----.. y
because we restrict ourselves to Cartesian coordinates. There are two kinds of products: the scaZar produat ... a • b = a b + a b + a b
x x
~ ... e
z z
and the arosB pl'Oduat
~
x
b
with coordinates
with a lot of well-known
[aybz - azby ' az bx - ax bz , ax by - ay bx ] properties. Then there is a differential operator
algebrai~
V = (1-, 1- , 1-) ox ' oy , oz Using this differential operator we can attack a scalar field tor field
~,
producing a vec-
!t ) ' v.. W1.th coord'mates (!t. ox' 2.1. oy 'oz and we can attack a vector field !, producing a scalar field :I: ... 3a 3a oa the divorg£ml!!e: V· a = oxx + 3Y" + az'" or a vector field
the
._.1.'
t
gr'cu:J;L.en :
:1:..,
Q x i with coordinates [3a 3a 3a 3a 3a _~]. oy'" - ozY 3x z oy~ This is the scheme we are going to generalize. In the e~erior algebra we can use any coordinate system: Cartesian coordinates, spherical coordinates, cylindrical
the aurL:
azX -
a,cv
II
367
coordinates, etc. For simplicity we shall, however, restrict ourselves to positively orientated coordinate systems. This fixes the components of the Levi-Civita form e to be IS E. 'k' We distinguish between covariant and contravariant components and we use thel~etric coefficients to lower and raise the indices. In exterior algebra we play with forms. As n=3 we have O-forms, I-forms, 2forms, and 3-forms. The scalars are represented by O-forms, but via the dual map we can also represent them by 3-forms! The vector fields are represented by I-forms, b~t via the dual map we can also represent these by 2-forms. Thus an arbitrary form in R is either associated with a scalar or a vector. To investigate the dual map we use formula (7.38). The results are comprised in the following scheme:
(7.69)
R3
The dual map in
Covariant components of
F
O-form:
Covariant components of
*F
3-form: *$ : <"$ ) 123 = vg$
I-form:
2-form: E3
E : (E I , E2, E3)
*E :
2-form:
B:
[-:"
B31
r. [-:' E2
0
-E 1
EI -"J 0
I-form: B12 0
-B23
-'" BZ3 J
*B : /g[B 2 3 , B3I, BIZ]
0
3-form:
O-form: 1 G *G : 7g
G : (Gh23= G
Most of the scheme is selfevident but you should be carefu1d when you dualize a three-form. A three-form is completely characterized by just one component, which we choose to be the 123- component. Applying formula (7.38) we get:
*G -_ JT 1 •
r-
vg
E.
ijk
'k G
lJ 1 . 'k Now we use that the contravariant components of £ ari. given by 72 ElJ (See exercise 7.6.1). This is where the mysterious factor 72 comes fr~. Raising the indices of c and lowering those of G produce the equ'va1ent formula 1
But here all permutations of reduces to
1
ijk
*G = 3T Tg E Gijk (ijk) give the same contribution so that the formula 1
1
1
*G = 7g E 23G123 = 7g G You should also observe that in Euclidian spaces of odd dimension we always have **F = F for a form of arbitrary rank. (Compare this with the scheme (7.44)). Armed with the dual map we can nOw investigate the vector analysis. In the following A, B and C will denote 1-forms. Let us first consider the wedge product: A A B is a 2-form with the components (A;Bi - ~Bi)' This obviously generalizes the us~a1 cross product. Although it is a '-form, A A B actually represents a vector whlch we ean find by dualizing:
368
II
If we use Cartesian coordinates, then /g = l,~ ani the expression above reduces to the components of the usual cross product a x b. ~onsequently * (AAB) is the strict generalization of the usual cross product ~ x b. Due to the fact that 1forms are odd forms, i.e. anticommute, we immediately recover the characteristic properties of the cross product: ~
~
~
~
axb=-bXa
~x!=O
Next we consider the scalar product between two I-forms:
~
~
This obviously generalizes the usual scalar product a b. More generally we can apply contractions between forms. Let. F be a 2-form, then F' A is the I-form characterized by the components F.. AJ. Here F represents a vector: F = *B, i.e. F
-
ij -
r.:g
lJ
E
ijk
Bk
F· A is actually characterized by the components
and therefore
k j
Ii E.1J'kB A which are the
=
. k
Ii E"kAJB 1.J ~
covariant components of
E:r:€!'cise 7. 7. 1 Problem: (a) Let F and that the scalar product
~
aX b.
G be 2-forms representing the vectors
~
~
and
a
~
b.
Show
~
represents the ordinary scalar product: a ' b. (b) Let F 4e a 2-form and consider the wedgeproduct F A B. Let F represent the ~ect~r a and show that the 3-form F A B represents the ordinary scalar product a ' b. Then we consider the trinle product , represents a scalar. We alre~dy know that x b. To find the meaning of (A A B) A the l23-component:
a
CA A B) A C' which A A B represents C we use exercise
But A A B is the dual of the cross product, i.e. We thus find
[CA
A
B)
A CJl23 =
/g[(a x b)3 C3
+
(A
is a 3-form, i.e. it the cross-product 7.2.1 from which we get
A B)12=
;g(a
x
b)3,
(a x b)lCl + (~ x b)2C2J
Dualizing this we finally get the scalar *[(A
A
B)
A
C J=
(a
x
b)iC.
1.
(a
and you see that [(A A B) A CJ generalizes the triple product x b). ~ But A A B is a 2-form, and therefore it commutes with C. We thus get
If we then dualize, this corresponds to the rule
etc.
... (a x
369 ... b)
~
(~ x~) •
=
b = (b
These rules are often stated in the following way: In a
to interahange dot and arOS$. Suppose now we have three I-forms: triple product:
A, Band
(A
A
C.
~)
x
II
...a
trip~e
product it is allowed
Then you can form another
B) • C
This is a~th~r one-form! A A B is the dual of the cross product a x b, but then [*(a x b) J • C is still another cross product, according to the preceding discussion! Thus (A A B)·C generalizes
It is wellknown that the triple product
... ... (a x b)
... x c
(~x
b)
x
t
satistifies the rule
In the exterior algebra this follows straight forward when you work out the nents off (A A B)'C '.\.B. - A.B.)C J
1.
J
compo~
j
1
From this you iDDl1ediately read off the formula
(A " B) • C
=
A( BJ C ) - (
AI C )B
If you compare this with the proofs in the conventional vector analysis, you will learn to appreciate the exterior algebra! Let us collect the preceding results in the following scheme
(7.7 0 )
Scalar product and wedge product Forms:
..""
Components:
IB )
i
Scalar product
<.!>
Contraction
F·A
j F .. A
Wedge product
A AB
A.B. - A.B.
Wedge product
F AA
Fij~ + FjkAi + FkiAj
~
(A
A.B 1-
1.J
0::
. .. 0
H
0::
E-< ~
l.
J
J
l.
--- ---------------- ------------- -----------------------------------Conventional vector Exterior algebra
analysis
...a ...b ...a x ...b
(A I B)
0
(1
~ H
."
b) ..
(1 " b)
0
H
E-< U
x
...
x c
(~xb).c = (txt)'b = (bxt).!
(txb)"t
=
(~.~) b -
AAB (A A Bl A C
...c
t(b·t)
-
(A A Bl
.C
A"B A C=CAAAB=BACAA (A A B)' C = ACB I C)
-
C>B
370
II
We now oroceed by investigating the differential calculus. In the exterior algebra we h~ve the exterior derivative d. It converts a scalar field ~ into a one-form with the components
d,
a,
a~
Thus
d~
generalizes the gradient ~~
a,)
a?' 3x3
(axl>
It converts a I-form A into the 2-form
dA with components
V x t.
But- this 2-form obviously represents the curl
Finally d converts a 2-form into a 3-form. If the 2-form represents a vector field, i.e. we consider the 2-form *A. then the 3-form d*A represents a scalar. It is this scalar we want to examine. As d*A is characterized by the components ai(*A)jk + aj(*A>ki + ak(*A)ij the 123-component is given by
Dualizing this we find that
*A
represents the scalar:
*(d*A)
=~ rg
a.(lgAi) l.
If you remenber that I:i = 1 in ~art~sian coordinates, you immediately see that d*A generalizes the divergence g. a. Notice however that we may express the divergence more directly by means of the co-differential 5. If A is a I-form we know that
-&A
1 rg
= ....,...
.
a.l. (/gAl.)
(Compare (7.64». This is in agreement with the result above since according to -5 .. *d*.
(7.59) we have
Exercnse 7.7.2 Problem: (a) Let F be a 2-f~rm r!pre~enting the vector a. Show that 5F rerepresents the curl V x a. (b) Let G be a 3-form representing the scalar field ,. Show that 5G represents the gradient ~
V,.
We know that the exterior derivative
d has some simple properties
rJ.2 = d(F" G) = d F"G+
(-1)
0
deg F
F"dG
Almost all of the identities involving grad, div and curl are special cases of these two rules. see the scheme two pages from thisl
371 (7.71)
II
The exterior derivative in R3.
O-form:
l-form:
4>
d4> : (a14>, d24>, 3 3 4»
l-form:
2-form:
A: (AI, A2, A3 )
dA,:
...al p
0
OJ bD
rl
al
... .....0 ...OJ
2-form:
+'
~
Ii
*A:
~,
A3 0
AZ _A1
-A'j
"Z A3- d3AZ
xxx
0
0
- - --------------------Conventional Vector analysis
xxx
0
i (d*A)IZ3 '" "i (y'gA )
AI
3-form: G : (GhZ3= G
"1 ~ -"zA1
xxx ",A,-3 A, 3-form:
4-form (Zero !)
dG
= 0
-------- ------------- -------------------------------Exterior algebra
Covariant formulas
Gradient: VtP
...al
i>o
0: 0
..... +' ....."
Divergence: 'i/ • a ~
~
A
a itP
dtP
- SA
n1 ai(lgAi )
Curl:
vx;: Laplacian: lltP
• k
dA
.rg E:ijkaJ A 1
- &H
.
~d i Vial.,)
The preceding discussion of the dual map, the scalar product, the wedge product, and the exterior derivative should convince you that exterior algebra is capable of doing almost anything you can do in conventional vector analysis. But exterior algebra is not just another way of saying the same thing. It is a much more powerful machinery for at least two reasons I 1. 2.
It works in any coordinate system, i.e. it is a covariant formalism. It works in any number of dimensions.
372
(A) l.
d2 = 0
(A)
If 'We apply
to a scalar field
'We get
.p.
d(d.p) = 0
(*) ~
II
As d.p represents the gradient of .p, the second curl . Thus (*) general ize s the rule
.
....0 ....'"
~
()
~
V x (V.p)
....
0.
'"
..:
2.
If we apply
(A)
represen ts the
.
0
=
d
to a vector field A •
we get
d(d A) : 0
(0)
As dA represents the curl of A, the second d represents the di vergence. Thus (0) generalizes the rule
V • (V
d (F A G) =
(B)
l.
x!) = 0 gF FA dG dF A G + (_l)de
If 'We apply
to scalarfields we get
(B)
d(,l/I) = (d,)l/I +
,
which generalizes ~
~('l/I) 2.
=
If 'We apply
~
(V.p)l/I +
(B)
ilJ,.pA)
.p
,
(d.p)AA+
=
and a l-form A 'We get
Which generalizes ~
~
~
V x (.pa)
.;
.
3.
.... ........'"
=
If we apply
ilJ"
(V.p) x a +
* A)
=
to a scalarfield
,
d ,A (*A) + .pd *A
,
and a 2-£orm *A we get
()
which generalizes
'"
~
~
~
~ • (.p!) = V.p
4.
If 'We apply
(B)
d(AAB)
=
+
a + ojlV
~
a
to l-forms we get dAAB-AAdB
,
which generalizes
V • (!
x
ib
=
(V
x 1) • E"
- !. (V
x
E") •
373
II
The only drawback in comparison with the conventional vector analysis arises when you discuss the equations of motions for particles. In conventional vector analysis Newton's equation of motion for a particle moving in a potential u(~) is simply given by d 2 :t
m= -Vu dt 2 This equation of motion can be solved very elegantly in various cases using the vector calculus. It corresponds however to the coordinate expression (7.72)
(7.73)
dt2
which is valid onLy in cartesian coordinates. To geometrize the formula (7.72)we must first extend the above coordinate expression to a aovaria:n.t formula! Exel'aise 7.7.3 Problem:
Show that the
covariant generalisation of (7.73) is given by 2 k m d x =
dt 2
-tnt:<
dx
ij dt
i
dx
j
_
li
dt
aU axi
3 where r ~. are the chris toffel fields in the Euc lidian space R • They sh6dld not be confused with the Christoffel fields in Minkowski space! (Hint: Use that the equation of motion (7.73·) extremizes the action t2 did j S =
![
t1
!mg .. (x) .....!.-......!.-. ~J
dt
dt
U (x)}! t
and copy the discussion in Section 6.7 ).
field
But a covariant formula explicitly containing the Christoffelfalls beyound the scope of the exterior algebra.
Worked exeraise 7.7.4 Problem:
Compute the Laplacian in spherical coordinates.
In the case of the Minkowski space the dimension is n = 4. Thus we have O-forms, I-forms, 2-forms, 3-forms, and 4-forms. Scalars are represented by O-forms and 4-forms. Vectors are represented by I-forms and 3-forms. Finally we have 2-forms at our disposal. They represent a new kind of concept, which did not exist in conventional vector analysis, but which plays an extremely important role in 4-dimensional geometry.
374
II
The effect of the dual map is comprised in the following scheme:
(7.74)
The dual map in Minkowski space:
a-form:
4-form: **$ = - $ *$ : (*$) 0123 =
r-g
~
1-form: A: (Ao, Ab A2, A3)
3-£orm: **A =A *A : (*A) 012 =
2-form:
2-form: a
F :
.;::g A3
FOI
F02 F03
-FOl
a
FI2- F31
-F02
-F12
-F03
F31
0
$
**F
~
a *F : r-g
F23
-F -F
-F23 a
-F
3-£orm:
1-form:
6 : (6)asv
*6
4-form:
O-form:
H : (H)0123= H
*H :
:
-F F
23 31 12
etc.
23
a -F F
31
F F
03 02
03
0
-F
01
12 F 02 -F 01 F a
=
**6 6 r-g[G1 23 , -C 23O , G301, _GO I2 ] **H
=
-H
-~
The only thing your should be careful about is when you dualize a fourform. The discussion of this is left to the reader (Compare the discussion following the scheme (7.69». The rest of the scheme should be self-evident. The minus signs associated with the double dualization are characteristic for Mi,nkowski spaces and you should be careful about them! Then we will investigatejthe differential calculus. First we look at a scalar field $ . It is converted into a vector field,
which we may naturally call the gradient. Then we look at a vector field. We can represent it by a I-form A. Then it is converted into a 2-form ,
375
II
which we may naturally call the curl. But we may also represent it by a 3-form *A. Then it is converted into a 4-form, i.e. essentially a scalar field. Dualizing this 4-form we get *d*A
=
-
fiA
Thus
d*A essentially represents the divergence. Finally we can look at a 2-form F which we can also represent by its dual form *F. The exterior derivative of F is a 3-form, which we may call the curl,
The exterior derivative of *F is similarly related to the divergence of
F: *d*F = - 5F (7.75)
The exterior derivative in Minkowski space:
,
O-form:
1-form:
d.p
:
(a o"
a2<1>· a3')
a1"
1-form:
2-£orm:
A: (A o• AI' A2 • A3)
dA: (aa AA - e A Aa)
2-form:
3-form:
F *F
: (Fall) :
!Fge: aByO
dF : (aa F lly + all Fya (d*F) 012 = a a (;.:g fl~
F Yo
3-form:
(*N 012
+ ay Fall)
etc.
4-form: =
a (d*fo)0123 = - ea(r-g A )
Fg A3
Concept: Gradient Curl
Exterior algebra
Covariant formula
d,
dA if
aa'
ea.
aa
F
AS -as Aa
+ all F ya + ay Fall
Ily
Divergence
- &A
kg
D' Alembert ian
- &F - &d,
a -/:; a (;.:g F -g a 1 aa( M-ga a.
aa (;.:g A")
'1
37b
7.8
II
ELECTROMAGNETISM AND THE EXTERIOR CALCULUS
In this section we will first exanplify the manipulations of the exterior calculus by studying electromagnetism in the Euclidean space R3.
The electric field strength will be represented by the I-form
where as the magnetic field strength will be represented by the 2form
There are several reasons for this, but let us just mention one of them. In Minkowski space the field strengths are represented by the 2-form:
[
F
~l -~l E2 -B3 E3
-::
0
~::l Bl
B2 -Bl
0
and here you clearly see that the electric field strength becomes a I-form while the magnetic field strength becomes a 2-form when we decompose F into space - and time - components;
_:;:_~:r--t---_~~~~_~
F
space I-form
space 2-form
1 I+'E B
E
If you remenber that B is a 2-form it 1s not difficult to translate Maxwell's equations: j Exterior calculus
conventional vector analysis
V •B = >0 ~
C<J
s::
ClB at +
0 •.-i .jJ
0
.,.j
Q
aE _ 2 c at
0
(7.76)
E = (\"
(7.77)
V
x
V
.E=
v
+
x B
=
I
-p
(7.78)
--1 Eo
(7.79)
Eo
-t
J
=
0
~+dE = at
0
dB
*
- &E
-c
2
SB
I = £;p
=
-1 J Eo
377
II
You can now reshuffle the Maxwell equations in the usual way as shown in the following exercise: Worked exeraise 7. 8. 1 Probe 1m:
Use the exterior algebra to re-examine the introduction of gauge potentials • the equation of continuity and the equations of motion for the gauge potentials.
Illustrative example:
The magnetia fieLd outside a wire.
Suppose we have a uniform current j in a wire which we identify with the z-axis. It is well known that the current produces a statiC magnetic field ~ circulating around the wire. According to Biot and Savart's law we get: ~
----L21TEoC2
(---=Y-; _x_; 0) x '+y 2 x 2+y2
At this level We do not have to distinguis between covariant and contravariant components. In the exterior algebra the magnetic field is described by the 2-form: ~
0
B
-x
0
.c+y2
B
-L
0
-L-
0
.c+y2
21T£ .,c2
x
...:..Y..-
X~2
x"+y2
y
O
*B : ~ [- ~ 21TEoC2
Fig. 131
x
or, if it is preferable, by its dual form x 2 +y'
x ; 0] X 2 +y2
which reproduces the original vector field directly. It is convenient to decompose Band *B on basic forms B=
---i-2 21TE o C
*B =
[--=.Y..- dy X2+y2
. [....=:."L~ X'+y2
dx
1\
dz +
+
x
dY]
X2+yZ
Then we introduce a new coordinate system, cylindrical coordinates, which has the symmetry of the problem. First we notice that
II
378 Cartesian coordinates and cylindrical coordinates are related by
x
PCOSql pSin
y
z
COSql dp Sinql dp +
dX i.e.
tlY dZ
=z
pSinql dql pCOSql dql
dz
Inserting these formulas we get B
. 1 --.J.- _ 21TE oC 2 p
*B
2 11~ oe2
dZAdp
and similarly
Using cylindrical coordinates, we havecons~Uy succeeded in writing Band *B as simpLe forms. To find a vector potential which generates the magnetic field, i.e. B = dA, we simply rearrange the expression for B (compare exercise 7.3.5): dz A d(lnp)
= ~2' (zdlnp) 1I E oC
From this we read off j 21TEoC2 Zdlnp
A
=
~z
21TEo~p dp
Ouside the wire we have a static magnetic field. Thus we conclude that outside the wire the Maxwell equations reduce to: dB Consequently
B
= 0 and
&B
= 0
is a primitively harmonic form.
Worked exercise 7.8.2 Problem: Show that B is almost co-exact, i.e. determine a three-form B
=
S so that
Ss
Show that S is necessarily singular along a Dirac sheet, i.e. a hal£plane bounded by the z-axis. Show that S can be chosen as a harmonic form.
Finally we can investigate the geometrical structure of Band *B' Here B is a simple form generated by the functions Z and lnp (for simplicity we forget ~ for a moment!) These functions produce the 211E oC honey comb structure shown on fig.13~Similarly *B is almost a simple form generated by the function ql, which, however, makes a jump somewhere between 0 and 21T (cf. fig. D2b). '!his exaIti'le ooncludes our discussion of electro-
magnetism and the three-dimensional geometry.
379
II
11
ql=-
2
y
-------Fig. l32a
x
Z-axis
Fig.132b
In the rest of this section we return to our main interest: electromagnetic fields in Minkowski space. The field strengths are represented by a skew symmetric cotensor field of rank 2
Therefore F is nothing but a differential form of degree 2. We may decompose it along the basic 2-forms associated with an inertial frame: (7.80)
F
+ B
z d:x
1\
dy + Bx dy
1\
dz + By dz
1\
dx
The gauge potentiaL is represented by a covector field A. Therefore A is nothing but a differential form of degree 1. We can decompose it along the basic l-forms associated with an inertial frame: (7.81)
NoW we want to compute the exterior derivatives dA and dF· We will use an al'bit.rary coordinate system. Then dA is characterized by the components d~AS - dSA~. previously we have introduced the gauge potentUll A~ through the relation (1.30)
valid in an inertial frame. But this can be geometrized as (7.81)
Now the components of F and dA must coincide in any coordinate system. Thus we learn that (1.30) not only holds for inertial coordinates but is valid in an arbitrary coordinate system:
380
Furthermore d~F~y
the
dF
II
is characterized by the components
+ aaFy~ + ayF~~.
In an inertial frame we know that
F~~
solves
~iaxwell-equation
o
(1.27)
This can be geometrized as (7.83)
So
F
is a closed form. Again we see that (1.27) not only holds for
inertial coordinates but is valid for arbitrary coordinates. The fact that the ansatz (7.82) solves the I-1axwell equation (7.83) now becomes
d2
a trivial consequence of the rule To
=
o.
finish the discussion of the electromagnetic field we should
also try to give a geometrical interpretation of the second Haxwell equation: (1.28)
F,
As it involves the divergence of
we are suggested to look at
SF.
In an arbitrary coordinate system this is a I-form, characterized by the aontravariant
components
In inertial coordinates variant components of
;=g SF
1.
and
Therefore we see that the contraJ
coincide in an inertial frame. But
then they are identical. Consequently the second Haxwell equation simply states that
SF = J
(7,84)
Observe, that as we know the components of
SF
ry coordinate system we can write (7.84) in a
and
J
aoval'iant
in an arbitraform:
(7.85)
If we take the co-differential of both sides of (7.84) we immediately get (7.86) But
-&J
(7.87)
o = SJ is the divergence of
J
so this equation is equivalent to
381
II
which reduces to the usual equation of continuity in an inertial frame. We have therefore succeeded in comprising the structure of the electromagnetic field into the following elegant form: (7.88)
Theory of electromagnetism Geometrical form
=0 ,sF = J &J = 0
r-Iaxwell's l. eq.
da. F f3y + di3 F ya. + dy Fa.f3 1 a (,t=gFa.i3) = Jr:I.
dF
Maxwell's 2. eq. continuity eq. Eq. of ga.uge r-otential
Gauge transformation
Covariant form
F
=
0
v=g a
1 /=a a. T-a -g da. ( -g ::r ) =
0
Fa.i3 = d a.Ai3 - df3 Aa. A ... A + da. X a. a.
= dA
A ... A + dX
Illustrative example: The monopoLe fieZd. Observe first that we can now work in arbitrary coordinates. A relation like Fa.i3 = da.Af3 - di3Aa. is also valid in spherical coordinates or even in a rotating coordinate system if you prefer! We have previously computed the field strengths of a monopole expressed in spherical coordinates (Section 6.10), cf. cq. (6.79). Therefore we can decompose the monopole field as: (7.89)
F
:fir SinB de
1\
d«)
This is a simple form which may be rewritten as F =-..5l. d(Cos6) 41T
1\
cip
AS it is a static field we can suppress the time coordinate for a moment, i.e. we look at space at a specific time to. Apart from the normalization factor, F Fig. 133 1s generated by the functions CosB and «). These two functions generate the honeycomb structure shown on fig. 133 , (i.e., CosS
= 0,
£, 2£, •••
and Tubes
0, £, 2£, ••• )
pointing in the radial direction
382
II
The fundamental tubes are defines by cose = +1, 0, -1 and = 0, 1, 2, 3, and the number of fundamental tubes emanating from the origin is 2x2~. Thus the number of fundamental tubes which cross a sphere around the origin is 4~. Observe also that the Etubes all intersect the unit sphere in the same area. To see this we ~
use the wellknown fact that a region n = {(e,~) lel~e~e2; ~1~~~~2} covers the area: 9 2 (j)2 -COSe2 ~2 r r Sine ded~ = r r d(-Cose)~ -cosel ~l e 1 ~l Therefore an arbitrary E-tube covers the area U=-kE ~=mE+E r dud~ = [-kE + kE + E][mE + E - mE] r U=ke:-E ~=m ThiS, of course, again reflects the fact that the monopole field is a spherically symmetric field! Now observe that because we have written the monopole field as the simple form F
=
*
Sine de
we can immediately find a gauge potential
1\
A
d~
,
generating the monopole field
Thus we can use the gauge potential (7.90)
A = -.3.. Co s e d~ 4~
Now we should be very careful because we have previously shown that the monopole field cannot be generated by a gauge r-otential throughout the whole space time manifold. (~ection 1.3l. Therefore we conclude that A = ~ cose cAp cannot define a smooth vector field throughout our manifold. The component ~ CosS is certainly a nice smooth function, so this must be related to the fact that the coordinate system itself is singular! We have several times seen that this coordinate system in fact breaks down at the z-axis, and therefore the gauge r-otential (7.90) 1s not defined on the z-axis. To investigate this a little closer we return to the Cartesian coordinates. Using that
383
II
(jl= Arctan Y
cose
x
we easily find
so that the gauge r-otential has the Cartesian components ..5l..[0 - 41f '
-
z~ r (x +yz)'
+
zx
r (X2+yZ)
,
0]
But these components become highly singular when we approach the zaxis! If we approach the pOint (O,O,zo) where Zo > 0 we can safely put: ~ '" 1. Thus A varies like r a. Aa.
"*
[0, - xl;y2
, XZ:y2 , 0]
and we see that
JAx J , JAy J ... ., when we approach the z-axis! Consequently A is highly singular on the z-axis in accordance with our general result: There aan exist no gLobaL gauge potential- generating the monopo Le fie Ld. However, if we are willing to accept a string, where the gauge potential becomes singular, then we have just found suoh a gauge potentiaL This string is, of couse, the famous "Dirac string"! (Observe, that in the preceding discussion we have "excluded" the origin from the spacetime manifold, because at the origin the monopole field itself is singular). You should also observe the following: The spherical coordinates not only breaks down at the z-axis, but the cyclic coordinate (jl also has a singularity at a halfplane extending from the z-axis. This is connected with the fact that (jl has to make a jump somewhere between o and 21f. Thus d(jl is not defined on the halfplane where (jl makes a jump. But if you return to Cartesian coordinates you find d(jl = - ~2 dx + x 2 Therefore d(jl we conclude: 1.
!
y2 dy
is perfectly well-behaved (except on the z-axis) and
itself is a coordinate function, which makes a jump at a halfplane, and the-refore it aannot be extended to a smooth function on the whole space surrounding the z-axis (Compare , section 1. 4)
384 2.
II
d~ does not make a jump at a halfplane and consequently it can be extended to a smooth I-form on the whole space surrounding the z-axis.
In what follows we will let d~ denote the smooth I-form defined on the whole space surrounding the z-axis. The singular gauge p:ltential A =:5l.. cose d~ is obviously not the on41T ly gauge p:ltential generating F. To get another example we observe that d(~) = O. Therefore we consider AI = A + ..5l.. d~ = -..5l.. cose d~ + ..5l.. ~ 41T
41T
41T
Although it formally looks like a gauge transformation, it is a singuLar gauge transformation because ~ is a singuLar coordinate function. Nevertheless A' generates the same field strengths. Observe, that: (7.9ll
A'
= -..5l..
41T
(Cose - 1) d~
and that the coefficient cose - 1 vanishes on the positive z-axis. This suggests that A' is onLy singular on the negative z-axis. To confirm this we return to Cartesian coordinates: A'
=:5l.. (~ - 1) 41T
r
If we approach the pOint proximation
(- xZ;y2 dx + X2~y2 dy)
(O,O,zo)
where
to get the asymptotic behavior of r - 1 ,,-
It fOllows that
A'
2z
~ - 1: r
X2
Z
we can use the ap-
X 2 +y2
+
Z
Zo > 0
+y2
2zo
varies like AI
,,-*
(~ dx - 2 zX
2
dy)
o
and the singularity on the positive z-axis has disappeared. Thus we are left With an semiinfinite Dirac string: The negative z-axis. This is the best result you can obtain! (If you take any closed surface surrounding the monopole, then it must contain at least one singularity of the gauge field. Otherwise you would contradict theorem 1 of section 1.3).
385
SOLUTI ONS OF "lORKED EXERC ISES: -
II
No. 7.5.6 Since *TAU is an n-form we only need to find the (12 ... n)-component. This is given by the expression n! (*T) _ (U) (n-k)!k! [12". (n k) (n-k+l)."nl
(7.92)
Now observe that al···an_kbl···bk E
£
a l ·· .an_kc l · .. c k
vanishes unless (b l " .b k ) is a permutation of relation holds (cf. exercise 7.4.3): I (n-k)!
(C " .c ) I k
Clearly the following
al···an_kbl···bk £
where we have used that there are (n-k)! permutations of (al ... a n _k ) , all contributing with the same value. Furthermore We can USe that TC1 .. ·Ck 1S skewsymmetric so that we get:
reduces to
Inserting these
We have thus shown
No. 7.5.10 For simplicity we discuss only the Riemannian case. We want to compare *(T'U) and *TAU· For definiteness we only compare their l ... (n-k+m) component, but the argument can easily be generalized to arbitrary components. The contraction T'U is characterized by the components fl· .. f
I
-T
m!
U
m
e l ·· .ek_mfl •· .fm
The dual form is then characterized by the components ~ ). (k
-m.
1 ---, £
m.
If we specialize to the
cI···cn_kgl···~el···ek_m
e l " .ek_mfl " .fm
(cl ... cn-kgl"'~) = (l •.. n-k+m)
*(T'U) *TAU
T
l ... (n+k+m)
is characterized by the components
f 1 ••• f m
component we finally get
= ~ T(n-k+m+l) ... nfl···fm U m!
U
fl ... f m
306
II
We must then show that (7.94) reduces to (7.93) except for a sign. This is done by brute force! We attack (7.94): Observe ~irst that (al ... ao-k) is an ordered subset o~ (1 ... n-k+m) and that (el •.. ek-mfl ... fm) is an ordered subset which is complementary to (al ... ao) i.e. they are mutually disjoint, and together they exhaust (l ••. n). Especially (el .•• ek-mfl ... f m) must contain the numbers (n-k+m+l ... n) . Now fix (bl ... bm) for a moment. Then (al ••. ~-k) is determined up to a permutation. All these permutations have equal contributions when we perform the summation, and we can therefore pick out a single representative shich we denote (al(b), ... ,ao-k(b)), since it depends on {bl, •.. ,b m} . Now that (al(b), ... ,an-k(b)) has been fixed we observe that (el .•• ek-mfl ... fm) is determined up to a permutation. Again all these permutations give the same contribution, and we need only pick up a single representative. Since (el ••• ek-mfl ... fm) contains (n-k+m+l, ... ,n) we can simply put e
l
= n-k+m+l
, .•. , e k m
=n
Furthermore we pick up a special representative for f , ... ,~ which we denote (fl(b), ... ,fm(b)) , since it too depends on {bl, ... ,tm}. wrth these preparations ('r .94) now reduces to ;,-
~ Sgn m.
[al (b) .•• a n- k (b )b l • •• bm] 1. •• n-k
n-k+1. .. n-k+m
E
al(b) ... an_k(b)(n-k+m+l) ... n~l(b) •.. fm(b
• (n-k+m+l) ..• n~l(b)···~m(b) U x'T b l ·• .b m where we only sum over b. Observe that (a (b) •.• ~_k(b)b •.• b m) is an ordered subset of (l, ..• ,n-k+m) and similarly (al(t) ... ao-k(b)fl(t) ••• fm(b)) is an ordered subset of (l, .•• ,n-k+m) . But then (fl(b), .•• ,fm(b)) is.a permutation of (bl .•. b m) and we can safely put fl(b) = bl' .•• '~m(b) = b m • Thls means that (7.94) can be further reduced to
i& I
m.
Sgn
[al(b) ••• an_k(b)bl ... bm] 1. •. n-k.
n-k+1. .• n-k+m
387
II
Let us focus on the statistical factor: (No summation!) al(b) ... an_k(b)bl···bm} Sgn [ l ... n-kn-k+l ... n-k+m Eal(b) ... an_k(b)(n-k+m+l) ... n bl ... b
m
This statistical factor is easily evaluated if we rearrange the indices of the Levici vi ta symbol al(b) ... an_k(b)(n-k+m+l) ... nbl···bm
+
al(b) ... an_k(b)bl···bm(n-k+m+l) ... n
This costs a factor (_l)m(k-m), but apart from this sign the combinatorial factor is now recognized as a complicated way of writing the number I ! We have thus finally reduced (7.94) to the form C_I)m(k-m);g T(n-k+m+l) .•. nbl···bm U m! b ... b l m If you compare this with (7.93) you deduce the following formula (7.95) *TAU = (_I)m(k-m)*(T'U) which is in fact valid for both Euclidean metrics and Minkowski metrics. The desired formula now follows by performing a dualization on both sides .
0
No. 7.6.3
Let us work it out for an Euclidean metric. Putting S &T = (_l)k(n-k+I)*d*T
we obtain (Observe that this actually holds for Minkowski metrics too). Here (n-k+l)-form characterized by the components (d*T)·· . (n-k+l)! 3[·(*T)· . 1 Jll···l n _k (n-k)! rJ ll··fn-k = __ 1_ Sgnlabl' .. bn-kj 3 (*T)
(n-k)!
1
= (n-k)!k!
For simplicity we investigate only the are easily generalized
.. . Jll·· .In-k
a
d*T
is an
bl ... b_k n
[ab l ... b k) E 3 (lgrcl··c k ) Sgn ji l ... i~=k b l ... bn_kc l ... c k a (l, ... ,n-k+l) - component, but the results
(d*T) 1. .. (n-k+l)
=
I Sgn[abl ... bn-k]E 3 (/gT (n-k)!k! 12 ... (n-k+l) b l ·· .bn_kcl.··c k a
cl
···
ck
)
For fixed a the ordered set (bl ... b n- k ) is determined up to a permutation, since a and (bl ... b -k) are complementary subsets of (l, ... ,n-k+l) . All these permutations give th~ same contribution so we need only specify a single representative (bl···b n _k ) = (bl(a), ... ,bn_k(a)) Once (bl ... b n- k ) is fixed the ordered set (cl ... c k ) is also determined up to a permutation. All these permutations give the same contribution too, and we need only specify a single representative. Since bl ..• b n- k are all different from a. we conclude that a coincides with one of the numbers cl ..• ck We can therefore always achieve that
388
II
The rest then can be chosen as c2 = n-k+2 , ... , c k = n With these preparations (7.97) now reduces to (d*T)l. . . (n-k+l)
ra bl(a) ... b (a)] Sgnl n-k g a (/gF a ( n-k+2 ) .. n) 12... (n-k+l) b l (a) ... b n_k (a)a(n-k+2) ... n a Using that g
~l (a) ... b _ (a~a(n-k+2) ... n
= (_l)n-k g
n k
we
~inally
obtain
(7.98) Similarly
(7.99)
ab l (a) ... b - (a) (n-k+2) ... n n k
(d*T)l ... (n-k+l) *s
is a (n-k+l)-form characterized by the (1, ... ,n-k+l)-component _ ~ jl···jk-l (*S)l. .. (n-k+l) - (k-l)! gl. .. (n-k+l)jl···jk-IS
;g S(n-k+2) ... n If we compare (7.98) and (7.99) and use the relation (7.96) we finally obtain the ~ollowing result
which we can immediately generalize to the result (7.65) No.
7.6.4
When we use Cartesian coordinates in an Euclidean space we do not have to distinguish between co-variant and contra-variant components. Furthermore we can put ;g = 1 If T is a k-form the derivative dr is then the (k+l)-~orm characterized by the components
~,• Sgn[~J ll· ~l""'~] •. lk Thus -&dT
is the
Hdr)
k-~orm
al"'~
acT
dl"'~
characterized by the components =
.l.. k!
a b
Sgn[C d l ·· j
.~] a
bal"'~
c
T
dl"'~
For a ~ixed c the ordered set (d ... ~) is determined up to a permutation. As all permutations give the same contribution we need only pick up a single representative ~ = dl(c), .•. ,~ = ~(c), and the above expression reduces to
where we only sum over
b
and
c.
Clearly this sum decomposes in the
~ollowing
way
389
II
It is the last term we must now get rid of. Since king one transposition we get
cfb
we can fix
dl(c) = b.
Ma-
Inserting this we have shown:
On the other
For fixed c the ordered set (d2 ... ~) is determined up to a permutation. All permutations give the same contributlon, and we need only pick up a single representative d 2 = d2(c), ... ,~ = ~(C) The expression above then reduces to
Adding the two results obtained we finally get: (AT)
No.
a l · .. ~
= (-d&T)
a l ... ~
+ (-&
3 3 T
a l ... ~
b b a l ... ~
o
7.6.9
(a)
w
= [f+ig][dx+iay] = [fdx-gay]
If we use that
*ay
= dx.
*dx
= -ay
+ i[gdx+fayl we immediately get
*[gdx+fdy] ~ [-gay+fdx] 3h dh dJM ~ dbAdz = a; dZMZ + dZ-
Consequently (c)
If
h(z)
W
. lS
dJM = ~~ dzAdz closed l. f and only ~f ~
dz
is a l-form, so that
dh = 0 3Z-
is holomorphic we get
db
= dh 3z
dz
'
which implies that
But Thus
h
(d)
Let
is a harmonic function and so are its real and imaginary parts. h(z)
be a holomorphic fUnction. Then
db
= 3z dh
dz
and that is a closed
390 ah az
l-form. According to (b) this means that use that
h
II
is a holomorphic function. We can also
is a harmonic function: 2
0" a h
azaz
a (ah) = a-z az
but this is exactly the Cauchy-Riemann equation for the function (e)
If we once more use that *dz
and
*dx = -dy
*dy = dx
ah az
we immediately get
-idz
Consequently the decomposition db = ah dz + a~ clZ az az
*
is a decomposition of db in an anti-self dual and a self-dual part. It follows that db is anti-self dual if and only if = 0 0
NQ.1.,1.4
We want to compute the Laplacian in spherical coordinates. We have previously computed the metric coefficients, cf. (6.36)
a
We also need the contravariant
components of
a
aif
df,
i.e.
[aif] -
1
a
a
1
a
a
rz
(\
a
1
r2Sin 2a
["
" ref
a f r
rzI
[
r2s~n2a
"qf
"a f a
We thus get
(1.100)
No. 7.8.1 The first two Maxwell equations can be solved using gauge potentials. The equation, dB = 0, is solved by the ansatz B = dA
,
where A is a vector field! Substituting this into (1.T1)you get
the second Maxwell equation
391
II
o , and this equation can be solved by the ansatz
where
,
is a scalar field. Thus we have rediscovered the gauge potentials $,A:
B
= dA
The last two Maxwell equations(7.78) and(7.79) can be used to find the equation of continuity. Rearranging them we get 1 ap
- Eo" at
and
from which we deduce that
~ at
=
SJ
i.e.
which is the proper generalization of the equation of continuity. Finally we deduce the equations of motion for the gauge potentials. Substituting eq. (7.101) into the two remaining Maxwell equations we get
{ {
&(1A + d,) at
+
....!..
alA -at
i.e.
1 a2 <~
acz +
p
EO
1 EO
= +-J
Sd), - ..2....
1 p EO
= -
1 a2 ~ - SA)
If
<"A)
satisfies the Lorenz condition
we thus recover the waVe equations
1 = EOC2 J
392
II
!'!.n. 7.8.2
First we observe that
(*)
B = &S
iff
*B
- d*S
Therefore B is co-exact in a space region same region. But
*B
n
if and only if
*B
is exact in the
= 21TEjoc2 d(jl
Apart from an irrelevant constant this is our old friend d(jl (See section 1.4) which isonly almost exact. Consequently B is only almost co-exact. From (*,**) we also see that
s
= -
j
21TEOCL
*(jl
It remains to determine the Levi-Civita form [. that ,.g = p (Compare exercise 6.6.3). Thus £
so we nave finally found
In cylindrical coordinates we know
= pdPAcIQ A dz
S
s
= -
21TEjocL
p(jl
dp A
A dz
and at the same time we have managed to write S as a simple form. In particular S is closed and it follows that S is a harmonic 3-form
- dSS
->tS = -dB
o
o
3£:
II
ehapter 8
INTEGRAL CALCULUS ON MANIFOLDS 8.1 INTRODUCTION The next problem we must attack is how we can extend the integral calculus to differential forms. Before we proceed to give precise definitions we would like to give you an intuitive feeling of the integral concept, we are going to construct. In the familiar theory of Riemann-integrals we consider a function of say two variables defined on a "nice" subset Then we d±vide this region cells
Ai
with area
U
of
U
into
f(x,y)
R2
z = f(x,y)
and in each
£2
cell, Ai' we pick up a point We can now form the Rie(xi'Yi) mann-sum: L f(xi'Yi)&
2
i
If
is "well behaved" and
f
U
is
/
a sufficiently "nice" region, then this sum will converge to a limit as £
~
and this limit will be indepen-
0
x-axis
Fig. 134
dent of the subdivision in cells. It is this limit we define to be the Riemann-integral:
lu
f(x,y)dxdy
Next we consider a two-dimensional smooth surface dinary Euclidean space sion 2 embedded in in
R3
As
n
dinates
(A1'A2)
surface
n.
R3
R3
in the or-
Similarly we consider a two-form
F
defined
is a two-dimensional surface, we may introduce coorwhich we assume cover
n,
i.e. we parametrize the
(See figure 135) .
Then we make a subdivision of the coordinate domain of area
n
i.e. a differentiable manifold of dimen-
&2
U
In each of these cells we pick up a point
into cells
(A 1 ,A 2 ). (i)
(i)
394
II
At the corresponding point
~(A 1 ;A 2 )
P; •
(i)
n
in
we
1.2
(i)
have two basic tangent vec-+
e
tors
(i)
F
-+
and
1
e
(i)
2
As
•
is a two-form, it maps
(e(i) 1 ,e) (d
into a real numSo we can
F(e 1,e.2 )
ber
(i)
(i)
form the sum -+
-+
L F(e 1,e 2 )& i (i) (i)
If
F
Fig. 135
x
2
n
is "well-behaved" and
converge to a limit as
&
-+
is a "nice" surface, then this will and this limit will be independent of
0
the subdivision in cells. It is this limit we define to be the integral of
F:
(8.1)
Jn
def -+ -+ 2 "" lim L F(e 1 ,e 2) & &-+0 i (il (i)
F
At this point you might object that
fn
F
cannot be a geometric
object, because we have been extensively using the special coordinates (;\1,1. 2 )
in our definitions. However, it will be shown that fn F does not depend on which coordinate system we choose, provided they define the same orientation on
n.
Of course the above formula is of little
value in practical calculations. Therefore we should rearrange it! 1 2 To each point (1. ,1. ) corresponds a set of canonical framevec-
(e 1 ,e2 )
tors
and therefore
F
generates the ordinary function
U ~ R
given by
Using this function we can
the integral of
conv~rt
F
into an ordina-
ry Riemann-integral: (8.2)
Jn
F""l1mLF(e1,e2)&2 & .... 0
i
(i)
(i)
We can also express the integral of of
F.
Here you should observe that
the whole of the Euclidian space
R3
F
F
in terms of the components
is defined as a two-form on
Let
-+
-+
-+
(i ,i 2 ,i 3 ) 1
denote the
canonical framevectors corresponding to a coordinate system on Then
F
is characterized by the components
R3 .
395
(e1 ,e 2 )
The canonical framevectors
II
on
n
(1 1 ,1 2 ,1 3 )
are related to
in the following way: and Using this we can rearrange (8.2) in the following way: 1
2
Fab(A , A )
'x a 'x
b
;;:'f;;? 0
"
i.e. (8.3)
We shall reconstruct these formulas from another point of view in a moment.
8.2 SUBMANIFOLDS - REGULAR DOMAINS In the general theory we have a n-dimensional differentiable manifold
M.
We want to characterize what we understand by a k-dimen-
sional surface
n
in
M
To do this we must fix some notations:
R(n,k)
n {x E R
x
=
1 2 (x ,x , .•. , xk , 0, ••• , 0) }
R~n,k)
n {x E R
x
=
2 (x 1 ,x , ... ,xk,o, ... ,0)
(See figure
,
x
1
~ a}
l36.~
n=3,k=2.
Fig. 136 We can then define the concept of a submanifold:
Definition 1 point
A subset P in
n n
=M
is said to be a k-dimensional submanifold if each has the following property:
There el1;i"sts an open
neighbourh()~d
V
of
P
and a coordinate
II
396
($,U)
system V
such that
$
maps
U n R(n,k)
bijectivelyonto
n n.
(See figure
137a-b
below.)
n=3,k=2
Fig. 137a
n=2,k=1. X2
U
Fig. 137b A coordinate system ted to
n.
n,
actually cover that we call
($,U)
with the above property is said to be adap-
According to the definition the adapted coordinate systems
n
hence they constitute an atlas. This justifies
n
a submanifold. As
troduce arbitrary coordinate systems on
itself is a manifold, we can in-
n,
as long as they are com-
patible with the adapted coordinate systems. Clearly the k-dimensional submanifolds is the eXact counterpart of the loose concept: a k-dimensional smooth surface in
M.
As a specific example you may' 2 consider M 8 • Then the 1 equator 8 , is a one-dimensional submanifold. As adapted coordinate systems we may use the standard coordinates on the
xl
sphere. (Compare the discussion in section 6.2. 138 too.)
See figure Fig. 138
Next we should try to characterize the domains of integration. From the standard theory of integration you know that these domains
397
II
should be chosen carefully, otherwise we will get into trouble with the b . f: R ~ R is a continuous function, then fa f(x)dx is
integral. If
well defined, while
f+OO
f(x)dx
-00
is somewhat dubious. In standard inte-
gration theory you would therefore start by restricting yourself to closed, bounded intervals
[a;bl
.
Then you will have no trouble be-
cause these are compact subsets of
R
and continuous functions on com-
pact intervals have many nice properties; especially they are bounded and uniformly continuous. Furthermore, you know that there is an intimate connection between integration and differentiation (the fundamental theorem of calculus) : b
t
df dx = f(b) dx
- f(a)
df to the values of f at the pOints dx b, which constitute the boundary of the interval [a;bl .
This relates the integral of a
and
Back to the general theory! To avoid troubles with convergence we will only consider compact domains.
(If
MS
RN
then a subset of M N R .) Further-
is compact if and only if it is closed and bounded in
more we should consider domains where the boundary itself is a nice "smooth surface". Otherwise we will not be able to generalize the fundamental theorem of calculus. This motivates the following definition: (See figure 139 too.) Definition 2 A compact subset
of a n-dimensional manifold
Q
M is called a
k-dimensional regular domain, i f each point P in Q has one of the following two properties:
1.
There exist an open neighbourhood Vof P in M and a coordinate system
($,U) on M suah that:
$: U n R(n,k) 2.
bijectiv~ly V
n n .
There exist an open neighbourhood Vof P in M and a aoordinate system
(~>U)
~: U n R~n,k)
on M such that:
bijective~y
V n nand
x 1 (P)
=0
The points with property 1 are called interior points of The set of interior points of
n
is denoted
int
n
n •
Clearly it is a
k-dimensional submanifold. The points with property 2 are called boundary points of
n •
398
II
M
z
P satisfies property 1. Q satisfies property 2.
(Observe that $-l(Q) is lying 2 on the X -axis, i.e. x 1 (Q) = 0)
Fig. 139
The set of boundary points is denoted
an
and it clearly is a
mensional submanifold called the boundary of ry point and
($,U)
n
If
P
(k-1)-di-
is a bounda-
is a coordinatesystem with property 2, then we may
use
as local coordinates on
an.
Thus we have decomposed! a k-dimensional regular domain into two disjoined components:
int
n
and
an.
The boundary
an
has an ex-
tremely important property! It is itself a compact subset:
Theorem 1
an
is a compact
{k-1) -dimensional submanifold.
Proof: The proof of this is somewhat technical and you may skip it. As n is compact it suffices to show that an is closed in M. Let (Pn)nEm be a sequences of points in an and assume that this sequence converges towards P, i. e. Pn'" P. We have finished if we can show that
P E
an .
399
II
As n is compact, it is closed in M and therefore we know that PEn But then P is either an interior point or a boundary point! Thus we must rule out the possibility of P being an interior point. Assume that P is an interior point. Then there exist an open neighbourhood V of P and a coordinate system (~,U) such that ~:U () R(n,k)
bijectively ..
V () n
Fig. 140 and
P
If follows that all the points in V n n are interior points. But PEn ~ P therefore there exist an integer N such that n>N implies tRat
n
P E V () n
Thus we conclude that tion! i
P
is an interior point if
n
an
This property of
an
an
in. But
a>J
N and that is a contradic-
an
an
is a com-
has property 1. Therefore you
are interior points, it follows that
boundary. Consequently
FOr a
~
itself is a regular domain! But as all the point in the
regular domain
Theorem 2
n
has important consequences. As
pact submanifold all the points in see that
n
(Cartans regu~ar
a(an)
an
has no
is empty:
~emma)
domain
n
the boundary
an
is also a regular doma-
has vanishing boundary
a (an)
o.
We shall include a point as a zero-dimensional regular domain, although it really falls outside the scope of our definitions. We do not assign any boundary to a pOint, so that the rule
aan = 0
still holds
in this very special case! As this point we had better look at some specific examples all using
M = R3
•
400
II
Examp l~ 1: If n is the closed unit ball, n
= {xl
then
intO
<xix> ~ I}, is the open unit ball,
= {xl
intn
<xix> < I},
0
and the boundary of
is the
two-sphere,
=
an
i.e.
{xl <xix>
= S2.
ao
= I}
,
Observe that
s2
is itself a two-dimensional regular domain without boundary
aan
=
0. Fig. 141
Examp Ze 2: If
n
is the northern hemisphere
82,
of
E R31 <xix>
n = {x then
intO
= 1, x
3 > oJ,
is given by,
intn = {x E R31 <xix> = 1, x 3 > and the boundary of n is the
oJ,
unit circle, an
= {x
Le.
1, x 3
E R31 <xix>
an
=
OJ,
sl ,
sl is itself a onedimensional regular domain without
where
boundary:
d(dF) = o.
a(an)
=
0
reminds us of the
They are, in fact, very intimately connected.
a differential form
cLosed form. the property
Fig. 142
aan = 0.
Finally we observe that the rule rule
ao=sl
F
had the property
dF
=0
In a similar way we will call a regular domain an
o ,
a closed domain.
If
, we called it a 0
with
This is in agreement with
the familiar expressions "a closed curve" and "a closed surface". At this point you should begin to have a feeling for the regular domains which are going to be the domains of the integration! There is, however, still a problem which we will have to face, and that is
the orientability of
n.
401
II
If we return to our intuitive discussion of how to integrate a twoform on a surface, it is clear that if we interchange the coordinates Al
and
then we also change the sign of the integral because ~i
-+i
~ F(e l ,e 2 )E
2
x
Fig. 143
But, if we interchange
Al
A2 , this simply means that we
and
construct a new coordinate system with the opposite orientation! This clearly shows that the integral depends on the orientation. Consider now a regular domain
0
in a manifold
M
Then int 0
is a k-dimensional submanifold which can be covered by adapted coordinate systems.
provided
int n
We say that the r~gular domain 0 is orientable is an orientc.ble manifold. (Compare with the
discussion in Section 7.4). The interesting point is that whenever
o
is an orientable regular domain, so is
orientation chosen on orientation on int 0
ao
Now let
int 0
aO!
In fact, the
induces in a canonical manner an
Tb explain this we choose an orientation on (~,U)
ao,
be an adapted coordinate system on
then: (xl, .•• , xk, ..• , xn) (xl, .•• ,xk,o, ••• ,0) xl < (0,x 2 , .•• ,xk,o, .,0)
°
serves as local coordinates on
M
serves as local coordinates on
int 0
serves as local coordinates on
ao
The local coordinates generate a positive orientation on ao if and onZy if the loaaZ coordinates (x l ,x 2 , •.. ,x k ,0, ••• ,0) generate a positive orientation on int Q. We have tried to exemplify this on This makes i t reasonable to define:
(0,x 2 , ••• ,x k ,0, ••• ,0)
fig.
144a-c.
402
II
k
Orientation of intll induced by a coordinate system.
Fig.
11111':--1,,--, 1I
l44a
by a coordinate k = 2
12 has a »hole«! Observe that the :s)inner« boundary
and the »outer« boundary get opposite orientations.
Fig. l44b
k dO
=
=1
,
I
I ,
I
III
Orientation of intO induced by a coordinate system.
~
P UQ
Q(+)
PH A point is negatively oriented when we move away, positively oriented when we move towards the point.
Observe, that in the case where
Fig. 144c
dO
reduces to a finite set of
points, we will have to construct a special convention, as you cannot introduce coordinate systems on a single point!
403
II
8,3 THE INTEGRAL OF DIFFERENTIAL FORMS With these preparations we are ready to define integrals of forms. , So let n be an orientable k-dimensional regular domain in Mn and let F be a differential form of degree k defined on Mn We assume for simplicity that
int
n
is a simpZe
manifold which can be
covered by a single coordinate system (not necessarily adapted!). We denote the coordinates of
int
n
by
(Al, ••• ,Ak) and those of
M
by
(xl •••• ,xk ••••• xn).
R
F
Fig. 145
The coordinates tated.
(Al •••. ,Ak)
are assumed to be positively orienk (Al, ••• ,A ) in U corresponds a set of
To each point
canonical frame vectors ~
~
el,···,e k As
F
number:
has degree
k, it maps (~l""'~k)
F(el, ••• ,~k).
ordinary function on
into a single real
. cons~der
-+
~
Thus we may F(el, •••• e ) as an k U • and therefore we can define the integral
as (8.5)
Jn F
def.
=
f u
-+
~
1
F(el, ••• ,ek)dA ••• dA
k
We may rearrange this formula by introduci~g the canonical frame ~s -t
~l'
on M.
-t
••. , ~n
404
II
(8.6)
Inserting this we get
Therefore we finally obtain: Definition 3
n
Let
=Mn
metrized
be an orientable k-dimensional regular domain paracoordinates (A1 , •• • ,)..k).
by
form of degree
k
(xl, .... xn) on
respect to coordinates integral of
\J
''i n J,
be a differential
F
F
with
al • • •ak
Then we define the
M
through the formula
F
-+ F(e l ,···, e k ) dAl ... dAk ~
fu
F
Let
characterized by the components
(8.7
Ju
F
a r .. a
ax
al
ax
~
k
al
~
dAl ... d).k
The formula (8.5) clearly shows that the integral is independent of the coordinate system (xl, ••• ,x n ) on
M
but we still have to
show that it is independent of the coordinate system ().l, ••• ,).k) on
n . Therefore we consider two sets of coordinates (AI, ••• ,).k) and (~l, ... ,~k) which both cover int n and which specify the same
int
orientation on positive.
-+
-+
F(el;···:e k ) (2)
n ,
int
i.e. the Jacobiant
i dA 1
= F(~i
(1)1 aliI
(2)
But F is a k-form and Therefore we conclude F (~.
; ••• ; ;;.
(1)
(1)
~l
~k
)
=
E.
1
(2)
~k) (2)
i dA k k
i (l)k aJ.1
.
1 · · · lk
F(E!l;···: (1)
-+
, .... , e
i l ,··· ,i k
Inserting this we get F(tl ;···;
Det
= F (e-+
is always
i (1)1
~
; ••• ; e. )
(If
i
i
aliI
k
~ ... dA
can only take the values
F (t ; ••• ; ~k) l
(1)
(a~~) d).J
First we observe that
(1)
all
k
1, ••• ,k.
405
II
Using the transformation rule for Riemann integrals we then obtain
~
I
e k ) Det (Olli) ---j I dA l •••k dA OA
(2)
Det(dll~) J
But
is assumed to be positive and therefore we can forget
OA
about the absolute value.
f F(el ;···; ek )
(8.8)
U2
(2)
n ,
d]Jl .•• dll
k
(2)
and this shows that int
We then finally obtain
IF n
=
f F(el ;···; ek ) Ul
(1)
dAl ••• dAk
(1)
is independent of the coordinate system on
as long as it generates a positive orientation!
up to this point we have only considered the situation where int Q is a simple manifold, i.e. it can be covered by a single coordinate system. We now turn to the general situation where we have to use an atlas (cj>i,Ui\EI consisting of overlapping coordinate systems, which together cover integral of a differential form over n.
IF
int
n.
We want to construct the
n
Now there is one case where the answer is obvious: Suppose there exists a special coordinate system (cj> o,Uol so that the restriction of F to n vanishes outside cj>o(U )' Then of course, we define o the integral as
fn Iu F(~ l·'···;~ F =
kldAl ••• dAk
o
Fig. 146 F ;,t 0 Next we want to consider the general case of an arbitrary differential form F • This requires, however, the introduction of a oechnical but important machinery: Recall that an open covering of a topological space subsets (OiliEr that covers M, i.e. M
=
M is a family of open
U O.
iEr
1
(Open coverings play an important role in the characterization of aompaat spaces. A topological space M is compact if an arbitrary open covering of M can be exhausted to a finite covering) • Then we consider a family of functions (WiliEr on M,
1/\ :
M~
R
We sfI¥ that such a family is ZceaUy finit6, when at an arbitrary point P E M only a finite number of the functions Wi have a non-zero value. Observe that this property guarantees tllat the sum
406
II
l: 1/J.
iEI
1
is well-defined in a completely trivial sense, even if the index set I is uncountable. (At each point P the sum reduces effectively to a finite sum). Consider once more a family of functions (1/Ji)iEI on our toplological space M. We say that it is a partition of unity if it is a locally finite family of non-negative functions such that l: 1/J. = 1 iEI 1 {i.e. for any point
P we have
L1/J.{P)
i
= 1)
1
Now we can filially introduce paracornpaat spaces. As you might expect,they are characterized by a suitable property of their open coverings: A topoboeioaZ spaae M is paraaompaat if we to an a:t'bi trary open covering {O i \EI Clan find a partition s~ah that 1/J i vanishes outside 0i' (We say that such a partition of unity is subordinate to the open covering. Observe that the partition of unity and the open covering have the same index set I). To get used to the concepts involved, you should try to solve the following useful exercise:
of unity {1/Ji)iEI
Exeraise 8.3.1 Introduction:
Let (
be a partition of unity subordinate to the open covering
(Ui)iEI and let (1/Jj)jEJ be a partition of unity subordinate to the open covering (OJ) jEJ . Problem: a) Show that (Uinoj)(i,j)EIxJ is an open covering. b) Show that (
open covering
Jh,J
.
(uinoj)(i,j)EIxJ
Now what have all this to do with Euclidian manifolds? Well it turns out that they satisty the following important theorem:
Theorem 3 An Euclidian manifold M is paracompact, i.e. to every open covering we can find a corresponding partition of unity. (For a discussion of theorem 3 including a proof, see Spivak [1970]). Okay, let us use this heavy artillery: Consider an atlas int
n.
Then
int
Q
is a submanif91d and
[
(
an open covering. Thus we
can find a partition of unity (1/Ji)iEI subordinate to this open covering. We use it to cut F into the small pieces: Fi = 1/J i F Here 1/J i F vanishes completely outside the coordinate system
Thus We can finally define:
~(1/JiF) 1
on
int Q
407
Definition 4
II
def.
(8.9)
This is a legal definition provided we can show that the number
fa
~
~i F
is
1.
independent of the atlas
($i,Ui)iEI we choose to cover
int
a
and also independent
of the partition of unity. This requires lengthy but trivial calculations, where We make extensively use of exercise 8.3.1. For details see Spirak [1970]. So now we can in principle calculate the integral of a differential form over an arbitrary regular domain.
Suppose that we want to integrate a differential form over a surface which is unbounded
•
Provided the surface is sufficiently nice this may
still be done in the usual way. Sufficiently nice means that 0 (in the toporcgicar sense). and that arr points in
property int 0
is closed
have either
or property 2 from definition 2. section 8.2. Then we
can still divide If
0
a
in two disjoint submanifolds: int 0
and
ao.
is a simple manifold.we can choose coordinates
(AI, ..•• Ak) which cover
int O.
We can then formally write down the
integral (8.10)
If the integrand
F(el, •.•• ekl falls of sufficiently rapid at "infinity".
the Riemann integral will converge and everything is nice and beautiful.
This is especially the case. if
has compact support. i.e.
F
M.
vanishes outside a compact subset of
F
Otherwise the integral is
divergent. If
int 0
is not simple, we must again use a partition of unity
can be chosen so that
~i
F has compact support on
int 0
is trivially well-defined and the complete integral provided
is absolutely convergent, i.e.
(~i)iEI' This
Therefore the integral
f F a
is well-defined
L f~· F i 0 1
LIN· FI i a 1.
<
CD
•
In that case the value of the sum is
independent of which partition of unity we use, and we can therefore define:
Jl
=
~ JO~i
F
We can also use the integral concept on non-regular compact domains. Consider e.g. the unit cube in
R3: 0
=
{(x.y.z) 10~x~1, o~y~l. O
The interior of the cube consists of all points with
O<x
O
408
II
Notice that the points with property (2) in exhaust
We still have the points on
(j(l
the edges, which form the sceleton of the cube. The interior is a nice submanifold, but the boundary is not a submanifold. The
oQ
integral over
is therefore apriori
oQ
not well defined. However,
is "smooth
piece by piece", and the edges are of a lower dimensionality than the faces. Therefore we can neglect their contribution to the integral. We can then define
JaQT = If
Fig. 147
i=l
is a regular domain, then it is composed of two parts:
Q
int Q
x
6 1:
and
Let us concentrate on the interior for a moment.
aQ.
It is a k-dimensional submanifold.
To each point
P
on
int Q
we
have attached a k-dimensional tangent space T (int Q). If we introduce coordinates (AI, .•. ,A k ) on int Q, they generaie a canonical frame
(~l""'~k) for the tangent space
Tp(int Q). Clearly
Tp(int Q) is a
k-dimensional subspace of the n-dimensional tangent space Tp(M). I f (xl, ..• ,xn ) denote coordinates on M, then they generate a canonical frame
-t-
-t-
(1 , ..• ,1 ) for the full 1 n
tangent space
T (M).
The
-+- p
-+-
frame (el, .•• ,e ) is related k -t-tto the frame (1 1 , ..• ,1 ) in n the following way -+-
(8.6) Now, if
F
F
1
a Hi
is a differential
form of degree
M, then
a
21L
-t-
ei
k
is a multilinear map on each of the tangent spaces Tp(M).
But we may also consider simply because spaces
When
F
Tp lint Q)
restriction of
same symbol F
Fig. 148
defined on
F
F
as a differential form defined on
int Q
defines a multilinear map on each of the tangent ~
Tp(M). This new differential form is called the to
int Q and it is convenient to denote it by the
F
is considered to be a differential form on
characterized by its components
M, it is
409 and we
may
II
decompose it in the following way
(7.8) In a similar way the restriction of
to
F
int Q
can be character-
ized by the new components
F(~. ;••• ;~. ); 1 < i. < k 1.1 1. J k
and we can decompose the restriction of
as
F
(8.11) since it is a k-form on a k-dimensional manifold.
(Compare exercise
7.2.3). This is very useful when you want to integrate
F!
By
definition: (8.12)
J
=
/
+
J/ ( e l ; •••
+ 1 k def. ;ek)dA A••• AdA =
+
+
J/ ( e l ; .•. ;ek)dA
1
..• dA
k
Thus the integration is performed simply by replacing dAIA •.. AdA k by the ordinary Riemannian volume element dAl ••• dAk! Observe, that dAIA ••. AdA k itself can be considered as a volume element in each of the tangent spaces
Tp(int Q) • (Compare the discussion in section 7.4).
Consequently the geometrical integral,
JQF (e
+
1
+
l; ••• ;e k)dA A•.. AdA
k
,
may be considered as a generalization of the Riemann integral defined for ordinary continuous functions on ordinary Euclidean spaces! To use this point of view in practical calculations we must have a method to find the restriction of
F(tl
is of course to compute
F
to
, .•• '~k)
int Q.
One way of doing i t
directly using the formula:
But in many applications it is preferable to use the decomposition
F
1
= k!
F
dx
a l •• .ak
al
A ••• Adx
ak
as a starting point. The question is then: of dx i to int Q ?
What is the restriction
E:r:ercise 8.3.2 Problem:
(a)
Show that the restriction of dx i •
(8.13)
(l
i
dx~~·...e... a (lA
dA a
to
int Q is given by
410 (b)
II
Insert the result obtained in (a) in the decomposition (7.8) of F, and show once again,by rearranging the formula obtained,that the restriction of F to int Q is given by -+- d 1 (8 .11) F = F( -+el, ••• ,e ) A A••• AdA k k
The calculations are especially easy if you use an adapted coordinate (xl, .•• ,xk , ... ,x n ) parametrizes M and (xl, •.. ,x k ) parametrizes int Q. Then the restrictions of dxl, ••• ,dxk , ..• ,dxn system, i.e.
k and int Q are simply given by: dx l = dx l , .•. ,dxk = dx k l n O. dx + = O, ••• ,dx So if you want to find the restriction of to
1. F to
k.' F
a l ·· .a k
dx
al
A ••• Adx
ak
int Q , you simply remove all terms containing
dxk+l, ... ,dx n
We conclude with one more remark on how to calculate the integral in practice. int Q
Suppose
is a nice orientable regular domain, where
Q
is not a simple manifold.
single coordinate system.
Then
int Q
cannot be covered by a
In that case we ought to use overlapping
coordinate systems, find a partition of unity etc. This is. very technical and very complicated. using our common sense!
In praxis we will compute the integral
Consider, for instance, the case,
Q
=
s2,
which is its own interior, i.e. int Q Thus
int Q
S2 .
is not simple.
But let us introduce spherical coordinates anyway!
Then we
exclude an arc joining the North Pole and the South Pole.
But it
constitutes a subset of lower dimensionality and therefore it does not contribute to the integral anyway. Thus we
x
comp~te
the integral in the following'
Fig. 149
simple way: Tl
2Tl
J F J J F(~8;~~)d8d~ Q
8=0
~=o
Worked exercise 8.3.3 Problem:
(a)
Let Q = S2 be the northern hemisphere of the unit sphere in Show that +
R3.
411 (b)
Let
Q = B3
II
be the closed unit ball in
Jdx
A
dy
A
dz
=
R3.
Show that
1
Tl
B3
8.4 ELEMENTARY PROPERTIES OF THE INTEGRAL The integral has, of course, several simple algebraic properties. It is obviously linear:
JQ (F+G) =
(8.14)
JQF + JQG
Now we have been very careful in our discussion of regular domains. To each orientable k-dimensional regular domain an orientable (k-l)-dimensional regular domain
Q aQ
we have attached Suppose
differential form of degree k-l, then the exterior derivative a differential form of degree
k.
F
is a
dF
is
Thus we can form two integrals:
and The question is then: What is the connection between these two integrals? Worked exercise 8.4.1 Problem:
R3. and let B be an arbitrary smooth
Let Q be the unit cube in two-form. Show that
f
Q
dB
f
aQ
B
It follows from exercise 8.4.1 that the two integrals are always identical when theorem: Theorem 4 Let
Q
Q
is a cube.
This is a special case of the famous
(Stokes' theorem)
be an orientable k-dimensional regular domain and let
differential form Of degree k-1,
(8.15)
J
be a
F
then
dF
Q
For a complete proof see Spivak [1970], but the formula can also be derived from the following naive argument.
The regular domain
be apprOXimated by a family of cubes: (Qi)i€I' cf. fig. 150
Q
can
412
II
The boundary of the cubes consists of
an
two kinds of faces: a)
The face lies at the boundary of
b)
The face lies in the interior of
v
n. n.
In case (b) the face always coincides
1
with a face from the neighbouring cube,
11'""\
n
and their contributions to the integral therefore automatically cancel. When we perform the integration, we can Fi
n
g.:l
with the boundaries of the cubes:
In. dF = ~ Ian. F ~ Ian F
fdF ~ ~
~
n
~
~
If the manifold
0 1"-./ 0 /\
but they will have opposite orientations
therefore replace the boundary of
i\
\......./
J
--
~
M is equipped with a metric, we obtain the
following important corollary: Theorem 5 (Corollary to Stokes' theorem)
n
Let
be an orientable (n-k+l)-dimensional regular domain and let
F
be a smooth k-form. then (_l)n-k+l
Ian *
F
Proof: From (7.62) we learn that
*
6 F
(_l)n-k+l d
*
F
and theorem 5 is now a trivial consequence of Stokes' theorem. Finally we have the rule of integration by parts.
0
We have
generalized the Leibniz rule in the following way (7.17)
F
where
n
be a
and
G are differential forms of degree
k
(k+£+l) -dimensional orientable regular domain.
In d
Ind(F A G)
FAG
+ (_l)k
and
£
Let
Then
In F A d G .
But we may transform the left hand side using Stokes' thaorem whereby we get
I
aQ
FAG
I
n
d FAG + (_l)k
I FAd G Q
This we can rearrange in the following way:
II
413
(Theorem of integration by parts)
Theorem 6
IanF A G -
(_l)k In F A dG
This, of course, is a direct generalization of the familiar rule: b
I
b
df g()d x x = [f(x)g(x»)b - If(X) Qg dx dx a dx
a
a
In many applications we will be able to throwaway the boundary term, either because
Q
= 0),
has no boundary (Le. an
forms vanishes on
or because one of the differential
an!
Let us discuss this new integral concept in a few simple cases: Consider first scaZar fields. represented as a zero-form. mensional regular domain.
A scalar field,
~
: M
~
R,
can be
It can then be integrated over a O-diBut a O-dimensional regular domain consists
simply of a point (or a finite collection of points). So in this case the integral reduces to
I
4>
4> (P)
P
i.e. it just reproduces the value of the function
4>
This is the
most trivial and most uninteresting case. A scalar field
4>
can also be represented as the n-form
(7.39)
M
But then it can be integrated over n-dimensional domains including itself.
Usually such a domain n can be parametrized by the coordinates (xl, ..• ,x n ) on M itself and the integral reduces to (8.18) where
In
U
*
4>
J
n
4>
;,g
dx
l
A ••• A dx
n
Iu
is the coordinate domain corresponding to
4>(X);,g dxl ••• dx n.
n
This shows
that we can generalize the ordinary Riemann integral in Euclidian space l n I 4>(X)dX ••• dX , n which is valid only for Cartesian coordinates, into the covariant expression
which is valid in arbitrary coordinates.
All you have to do is to
replace the Cartesian volume element with the covariant volume element ;,g dxl ••• dx n • When we integrate a scalar field 4> on a manifold M
414
g
with metric
II
we always use the covariant integral (8.18).
JQ
=
$
As an application we can choose
JQIcJ dxl ••. dx
£
then represents the volume of
Q
1.
The integral
n
since
E
= ;.g d
xlA .•• Ad xn
is the
volume element in the tangent space, (compare the discussion in Section 7.4) (8.19)
Vol [Q]
=
J ;.g dxl .•. dxn Q
Here
Q
is restricted to bean n-dimensional domain, but the formula
(8.19) can easily be generalized to cover the "volume" of arbitrary k-dimensional domains. i.e. the length of a curve, the area of a twodimensional surface etc.
To see how this can be done we consider an
orientable k-dimensional regular domain fold way
The metric
M Q
on
Q
in our n-dimensional mani-
induces a metric on
M
becomes a k-dimensional manifold with metric.
be a parametrization of
and in this Let p,l, •••• Ak )
The intrinsic coordinates (Al •..• ,Ak)
Q. -+-
-+-
generates a canonical frame (el •••• ,e ) in each of the tangent spaces k (compare figure 148). The induced metric is characterized by the intrinsic components
g(~i;~j) The k-dimensional volume of
Q
is therefore given by
)
(8.20)
Volk[Q]
=
JI Q
1
-+--+-
Det[g(e.;e.)] dA •.• dA ~
k
J
This can also be reexpressed in terms of the extrinsic components of
g:
The volume formula is then rearranged as
JQ/ Det [g ab
(8.21)
a
b
ox ---j ox ] dA 1 ••• dA k ---. oA~ 01..
E:x:ercise 8.4.2 Consider a curve r connecting two points P and Q Show that reproduces the usual formula (6.40) for the length of a curve.
Problem:
Next we consider vector fields. represent
t..
by a one-form
A.
dimensional domain, i.e. a curve
r
with the parameter
Let
t..
be a vector field.
VOll[r]
We can
It can then be integrated over a one-
r
If we parametrizes this curve
A, (cf. figure 151), the integral reduces to
415
(8.22)
A2 r J A=J
r
I
< A
AI
Observe that in a
~ > dA
=
JA2
II
d~i
Ao ---d~ dA
"I
l.
1\
Euclidian
R
space
Q ->-
->-
dr dA
A·-
p
and the integral (8.22) therefore generalizes the usual line integral
12 .. Fig .151
We shall also refer to Suppose
A
fA
as a tine-integraL
r
$
is generated by a scalar field
i.e.
d$ ,
A
then Stokes' theorem gives
Jr d$ The boundary
ar
= Jar $
consists of the points
negative oriented and
Q
P
Q, where
and
is positive oriented.
P
is
By convention we
therefore put
Jar
$
= $ (Q)
-
$ (P)
Stokes' theorem consequently produces the formula (8.23)
which generalizes the usual theorem of line-integrals
(sec. l.l}It
does not, however, use the full strength of Stokes' theorem. If we use that
Jr
d$
=J
~
A i 2 dx AI axl. dA
dA
=
J"2 AI
~
dA
1\
we irrmediately see that it is a simple reformulation of the fundamental theorem of Calculus. Worked ezercise 8.4.3 Problem:
We use the notation of exercise 7.6.9. Especially manifold in R2, which we identifY with C, (a) Let r be a curve in M Show that
M is a two-dimensional
Jrh(z)dz = frh(z)dz where the integral on the right hand side is the usual complex line-integral.
416
II
(b) Consider an analytic (holomorphic) function curve connecting A and B . Show that
h(z).
Let
r
be a smooth
Ih'(z)dz = h(B) - h(A)
r
(c) Consider a two-dimensional regular domain Q bounded by the closed curve r . Let h(z) be an analytic (holomorphic) function in M. Prove the following version of Cauchy's integral theorem: fh(z)dz=O
r
A
A vector field
can also be represented by a (n-l)-form
Then it can be integrated over a (n-l)-dimensional domain. dimensional surface
Q
is usually called a hypersurface.
If
an orientable hypersurface then it allows a normal vector field
i.e. at each point
P
to the tangent space
on
Q
the normal vector
Tp(Q).
Let us parametrize the surface Q with the parameters (AI, ••• ,A n - l ). They generate the canonical frame vectors -+-+(e l , ••• ,e n _ l ). Furthermore we let ti denote the normal
...
ti(P)
Q
n
is I
is orthogonal
r
...
vector n = elx ••• xen_l (Compare the discussion in
Fig. 152
section 7.4). According to exercise 7.5.4
*A (-+-el,···,e-+-) ~ = A' n_l
*A.
A (n-l)-
we have
(-+-+-) elx ••• xen_l
The integral of *A therefore reduces to
fQ* A -- fu*
(8.24)
Thus
I* A Q
-+-) ,1 d n-l -A (-+el,···,e n _ l dA ••• A
fuA7,-+-n d,lA •.• dA,n-l
simply represents the fZuz of the vector field
the hypersurface
Q
A
through
For this reason we refer to the integral of
* A
as a fluz-integral. Having introduced flux-integrals we can now explain why theorem 4 is called Stokes' theorem. If S is a surface in R3 bounded by the curve r theorem 4 gives us fs dA = Ir A The right hand side is the line integral of A along
r.
We also have
d A = *(ilxA) (Compare the discussion in section 7.7). The left hand side therefore represents the flux of the curl ilxA. Consequently theorem 4 generalizes the classical Stokes' theorem from the conventional vector calculus, cf. (1.16).
417
n
Let
be an n-dimensional regular domain bounded by the hyper
an
surface
II
A
If
divergence of
-+-
A .
is a vactorfield, then
&A
-
represents the
We can integrate this divergence over
and get
the volume integral
According to the corollary to Stokes' theorem this can be rearranged as
=
(8.25)
(_l)n-l J
on
*
A
up to a sign we have therefore shown that the integraL of the divergence is equal to the flux through the Qoundary. This is the gene-
ralization of Gauss' theorem to an arbitrary manifold.
We can also
work out the corresponding covariant coordinate expression.
Using
(8.24) and (7.64) we get
(8.18) ,
a. (/g Ai)dnx
(8.26)
Jn
~
where the normalvector (8.27)
n.~
...
n
= Ig
= JAin. on
~
dAl •.. dA n - l
is characterized by the covariant components
e:. .
~Jl'"
j
n-l
We can also reproduce the integral theorems from the two-dimensional vector calculus. Consider for instance a region n bounded by a smooth curve r (see figure 141). Then it is well-known that the area of n is given by the line-integral
(8.2 8) Area (Here to
;
*;
'f
-+- -+- . [n) = ~ r*roar
r
is the vector orthogonal
and with the same length. Remember
also that
x
~*;oAt represents the area of
the triangle spanned by ; and 1+6;). It is convenient to write out this formula in Cartesian coordinates
(8.29) Area
[n)
= ~Jxdy-ydx
Fig. 153
To deduce this formula we use that dx. A dy is the "volume form" in R2. Consequently Area [n] i.e. Area [n) An application of Stokes' theorem now gives
Area [n)
= ~Jrxdy
- ydx
which we immediately recognize as being equivalent to (8.29).
418
II
We conclude this section with a discussion of simple forms, where we can give a naive interpretation of the integral. For simplicity we consider the threedimensional manifold M = R3. If we choose a coordinate system on M, then the coordinates (Xl, x 2 , x 3 ) generates several simple forms. for instance: dxl.
dx l
A
dx 2
•
dx l
A
dx 2
A
dx 3
•
dxl.
Consider first the basic one-form Q. (See figure 154).
Let
r be a smooth curve from P to
p
a Fig. 154
Ir
We want to interpret the integral [a;a+ll. [a+l;a+2l, ••.
xl
We divide [a,bl in unit intervals
Then the integral is roughly given by:
Jrdxl The coordinate
dx l
"" L
< dxll
i
"tli
>
generates a stratification characterized by the surfaces: xl = -1. xl
O. xl = +1, .•.
According to our analysis in section 7.2 , the number
<
dxll ... e i > l
can be inter-
preted as the number of surfaces intersected by the curve segment corresponding to the i'th unit interval. Therefore the total integral can be naively interpreted as the number of surfae~8 intersected by r. This can be demonstrated rigorously using Stokes' theorem. It immediately gives us
JIl
dxl
= J xl = xl(Q) or
- xl(p)
and the number xl(Q) - xl(P) cle~ly represents the number of surfaces intersected by r. Observe that if r is a ciosed curve, then each surface is intersected twice from opposite directions. Therefore
fr
dx
l
=0
This is also in accordance with Stokes' theorem
Next we consider the basic form dx l A dx 2 . (See figure 155).
Let
Il
be a smooth surface in
M.
419
A2
II
U
AI
Fig. 155
x
1
We want to interpret the integral
III We divide
U into unit squares.
I
.bel A
dx'
Then the integral is roughly given by:
dxl
dx 2 "" 1: "-A ....1
A
i
Il
A
dx2 ( +e i l'. +e i 2 )
The coordinates (x l ,x2 )
generate a honeycomb structure in M According to our I 2( +i +i ) analysis in section 7.2 the number dx A dx e ; e can be interpreted as the 2 l total number of tubes intersected by the surface segment corresponding to the i'th square. Therefore the integral can be naively interpreted as the number of tubes intersected by Il. This can also be justified more rigorously by the following argument: For simplicity we assume that we can use (x l ,x2 ) as adapted coordinates on Il, i.e. Il is parametrized on the form:
x3
= x 3 (x l ,x2 ).
III dxl
A dx
2
=
Then we get
Iu dxl ax2
U is an open subset of the xl -x2-Plane. Tubes in M correspond to unit cells in the x l -x2-plane. The number of tubes intersected by Il is equal to the number of unit cells contained in U. On the other hand f ax l ax2 U is equal to the area of U and this number also represents the number of unit cells contained in U. If Il is a closed surface, then
Here the coordinate domain
In dxl
A
dx
2
=
a
because each tube is intersected twice with opposite orientations. This illustrates Stokes' theorem for the basic form
dx l A dx 2 ,:
o
II
420 Exercise 8.4.4 Problem:
Let
Q
= S!
be the northern hemisphere of the unitsphere in
that the "number" of tubes intersected by
S!
is
R3.
Show
Compare this with exercise
Tl.
8.3.3. Exercise 8 . .4.5 Problem:
2 Consider the basic three-form dxl A dx A dx3 The coordinate functions xl,x2 and x 3 generate a cell-structure in M. Let Q be a
3-dimensional regular domain in
f
M.
Show that
2 l 3 dx A dx A dx
Q
can be interpreted as the number of cells contained in
8.5
THE HILBERT PRODUCT OF TWO DIFFERENTIAL FORMS Consider two differential forms
We have previously (sec. 7.5) T
Q.
and
U, cf. 1
T
U
and
of the same degree
k
introduced the scalar product between
(7.52-53):
(Tlu)= k: T
i l ··· i k
~*(*T
U..
~l·· ·~k
A
U)
This is the relevant scalar when you look at a specific point
P,
i.e.
when you want to introduce a metric in the finite dimensional vector space
k Ap(M).
Globally the differential forms of degree dimensional vector space
Ak(M)
k
span out an infinite
and we would like to associate an
inner product with this space.
To motivate it we consider real-valued smooth functions on a closed interval [a;b].
Here we have the natural inner product: < fig>
.fbf(X) g(x)dx , a
=f
*f A g
[a,b]
In analogy with this we consider the n-form (7.50)
*T
1
A U = -- T
i l ·· .i k
U
il ••• i k
k:
,
If it is integrable, we define the inner product in the following way: def. 1 i l ·· .ik 1 n A U = --k' T U . . Ig dx ••• dx (8.30) < T I u >
f
M •
~l·· ·~k
We shall refer to this inner product as the Hilbert product. compact, then the Hilbert product is always well-defined. If compact, the integral is not necessarily convergent.
If M is M is not
II
421
It is, however, always well defined if one of the differential forms has compact support. Consider a Riemannian manifold. with compact support.
Let Ak(M) denote the vector space of k-forms o The Hilbert product defines a positive definite metric in
this infinite dimensional vector space.
Thus Ak(M) o complete it and obtain a conventional Hilbert space
is a pre-Hilbert space. We can
~(M) called the Hilbert space of square-integra?le k-forms •. An element of the form 1 l.l l.k T = k' T. • (x) d x A ••• A d x • l.l· .. l.k
~(M) is on
with measurable components
is convergent. For a pseudo-Riemannian manifold things are slightly different. Here the scalar product (Tlu) is indefinite and therefore the Hilbert product is indefinite too. If we complete
Ak(M) o
we therefore get a Hilbert space with indefinite metric.
In the following we shall always assume that all the differential forms we are considering have a well defined Hilbert product.
Observe
first that up to a sign the dual map is a unitary operator, i.e. it preserves the inner product between two differential forms:
Theorem ? (a) The dua'l map * is a unitary operator on a Riemannian manifo'ld: <* T I * u > = < T I u > (b) The dual map * is an anti-unitary operator on a manifold with Minkowski metric: <* T I * u > = -< T I u > Proof: This follows immediately from the corresponding local property (Theorem 9, section 7.5): < *T
I
*u > =
t
(*T
I
±
*u)'
Then we finally arrive at a most important relationship between the exterior derivative -form, U
d
and the codifferential
5.
a k-form and consider the inner product: <
U
I
dT >
=
I*U
A dT
M
This can be rearranged using a "partial integration":
Let
T
be a
(k-l)
422 d(*UAT)
=
II
d*U A T + (_l)n-k *U A dT
From exercise (7.6.1) we know that
d * U
=
(_l)n-k+l * IS U •
Furthermore
JM d(*u either because
M
A
T) =
JaM
*U
A
T = 0
is compact without boundary or because
"vanishes sufficiently fast at infinity".
J * U A dT = J *
(8.31)
& U A T
i .e .
Thus we see that due to our sign convention, 6 operator of
d.
*U A T
Therefore we obtain: < UI d T > = <
& U IT>
is simply the adjoint
It should be emphasized that this holds both for
Riemannian manifolds and manifolds with a Minkowski metric. This has important consequences for the Laplacian operator.
M be a Riemannian manifold.
Let
Using (8.31) we can rearrange the
Laplacian as
...
where
&'
denotes the adjoint operator.
B~rmitian operator.
positivQ k-forms.
Consequently
-A
is a
The above argument applies to arbitrary
In the special case of scalarfields it is well-known from
elementary quantum mechanics, where the Hamiltonian, H
=-
i'J2
2mA ,is a
positive Hermitian operator reflecting the positivity of energy! We can also use (8.31) to deduce an important property of harmonic forms on a compact manifold.
Theorem 8 Let
M
be a compact Riemannian manifold.
A k-form
T
is harmonic
i f and only if it is primitively harmonic, i.e. A T
=0
iff;d T
= &T =
0
Proof: Let us first observe that <-ATIr> =
T
is harmonic, i.e. AT
0, then the left-hand-side vanishes
automatically. But the right-hand-side consists of two non-negative terms:
Hence they must both vanish: < & T I & T > =
II
423
But as the inner product is positive definite, this implies
&T
=d
T
0
=0
Consider for instance a zero-form, Ii
q,
vanishes automatically and
i. e. a smooth function,
:M~R.
constant. Consequently
T
Then
vanishes if and only if it is
d
This
Liouvillestheorem in complex analysis.
M is not compact the above consideration breaks down (unless
itself has compact support) . Then you should be more careful! E.g. in
q,
potential theory we are interested in the electrostatic potential
E
generating a static electric field
q,
reduces to the Laplace equation
A
The equations of motion for (cf. (1.19»:
= 0
in space-regions where there are no electrical sources.
A typical
problem starts with a space-region, which we represent as a three-
n
dimensional regular domain
q,
potential
n.
the interior of more since
=R3.
on the boundary of
We have been given an electrostatic
n
and we want to reconstruct
But observe that int
n
is not compact!
q,
in
Further-
does not necessarily vanish on the boundary we cannot
neglect the boundary term. We will not go into any details, but the starting point is Green's identities which you can have fun working out in the following exercise:
worked exercise 8.5.1 Problem:
Let n be an n-dimensional orientable domain in an n-dimensional Riemannian manifold n. Deduce Green's identities:
(8.32)
(1)
<
(2)
<
(8.33)
dq,1 d~>
<¢IA~>
+
fan*[¢d~]
= (_l)n-l
Use Green's identity to show the following property of harmonic functions: If
q,
=0
on
on
then
¢
=0
in
n.
We can also investigate singular solutions to the Laplace equation
in
R3.
For instance we know that
q,(i) = 1r
is a harmonic function
(generating the Coulomb field!) But it is not primitively harmonic, as it is not constant.
Thus it "violates" theorem 7. This is no disaster M R3 ,{O} which is not compact.
since it is only well defined on observe also that Hilbert product
q,
and
d¢
=
are not square-integrable on
M. The
424 <
EIE >
=
J
!4 dxdydz
II
= 4TI
R3 ,{O}
J~ ~ 0
diverges at the origin, reflecting the infinite self energy of a point-charge!
If you try to repair the proof of theorem 8 through a
regularization procedure, i.e. put
then you should observe that the "partial integration" leading from
I
I
<& d ~ ~> to
is no longer valid because the differential
forms no longer vanish "sufficiently fast".
Worked exercise 8.5.2 Problem:
be a smooth scalar field on M£
~
Let (a)
Show that = - < d ~ Id P
+
f
~
*d ~
dM£ Put
~(;t)
< d ~ Id
= 1.r
. Show by explicit computation that
=J
P
~
* d ~ = 4TI
dM£ Whenever you work on a non-compact manifold, you should
Moral:
look for "boundary" contributions.
8.6 THE LAGRANGIAN
FOR~ALISM
AND THE EXTERIOR CALCULUS
We will now investigate the Lagrangian formaLism
as an important
example of how to apply the integral formalism. In a field theory the
equat~ons
principle of least action.
L, which is a scalar field:
Lagrangian density S
=
of motion are determined by the
The action itself is constructed from a
JQLE = JUL(x)i=g dxo dx l dx 2dx 3
This scalar field
is constructed from the fields and their
L
derivatives. In the simplest case of a scalar field function of
~
a~~.
and
~
it is thus a
Using the covariant action we can now
construct the covariant equations Of motion:
They are obtained using
the now familiar variational technique: We replace the scalar field ~
by
~+£~,
where
~
vanishes on the boundary of
Q.
425
II
This generates the new action
S(E)
= JU L(~+E~,
0
~+E3 W);:g
].l].l
which is extremal when
°
~~IE=O
=
JU[(;:g
J (~~
=
(;:g
+
i.e.
o(~L~)o].lW);:g ].l
W +
U
~~)lji
= 0,
E
4 d x
d
d
4
x
(~L ~) lOlllji Jd 4 X ].l
Using partial integration on the last term we get
o=
fJ
I-g
As
lji
is arbitrary, this is only consistent if:
~~
-
d].l
~~ = ?:g
(8.34)
~) ]
(/-g
d].l
(;:g
0
4 lji d x
(~\)) ].l
If we have several fields
~a'
then each of the components have to
satisfy the appropriate Euler equation!
Thus we have shown:
Theorem 9 The ~a
aovariant equations of motion for a aolleation of saalarfields
are given by
(8.35)
Although it is much more dubious,the
covariant equations of
motions (8.35) are actually valid for vector fields, interpret the index
a
too, when we
as a spacetime index!
Exeraise B.6.1 Problem: Consider the Lagrangian density for the electromagnetic field L
,,_! 4
F
F].l\!
].l\!
Re-derive the covariant Maxwell equations (7.85). The geometrical point of view can in fact be used to throw light on one of the more subtle points in the derivation of the Euler equations. To pass from
to
we used a "partial integration", but we neglected the boundary terms, because vanishes at the boundary", This is justified in the following exercise:
"1j.I
II
426 Worked exeraise 8.6.2 Problem:
a)
Show that
()L ~
are the contravariant
components of a vector field A.
II
b)
( 8.36)
Let
$
be a scalar field vanishing on the boundary of
Jn(~ A~)d ~
4 $ d x =
-Jn3~ ~ A~)$
n . Show that
4 d x
As we have seen in section 6.7 we can also write down the equations of motion for a free particle in a
covariant form.
Consider now the
case of an electrically charged particle moving in an electromagnetic field.
Combining (1.26) with (6.46) we are led to the following co-
variant equations of motion
(8.37 )
We would like to derive it from a Langrangian.
Now we have previously
determined the non-relativistic action for an electrically charged particle interacting with the electromagnetic field: (2.13)
S
This suggests that for a system consisting of an electrically charged particle and the electromagnetic field we should use the following reZativistia action
S Sp
= Sp
+ SI + SF
-mt/-g~s~(
with dx S dA dA
Ai A2 (8.38)
SI
q
SF
- 1-4
ll
J All dx w:- dA A1 Jn
Fll\!
Fll\!
;=g
d 4x
Here the dynamical variables to be varied consist of a)
The position of the particle: Xll(A) .
b)
The gauge potential: All(X).
427
II
We leave it as an exercise to the reader to verify that the invariance of the action (8.38) under these variations actually leads to the desired
covariant equations of motion!
worked exercise B.6.3 Problem:
a)
Show that the interaction term can be rearranged as (8.39)
8
r
= In All Jll
r-g
d
4 x
(Compare this with exercise 3.10.2). Show that the relativistic action (8.38) leads to the following equations of motion:
b)
d 2 x C!
C!
(8.37)
m d,2
(7.85)
-L a (Fg Fg v
= qF
ax~
FllV)
axll axV
C!
8 dT -
mfllV
a:r dT
Jll
It is also interesting to see that the Lagrangian formalism can be thrown into a purely geometrical form.
This means that we can discuss
the equations of motion completely without introducing a coordinate system! Of course. the first step consists in reexpressing the action in a purely geometrical form.
The KZein-Gordon field:
Let us look at some specific examples: ~,where
This is a scalar field
the
action is based upon the Lagrangian density (3.49). The
covariant action is then given by S
=
fnL
e:
= fnb(all~) (all~)
-
~ m2~2];:g
4 d x
which we rearrange as (8.39)
S
=-~
=
f-~ * d~Ad~ n
-
~m2 * ~
A
~
(Compare the discussion in section 8.5). This was the first step. Next we perform a variation ~ ~ ~
where
1}i
+ e:1}i
is a scalar field, which vanishes on the boundary of
n.
We then get S(e:) =-~
=-~
o
dS =-
428
II
( H ere we have used that 1jJ vanishes on the boundary to throwaway the boundary term coming from the partial integration). But this is only consistent if ~ satisfies the equation (8.40) -Sd~ = m2~ , which is the geometrical form of the Klein-Gordon equation: This is a one-form A , and the action is based The Max~eZZ fieZd: upon the Lagrangian density (3.50) which leads to a covariant action given by
This is rearranged as (8.41)
Then we perform a variation,
A where
S(g)
-~
-~
From
A+E:U
is a one-form, which vanishes on the boundary of
u
£:2
- €
this we immediately get dS o = cr---= -
€I€=o
( Here we have used that U vanishes the boundary of n to throw away the boundary term coming from the partial integration). But this is only consistent if (8.42 )
Le.
-OF
o
which is nothing but the Maxwell equations! Exercise 8.6.4 Problem:a)Consider the massive vector field by
A
Show that the action (3.51) is given
(8.43) b)Perform the variation tion
A ~ A+£:U
and deduce the following equations of mo-
(8.44) c)Show that they are equivalent to
(8.45)
cA= m2 A
-6A=O
We may summarize the preceding discussion in the following scheme:
429
FIELD (8.39)
~
(8.41)
Maxwell
( 8.42)
A
(8.43)
Massive vector
(8.44)
Equation of motion
Action
Klein-Gordon
(8.40)
II
S
= f~~(*d~Ad~)-~m2(*~A~)
S
=
S
=J
-Is, ~ (*dA) AdA
-6d~ = m2~
-odA
-~(*dAAdA)-~m2(*AAA) -odA
n
=
0
= m2 A
A
8.7 INTEGRAL CALCULUS AND ELECTROMAGNETISM As an other example of how to apply the integral calculus we will use it re-express in a geometrical form various electromagnetic quantities like the electric and magnetic flux through a surface and the electric charge contained in a 3-dimensional regular domain. We start out peacefully in 3-space to get some feeling for the new formalism. Remember that tially
B
E
is a I-form, but
B
is a 2-form. Essen-
is the dual of the conventional magnetic field. Now let
n
be a 3-dimensional regular domain. From the discussion of fluxintegrals (8.24) we get B
The magnetic flux through the closed surface
an .
(8.47)
Jan*E
The electric flux through the closed surface
an .
(8.48)
f;p
The electric charge contained in
(8.46)
fan
n
We can now use the integral calculus to deduce some wellknown elementary properties: Example 1 If n ~ke
c~ntains
no singularities, then the magnetic flux
an
closed surface
vanishes.
This follows from an application of Stoke's theorem:
~ (due to (7.76».
J an B = JndB =
0
0
~
through
430 Example 2
II
(Gauss' theorem)
The electric flux through the closed surface electric charge contained in
an
is
n.
This follows from an application of theorem 5 (Corollary to Stokes' theorem). From the Maxwell equation ;
o
x
[the charge]
=
(7.78)we get
£lJ *p 0
n
=
-J n*QE
Example 5:
This time we consider a static situation. Let face with boundary to
S
be a sur-
r . Then the flux of current through S is equal
£oc 2 times the circulation of the magnetic field along Observe first that the Maxwell equa-
r.
tion (7.79) reduces to &B
=
1 £oc2 J
for a static configuration. The proposition then follows from an application of theorem 5 (Corollary to Stokes' theorem) :
Fig. 156 We conclude the discussion of electromagnetism putation of two important integrals:
in 3-space with the explicit com-
a) Consider the spherically symmetric monopole field. Let 52 be the closed surface of a sphere with radius r . Then polar coordinates (r,8,~) are adapted to the sphere, and we can choose (8,~) to parametrize it! We Can now compute the magnetic flux through the closed surface 5 First we observe that the monopole field B is gi ven by (cf. (7. 89) ) : B
= t,;- 5in8d8AiAp
Then we immediately get: 2Tf
(8.49)
Tf
J52B = tL ~Of 8=0J 5in8d8d~ = g Tf
Fig. 157
431
II
b) This time we consider the magnetic field around a wire with current j. Let r be a circle around the wire with radius p. Then the cylinder coordinates (p,~,z) are adapted to the circle, and we can choose ~ to parametrize the circle. We can now evaluate the line integral of the magnetic field along the closed curve r. For this purpose we must find the dual form *B which is the one-form representing the magnetic field. This has been done previously (cf. section 7.8) *B -
-L
- 21TE: o c 2
z
dip -t
J
Thus we get •
2'TT
fr *B = 21T~ c 2 fd~ oo
x
•
=
E7 n
Fig. 158
0
Then we proceed to consider electromagnetism in Minkowski space, i.e. 4-dimensional space-time. Here it is more complicated to express suitable quantities, so we shall adopt the following terminology: Suppose we have chosen an inertial frame S. Let (xO,X 1 ,X 2 ,X 3 ) denote
suppressed~
One dimension
the corresponding inertial coordinates. We say that the three-dimensional submanifold
F,; is a space slice if it is a subset on the form F,; = {XEMlxo=tO}
i.e.
F,;
consists of all the spatial to . Ob-
points at a specific time
serve that the spatial coordinates (X 1 ,X 2 ,X 3 ) are adapted to F,; The fundamental quantities describing the properties of the electromagnetic field are the field strengths A , and the current
J . Now let
n
Fig. 159 F
and *F
,the Maxwell field
be a 3-dimensional regular domain
contained in a space slice relative to the inertial frame
S • Then we
can form the following integrals which we want to interpret:
Ian F
Ian*F
Observe first that the restriction of
fn*J
and dxo
to
n
vanishes. Conse-
quently the integrals involve only the space-components of the integrands. But (8.
50~
F
and
F
=
*F
are decomposed as
[*]
and
*F
[*J
432 So the restriction of (respectively striction to
F
II
(respectively
*F) to
an
is given by
B
*E). Similarly the dual current has the following reintn
*JI~
..
=
(*J)
123
dX ' Adx 2 Adx 3
= -JOdX ' Adx 2 Adx'
We can now generalize the results obtained in (8.46)-(8.48) to the following Lemma 1 Let
n
be a J-dimensional regular domain contained in a space slice.
Then (8.51)
JanF
r
(8.52)
Jan
*F
-In*J
(8. [, J)
The magnetic flux through The electric flux through
=
an an
The electric charge contained in
n
Exercise 8.7.1 Introduction: Let n be a. 3-dimensional regular domain obtained in a space slice and let F be smooth throughout n. Problem: Use the 4-dimensional integral formalism to re-examine the following well known results: a) The magnetic flux through an is zero. b) The electric flux through an is equal to the electric charge contained in n (As usual we have put EO = c = 1 ).
ILLUSTRATIVE EXAHPLE:
MAGNETIC STRINGS IN A SUPERCONDUCTOR
We have earlier been discussing some of the features of superconductivity, especially the flux quantization (see section 2.12). Recall that in the superconducting state of a metal, the electrons will generate Cooper pairs. These Cooper pairs act as bosons and we can'therefore characterize the superconduc~ing state by a macroscopic wave function
~
called the order parameter of the superconducting metal. The square of the order parameter,
1~12,
represents the density of the Cooper pairs.
We want now to study equilibrium states in a superconductor. Consider a static configuration ~(~), where ~(~) is a slowly varying spatial function. In the Ginzburg-Landau theory one assumes that the static energy density is given on the form: (8.54)
H
433
II
Here a is a temperature dependent constant (8.55)
a
=
aCT)
=
a ~c T
(with a positiv)
c
where Tc is the socalled critical temperature. The constant y is just inserted to normalize H to be zero at its global minima. The equilibrium states are found by minimizing the static energy. This leads to the Ginzburg-Landau equation (8.56)
Now consider the potential (8.57)
It has the well-known shape shown on figure 160. Above Tc the vacuum
T>T
c
T
c
Irn1jJ Re1jJ
Re1jJ
Fig. l60
configuration is given by 1jJ = 0 i.e. the metal is expected to be found in its normal state where the
Cooper pairs are absent. Below Tc the configuration 1jJ=O
becomes un-
stable and we get a degenerate vacuum, the superconducting vacuum, with a temperature dependent density of Cooper pairs given by a
- B
a T - T
-.~
B
c
All this is in good accordance with the experiment! 1jJ-axis
The coherence Zength:
Consider a specimen, a semiinfinite slab, bounded by the y-z-plane. For negative x we are in the normal region, where
1jJ=O , and for positive x we are
Nannal region
Superconduc-
ting region.
in the superconducting region. So for sufficiently large x we expect the order parameter to be in its vacuum state
x-axis Fig. 161
434 IljJl =
II
~ If
The coherence length, s(T), is the characteristic length it takes the order parameter to rise from its normal vacuum, ljJ=O, at the boundary to its superconducting vacuum, 11jJI=V-;, inside the slab. The problem is essentially one-dimensional so the Ginzburg-Landau equation (8.56) reduces to
with the boundary conditions ljJ(O) = 0
H:o)
r=i
= If
The solution to this problem is simply a "half-kink", cf. (4.23) if
x>O
if
x<:O
ljJ(x)
Consequently the coherence length is given by (8.58)
o
S{T)
Then we consider what happens when we put on a magnetic field B. The order parameter will now couple minimally to the magnetic vectorpotential, i.e. we must exchange the exterior derivative d with the gauge covariant exterior derivative D
=
d -
ifiA
(where A is the magnetic vector potential) • Exercise 8. 1'.2 Problem: Show that
(8.59) Notice that it follows from exercise 8.7.2 that the square of the gauge covariant exterior derivative does not vanish in general. When including a magnetic field the static energy density (8.54) is modified to (8.60)
H
435
II
Worked exercise 8. 7. J Problem: Show that the equations of equilibrium configurations are given by
(8.61a) (8.61b)
D*D1jJ
*
(Shjll 2 + a)ljJ
[$D1/J -
i.e.
E
ljJD1jJ]
i.e.
Let us now assume that the order parameter is in its ground state (characterized by a vanishing energy density), i.e. (8.62)
DljJ
=0
U(ljJ) = 0
USing exercise 8.7.2 we then immediately get (8.63)
o
i.e. either B vanishes or ljJ vanishes! In the normal state ljJ=O , so here the magnetic field can propagate freely, but in the superconduc-
r=a
ting vacuum IljJl=YIf ' and B has to vanish. Consequently the magnetic field is expelled from any region where the order parameter is in the superconducting vacuum! This, of course, is the famous Meissner effect. Observe too, that if the order parameter is in its vacuum state, then the super-current,given by (8.64) automatically vanishes.
ljJ-axis The penetration length:
We can now introduce the second characteristic length in superconductivity. Again we consider a semi-infinite slab bounded by the y-z-plane. We now apply an external magnetic field B and we want to find
/
out how far into the superconducting region x~~-ax~i~s~==~~~E=====§;~~~~~ ~ A(T) Fig. 162 that the magnetic field penetrates. This ~ time we use the Ginzburg-Landau equation
..-::
for the magnetic field:
In this equation we can approximate IljJl with its equilibrium
value~
This does not mean that the order parameter is in its superconducting vacuum since the gauge covariant derivative DljJ need not vanish. EspeCially we can still have super conducting currents floating around. Taking the exterior derivative on both sides we now get:
II
436
- d&B
=
~2
-flfl2B
Here we have used that Wis proportional to exp{i~} so that ~d~ is proportional to d~ , i.e. it is a closed form. Since furthermore dB vanishes on account of Maxwell's equations we arrive at the Laplace equation: (cf. 2.77) (8.65)
Again the problem is essentially one-dimensional and the Laplace equation therefore reduces to 2 .. d 2Bij - ( x ) =-~2 B~J (x)
dx 2
Since B is a closed form we furthermore have dB23
ax
o
This leads to the following solution: ; Bl2 (x) = Bl2 (O)exp[
-ljix1
Consequently the penetration length, A(T), is given by (cf. 2.78) (8.66)
t:Sfl A(T)=I/-'=-Ct
q
(Notice that the above considerations are strictly speaking only valid when tIT) « AIT) ). We have now introduced two characteristic lengths, the coherence length tIT) and the penetration length AIT), both of which depends on Ct and hence on the temperature. We can now form a third parameter, their ratio (8.67)
K =
li'!l HT)
=
fl Vj3 q
which is temperature independent and is called the Ginzburg-Landau parameter. Recall that there are two different types of superconductors, type I and type II. If we apply an external magnetic field, we know that the suoerconductivity is destroyed for sufficiently strong magnetic fields. In a type I superconductor there will be a critical field strength, Bc' above which superconductivity breaks down and the magnetic field is uniformly distributed throughout the metal. In a type II superconductor there will be two critical field strengths, BCland BC2 When we pass Bc
II
437
superconductivity will not be completely destroyed but rather the magnetic field will penetrate into the metal in form of thin magnetic strings, vortices.
First when we pass BC2 the superconducting regions
break down completely. We want now to investigate the structure of a magnetic string in a type II superconductor. We will assume that ;(T) «
A(T). Thus a single string
consists of a hard core of radius
; (T)
where the density of Cooper pairs vanishes. Inside this hard core we have therefore re-established the normal vacuum. Outside the hard core the magnetic field falls off and it vanishes essentially in a typical distance
A(T). Thus we have
also a soft core:
; < P < A . Observe
~ '< ; . ~ soft core
ACT)
that we have circulating supercurrents in the soft core which prevents the magnetic field from being spread out ( the Meissner effect ).
V
Finally we reach the
sUperconducting vacuum outside the soft
H
core, where neither the magnetic field
hard core
nor the order parameter contributes to
;(T)
11jJ1
.. Fig. 163
the energy. Remark: Consider a magnetic string with flux ~ and let us for simplicity assume that we have a constant magnetic field inside the gtring. Then
~o Bo'TTP2 where p is the radius of the string. Thus the magnetic energy stored per unit length is given by ~2 (8.68) J = ~'B02'TTp2 = --2 2'TTp2 Thus we can diminish the energy of a magnetic string with a constant flux by spreading out the string. If nothing prevents it a magnetic string will thus grow fat. In the case of a solenoid it is the mechanical wire, in which an electric current flows, which prevents the magnet~c field ~rom spreading out. In a superconductor it is the Meissner effect which prevents the magnetic field from spreading out.
Now let us try to compute the energy per unit length in a single vortex line. We shall neglect the hard core which by assumption is very thin. Thus we concentrate exclusively on the soft core. Let us denote the region outside the hard core with
n .
We start by rearranging the
expression for the static energy:
Hn = But
~
1jJ =~exp[i~]
+
~
+
iJ~I1jJ12
+
~}2
E
outside the hard core so we can neglect the pot en-
438
II
tial energy. Furthermore the supercurrent reduces to
&B)
(=
and the gauge covariant exterior derivative of the order parameter similarly reduces to DljJ = qiexp {iql}[
g~t,
~Al
by direct inspection, that
;
-
H
;A2(T)[
<&BI&B>n1
~A2(T)(Jn*BAd&B
*&BAoSBl =
+
~A2(T)Jn
U sing Stokes' theorem this is rearranged as
(8.69) Thus all we have got to know is the magnetic field and its derivative on the boundary of the hard core! To proceed we must determine the field configuration more accurately, i.e. we must in principle solve the Ginzburg-Landau equations (8.6la-b). It turns out to be impossible to write down an explicit solution representing a magnetic string, so we shall be contend with an approximative solution. We shall concentrate on a cylindrical symmetrical string. Notice that the flux is necessarily quantized (2.80) where n is related to the jump in the phase of the order parameter when we go once around the string~ For a string with n flux quanta we therefore use the ansatz: o lim ~(p) P -.. 0 (8.70a) ljJ(r)= ~ ~(p)exp[inqll with { lim
1
p -..
(Notice that the boundary condition at p=O, which is in accordance with our general description of a magnetic string as shown on fig. actually removes the singularity of exp[inqll at the origin). The exterior derivative of the order parameter is given by
439
d~
~{~ dp
=
+
II
ind~}
outside the soft core it is therefore approximately given by d~
...
1jJ{ind~}
Since the gauge covariant exterior derivative of the order parameter must vanish outside the soft core, the magnetic vector potential is asymptotically given by A ... n~d~ q
This suggests the following ansatz for the magnetic vector potential lim A(p) 0 with { p.... 0 (8.70b) lim A(p) 1 p .... '"
(The boundary condition at p=O removes the singularity of d~ at the origin) • We proceed to determine the static energy density for this type of configuration. Notice that (dpldp) = 1 ;
(dpld~)
= 0 ;
(d~ld~)
1
(dp"d~ldp"d~)
= ;i2 ;
1
;i2
(cf. the exercises 6.6.3 and 7.5.7). ISing that (8.7la)
D~
(8.7lb)
B
~{ ~ ~(p)dp +
in[l - A(p)ld~}
nfl A' (p)dp"d~ q
the expression (8.60) for the static energy reduces to H
=
;n2~~.(~')2 _ ;~.(~')2 _ ;~n2
(1
~2A)2 ~2 + ~~[~2_ lF
(Notice that ~ is negative:) As expected this only depends upon p so that it is manifestly cylindrical symmetric. It follows that the static energy per unit length is given by (8.72)
J[~,Al
2'TTfJ[
o 2Tff{~n2~~,<~')2- ~!:!p.(
o Next we functions is to plug (8.6la-b). *dp
=
q
p
8
8
48
p
want to find the equations of motion for the unspecified
;
*(dp"d~) = ±dz p
*d~
(cf. the exercises 6.6.3
and
;
* (dz"dp) = pd~
7.5.5). It follows from (8.7la-b) that
440
II
D*D1jJ
Thus we obtain the following equations of motion A"
)?~T)<jl~[l-Al
(8.73a)
p[pl
-
(8.73b)
(p<jl')'
-a[<jl2-l1<jl + n 2 (l-t) 2<jl p
There is, however, a quicker way to obtain these equations. We know that the cylindric symmetrical equilibrium configuration must extremize the energy functional J[<jl,A] given by (8.72). The corresponding EulerLagrange equations are given by
Thereby we recover the equations of motion (8.73a-b). One should however be very careful when using this strategy. It only works because in this particular case the ansatz (8.70a-b) is in fact the most general cylindrical symmetric ansatz. In general a variation among a restricted subset of configurations need not extremize the action against arbitrary variations. We must then show that the second order differential equation (8.73) possesses a solution with the appropriate boundary condition. This is a somewhat difficult task. uSing advanced analysis ( Sobolev space techniques) it can be shown that the energy functional J[<jl,A] , which is positive definite, has a smooth minimum configuration satisfying the appropriate boundary conditions. (For details see Jaffe and Taubes(1981») In praxis one is often contend with computer simulations. Notice that <jl (p) " 1 A(p) " 1 is a trivial solution to equation (8.73) allthough it breaks the boundary condition at the origin, As a consequence the corresponding string is singular at the origin. Neverthe the less the energy density vanishes outside the origin. It is called a vacuum texture and obviously represents an infinitely thin string carrying n flux quanta. Although we cannot solve the equation of motion explicitly we can easily determine the asymptotic behaviour. In the limit of large p the equations of motion simplify considerably. If we introduce the functions f (p)
= l-A(p) IP
and
g(p) = /-a:{l_<jl(P»)
it is easy to show that asymptotically they solve the differential
441
II
equations f "( p)
...
1
),2(T)f(P)
g"(p)
and
p .... '"
This implies the following asymptotic behavior f (p)
... p....'"
Cl exp [-
-rf.ri 1
and
in accordance with our previous discussion of the coherence and penetration length. Okay, at this point we return to our estimate (8.69) for the static energy. Using the equation of motion (8.73a) it follows that n2h2 1 *BASB = q2),2 (T) pA' (p) [l-A(p) J
1fn2~{:!:AI
J ...
q
p
(p) [l-A(p) ]cp2 (P)}I
p=i; (T)
To proceed further we notice that the equation of motion for A(p) does not contain n explicitly! (It does contain n implicitly through the function CP(p) .) If we impose the condition CP(p)=l we get a vacuum texture for the order parameter, but we can still retain the boundary conditions for A(p). The solution to the reduced equation of motion (8.75)
A'
p[;i]
I
1
'" - T2(T) [I-A]
with
lim p.... O
A(p)=O
and
lim
A(p)=l
p.... '"
is then likely to reproduce the vortex outside the hard core. In this approximation A(p) is completely independent of n. It follows that J is proportional to the square of n: (8.76) AS a consequence a vorte~ string with a mUltipZe fZu~ is unstable. E.g. a single vortex string with a double flux quantum has twice the energy of two widely separated vortex strings with unit fluxes. This shows that when the magnetic field penetrates into the superconducting region it will generate a uniformly distributed array of vortices all carrying a single flux quantum. Remark: Actually equation (8.15) can be solved explicitly using Bessel fUnctions. The expression (8.14) can then be estimated using the known asymptotic behaviour of the Bessel fUnctianin the limit of small p. In this way one obtains J n21~[K] ( 8.11) ...
q2),
(Tl
where K is the GinZburg-Landau parameter (8.61). For details see de Gennes (1966)
II
442
8,8
THE NM1BU STRING AND THE NIELSEN-OLESEN VORTEX
The soliton was introduced as a smooth extended version of a relativistic point particle. Now we will show that a similar interpretation is possible for a vortex string. first we must introduce a suitable generalization of the relativistic point particle known as a relativistic string or a Nambu string. .... A point particle is characterized by its position x at a the time t. In the four-dimensional space-time it sweeps out a time-like curve, the worZd line, cf. fig.164~ The action of free particle is particular simple being proportional to the arc length of the world line: dx].Jdx v S =-m V-g].JvdA dA dA
I/
I:B
Varying the world line we then obtain the equation of motion,which inertial frame reduces to 2 u XO x md 0 dT 2 XO
xO=t
Xl
0
Y""d
\,
eT
lino
x2
.... eo
in
an
~:~:"""
world sheet
~\Js
x2
~
Fig. l64a In a similar way a relativistic string will be characterized by its position ~(o) at the time t, where ~=*(o) parametrizes the spatial curve representing the string. It will be convenient to assume that the string is finite and that its endpoints correspond to 0=0 and o=rr In the four-dimensional space-time the string sweeps out a time-like sheet, the worZd sheet,cf. fig~64b. The action will be chosen to be proportional to the area of the sheet. If we parametrize the sheet by x].J = X].J(O;T) where 0 is a space-like and T is a time-like parameter, it follows that the induced metric on the sheet is characterized by the components h ........ dx].Jdx v ........ dx].Jdx v 00 = g (eo;e o ) = g].Jvdo do hOT = g (eO;e T ) = g].Jvdo dT ' etc. From the induced metric we then get the area element
443
II
(8.77)
We then define the action to be ." a=7T,--_ _ _ _ __ (8.78) S /(h a .)2_ haa h •• dad. '1 a=O
2;aJ J
FLom the action we get the equation of motion by varying the world sheet:
Notice that there are no restriction at the endpoints a=O and a=rr • Thus we get not only equations of motion for the string but also boundary conditions for the end points. The general covariant equations of motion are extremely complicated, so we shall restrict ourselves to an inertial frame where the metric coefficients reduce to
=
g].lV(X)
Tl].lV
Then the Lagrangian density depends only upon dx].l = x,].l and da We now obtain the displaced action
S(e:)
r
r~(~~].l
'1
+
a=O
We must then demand 3L a ].l
+ -].l2X ldad. ax d.
'2
As Y tion (8.79)
a=rr
J J {aa '1 a=O a
dL a~
+..£.. a.
aL } ].l d ax].l y ad.
+
is arbitrary this leads on the one hand to the equations of moa aL + d aL aa d~ a. dX].l
=
0
On the other hand it leads to the edge conditions: (8.80)
~~
=
la=O
;~~
= 0
la=7T
Let us introduce the following abbreviations for the conjugate momenta: I
444
II
(B.Bl)
Then the equations of motion supplied with the boundary conditions simplifies to (B.B2)
Observe that we also have the following identities at our disposal: (B.B3)
P ~·· TA~
=
p~x· a ~
= p2 + T
~2 = p2a
4n 2 a
+
~~
~
= 0
which are trivial consequnces of the explicit expressions for P~ and P~. Notice especially that• the edge conditions Lffiplies that P~=O at the eda ge. As a consequence x 2 vansihes at the edge, so the end points move with the speed of Zight. We may also introduce the four-momentum of the string: a=n
p~ =
(B.B4)
f
p~da
(T
fixed)
a=O
The four-momentum is conserved (as it ought to be) a=ndP~
f-
dT
a=7Tap~
Tda
a=O
--ada da J a=O
=
{p~(a=O) a
-
P~(a=n») a
o
Similarly we may introduce the angular momentum of the string: a=n
J~\l =
(8.B5)
f [x~P~
a=O Exel'oise 8.8.1 Problem: Show that the angular momentum is conserved, i.e. ~~\l ~ 0 dT .
Exeroise 8.8.2 Introduction: Problem:
(a)
Consider a rigid rotating otring parametrized by XO~T ; xl= A(a-!n)CoswT ; x 2= A(a-~n)SinwT ; x 3= 0 Show that it solves the equation of motion provided 1 = !AwT
(b)
(Hint: Show that this represents the edge condition) Show that the spinning string has the momentum pO = ~ pI: p2= p3= 0
(c)
Show that the spinning string has the angular momentum J12:
A2 n 2
16a
;
J23
= J31
= 0
445
II
from the above exercise it follows that a spinning string behaves like a particle with the rest mass M
=
J
=
Arr
4a
and an intrinsic spin
A2rr2 16a
aM
2
Notice especially that the spin grows ZinearZy with the square of the mass.
Now what is the relevance of such relativistic strings in high energy physics? If we plot baryons (i.e. strongly interacting particles) in a diagram with the mass square on one axis and the spin on the other J(spin)
axis, then the baryons with the same isospin, strangeness etc.
9!
fallon straight lines! These
8!
are the so-called Regge trajecto-
6l
ries. This remarkable property
5!
suggests that we consider the baryons to be somehow composed of relativistic strings. Okay, with this simple minded
b./
Regge trajectories
71
///A
4!
212
°
0/0
3l
n
/
° ./
~/
° ./ //
Fig. 165
/"
remark I hope to have convinced you that relativistic strings,
2
4
6
8
10 M2(GeV2)
Nambu strings, are very interes-
ting objects. As in the case of the point particles one would now be interested in smooth extended solutions to field theories that behave like thin strings.
ILLUSTRATIVE EXAMPLE: THE NIELSEN-OLESEN VORTEX. The first example of a string-like solution in a classical field theory was found by my respected teachers, Holger Bech Nielsen and Poul Olesen, in 1973.*) They considered a relativistic field theory based upon a complex charged scalar field coupled minimally to electromagnetism. In analogy with the
~4-model
they furthermore included a non-trivial potenti-
al energy density given by
U(~)
=
i(I~12 - r2)2
Models with such a potential term are generally referred to as Higgs' modeZs and the corresponding scalar fields are known as Higgs fieZds.
*) Vortex-line models for dual strings, Nucl. phys. B61 (1973) 45
II
446 The model based upon the Lagrangian density (8.86)
, with
where the Higgs field
~
D~
=
d~
-
ieA~
is coupled to an abelian gauge field, is spe-
cifically referred to as the abeZian Higgs' modeZ. The associated field equations are highly non-linear (8.87a)
SF
(8.87b)
D*D~
ie = :f[~D~
=
-
-
~D~l
(AI~12_ Jl2)~ £
1:.....3 \I (V~FJl\l) g
i.e.
V-g
i.e.
In this model we now proceed to investigate the purely static configurations, where furthermore the electric field is absent. Such a configuration
is represented by a spatial function ~(~)
= A(~).d~
tial one-form A
and a spa-
. Iurthermore it corresponds to a solution
of the equations of motion (8.87a-b) precisely when it extremizes the static energy, which in the present model is given by (8.88) As a consequence the theory of static equilibrium configurations in the abelian Higgs' model is completely equivalent to the GinzburgLandau theory for superconductivity, provided we make the identification:
(This follows immediately from a comparison of
(8.60) with (8.88).)
The first important observation is that the abelian Higgs' model possesses no soliton solutions in the strict sence, i.e. there are no non-trivial stable static finite-energy configurations. This follows from an argument which is typical for gauge theories: A static finite-energy, configuration must satisfy the boundary conditions (8.89)
lim B r ......
=0
lim r+co
D~
=
a
lim I~ r+co
I =
fA
Asymptotically the Higgs field is therefore completely characterized by its phase factor
In general the phase, ~(~), need not be single valued but can make a quantized jump wed!
2nn. But in the present case such a jump is not allo-
447
II
To see this, suppose the phase makes a jump when we go once around a distant closed curve
r . This distant closed curve can be shrunk to
a distant point, while the shrinking curve remains distant!
(Topolo-
gically it is a closed curve outside a ball and such a curve can clearly be contracted without intersecting the ball.) Notice that by continuity the phase must also make the jump
2nn
when we go once a-
round the shrinking curve. As the curve shrinks to a point we thus produce a discontinuity in the phase factor
exp[i~(~)l. But such a dis-
continuity in the phase factor contradicts the smoothness of the Higgs field! Since the asymptotic phase
~(~)
exp[i~ (~) 1
=
is singlevalued the phase factor
-
is in fact trivial, i.e. it is single valued throughout the whole space. Thus we can remove this phase factor completely by a gauge transformation. To conclude we have therefore shown the existence of a ge where
gau~
is real!
In this particular gauge the boundary conditions (8.89) reduces to (8.90)
lim A(~)
lim
= 0
r+'"
=
0
r .... '"
Consider now the deformed configuration
It satisfies the boundary condition (8.90) for any choise of E. At E=O it reduces to the vacuum configuration, while for
E=l
it reduces to
the given configuration. Consequently we can find a one-parameter family of finite energy configurations which interpolate between the vacuum configuration and the given configuration. This shows that a sta-
tic finite-energy configuration cannot be stable since we can "press" it down to the vacuum. To obtain interesting configurations we must therefore relax the boundary conditions (8.89). Rather than looking for point-like configurations we will now look for string-like configurations. Consequently we will concentrate on configurations which are independent of z. This time we therefore impose the boundary condition that the energy per unit length along the z-axis is finite. This restriction implies the following boundary conditions: (8.91)
lim B p.... '"
=
0
lim p.... ..,
D
=
0
lim p.... '"
I
448
II
where p is the distance from the z-axis. As before we conclude that asymptotically the Higgs field is characterized by its phase factor: cj> (r)'"
p-+co
If exp[ iIP (r) 1
This time, however, there is nothing to prevent the phase of the Higgs field to make a jump curve.
2nn
when we go once around a distant closed
A distant closed curve cannot be shrunk to a point without
intersecting the z-axis.) Notice,however, that if the phase makes a non-trivial jump then the Higg's field must necessarily vanish somewhere inside the string corresponding to a discontinuity in the phasefactor
Since the gauge covariant exterior derivative of the Higgs field vanishes outside
the string we get as usual dIP - eA
...
0
p-+co
Consequently the jump of IP is related to the magnetic flux in the string: ~
lim Po -+(0
B P
J
lim Po-+CO
f A P=Po
lim Po-+
ef dIP CO
21T ne
P=Po
This is also in accordance with the equivalence between the abelian Higgs' model and superconductivity. The number
n
can consequently
be identified with the number of flux quanta in the string. Since the flux is quantized it is impossible this time to interpo-
late between the vacuum configuration and a configuration with a nontrivial flux. A string with a non-trivial flux is thus topologically stable!
We proceed to look for configurations which minimize the static energy per unit length. As in the Ginzburg-Landau theory we shall assume that the penetration, length is considerably larger than the coherence length, i.e. in the present case we assume (8.92)
We can then carryover the conclusions obtained in the previous section concerning magnetic strings in a type II superconductor. In the sector conSisting of configurations carrying a single flux quantum the groundstate is the cylindrical symmetrical configuration. In the abelian Higgs' model this is known as a Nielsen-Olesen stping. In a sector consisting of configurations carrying multiple flux quanta there can be no exact ground state. This is because a cylindrical symmetrical configuration is unstable since its energy grows with
449 the square of the number of flux quanta, cf.
II (8.76). As in the sine-
Gordon model one can however construct approximative groundstates consisting of n widely separated Nielsen-Olesen vortices. Notice that a string-like excitation such as the Nielsen-Olesen vortex cannot itself represent a physical particle since it has infinite energy due to its infinite length. So if the Nielsen-Olesen vortex is going to be physically relevant we must find a way to terminate it. This can be done by including additional point particles in the model, which then sit at the endpoints of the string. As the Nielsen-Olesen vortex carries a magnetic flux these additional particles must necessarily be magnetic monopoles (or anti-monopoles).
Fur-
thermore the magnetic charge must be quantized since the flux carried away by the string is necessarily quantized. As we shall see in the next chapter that is fine: Magnetic charges are in fact quantized according to the rule g
211 = ne .
Suppose then we introduce point-monopoles in the abelian Higgs' model. Notice first that it is impossible to introduce just a single monopole. This is because a single monopole is characterized by a long range magnetic field and that cannot exist in the abelian Higgs' model due to the Meissner effect. Let us clarify this point: You might object that the Meissner effect could just squeeze the magnetic field into a thin string extending from the monopole to infinity. But being infinitely long the string would carry an infinite energy and that is not physically acceptable. (The Coulomb field created by an ordinary charged particle is acceptable because the energy stored in the field outside a ball containing the particle is always finite. The infinite self energy of the electron comes from the immediate neighbourhood of the electron, not from infinity.) This means that we can only introduce monopoles in terms of monopole-anti-monopole pairs. A magnetic field line extending from the monopole can then be absorbed by the anti-monopole. Suppose then that we try to separate such a monopole-anti-monopole pair. This will necessarily create a thin magnetic string between the monopole and the anti-monopole. The string has a typical thickness given by the penetration length, i.e. the radius of the string, r o ' is
f'A
R<--
e
Since the string carries the magnetic flux
g, where g is the magne-
tic charge of the monopole, the magnetic field strength will be of the order B
R<
iT?-o
Thus the magnetic energy stored
in the string is of the order
450
E ~ ~B2(wr~~) where i
II 2
=
~ w;z
i
o
is the distance between the monopole and the anti-monopole. It
will therefore require an infinite energy to separate the monopole-antimonopole, i.e. it is physically impossible to separate them and thereby produce free monopoles (or anti-monopoles). One says that monopoles are confined in the abelian Higgs' model. This is to be contrasted with ordinary electrodynamics. There one may also consider bound states consisting of two oppositely charged particles, say hydrogen or positronium. But in that case you can easily knock off the electron and thereby produce free electrons. To summarize: (a)
In standard electrodynamics the binding energy for a bound state consisting of oppositely charged particles is given by the Coulomb potential VCr)
= -
~! 4WEo r
and it only requires a finite amount of energy to separate them. (b)
In the abelian Higgs' model the binding energy for a bound state consisting of a monopole-ant i-monopole pair is given by the linear potential 2
VCr) = l~ r
o
and it requires an infinite amount of energy to separate them, i.e. the monopoles will be permanently confined. Returning to the monopole-anti-monopole pair in the abelian Higgs' model we see that the potential generates an attractive force (which is independent of the distance for large distances). To prevent the pair from collapsing we must therefore put it into rapid rotation. In this way we produce a composite particle consisting of a spinning magnetic string with a monopole at bne end and an anti-monopole at the other. lor a sufficiently long rapidly spinning string we have thus come back to the Nambu string. Now what is the relevance of these considerations in high-energy physics? Ebr various theoretial reasons hadrons (i.e. strongly interacting particles) are generally thought to be composite particles, the constituents of which are called quarks. This so-called quark model has had a considerably succes in explaining the observed spectrum of hadrons (including some predictions of hitherto unknown hadrons) .
But there is one great puzzle concerning the quark model: A free
451 II quark has neVer been observed as the outcome of a scattering experiment in high-energy physics.
(Some famous solid-state physicists claim
to have observed free quarks in a very beautiful but delcate experiments*! but their observations have not been confirmed by other groups and they are not generally trusted.) Thus quarks seem to be confined. How can we explain this quark confinement? One possible idea looks as follows: The quarks interact through the strong interactions, and it is customary to introduce a field, the so-called colour field, which is responsible for this interaction (in the same way as the electromagnetic field is responsible for the electromagnetic interaction.) In analogy with the electromagneitc field the colour field comes in two species known as the colour electric field and the colourmagnetic field. The quarks carry colour electric charges and thus act as sources for the colour electric fields. One then adds as a basic hypothesis that in the quantized theory the coloured vacuum will act like a superconducting vacuum, but in contrast to the abelian Higgs' model we are supposed to reverse the role of electric and magnetic fields,
i.e. this time it is the colour electric field which is expelled due to the Meissner effect.
(0£ course this must eventually be proven di-
rectly from basic principles if we are really going to trust the argument. This is the real hard part of the game and only little progress has been made due to great technical difficulties in the quantum theory of coloured fields.) Okay, suppose the above hypothesis concerning the coloured vacuum is correct. Then we can verbally take over the arguments concerning magnetic monopoles in the abelian Higgs' model.
free quarks cannot
exist because the colour electric field is squeezed into a finite region. It is however possible to collect
them in a quark-ant i-quark
pair, and, due to a greater complexity in the structure of coloured fields, it is also possible to put three quarks together in such a way that the net colour charge is zero.
(Incidentally this is where
the name "colour" comes from: There are three basic colours in nature - red, green and blue - and if you "add" them you get white, i.e. the net colour vanishes:) Thus the following picture arizes: Space is divided into two regions. The dominating region filling up space almost everywhere consists of the superconducting coloured vacuum. But here and there we find small regions filled with quarks and colour electric fields. These "normal" regions, where the superconductivity breaks dowvn, are called bags, and they act as prisons for the quarks. Thus a hadron is an *) G.S. LaRue, W.M. Fairbank and A.F. Hebard, "Evidence for the existence of fractionally charged matter", Phys. Rev. Lett. 38 (1977) 1011
452
II
extended object consisting of a bag containing
either a quark-ant i-quark
pair or a couple of three quarks. BAG WITH TIIO QUARKS
BAG WITH THREE QUARKS
(meson)
(baryon)
\
"
Superconducting vacuWll
Fig. 166
I -~
Consider a quark-ant i-quark pair. If we try to separate the quark from the anti-quark we necessarily produce a thin colour electring string between them and just as in the abelian Higgs' model it can be shown completely elementary that the energy stored in the string is proportional to the length of the string.
Thus we arrive at the typi-
cal linear quark potentiaZ:
VCr)
=
kr
Finally we can excitate hadrons by rotating
the bags. Such rapidly
spinning bags will be elongated and we therefore expect them to behave like spinning Nambu strings! So I hope you now see the importance of the abelian Higgs' model. It serves as a theoretical laboratorium where we in an elementary way can test the consistence of various properties of the model, before we try to examine these properties in the more complex models which are thought to describe systems, that arp. actually found in nature.
We conclude this section with yet another argument which supports the interpretation of a Nielsen-Olesen vortex as a smooth extended Nambu string. First we rearrange the action of the Nambu string slightly. Consider an inertial frame and let us identify the parameter T with the corresponding time in the inertial frame, i.e. we put T=t • The string is therefore parametrized as x = x(O",t) The arclength of the string is given by ds = The tangent vector
~~
jdXdX dO"dO"
dO"
thus becomes a unit vector. The velocity of a
point on the string is given by
453
II
-+
ax at
-+
v
In our case only the component perpendicular to the string is physically relevant. Evidently it is given by -+
-+
-+-+
-+ ax ax axax v.! = at - as (asat)
=
-+2 v.L
(ax)2 _ (axax)2 at asat
Next we observe that metric coefficients are given by 3s (3X3X) ao asat
;
h TT=
-1 +axax atat
(sing this we can rearrange the area element as follows
V-h = v' (hOT )
-
h
00
h
TT
Consequently we finally obtain (cf.
(6.49)):
t2
(8.93)
s
- .1:...J )1 _ ~212rro
dsdt
tl
Next we consider a thin Nielsen-Olesen vortex string moving around in space. It could be a static vortex which we have "boosted" into a uniform motion or it could be a more complicated motion specified by some appropriate initial data. The vortex will be a solution to the equations of motion derived from the Lagrangian density (8.86), but we are not going to specify the potential in this argument. All we use is that the model aloows stringlike solutions, where the fields are almost in the vaccum state outside the thin string. The deviations from the vacuum state will be exponentially small and we shall simply neglect them. Thus the Lagrangian density vanishes outside the string and the Lagrangian density therefore acts as a smeared out a-function. We can now find an aprroximative expression for the action of the string.
First we introduce a rest frame for a small portion of the
string. The rest frame moves with the velocity
~
, i.e. perpendicular
to the string. The coordinates in the rest frame will be denoted to,xO, yO,zO
and notice that ZO
is simply the arc length for the string!
In the rest frame the string reduces to a static string. The Lagrangian density thus becomes equal to minus the energy density, L = - H
If J denotes the energy per unit length in the rest frame we have therefore shown
But then we get the following expression for the action per unit length
II
454 S
= J LdxOdyOdzOdtO = J LJi-vl
dxOdyOdsdt
where we have used the transformation formula
dtjl-~~
dt ° =
Consequently we have deduced the following approximative expression for the total action (8.94)
But that is precisely the action of the Nambu string!
furthermore
it allows us to re-express the slope of the Regge trajectories, a, in terms of the energy density per unit length, J: C/.
1 271J
=
5
SPACE
~rest frame
DIAGRAM
,
z
,/
-+
'~ \
Vortex
y
x
Fig. 167
8.9 SINGULAR FORMS We can also extend the exterior algebra to include singular differentiaZ forms. A complete discussion falls beyond the scope of these notes. First we will discuss the naive point of view generally adopted by physicists and then we wfll give a brief introduction to the framework of distributions, which is a concise formalism adopted by mathematicians. In the naive point of view a differential form of degree
1 T = -k'• T.1
i
.
1
.•. 1
(x)dx
l
i A • ••
Adx
k
k
k
is called a singurar form if the coordinate functions
T. 1
l
···l.. k (x)
are
not smooth, i.e. they possess singularities like the discontinuity in the Heaviside function or the singularity of the
a-function.
We will assume that the usual theory of exterior calculus can be ex-
455 II tended in a reasonable way to include singular forms, so that we can use rules like
d2
=0
or Stokes'theorem even if singular forms are
involved. The first thing we will generalize is the
a-function:
Definition 5 If
M
is a manifold with metric
peaked at the point
Po
g, then the
a-function,
ap
,
is defined to be the singular scalar fie9d
R
apo : M ~ characterized by the property that =
fM
~ap /gdx
1
A••• Adx
n
0 ~
for any smooth scalar field
.
We can easily work out the coordinate expression for dinates the scalar field
apo ~(x)
function, which we denote
a ' In coorpo is represented by a singular Euclidean From the integral property of
ap
we then get
o
But then
i.e.
C!-(x)
=
_i_on (x-x )
,fg('Xj"'
0
where
an (x-x o ) is the usual Euclidean a-function. Using singular forms we can now associate an electric current
J
to a point particle. This is a highly singular one-form, which vanishes outside the world line of the particle. Using an inertial frame we saw
in section 1.6 that the electric current was characterized by the contravariant components (1. 34)
The covariant coordinate expression for the
a-function suggests
that we define the components of the current to be 1 dx ll dT J 11 (x) = q - - , a (X-X(T»(8.96 ) I-g(x) dT
f
where this expression is valid in an arbitrary coordinate system. To justify that this is the covariant expression representing a singular vector field we must show that
Jll(x)
transforms contravariantly.
Observe that a general expression like
f~ (x-x (T» where
~
x dd :
dT
is a scalar field, does not transform contravariantly! That
everything work out all right is due to very special properties of the
II
456
8-function. So let us introduce new coordinates
(yl, •.. ,yn)
and see
what happens. In the new coordinates the current is characterized by the components
But
= __1_0 4
__ 1_8 4 (y -y(T)) I-g (~ 0
(x -X(T))
I-g (XJ
0
since it represents a scalar field and therefore we get JJ.l(yo)
qJ
=
8 4 (x O -X(T))
Now we use that replace
8 4 (x O -X(T))
1
I-g(xo )
(2)
~ dXV(X(T))
is peaked at
~ dXV(X
by
*~(X(T))d;TV
dT
Xo . Consequently we can
and then move it outside the integral
O)
whereby we obtain
J
v 1 _ 8 4 (x -X(T) )~d dx T )q _ _
~ (x
Jll(y ) (2)
dX V
0
l-g(1:)
0
0
~(x V dX
T
) JV(X ) 0
(1)
0
So everything is okay!
Worked exercise 8.9. 1 Problem:
Show that the singular current i.e. 5J = 0 •
J
with components (8.96) is conserved,
In a similar way we can introduce a singular 2-form
F
representing
the electromagnetic field generated by an electrically charged point F
particle. The singular forms
and
J
then obey the Maxwell equa-
tions: dF = 0
ILLUSTRATIVE EXAMPLE:
HOW TO HANDLE SINGULARITIES.
You might wonder how we can control the singularity of
F. It is
infinite at the position of the particle. How can we differentiate a function which is infinite at a single point like the function
I?
x
Usually this is done by a regularization procedure. Consider the static Coulomb field as an example. For simplicity we work in the ordinary Euclidean space
R3
and use Cartesian coordinates
etc. The Coulomb field is characterized by the components Ei _ - L -41TE
O
xi
'?'
We regularize the field making the exchangement r ~
f:rZ+ET
II
457 In this way we get the regularized field strength
which is now smooth throughout the complete Euclidian space. If we differentiate this,we get
Consequently the regularized electric field corresponds to a smooth charge distribution
which asymptotically vanishes like around s-3. Q(E)
r=O
2r -5. s
For small
it is peaked
pes)
Observe that the total charge 4n
s
where it varies like
foo
p(s)r 2 dr
o E« 1
q
is independent of
s.
(Use
r=sx).
Thus we finally obtain the desired result lim
peE)
= q03 (x)
i;2m which we conclude that Fig. 168 Observe that the electric field panied by a magnetic field
B(s).
E(s) in this example is not accomThe magnetic field can be obtained
from the equation
But the curl vanishes automatically as a consequence of the spherical symmetry.
0
Finally we will sketch how to introduce singular forms within the framework of distributions. Consider first the case of ordinary functions on the Euclidian space Rn. Here a function f is ~haracterized as a map Rn ~ R , i. e. each point x in Rn is mapped into a real number f(x) • We want now to construct singular (or generalized) functions. This is done in the following way: First we choose a test space. In the theory of generalized functions this test space consists of all smooth functions with compact support, and it is denoted D(Rn ).
458
II
The functions in the test space are called test funations. We can now formulate the following definition Definition 6 A generalized function (i.e. a distribution) T is a Zinear functional on the test spaae, i.e. T is a linear map:
Now, what is the connection between ordinary functions and generalized functions? Apparently they are defined in 2 completely different ways. To understand this connection we observe that any ordinary continuous function f : RO ~ R generates in a canonical fashion a linear functional D(Rn ) ~ R which we denote by T to f avoid confusion. This linear functional is defined through the Hilbert product Tf [$ 1 =
(8.97)
-+ f n f (xl
-+
n
$(x)d x
R
where $ is any test function. The integral is well-defined becaUSe $ has compact support. Thus any continuous function f generates a generalized function T , f which by abuse of notation is usually denoted f too. But the conVerse is not true. Consider e.g. the 8-function. It is defined as the linear functional O[$l = $(0)
In analogy with (8.97) this is often rewritten in the more informal w~
f
o(x) $(~ldnX = $(0)
But there exist no ordinary function " Ii (x) " satisfying this integral identity, so strictly speaking it makes no sence! Ok~, so much for the generalized functions. Consider a manifold M with metric g. We want to introduce singular k-forms. As the test space Dk(M) we use the set of all smooth k-forms with compact support. They are called test forms. We can then define:
Definition 7 A weak k-form T is a linear functional on the test space i.e. T is a linear map
T
Dk(M)~R
Let us fir.st check that an arbitrary smooth k-form T can be represented as a linear functional so that the weak k-forms include the smooth k-forms. Let T be a smooth k-form, then the linear functional is defined through the Hilbert-product (8.99 )
T[Ul
def
=
fk 1
M
T
i l · •• i k
(x)
u. J. l
. (x)
•· '~n
vg d nx
where the Hilbert product is well-defined because the test form support.
U has compact
Let us discuss some specific examples of weak forms. First we consider the electric current J associated with a point particle in Minkowski space. In the naive approach it was characterized by the contravariant components
459
L= q Jr -.r--::r='-g\X}
.J.l( )
x
J'
Let
<5
4
U be a test form.
<Jlu>
=
t q
q
q
J v'-~(X)
f :~ [f
f :].1
<5
dx].l
(X-X(T)) . dT
II
dT
Then formally we have:
4
(X-X(T))
U].I(X)
<5
:].1 dT
UjJ (x)
Fg"(XT d 4x
4(X-X(T))d 4X] dT
M
Uj.l(X(T))dT
=q
fr U
Within the frame work of distributions we therefore define the singular current as the linear functional: (8.100 )
where r is the worldline of the particle! Next we consider the singular Coulomb field in the ordinary Euclidian space R3. Its components are infinite at the origin and we must therefore define the Hilbert product through a limit procedure. Let us put M = {~I " ;t iI > d. Then we represent E by the linear functional E (8.101)
ElUl def. = l~m ~ E+O
0
J
xi
...
3
~ Ui(x)d x
ME r
(That the limit is well-defined follows easily if you work out the integral in spherical coordinates!) Now if the weak forms are going to be of any use, we must be able to extend at least part of the exterior calculus. The differential calculus is the easiest one to extend: (a)
The
e~terior
dBPivative:
Suppose first that T is a smooth k-form. functional dTlul =
Then
dT
~s
represented by the linear
where we can throw away the boundary term, because the test form U has compact support. But test forms are always smooth, so if T is a weak form, we can immediately generalize the above result and define dT to be the linear functional (8.102)
dTlUl d~f. Tl&Ul
(b) The ao-differentia.Z:
Then
&T
is represented by the linear
T is a weak form, we can therefore define
&T
to be the linear functional
Suppose first that functional 5Tlul = When
T is a smooth k-form.
<&Tlu> =
(8.103)
Exercise 8.9.2 Problem: Let M be a Riemannian manifold. Show that the Laplacian of a weak form T is represented by the linear functional (8.104)
460
II
Exercise 8.9.:3 Problem: Show that we can extend the dual map to weak forms in the following way: If T is a weak k-form, then *T is the weak (n-k)-form represented by the linear functional (8.105)
*T[ul d~f { (-I)k(n-k)T[*Ul on a Riemannian manifold -(-I)k(n-k)T[*Ul on a manifold with Minkowski metric.
Having generalized the differential operators we can easily check some of their fundamental properties. For instance we get 2
d T = 0
and
&2T = 0
even if T is weak. This fOllows immediately from the definitions (8.102)and (8.103) and the corresponding properties of the test forms. [E.g. we obtain d 2T[Ul = T[&2Ul = T[Ol = 0 so that
d 2T
is the zero-functional and similarly for
&2 T .l
Worked exercise 8.9.4 Problem: (a) Let J be the singular current (8.1.00) associated with a point particle in Minkowski space. Show within the framework of distributions that it is conserved i.e. ~J = 0 • (b) Le~ E be the singular Coulomb field (8.101)in ordinary Euclidean space R3 . Show within the framework of distributions that -&E
=~ Eo
8(x)8(y)8(z)
Observe that the space of weak k-forms incorporates not only all smooth k-forms but also all orient able regular k-dimensional domains! If ~ is a k-dimensional orient able regular domain we represent it as the linear functional
mul
(8.106)
= f~
Since test forms have compact support we can even allow ~ to be an unbounded noncompact domain. This is a very powerful generalization of the usual 8 -function: A O-dimensional regular domain ~ consisting of the single point Po is represented by the linear functional ~[~l
= fp o ~ = ~(Po)
i.e. it generates the usual 8 -function. Using this notation we observe that the singular four-current (8. 100) associated with a point particle is essentially identical to the worldline of the point particle since i.e.
J =
qr
Observe that for regular domains we have introduced the boundary operator,a:~a~ • This can now be interpreted as a differential operator! The boundary of ~ is represented by the linear functional
Using Stokes' theorem and (8.102)this can be rearranged as a~[Ul
= f~du = ~[dul = &~[Ul
so that the boundary Operation coincides with the co-differential! The important property aa~ = ~ is now seen as a special case of the rule &2T = 0 Coming this far you might think that everything can be done using singular forms. But that is not true! When we try to extend the wedge product or the integral calcu-
461
II
lus we run into trouble: Consider the wedge product of a smooth k-form T represented by the linear functional
and a smooth m-form
S . It is
TAS[U] d~f'
(8.107)
but if S is singular too, then this will not work, because U'S is then no longer a test form. If we introduce a suitable topology on the testspace vk(M) and use limit procedures then one can sometimes extend the wedge product to a pair of weak forms or similarly extend the integral to m integral of a weak form over a regular domain, but it is not always possible! These deficiencies of the extended exterior calculus should not be underestimated. "Distribution" is not a magic word you can use to justify any calculation you want to perform with singular quantities. Consider e.g. the derivation of the electromagnetic energy momentum tensor in section 1.6. Here we discussed the electromagnetic field generated by a collection of point particles. The equation (1. 40)
had a central position in the argument. Consider the right hand side. Here the electromagnetic field strength Fa is singular at the position of a particle and so is the current JY. But then theIr product is not well-defined, not even in the sense of distributions! Consequently the argument in section 1.6 is only a heuristic argument of didactic importance, not a proof in the strict sense. Weak forms were introduced by de Hham. Among other things he was motivated by the electric current associated with a point particle. He therefore used the name "currents" for weak forms in general. That is however misleading in a physical context and I have therefore adopted the name "weak form". If you want a more rigorous treatment you should consult de Rham [1955] or Gelfand and shilov [1964].
SOLUTIONS OF WORKED EXERCISES: No. 8.3.3 (a) We use spherical coordinates adapted to
2 • S+ .
S! = {(r,e,\p) Ir=l; o<e<1!.· - -2' 0::~27f} y = rSineSin
SineCoslIdr +
rCoseCos~e
-
rSineSin~
dY
SineSin~r
rCoseSin~e
+
rSineCos~
When we restrict dx zero, whereby we get
+
dY to S! we put
and
dx
cosecos~e
-
SineSin~
dY
CoseSin~e
+
SineCos~
r
equal to
1
and
dr
equal to
462 i.e.
II
dxNly = Cose'Sinedelld4l
Then we replace the geometrical volume element lume element ded~ and finally get
I
dxNl:y =
s!
f
delld4l with the "Riemannian" vo-
t71 CoseSin6'dedlP = 71
e=o ~O (b) Spherical coordinates are also adapted to B3 = {(r,e,~)lr~l
B3 :
o~e'~71
0~~271}
Observe that dxNlyNlz is the Levi-Civita form in coordinates we can therefore rearrange it as
R3 . In terms of spherical
dxNlyNlz = /'idrNlelld4l Using (6.36) we now get g = r4 Sin z e
i.e. drAd6A~
When we replace the geometrical volume-element
volume-element
drded~
with the "Riemannian"
we finally arrive at
I1
I7I
r=O e=o
I2:2sinedrded~ = ~ ~O
which we recognize as the volume of the unit ball. No. 8.4.1 A smooth 2-form B
can be decomposed as
B = B3(x,y,z)dXAdy + B1(x,y,Z)dyAdz + Bz(x,y,z)dzAdx Due to linearity it suffices to verifY Stokes' theorem term by term. We have thus reduced the problem to the consideration of a smooth 2-form on the form B = B(x,y,z)dxAdy
We then get
Cartesian coordinates are adapted to the unit cube and the two integrals can now easily be computed. (1)
IadE = IIl~ZAdxAdy
I I
f~Zdxdy
f f x=O y=O z=O
I
J [B(x,y,l)-B(x,j,O)ldxdy
x=o y=O The boundary all consists of six surfaces but only two of them contribute, since dx=o (or dy=O) along surfaces where x (or y) is constant. The bottom and the top of the cube are oppositely oriented: The Cartesian coordinates (x,y) are positively oriented on the top but negatively oriented on the bottom (see fig. 169). We now get I
(2f
B = JaIlB(X,y,Z)dxNly = all
I
J J B(x,y,l)dxdy -
y .; /'
'" '" Fig. 169
x I
I
I
J B(x,y,O)dxdy
x=o y=o x=o y=O from which we see immediately that (1) and (2) are identical.
0
II
463
No. 8.4.3 (a) Let r have the parametrization z(t) = x(t)+iy(t) , a
~
[ax + ik]dt at at
When we want to integrate the one-form h(z)dz dz to r, dz
~ at« h'>t = [ax at
we must find the restriction of
+ ~at ·~]dt
and then exchange the geometrical volume-element dt with the "Riemannian" volume-element dt • Both integrals therefore reduce to ([f(X,Y)+ig(X'y (bl Notice that
)][~~ + i~]dt
~~ dz + ~~ dz
dh(z) =
but h(z) is holomophic and therefore ~~ 0 (Cauchy-Riemann's equation). Consequently we get dh(z) = ah dz az The formula is now a simple consequence of Stokes' theorem
=
fr
h'(z)dz
= J dh(z) = J
r
ar
(c) According to exercise 7.6.9 h(z)dz
= h(B)-h(A)
h(z)
is a closed form. Therefore we get
r d[h(z)dz] = 0 Jr h(z)dz = fa~h(z)dz = J~ No. 8.5.1
J a~
* [cpd1/l]
JanCPA*d1/l = Jd[cpA*d1jJ] ~
= J dcpA*d1/l ~
+
o J cpAd*d1/l ~
J~dCPA*d1jJ - f~CPA*A1/I = (-l)n-l[f~*d1jJAdCP
+ fn*A1/I ACP]
(-1)n-l[
and 1/1
J * [1/Idcp]
you get
= (-1)n-l[
+
a~
Then you subtract these two identities and get (8.33):
J *[cpd1jJ] - f a~
If we put
Cp=1jJ
an
*[1/Idcp]
= (-1)n-l[
-
<1jJIA~J
'1']
in (8.32) we finally get
* [cpdcp]
= (-l)n-lJ
an If
cp is a harmonic function that vanishes on the boundary,we conclude
464
Therefore d~=O so that must vanish throughout ~
~
II
is constant, but since it vanishes on the boundary it
U
No. 8.5.2 (a) This is simply the result of exercise 8.5.1 formula (8.32) when we put - :5-(3)dx + ~ + ~dz) (b) d~ r r r r *d~
~=~
- :5-(4:YAdz + "ZdzAdx + ~Ady) r
r
r
r
Using this we get = - .l;.[xdyAdz + ydzAdx + ZdxAdy] r
- lsin9d9Adlp r
The integrals are now easily computed:
- -IJ £
aM
Sin9deAdlp = - -1 E
E
Jrr J2rrSin9d9d~
9=0
=
_ 4rr E
~O
No. 8.6.2 (a) The Lagrangian density is a scalar field. Consequently
~=1h+ aL a£IE=O a>'" alai)
"'" 0jl'"
is a scalar field too. From exercise 6.9.3 it now follows that field,while
~ a (ajl> )
~~ is a scalar
are the contravariant components of a vector field.
(b) The left hand side is the Hilbert product of
A
and
dW :
J~ArJ.(arJ.~);.:gd4x =
But here the boundary term vanishes since
~=O
an. Using that
on the boundary
d*A = *&A (see 7.62) we finally get
f ArJ.(arJ.~)/=gd4X =
~
Moral: When one of the factors vanishes on the boundary then adjoint operators". 0
No. 8.6.3 (al
SI =
qJ:2~ ~~dA J
=
and
d
qJ:2[J~Av(X)04(X-X(A))d4X]~:dA
A~(Xl[qfA2~~ ~4(X-X(A))dA]/=gd4X Al I-g
~
n&
J A (xl J ll(x)/=gd 4X n ll
are
465
(bl The dynamical quantities to be varied are the position of the particle, and the gauge potential, All (xl : x~(Al
x~(Al+EY~(Al
+
A~(xl + A~(X)+EB~(x)
B~
=0
on
all.
4 steps:
We then proceed in
1. Variation of Sp: This has been discussed in section 6.7: T2 d2 xa dSp a dx~ dxv = (~+ mr ~v---;rr ---;rr)Ya(T)dT IE=O TI 2. Variation of SF: This has been discussed in exercise 8.6.1:
a.e- J J
F- = -dSa [;:gF~vlB (x)d 4 x de: 1 E=O Il ~ v
3. Variation of s:} with respect to the particZe trajectory
q(>~[X(A)+EY(A)l(a:: + e:~)dA i.e.
(where we have performed a partial integration on the last term) A
J
2
q Al
[~ _ lfuL]dx~ ax~ dA
axv
v dA Y
A
q
J
2Fv~iiJ\ dx~ V Y dA
=
Al
q
IT2 ~S---;rr dx S dT Ya TI
where we have re-introduced the proper time
T
4. Variation of SI with respect to the field: Using the result obtained in a) we easily get I -dS d-- = s 1 s=o
J ;:gJv(x)B Il
(x)d 4 x
v
Collecting all the results we finally obtain
o = dS
dSls=o
Ya and Bv are arbitrary, so this is only consistent if (8.37) and (7.85) are satisfied.
But
0
466
II
No. 8.7.3 (a) To get the equations of motion for B we perform the variation A~A+e:U
i.e.
We then obtain H(e:) = !
+~2e:2
o =
+
~(~D1/J
, we can easily rearrange
~»
-
This leads to the equations of motion
- &B = ~(~D1jJ
-
1/1~)
(b) To get the equations of motion for 1/1 we perform the variation 1/1
~
1/1 + e:$
i.e. We then obtain
Thus we must demand dH o = ~e:=~ !
JaI1/l12(1/I¢ au - + 1/1¢) - e:
= l
o =
.rg d 3 x
so that we can rearrange it as
o = <$I&D1jJ + ~W*D1jJ) + This leads to the equation of motion - &D1jJ -
~CAA*D1jJ)
Dualizing it we finally get
=
2al~121/1
2al~121/1>
467
au 2alwl2wE
9 d*0IJ! - iflAII*OIJ!
=
=
II
9 Cd - itiA)*OIJ!
=
o
D*OIJ!
No. 8.9.1 The verification of the current-conservation is done by a brute force calculation:
-= a (Fg Jl1) =....!L a v-g 11 ;:g 11 1
J- I i4 l1 dx dT
(X-X(T))dT
=_....!L ~
l1 rdx -a- tS 4 (X-X(T))dT dT dXl1(T)
J
but using the chain rule, this can be rearranged as
J~ tS 4
=_.....9....
I=g
dT
(X-X(T))dT
= _....!L [tS 4 (X-X(T) )]T=_ = 0
;=g
since
x
T=-OO
is fixed and Ixo(T) I
+
00
as
ITI
+
00
•
0
No. 8.9.4 (a) Let $ be a test-form of degree 0 (i.e. a smooth scalar field with compact support). Using (9.63) and (9.66) we get
&J[¢] = J[d¢] = q frl¢
= q[¢(T=-)
- ¢(T=-OO)] = 0
since ¢ has compact support. Consequently &J is represented by the O-fUnctional. (b) Let ¢ be a test form of degree o. Using (8.101) and (8.103) we get
= -lim <Eld¢>
-&E[¢] = -E[d¢]
(*)
E+o
ME
But observe that i E=~E----L..dW 471Eo r3 471EO
where
wei) =1. r
Inserting this into
-& E[¢] = ~
lim o E+o
Now we use Green's identity
= ~ lim 7IEo E+o
W is
But
(*) we get
(8.32) and rearrange it as
[f *¢dW -
aM
E
harmonic in
ME
so the last term vanishes. In the first term we integrate
over a sphere shrinking to a point. In the limit we can therefore safely replace $(~)
with
¢(O):
Here we obtain
*dW = - L(xdylldz r3
so that we get (ME
+ ydzlldx + zdXNly) = -sin
6d611d$
is outside the sphere, so it gets the opposite orientation):
468
II
But then we have shown , -&E[CP] = ~ o
which means that
-&E
cp(O) ,
is represented by 3- times the a-functional. £ o
0
469
II
chapter q DIRAC MONOPOLES 9,1 MAGNETIC CHARGES AND CURRENTS Previously we have only been considering conventional electrodynamics according to which magnetic charges are excluded.
This is in
accordance with experiments, where electric fields are always generated by electrically charged particles, whereas magnetic fields are generated from electric currents. Never-the-less there is apriori no reason to exclude magnetic charges, and as shown by Dirac*\n 1931 their existence would have interesting theoretical consequences. Especially the magnetic charge will necessarily be quantized due to quantum mechanical effects. Current ideas about the fundamental forces and the origin of the universe also strongly suugest that magnetic monopoles were in fact created in the early history of the universe. Most of these monopoles would have annihilated each other again ( in the same way as matter and anti-matter annihilate each other ), but a small fraction may have survived, allthough they will be extremely difficult to detect mainly due to their large mass. Even if magnetic monopoles do not exist the underlying mathematical model is very interesting and it has had a great impact on our understanding of gauge theories. Let us consider a pOint particle which serves as a source for the electromagnetic field. We expect a singularity at the position of the particle (like the singularity of the Coulomb field). Therefore we cut out the trajectory from Minkowski space, i.e. if the point particle then the basic manifold is L
r is the worldline of
= M'
f. We know then
that the electromagnetic field is represented by a smooth 2-form F on L. Furthermore F satisfies Maxwell's equations (7.88)
iF
=
0
Consider now the magnetic flux through a closed surface surrounding the point particle. To fix notatiOn let S be an inertial frame and consider a three dimensional regular domain
n
contained in a space
slice relative to S, cf. fig.170. We have then preciously shown that the magnetic flux ~ through the boundary
an
is given by
fanF
(cf.
(8.51)
*) ~antized singularities in the electromagnetic ~ield, Proc. Roy. Soc. A133,(6o).
).
470
II
Space time diagram - One dimension suppressed.
XO,~~
We want to show now that, as a consequence of the Maxwell equations, the magnetic flux has the following three basic propert:ies:
1} It is the same for aZZ cto-
I
/ ___~.
~(-
x2
sed suzofaces SUI'rOunding the point particle. As a consequence it can be interpreted as the magnetic charge of the pointparticle (in the same way as the electric flUX throough a c Zosed suzoface
xl Fig. 170
reproesents the electric charge within the suzoface).
It is independent of time, i. e. the magnetic charge is consel'Ved. 3} It is independent of the obserovero, i.e. the magnetic charge is a Lorentz scalar (oro to be proecise: a pseudo scalar, since the flUX depends upon the orientation).
2}
Space time diagram - One dimension suppressed
F:ig. l71b
Fig. I7la
To show property 1 we consider two closed surfaces Q 1 and Q2 contained in the same space slice, cf. faces will be denoted
fig.17~The
region between the two sur-
W. Notite that its boundary, aw, consists of the
two surfaces QI and Q2, but that W induces opposite orientations on Q 1 and Q2. From Stokes' theorem we now get: o
= fW dF = faw F =
fQ2F -
fnlF
:i.e.
~I
= ~2
By repeating the same argument we can similarly show property 2 and 3.
(This requires the consideration of two closed surfaces contained :in
diffrent space slices as shown on fig.17Ib). A point particle can of course in principle carry both an electric charge q and a magnetic charge g. If the part:icle carr:ies both electric and magnetic charge :it is called a dyon.
471
II
We want now to enlarge our considerations and consider a system where we have included a smooth distri.I:AItion of magnetic charges and currents. We kIx:1N that the electric charges and currents may be comprised in a smooth I-form which acts as a source for the electromagnetic field through the Maxwell equation,
d*F =-*J.
This equation has two important con-
sequences: Taking the exterior derivative at both sides we get, d*J
a.
=
0, which
is the continuity equation, i.e. the conservation of electric charge. b.
If
~
is a three-dimensional volume contained in a space slice
then the electric flux through
contained in
is equal to the eZectric charge
~:
Q
-f n
*J =
In
d*F =
fd~ *F
With this in mind we now introduce a magnetic four-current k
O
k
is the magnetic charge density and
=
1
2
(k ,k ,k
3
)
K
where
the magnetic
current. The magnetic four-current is going to act as a source for the electromagnetic field. Consequently we must modify the ordinary Maxwell equation,
dF
= o.
Now, if we insist that the modified equation
should guarantee the conservation of magnetic charge, and the identity of the magnetic charge contained in a volume with the magnetic flux through its surface, then we are forced to give it the following form: (9.1)
dF=-*K
In this form the symmetric role played by the electric and magnetic charges is displayed very clearly.
Exercise 9.1.1 Problem:
(9.2 (9.3
)
Prove the following equivalences _1_ 0a (Fg~aB)
-k B iff
daFsy+dsFya+dyFaB
_1_ d (FgF aB )
jB iff
daFBy+dBFya+dyFaB
r-g
r-g
a
- -
*
*
*
FgEaByokO
;.:q
.0 -gEaByoJ
--
According to exercise 9.1.1 we can translate ( 9.1)
into the equi-
valent covariant formula: 9.4
1 d (r-g~ll\!) Fgll
To summarize, we have established the following scheme for the extended Maxwell equations involving both electric and magnetic charges:
472 Conventional vector analysis V·s
Covariant expressions
= Pm
Geometric formula
aaFSy+aSFya+ayFaS = _ FgE
)
( 9.5
II
...
r
as + VxE = -k at
aSyo
KO
dF =-*K
_l_a (Fg;aS) = S K ] r-ga
L
*
V·E = ~ E P
[6F
=-K]
*
* aaFsy+aSFya+ayFaS =
0
( 9.6
_ r-gE
)
aE at
C2~xS =-~ J EO
-t-
JO aSyo [_l_a (FgF aS ) = Ja] FgS
d*F =-*J
[6 F =
J]
Using singular differential forms we can also associate electric and magnetic currents with point particles. These currents are then represented by singular I-forms
J
and
K
defined throughout the
whole of Minkowski space. Similarly the electromagnetic field is represented by a singular differential form
F
satisfying the modified
Maxwell equations
- dF = *K - d*F *J on the whole of Minkowski space. The components of the singular currents are comprised in the following scheme:
Conventional vector analysis p
( 9.7
-t-
)
J
PM ( 9.8
)
+
= =
Covariant expression
qo3(~-i(t) q~o3 (~-i(t»
... ... = gM o 3 (x-x(t» +
3
-7
+
ll
JIl
1 o"(x-x(T»dT dT --= q JdX
KIl
1 (f[ --o"(x-x(T»dT = gM JdX
r-g
k = gMvo (x-x(t»
Il
r-g
Within the framework of distributions the currents are simply given by J
( 9.9
where
r
= qr
and
is the world line of the particle.
Next we consider the dynamics of charged point particles. They interact with the electromagnetic field and thus experience forces. To
473
II
simplify calculations we will restrict ourselves to inertial coordinates in what follows. We then already know that an charged particle experiences the Lorentz force so that its equation of motion is given by (1. 26)
But if the particle is magnetically charged too it should experience a force analogous to the Lorentz force. It is this force we are going to determine. Although you might guess it correctly using a symmetry argument (see the illustrative example at the end of this section) we derive it from a more general argument. The interaction between
will
the magnetically charged particles and the field should be constructed in such a way that the energy and momentum of the total system, i.e. fields and particles, are conserved.
(Compare the discussion in section
1. 6)
Therefore we consider a system consisting of a field and ticles. The qn
n'th
N
par-
particle is supposed to carry the electric charge
and the magnetic charge
gn
(so we admit the possibility of
dyons). We have previously determined the energymomentum tensor of this system (section 1.6)
• The contribution from the particles is
given by N L
(1.36)
n=l
dx S n Jp n a(T)----d o'(x-x (T»dT T n
and the contribution from the field is given by (1.41)
We must choose the interaction in such a way that 3 TaB
S
P
-3 TaB
S
F
Okay! Let us get started. As in section 1.6 we get
N (1.39)
L
n=l
dP a
Jo'(X-X (T»~ n
de
1"
Then we look at the contribution from the field:
(*) Fa JY + (3 Fa )F YS +
B
Y
where we exchanged
3 F YB
B
with
JY
Y
according to the Maxwell equa-
tions.
Worked e~ercise 9.1.2 Problem: Show that the generalized Maxwell equations (3 Fa )F YB + ln aB 3 [FyoF 1 B Y __ ~ _ J _ _ yo
474
II
According to exercise 9.1.2 we can rearrange the expression Fa JB_Fa KB B F B B We then insert the expressions 9.7,9.8) netic currents and finally arrive at
(*)
as
_ d TaB
for the electric and mag-
J
N [ dx B TaB = L q Fa _n_ B F n=l n B dT Comparing this with dBT aB p you see that energy-momentum is conserved, provided the following identity holds dpa dxB n Fa n (9.10) dT qn BdT
-a
Thus magnetic charges gives rise to a force very similar to the Lorentz force except that we have interchanged the role of electric and magnetic fields. We may summarize this in the following scheme:
Forces arising from the interaction between charged particles and the fields Lorentz invariant equations of motion
Conventional vector analysis
(9.11 )
(9.12 )
. ..
d 2 x a = FaB~B q dT
F = q(E+~xB)
F
m--crr
d 2xa = *a dx B -gF BdT ~
= g(B-~xE)
We can also write down the equations of motion
(9.11,9.12)
in co-
variant form. As we have seen in section 6 •. 7, th. 4 this requires the introduction of Christoffel fields corresponding to the fictitious forces present in a non-inertial frame. Thus we are led to the following covariant equation of motion dx\.l dx v d 2 x tl a *a a dx B (9.13 ) (qF B- gF B) dT - mr m~ dT dT \.IV
(Compare this with
(6.46)
)
.
ILLUSTRATIVE EXAMPLE: CHARGE ROTATIONS Having completed the discussion of magnetic charges you should observe a curious fact. The equations of motion for the fields and particles exhibit a special kind of symmetry: They are invariant under the combined transformations: (9.14 ) (9.15 )
Cosa [ Sina
-Sina
Cosa
-Sina
[ Sina
Cbsa
Cosa
475
II
(Observe that *F' is really the dual of F' Observe also that the transformation rule for the charges is a consequence of the transformation rule for the field, when we identify charge with flux!) To check this symmetry we first rearrange the equations of motions for the particles:
You can now easily check the invariance by broute force. Then we observe that because the world lines of the particles are unaffected by the transformation the currents will transform in the same way as the charges:
~
[
~:
] -+[
[~~::
]
-Sino. Coso.
][ ~ ]
When we now rearrange the Maxwell equations on the form
d[ *FF ] = f-*K ] L~J
= _
* [K]
J we immediately see that this is invariant too because and [ K] transform [ *FF with the same matrix. J This symmetry is usually referred to as charge symmetry. Decomposing t~e tensor F we obtain the following charge transformation of the field strengths E and
J
(9.16)
[: ]
[
+
!]
C~so.
B
-Sino.] [ Coso. E
S~no.
Observe that the energy and momentum of the system are unaffected too by charge transformations. It is enough to check this for the electromagnetic field, but here it is obvious as the energy and moaentum densities are given by ( 1. 4) 2
E
=
E.o "2
+2
;t2
(E + l:l)
and
-+
g
-+
= EoE )(
B+
The meaning of charge symmetry is presently not well understood, but it has one amusing consequence which is in fact responsible for its name. On the classical level it is completely impossible to distinguish between electric and magnetic charge. What you call electric and what you call magnetic is only a matter of convention. You can redefine all charges by performing a charge transformation. If you e.g. choose 0. =:! you find: 2
qn
+
g'n
=
gn
i.e., what was before called electric charge is now called magnetic charge and vice versa ..
Exer-aise 9.1.3 PrOblem:
Introduce the complex valued differential forms: F
(9.17)
(9.18)
F +
iF
(Complex field strength)
J
K + iJ
(Complex current)
Q
g +iq
(Complex charge)
a) ShOW that the complex field strength is anti-self-dual, i.e. *F = -iF
b) Show that the equations of motion can be rearranged as B d 2 X o. (9.19) dF =-*J Re(QiFo. B)dd~ ~ (9.20)
c) Show that the charge symmetry transformations reduce F + F' = eio. F J -+- J' = eio. J
to
Q + Q'
476
II
Show also that the Lagrangian of the free electromagnetic field is not invariant under charge rotations. (This is another example of a symmetry of the equations of motion which is not present at the Lagrangian level. Cf. the discussion in section 2.2)
9.2 THE DIRAC STRING When we want to include magnetic monopoles in the Lagrangian formalism we immediately run into difficulties. The electromagnetic field strength
F
is no longer derived from a global gauge potential
A,
i.e. it is no longer exact. If we insist on introducing gauge potentials it can only be done at the expence of singularities in the gauge potentials(i.e. Dirac strings, cf. the discussion in section 7.8) which were not present in the electromagnetic field strength. Somehow we must learn how to handle these singularities. ILLUSTRATIVE EXAMPLE: THE DIRAC STRING AS A PHYSICAL STRING Consider once more the pure monopole field
It can be derived from the gauge potential (cf.(7.91)): ~
r(r-z)
-K(Cos8 + l)9tp
with
o However, A is singular at the origin (r=O) and on the positive z-axis (r=z). The singularity at the origin reflects the singularity in the pure monopole field. The singularity at the positive z-axis constitutes the string, which was not present in the pure monopole field. We will now investigate this singularity and propose a PhYsical interpretation. To be able to control the singularities we regularize the gauge potential ~
R(R-z)
with
K
o
R
477
II
A
Observe that is smooth throughout the whole of space. Bowever it has a tendency to "peak" at the origin and on the string. The corresponding electromagnetic field is given by
They satisfY Maxwell's e~uations, provided we introduce a current J£ ~XB = 1 7 £ £C2 J£ o We may consider this to be an (extremely idealized!) model of a semi-infinite solenoid generating the magnetic field! In the limit £+0 the semi-infinite solenoid becomes infinitely thin and we Fig. 172 have a string of concentrated magnetic flux, which is spread out at the origin as the monopole field. Thus we have modified the originally pure monoPole+fi:ld by superimposing the string with the magnetic flux. The singularity of the A-fleld at the position of the string now reflects the singularity of the magnetic field in the modified monopole field! Observe, that if we compute the magnetic flux through a closed surface, we now get zero, because the flux of the monopole field is exactly compensated by the singular flux through the string.
Worked exercise 9.2.1 Problem: (a) Show that the regularized magnetic field is given by
B - ~ [; _ £2 (2R-z) kl £ - 4n RT R3(R-z)2 J (b) Show that the modified monopole field is given by ~
+
B = £+0 lim B = 4":& £ n r
Fig. 173 A
~a(x)a(y)8(z)k
where 8(z) is the Heaviside stepfunction and along the z-axis.
k
is the unit-vector
The associated current j which is responsible for the concentrated magnetic flux along the string is extremely singular: a (x) a ' (y) £
:23
=
o
~xB =
-gM
-a' (x) a (y)
[
0
where a' (x) is the derivative of the a-function. The magnetic field associated with the string has some nice properties. It corresponds to a n8ga~ive magnetic charge -9M at the origin, since implies
BSTRING
= -9MO(X)6(y)0(Z)k
~.BSTRING = ~a(x)6(y)d~~Z) = -gMa (x)6(y)6(z)
This cancels exactly the positive magnetic charge
9M
associated with the pure mono-
478
II
pole field! Thus when we consider the modified monopole field (9.23) we see that the magnetic monopole has disappeared.
0
Consider the electromagnetic field
F
produced by a magnetic mono-
pole and an electrically charged particle. This field has the following properties
- dF = * K fnF = gM
1) (9.24)
where
2)
n
(where (where
is any sphere surrounding the monopole. Both of these pro-
perties make it impossible to generate
A.
K is the magnetic current) gM is the magnetic charge)
F
from a global gauge potential
Dirac cured these diseases in much the same manner as in the pre-
ceding example: First he chosesa string extending from the monopole to infinity. This string sweeps out a two-dimensional sheet
L
Worldline
in spacetime. The string is
chosen completely arbitrarily, except that it is neVer aZ}owed to cross the
____---,of
the electrically charged _____-------,particle
trajectory of the charged particZe.
This is known as Dirac's veto and it will play a crucial role later on! To cure the diseases of
F
mentioned abo-
rg
ve Dirac now introduces a singular electromagnetic field, which onZy lives
Fig. 174
on the string. From a physical point of view Dirac's veto is then clear. If the string crosses the trajectory of a charged particle, the particle would be strongly influenced by the singular electromagnetic field associated to the string. But the string is a fictitious object, which must not disturb the electric charged particles. Now, although the sheet
L
is not compact and consequently not a
regular domain in the strict:sense, it can still be decomposed in an interior region and a boundary. The boundary is simply the trajectory of the magnetic monopole. The interior region is a two-dimensional submanifold, submanifold,
int L , a~
=
and the boundary is a closed one-dimensional
rM
Fix a coordinate system
(xO,X i ,X 2 ,X 3 )
on Minkowski space. Then we a on int E,
may consider the coordinates as scalar functions p~xa(P)
Furthermore we may consider the
-.L a~ (x-x(P»
r-g
a-function,
x
479
II
as a scalar function on int~ Using this we can construct the following two-form on
Space time diagram
int~
int~
Being a two-form we can then integrate
[XO (P) , ••• J
it over the two-dimensional domain int~
(9.25)
and obtain the quantity
)(
p
*as (x) = -gM f l i t (x-x(p»dx ~ ~a r-S S 1:-a "aX
r-g
x
The integral obviously depends on the point x in the a-function and on the indices
a
S.
and
h Fig. 175
We have written
the quantity in the form
Sas(x)
,
because it forms the contravariant
components of a tensor. This is shown in exactly the same way as for the singular current (8.59) of section 8.8. The integral 1
-gMf ~-
r-g
..
~
<5 (x-x (P) )
~a
~B
dx "dx
is a geometrical quantity, independent of the coordinate system we choose on
int~
parameter
Al ,
Usually we parametrize the sheet by a space-like where
-OO
and a timelike parameter
A2 ,
Using such a parametrization we can rearrange the integral as
(9.26)
* B
sa (x)
=
1
It
~
I
2
-gMf u--
r-g
[axa ax B- aP" axa ""fiT axB] cIA ""fiT aP"
I
dA
2
U {(A I ,A 2 ) I AI
dual of another 2-form 5 . As -5 = *(*5) terized by the covariant components:
this 2-form
5
is charac-
(9.27) ~o
ax ..,
I .. ,2
aP"UI\UI\
Remark: Within the framework of distributions the sheet ~ itself can be considered a singular 2-form. This 2-form is closely associated with 5. If U denotes an arbitrary testform we get through a formal computation:
480
II
~fR,;aS(X)Ua8(X);=gd'X
<*slu>
,
-
1
2
-HR,gMf uo (x-x(A ,A »
[axa ax8 axa ax 8] ·~i>:TW-WaIT Ua8 (x)dA
1
2,
dA d x
Within the framework of distributions we therefore have
(9.28) S
We summarize the main properties of the singular form
in the
following lemma Lemma 1 (Dirac's lemma) The singular form 1)
S
has the following properties:
It vanishes outside the string.
(9.29 )
2)
dS
(9.30)
3)
fns
=
*K
=-gM
n
for any cZosed surface
surrounding the mono-
pole. Proof:
(1)
If
x
lies outside the sheet
o'(x-X(A 1 ,A 2 »
(2)
To check the relation _1_
a
s
(Fg a8)
;=ga
L,
then the
o-function
automatically vanish. (9.2~
we
u~e
that it is equivalent to
= KS
(chain-rule)
(Stoke's theorem) = (3)
Finally we must
CCIlIpUte
rounding the monopole.
th= flux through a
closed surface sur-
But a closed surface
n
the monopole is the boundary of a regular domain
surrounding
W con-
481
II
taining the monopole. Consequently we get using (9.29) Ins =
Worked Problem:
laws
Iw ds
=
=
-lw*K
=-gM
D
e~ercise
Remark:
9.2.2 Prove (9.30)by an explicit computation of the integral.
We can also reformulate (9.29 ) as
5*S = - K Within the framework of distributions the boundary operation coincides with the codifferential and we therefore get &*S = -g~l:
= -gMal: = ~rM
= - K
F.
Using Dirac I s lenuna we can now "cure the diseases" of
It was
generated by a magnetic monopole and an electrically charged particle and therefore had the properties dF=-*K
(9.24)
InF=gM
We now choose an arbitrary string extending from the monopole to infinity. Associated with this string we have a weak form
S
with the
properties:
dS
(9.29,30)
= *K
Consequently we see that
F + S
is exact, although singular, and we
therefore can find a global gauge potential A, The gauge potential A
which generate
F + S •
will, of course, be singular too. It will be singu-
lar at the position of the monopole, reflecting the singularity of and it will be singular at the string, reflecting the singularity of S • Formally
S
represents a concentra-
ted magnetic flux flowing towards the monopole. Hence we may formally interpret
F + S
as a concentrated magne-
tic flux flowing towards the monopole position along the string and then spreading out to produce the monopole field. However, it should be emphasized that
S
has no physical meaning. The
position of the string can be chosen completely arbitrarily and the introduction of
S
is a purely formal
trick, which cures the diseases of
F!
Fig. 176
F ,
482
II
9.3 DIRAC'S LAGRANGIAN PRINCIPLE FOR MAGNETIC MONOPOLES We are now in a position where we can state the Lagrangian principle of
Dirac~)Dirac's
(9.32)
S
action consists of three pieces:
= SpARTICLES
+ SINTERACTION + SpIELD
where SpARTICLES
SINTERACTION
SpIELD
q
Jr
A dxpa dA
a dA
q
=
Notice that the interaction term only contains a coupling between the electrically charged particle and the field! However, if we look a little closer at
SpIELD
we see that it contains information about the
monopole trajectory because (9.33)
P
~v
=aA ~
v
-aA-S v ~ ~v
Hence when you vary the trajectory of the monopole, you will have to vary the sheet which terminates on the trajectory. But that will force S~v to vary, and thus SpIELD contains the coupling between the monopole and the field. That the above action in fact gives the expected
equations is the contents of the following famous theorem: Theorem 1 (Dirac's theorem) All equations of motion for a system consisting of monopoles, electrically charged particles and the electromagnetic field can be derived from Dirac's action, provided you respect Dirac's veto, i.e. netic strings are never
allo~ed
the mag-
to cross the worldline of an electri-
cally charged particle. Proof: The proof is long and technically complicated and you may skip it in a first reading. First we list the equations of motion which we are going to derive d2 x a dx ~ dx v a dx S e + ra e e ] _ qP 6--e(1) ( 3) me [ --;rrzdF =-* K ~ v liT dT
---err -
(2)
d*F =-* J
(4)
Then we list the dynamical variables to be varied:
*) The theory of magnetic poles, Phys. Rev. 74 (1948) 817
483 (a) (b)
(c)
Xea(A) xma CA) All (x)
II
Trajectory of electrically charged particle. Trajectory of monopole. Gauge potential.
Observe, that the string coordinates X(A I ,A 2 ) are not considered as dynamical variables. They are fictitious coordinates and shOuld be completely eliminated in the end of the calculations. Okay! Let us go to work: Eq. (1) is not a dynamical equation. It is a purely kinematical equation which is built into the model from the beginning: F + S = dA but by construction we have:
dF + dS = dS = *K
dEIE=o
dF
A A + EB Il Il Il fni=gJa(X)Ba(x)d'x
Eq. (2). Performing the variation
dS I
0
~
-dS
we get:
which leads to the desired equation of motion:
X
Eq. (3) follows from the variation of the trajectory Il(x) . Performing the variation XIl(A) ~ XIl(A) + E~(A) we get from a previous Ealculation (exercise 8.6.3) :
dS I dEl E=O
r
2
-a) dA ([aAa_S]dXCX + A 3L. BY dA adA ax aAa dx B a dxa ~S - 8 (IT' Y - ax8 (IT' ya]dA q J[aA ax dAS)dXCX -B - - - - - Y dA qJC) dXa dA q
Al
From which we conclude: v 8 dx ll dx \ d2 B me g a8 ( ~ + r Ilv
X
484
II
sO we got an extra term. This is the first time We use Dirac's veto! According to Dirac's veto the sheet will never cross the trajectory of the electrically charged particle. Thus S B vanishes on the trajectory of the electric charged particle and we simply throw i% away!
Eq. (4) follows from the variation of the monopole trajectory and it is by far the most complicated step. ' Performing the variation! ~ll (A) ~ ~ll (A) + ~ (A) we get: d2X B dS p m d£!£=O =-~gaB~+
J
(
So this caused no problems! But then we must investigate Final
To do that we must proceed carefUlly. First we observe that when we vary the monopole trajectory, we must vary the sheet too: ~(Al,A2) + ~(Al,A2)+£YV(Al,A2)
The variation ~(Al,Al) should satisfY the boundary conditions: O=YV(A 1 ,A 2 )
on the initial and final space-slice
YV(O,A 2 ) =~(A2)
i.e. the deformed sheet terminates on the deformed monopole tra177 jectory! but apart from these boundary conditions we can choose ~(Al,A2) arbitrarily. It will also be convenient to rearrange the expression for SFIELD! We have previously found (cf. (7.56)):
So we may rearrange the field action as follows
the calculation is simplified to the computation of dS
F d£!£=O
'" d [ J JFllVd£!£=o gm a 61+6 2+6 3
where
4
~
~ (d (Xll+£yll)
(x-x-£y)
dAr
485
63 =
gmJF~vJO"(X-X)~ ~
In the first term
61
II
d;l.ldA 2d"x
we rearrange the integral
Consequently we end up with
( *)
61 = 9
We then attack
aF* v
J
~
---.l!..':'. (x) m axa
ax~ axv "5:T ~ y<Xd;l. IdA2 a
A
62 :
= 9mJ[J;~v(X)O" (X-X)d"XJ~ ~
62
*
1
dA dA
2
~~aXV
= 9mJF~V (x)
arr a57 dA 12 dA
Using that
we may rearrange this as follows (observe that we cannot use partial integration due to the term
(fXJ ') all.
In exactly the same way we can rearrange the
( ***)
a * ~ ax~ 63 = gm aTZ"(F~vy )tfrdA 1 d;l.2-
J
Consequently 61,62 and 63 have been split into two groups of similar terms. Three of the terms obviously match together. Performing the following index substitutions: 2. term ( **) a + ~ v + v ; ~ + a 3. term ( ***) ~ + ~ a + v ; v + a these may be rearranged as
* * * a~~ x 9mf (a a F~v +a ~ Fva +a v Fa~ ) arr
,~v~ ~T dA 1 d;l.2 a;l.·~
but according to the Maxwell equation already established in (2) we knOW that: * * * ~ B aaF~v+a~Fva+avFa~ =-v-gea~vBJ
J
But vanish~s on the string, due to Dirac's vetd Thus the three terms cancel automatically, and we are left with two remaining terms:
dS F
dele=o
Ja
* ~~
ax\!
1
= gm arr(F~vy )~;I. dA
2·
f
a * ~v ax~ 1 2 + gm aTZ"(F~vy )~A dA
486
II
because the boundary of the sheet is nothing but the monopole trajectory. Believe it or not: We are through! dSp. . f dSp Combining the above expression for a:;:;r-- Wl th the preVlous one or a:;:;r-~IE~ ~IE~ we get
From which we deduce
d 2 iS mmgaS [ ~ +
S
di~ diVl
*
diS
r ~v-~ (f[" = -gmPaS(f[
o
This concludes the famous proof of Dirac's theorem.
9.4
THE ANGULAR MOMENTUM DUE TO AMONOPOLE FIELD Consider a system consisting of a pure monopole and a pure electric
charge and suppose that they are at rest at some time
...
...
t=O •
B
E~g
t
/
-.... ,
"q -.-~
--
...
---
E
~
g ... B
Fig.
178
Then they will stay at rest forever! The electrically charged particle will experience the Lorentz force: (9.11)
where and
...... E,B
;;xB=O
F
x
B)
because the particle is at rest. Similarly the monopole
will experience the force: (9.12)
q(E + V
are the field strengths produced by the monopole. But
...
E=O
487
II
which vanishes for the same reason. Next we observe that the chargemonopole pair create an electromagnetic field with a non-vanishing momentum density (1. 42)
Ex B
In fact,
is circulating around the axis connecting the mono-
pole and the electric charged particle. For a monopole at rest at the origin the momentum density is responsible for an angular momentum density -t-
J
-+-+
rxg
and consequently the electromagnetic field created by the charge monopole pair carries a total angular momentum given by
J =
Jjd3~
=
Jrxgd3~
(Bypassing we observe that the angular momentum is independent of the choice of the origin as long as it is placed on the axis connecting the monopole-electric charge pair. This is due to the fact that
-+
g
is
perpendicular to the axis). If we introduce a Cartesian coordinate system with the monopole at the origin and the z-axis pointing towards the electrically charged particle, then a computation, which we leave as an exercise, shows that the components of this angular momentum are given by -gMq
J3
(9.33 )
= 41T
Worked exeraise 9.4.1 Problem:
Compute the angular momentum of the electromagnetic field created by a charge-monopole pair.
So the electromagnetic field does carry a finite angular momentum along the z-axis and the size of the angular momentum is independent of the separation between the two particles. This is the first hint that quantum mechanically the magnetic charge charge
gM
and the electric
are not independent of each other. In fact, we expect some-
q
thing like the following naive quantum-mechanical argument to be valid: The size of the angular momentum along the z-axis should be quantized as
n
~
, where
n
is an integer.
(At this point we cannot ex-
clude half-integer spin, because the angular momentum carried by the field is not of the conventional orbital angular momentum type). Thus we expect (9.34)
n
integer
(Dirac's quantization rule)
But this is very beautiful, because this could provide us with an explanation of why electric charges are quantized in multiples of electron
488
II
charges. If there exists just one monopole, then, continuing our very speculative line of argument, every electrically charged particle would interact with this monopole. Due to the quantization of the angular momentum created by the charge-monopole pair the electric charge must therefore be an integer multiple of 2nft gM
The above considerations can be extended to dyons too. Since the angular momentum is invariant under charge rotations, we can rotate one of the dyon-charges into a pure monopole charge (cf. the illustrative example in section 9.' ): (g, ,'1,) ... (gi ,'1i) = (~,O) (g2'~) ... (g2,«.2) It follows that the angular mome~tum.dens~ty comes from the magnetic field created by the first dyon and the electrlc fleld E2 created by the second dyon. As above we therefore get
S,
1\
where rO is the unit-vector pointing from the first to the second dyon. In terms of the original variables the angular momentum is given by j = g,'12 -g2 '1, ~
4n
0
and we therefore expect the combination (g,'12-g 2 '1,) to be '1uantized.
Now we want to investigate a little closer the angular momentum carried by the electromagnetic field. For this purpose we consider a static spherically symmetric monopole field. You may think of it as being created by an "infinitely" heavy monopole placed at the origin. We then want to investigate the motion of an electrically charged particle in the monopole field. We shall treat the situation in the non-relativistic approximation because then we can later on quantize the motion of the electric charged particle using the Scrodinger equation (Cf. the discussion in ch. 2). Newton's equation of motion is given by d2
r
~
...
B
r vXro
qgM ~
= qvx = IGT
Introducing the parameter, I< = 4 gM , which eventually will become the 4n angular momentum of the electromagnetic field, we rearrange this as (9.35) We proceed to analyse the consequences of eq.
(9.35) in 5 steps.
Step 1: In the first step we look for constants of motion. We multiply both sides of eq. (9.35) with ~, thus obtaining:
o
i.e.
i.e., the kinetic energy is conserved. This is because the monopole
489
II
field performs no work on the particle, since the force
B
is always
perpendicular to the velocity. Thus we have found the following aonstants of motion:
( 9.36)
v (= speed) and
~:
T = ~mv2(= the kinetic energy)
Then we multiply both sides of eq.
(9.35) with
r,
where-
by we obtain
But now we can use that d
2(r2) _ d 2r2
~-~
Integrating this we get r2 =v 2 t>+at+13
(9.37)
Let us fix the time-scale so that at
t=o
the particle is as close as
possible at the origin, i.e. r(O) = d(= distance of closest approach) Then eq.
(9.37) reduces to
(9.38) This has an important consequence. If the speed
vio , then the par-
ticle comes from infinity and returns to infinity. Thus there are no bound states.
[The case
v=O
where the electric charge stay at rest is
a very special situation. In fact, it is unstable: The slightest perturbation and the electric charged particle will move to infinity! It is therefore irrelevant when we investigate the scattering of charged particles in monopole fields.] Then we consider the orbital angular momentum
~:
t =
rxmv
Since the monopole field is spherically symmetric we expect that there should be a "total angular momentum" which is conserved. First we observe that the orbital angular momentum is not conserved:
d ......
. . ...... dv
...... r
rv-(;.v)r r2
dt(rxmv) = vxmv+rxrnat = KrX(vxr') = K where
rr
r =
is the unit vector pointing towards the charged particle.
To rearrange the right hand side we observe that d
A
d ( ... ,
... ... dr
rv-r -dt
d~ = dt ~) = ----r~~
so that we finally obtain:
490
d ~~) 'dt(rXmv
II
dr Kdt
But then we have found the following constant of motion: ...
~
j
(9.39)
rxmV-Kr
This means that the total angular momentum of the system must be iden-
j . The second term,-K~, is, of course, nothing but the
tified with
angular momentum of the electromagnetic field. Step 4:
t
nents
~
Since we have decomposed and
Kr
into two orthogonal compo-
J
we immediately get
(9.40)
But
is a constant, and
K
is conserved. Consequently
J
served too. So, although the orbital angular momentum
t
t
is con-
is not con-
served, we find that its size is conserved! This has a curious consequence: Consider the system at t(-oo)
t=-oo
= mv(-oo)b
and
and
t=O , then we get t(O)
mv(O)d
Particle Distance of closest approach Impact parameter (distance of closest approach for a free particle)
Tangent at infinity
Fig. 179
where
b
is the impaat parameter. But
t
and
v
are conserved. Thus
we finally obtain: b
(9.41)
=
d
Observe, that if we point directly towards the monopole, i.e.
b=O,
we will hit it! But this is a very exceptional situation: If we point directly towards the monopole, the electriCally charged particle Will
IlOve
freely because ...
~
vxB
0
But the slightest perturbation and the particle will react to the force from the monopole field and no longer hit the monopole. Consequently this is an unstable situation and we shall neglect it.
491 ~:
tion (9.39) with
r
we get
(9.42) But
J
II
Finally we observe that multiplying both sides of the equa-
i.e. is conserved and
K
+
~
Cos(J,r)
-K
=J
is a constant. Thus the angle between
~
J
and
is conserved and therefore the partiale is aonstrained to mave on a aone with j as axis:
g
Worked exeraise 9.4.2 da Problem: Compute the differential scattering cross section dn for the scattering of electrically charged particles in a monopole fielu.
9.5
QUANTIZATION OF THE ANGULAR MOMENTUM As we have verified, the magnetic strength
gM
of the monopole en-
ters into the expression (9.39) for the angular momentum of a charged particle in a monopole field. When we quantize the motion of the charged particle we know that the angular momentum becomes quantized, and we expect this to liminate the possible values of the possible values of of angular momentum:
K A
K
•
To determine
we must therefore determine the operators A
A
J l ,J2 ,J 3 . For a charged particle moving in a magnetic field we have previous-
ly determined the Hamiltonian (Cf. eq.
H=
(9.43 where
A
(2.65)):
JL(-i~V-qA)2 2m
is the gauge potential generating the magnetic field. Con-
sequently we must first find
~gauge ~tliU
~
A
producing the monopole
492
field. But here we run into the wellknown trouble that no globally defined gauge potential can produce l:he monopole field, i.e. A will necessarily contain singularities. However, as we have seen in section 9.2, we may concentrate the singularities on the z-axis. We can still make a choice: The singularities form a string which can be either infinite or semi-infinite. As we shall see, the actual choice of the string has consequences on the quantum mechanical level and we shall therefore work out both cases. The semi-infinite string was originally introduced by Dirac, and we shall speak about the Diraa formalism. The infinite string has been especially advocated by Schwinger and we shall speak about the Sahwinger formalism. Let us collect the appropriate formulas in the following table: (9.44)
Sahwinger formalism:
'-/
A
(9.45)
Diraa formalism:
- ~cose,,1P
A
41T
zy _ --Y--_r(r:z) qA
= K
=
K
[
r(r-z)
o
Observe that the Hamiltonian (9.43) becomes a singular operator, but there is nothing to do about this for the moment. We will have to live, for some time with the singularities along the z-axis and we will have accept similar singularities in the wave function w(r,t). Next we construct the operator corresponding to angular momentum. On the classical level we know that ~ ~ ~ 7 j = ~ rxmv - Kr =; r x (p-qA) Kr (9.46) This corresponds to the operator (9.41) ~ r x (-ifiV-qA) - Kr = -ifirxV - ( rxqA + Kr)
*0
A
Consequently we have found the following candidates for the angular momentum operators:
493
Schwinger formalism:
(9.48)
J
2
31 =-ifl(Y~ ay az - z~)
- x~) =-ifl(Z~ ax dz
Kxl:y2
32
1\
J
Dirac formalism:
KX 2 +y2
l =-ifl(y,}Z
1\
(9.49)
- Z~) ay -
1\
J
II
3 =-ifl(
xr
=-in( z-1ax
- x~) dZ
~3 =-ifl(, x~ "y - Ya"x)
x"~ - y"~)
K-L
r-z
- K~
r-z
+ K
Before we continue the discussion we recall some general aspects of the angular momentum. According to Dirac the components of the angular momentum are always represented by three Hermitian operators ~1'~2'~3 satisfying Dirac's commutation relation: (9.50)
If the system is spherically symmetric they must furthermore commute with the Hamiltonian, i.e. [ft;~i]
(9.51)
=
0
This guarantees the conservation of the angular momentum, since according to the quantum mechanical analogue of (2.61) we have ifl
d~i
dt
=
[ft;~.] ~
= 0
In many problems with spherical symmetry we can use the operators of orbital angular momentum (9.52)
which trivially satisfy Dirac's commutation rules (9.50). But we cannot use them as angular momentum operators in this particular problem because they do not commute with the Hamiltonian (9.43). On the contrary the above operators (9.48-49) not only satisfies Dirac's commutation rules (9.50), but they commute with the Hamiltonian (9.43) as well. The verification of this is left as an exercise: Exeraise 9.5.1 Problem: (a) Let
f(X)~ be a differential operator. Show that: [f(X)a"x;g(x)]
= f(X)¥X
i.e. the commutator is a multiplication operator. (Hint: Let both sides operate on a testfunction ~(x)). (b) Show that the operators of angular momentum «9.48-49) satisfy the commutation rules with (c) Show that the operators of angular momentum «9.48-49) satisfy the commutation rules (9.50) and (9.51) as required by Dirac.
494
II
We pr=eed to investigate the sp:ctrum of the angular m:mentum. Using the commutation relations (9.50) one can determine the possible eigenvalues for the angular momentum operators.
(For details see e.g. Schiff [196S]).
First we introduce the square of the total angular momentum: A
It follows that
J2
we can diagonalize states (9.55b)
A
A
A
J2 = J~ + J~ + J~
(9.54)
commutes with J2
and
J
J
1
,J2
and
J
3
,
and consequently
simultaneously. If we label the eigen-
3
we get:
1jJjm
=
(9.55b)
=
rnfI1jJ.
Jm
j (j+l)f'1.21jJ.
Jm
the possible eigenvalues are given by
Furthermore it (9.56)
m
-j,-j+l, .•• ,j-l,j
j
O,t,1,t,2,f,···
Finally it is preferable to introduce the raising and lowering operators (9.57) They satisfy the commutation rules (9.58) and as a consequence the operators J+ and J genfunctions with a fixed j.
fl/j (j+l)
m(m+l) 1}!.
= fi/j(j+l)
- m(m-l) 1jJ.
=
(9.59)
connect the different ei-
Jm Jm
In the particular example of a charged particle in a monopole field we can now use the explicit form (9.50-51) for the angular momentum operator to determine which of the possible eigenvalues (9.56) that are actually realized. Recall that classically (9.40) j2 = 12 + K2 This equation has a quantprn mechanical analogue We have decomposed the quantum operators in the following way:
~
=~
x
(-iflV-qA) -
K; = t - Kr
But then we can square this operator relation:
Worked ea:erais!il,. 9. 5. 2
Problem: Let Show
L b.lO. the operator
'r, = ~x(-iI'lV-q.Jt)
thatt~e operat~r4Products
L'r
both vanishes.
and
r'r,
495
According to exercise 9.5.2
II
both operator products
automatically vanishes. The following operator relation has therefore been established:
~2
(9.60)
Here
K
~2
/I
j2
is a C-nurnber and
r?
and
therefore have the same
eigenfunction~:
= [j(j+l)h 2-K 2 l1jl.
r?1jI.
)~
)m/l = L2+L2+L2 /I
/I
is a positive operator, becau/lse But 1 2 3 are Hermitian operators. Thus the eigenvalues of L2 and we conclude
i.e.
(9.61)
j(j+l) .::
/I
L ,L 1
/I
2
and
L 3
are positive
2 (..:-K) 11
This is our main result because this shows that in a monopole field j=O
is not allowed. Consequently the minimal value of
j
is non-tri-
vial and the system therefore possesses an intrinsia spin! To determine the exact spectrum, i.e. the allowed values of m
eigen~unctions
together with Xhe
expressions for
j2
and
J
3
j
and
Yjm , we must now use the explicit
It is convenient to use spherical coor-
•
dinates. After a long but trivial computation you obtain the following explicit expressions:
Schwinger formalism:
J~2
-
_~2r __l__ JL(s;ne~)
-
11
LSine ae
~
3e
(9.62) J
3
Dirac 3:2
formalism: -
l'l2[
I a ( . eJ,. +_1 32-J_2iflK___I ___ JL+2lC2 ___ 1 ___ Sine ae s~n-ae) SinTGW . I-Cose 3IP . l-Cose
(9.63)
-ifl JL +
alP
K
The Schwinger formalism is the most complicated one. But let us start with it to get a feeling for the general machinery. We look for eigenfunctions of the form gate the eigenvalue equation (9.55a) i.e.
J31j1. flml/J. 3)m :lm -ifl a\llY jm (0,1P) = mflY
jm
(0,\Il)
yjm(e,IP)
• First we investi-
496
II
But this shows that we can factorize y, in the following way )m IP (9.64) Y (0,1P) = Pjm(COs0)e i m jm Since Yjm (0,1P) ,must be a smooth function we demand that m is an integer, so that e 1mIP is periodic with the period 2n You should observe that
e imlP
is still singular at the z-axis where it is discon-
tinuous. (If we approach the z-axis along the line ~O but if we approach the z-axis along the line IP-~ then
then eimlP=l eimlP=_l)
Usually we get rid of this singularity by demanding
O=Pjm(Cos0) on Yjm itself will be smooth. In our models, however, we have singularities on the z-axis from the beginning since the z-axis, because then
...
the gauge potential A
itself is singular on the z- axis. We shall there-
fore neglect the problem. Because
m
can only take integer values we
see that onl.y bostmia states are possibl.e. Then we must use the second eigenvalue equation the expression (9.64) and introducing (9.65)
x=Cos0
m2+2nix+ (K) 2 l-x 2
2 [ - :x [ (1-x );x ] +
(9.55b) . Inserting
we obtain: ] j(j+l) Pjm(x)
-
o
We should then look for regular normalizable solutions of this equation on the interval
[-1,+11. It can be shown that eq. (9.65)
regular normalizable solutions if and only if
m
both integers or both half-integers. Furthermore
and j
has
K
Ii are either is constrained
through the relation
)'
(9.66)
~
hl 1'1
in agreement with the previously obtained result (9.61).
Wopked exepaise 9.5.3 PrOblem: Consider the differential e~uation (9.65) on [-1,1]. Determine under what circumstances it has regular normalizable solutions and show that such solutions are given by -Hm+!'5.)
PJ'm(x)
~ (l+x)
= N.·(l-x) Jm
In the Schwinger formalism integer valued. Consequently
K
-Hm-!'5.)
j-m
~ ~(l-x)
(j +!'5.)
dxJ-m
(j _f.)
~ (l+x)
we know from the beginning that
~]
m
is
must be integer valued tOQ' To summa-
rize we have obtained the following results:
Theopem 2 The Sahwingep
fopmal.ism ,
l.eads to a bosonia speatpum: K
m')'K
ape al.l. integeps
Fupthepmope IKI is the intpinsia spin of the state. The eigenfunations ape given by the fopmul.a
497 (9.64)
II
Pjm(COS0)eimq)
where
(9.67 )
Njm(l-x)
-l:;(m+~) 11
(l+x)
-l:;(m-~) l1d j-m[
j-m (I-x)
(j+~) fl
(J"-~l fl
(l+x)
dx Then we briefly discuss the Dirac formalism
. Here the spectrum is
more easily obtained. From the eigenvalue equation
(9.55a) we get
-i ~Y" alP Jm (0,n) ,0/ = (m-~)Y" fl Jm (0,n) ,0/ But this shows that we can factorize
Yom as follows: )
(9.68) Since
K
i(m--)IP
Yjm (0,1P) = Pjm(COs0)e ei(m-fi)1P
fl
has to be periodic we conclude that
ger valued. But from the general theory, we know that
is inte(m-K) m is either
half-integer valued or integer valued! We have now two possibilities: Either
(j,m'K)
are half-integer valued, or
(j,m'K)
are all integer
valued. Thus we have the possibility of a fermionic spectrum! To check that all the listed combinations are admissible we must investigate the second eigenvalue equation (9.55b). Inserting the expression (9.68)
and introducing
tion (9.69)
[
_
But that is
~[ l-x 2 ~] ax ( ) ax exactl~
x=Cos0
we obtain the differential equa-
2 K (K)2
+ m +2ll1fix+.fi" I-Xi
] -j (j+l) Pjm(x) = 0
the same equation as the one we analyzed in the
Schwinger formalism .! (See eq. (9.65) ). But there we found that K there existed regular normalizable solutions, provided m and fi
were both half-integers or both integers! Thus they are all admissible and we conclude Theorem
:3
The Diraa
formalism
has both a fermionia and a bosonia speatrum:
are all integers or all half-integers Furthermore
IKI
is the intrinsia spin of the state. The eigenfunations
are given by the formula: (9.68)
where (9.67 )
PJ"m(x) = N" (I-x) Jm
-l;(m+K)
(l+x)
-l:;(m-K) d j - m [ (j+K) (j-K)] --"- (I-x) (l+x) dxJ-m
II
498
9.6
THE GAUGE TRANSFORMATION AS A UNITARY TRANSFORMATION As we have seen, the choice of the gauge potential
A
has consequences
on the quantum mechanical level. This may be somewhat surprising: Usually the choice of a gauge potentialis unique up to a gauge transformation and gauge transformations leave physics unchanged. When discussing monopole fields you should, however, be careful! The crucial point is that here different choices of the gauge potential need not be related through a global gauge transformation. E.g. the spherically symmetric monopole field has been represented by the two gauge potentials:
(9.44-45)
A2 =-*(COS0+l) dIP
and
Formally we have
A2 = Al - *dlll but
III
is not smooth throughout space time. It makes a jump somewhere
between
0
and
2rr
and therefore
Al
and
A2
are not related
through a global gauge transformation. On the quantum mechanical level ,things behave a little different. Here gauge transformations are represented by unitary transformations, as we will now explain: Suppose the state of a quantum mechanical system is represented by ~(r,t)
the Schrodinger wave functions
and the various physical quan-
PI,P2,P3,
tities are represented by Hermitian operators
etc. If
U
is a unitary operator then the same state can be represented by the transformed wave function
~ '(r,t) = U~(r"t) and the various physical quantities by the transformed operators
p;
PI' = Ul? IU- I
=
UP 2 ij- I
The transformation: (9.70)
is called a unitary transformation and it leaves all matrix elements invariant ~
~
<~2Ipl~l> ~ <~2Ip'l~i>
=
~t~t~
<~2Iu UPU ul~l>
=
~
<~2Ipl~l>
Consequently a unitary transformation leaves physics unchanged, which justifies that
~.
= U~
represents the same state.
Consider the Hamiltonian
H
=
Jl(-inV-qA)2+q$ 2m
representing a charged particle moving in an electromagnetic field. We now introduce the following unitary operator
II
(9.71)
It generates a unitary transformation which transforms the wave function in the following way
.g
(9 . 72)
(+
l.nX r,t
1}J(~,t) ~ l/J' (~,t) = e
)
+
l/J(r,t)
Furthermore the operator, P = -inV, representing the conjugate momentum transforms according to the rule
~
itx(r,t) (9.73)
(-i~v)e
e
-i* X(~,t)
-qVX- inV = P-qVX Finally the operator
.'" a l.I1 at
representing the energy transforms accor-
ding to the rule i*x(f,t)
... a
l. I1at'
(9.74)
e
. a -itx(~,t) (l.nat)e
... a
lx
qat + l.Ilat'
l/J' (~,t)
Using this we see that the transformed wave function
solves
the transformed Schrodinger wave equation: ifl.;tl/J' =
{2~[-ifl.V-q(A+VX) ]2+q(~
+
tt) }l/J'
But this is obtained from the old Schrodinger equation provided you perform the substitutions:
A~A+VX
(9.75)
and
This justifies our claim that quantum mechanically gauge transformations correspond to unitary transformations! There is however one important difference between the classical situation and the quantlw mechanical situation: Classically we have always demanded that
X(~,t)
should be
a smooth function. Quantum mechanically this is no longer relevant. All - + . gox (t: t) we should demand is that U(r,t) = el.rr ' is a smooth function. Therefore the jump of
X
need not be single valued, but can make jumps provided *X
2~ ! Thus the unitary transforma-
is proportional to
tion is slightly more general than its classical counterpart. With this in our mind we return to our discussion of the spherically symmetric monopole field. The choice of the two gauges (9.44)
Schwinger formalism:
Al = ~osedlP
(9.45)
Dirac formalism:
A2 =~(Cose+l)dlP
suggests that we look at the following "unitary operator" .K
-l.fi:IP e This "unitary operator" is not well-defined unless it is a periodic (9.76)
U
function of
~,i.e.
K
b
500 is an integer! When
II K
h
is an integer we have
previously seen that both choices of the gauge are admissible and therefore we have shown that the Schwinger formalism and the Dirac formalism are unitary equivalent in this case. When
hK
is a half-integer we have previously seen that only the Di-
rac formalism is admissible. The unitary operator
(9.76)
which formal-
ly transforms it into the Schwinger formalism breaks down. Observe, however, that for instance
U = e 2ifi~
is well-defined in that case. It
corresponds to the "gauge" transformation -lr(COse+l) d~ ~ -lr(COse-l) d~
i.e.
where we have moved the semi-infinite string from the positive z-axis to the negative z-axis. Thus, although we cannot convert the semi-infinite string into an infinite string, we can still move the string around, using unitary transformations. Finally you should observe that although the unitary operator (9.76) is periodic for suitable values of
K
,
it is still singular at the z-
axis. This is inescapable, because the gauge potentials themselves are singular at the z-axis.
9.7
QUANTIZATION OF THE MAGNETIC CHARGE
We conclude this chapter by presenting two other arguments for the quantization of the magnetic charge:
1. argument:
On the classical level the position of the string is cho-
sen completely arbitrary. Consider two di'fferent strings
YI
and
Y2 .
They give rise to two different gauge potentials: AI: (-'h, All and A2 : (-
= 2~(-inv-qAI)
+ q
respectively
H
2
Thus we get two quantum mechanical descriptions of the
..
~ ~l1at ,
respectively
~
system:
in~ at = H2 '/''t'
But since the position of the string is of no physical importance, the two situations must be indistinguishable, i.e., they must be unitary equivalent. The two gauge potentials are connected through a singular gauge transformation, A2-A I=dX , where X is singular along the strings. Thus the two descriptions are only unitary equivalent provided eKp[~
501
II
is a single valued function outside the strings. The jump of X must therefore be quantized, i.e. (9.77)
This jump can now be computed using the following trick: Consider a sphere
s2 which we split into two semi-spheres 2 2 s+ and S_ by the equator y, cf. fig.18l Then we get
fYdX = fYA2 -fYAl = fYA2 f-yAl +
Using Stokes' theorem we therefore get
fydX
=
JS!dA2
+
J dAI
J
=
s:
F =
g
S2 (Notice that A2 is smooth on the semisphere
s:
and similarly for AI). Thus
the jump of
is simply given by the
magnetic charge! The ·quantization rule (9.77) can therefore be rearranged as
1r = n21T which is precisely equivalent to (9.34)! 2. argument:
While the first argument was very formal the second argu-
ment will be more physical. It is based on path-integral techniques and is closely related to the Bohm-Aharonov effect. Consider two paths with the final point
r
and
l
B
r
which connect the initial point
2
A
in our space diagram (see fig. 182). Suppose
furthermore that the paths pass on each side of the string as shown on the figure. Then
r
l
is hit by the string.
and
r
form the boundary of a surface
2
n
which
(Strictly speaking we should make a space-time
diagram, but it is not easy to "visualize" this situation in 4 dimensions). The amplitude
K(BIA)
for a charged
particle to propagate from
A
to
B
is
~r(BIA)
the sum of all the amplitudes
Space diagram (time suppressed)
for the particle to propagate along some specific path K(BIA) =
r
,cf.
J~r(BIA)D[r]
(2.17).
2:.S (B;A)
=
Ie~
r
D[r]
When the charged particle moves in an external electromagnetic field it picks up a change in phase given by: .!:gJ A dx a (2.33) efl. r a
A
Fig. 182
502 The contrib.ltion fran
r 1 am r 2
will interfere
II
am
this interference is given by:
e
(9.78)
Using the notation of section 9.2
the integral in the exponent can be
rearranged as follows
~rA
= ~dnA = fndA = fnF+S = fnF+fns
Consequently the change in phase due to the external field is given by
.!.sf F .!.sf S e fl n • e fl n
(9.79)
The first term is Okay. The charged particle ought to be influenced by the monopole field, but the second term must not contribute, since otherwise we could experimentally determine the position of the string using the Bohrn-Aharonov effect! We are therefore forced to put
.!.s f
(9.80)
But fns through
e fl.
S
n =
1
is the magnetic flux concentrated in the string which passes i.e. it is equal to g. Consequently eq. (13.42) implies
n,
(9.34) It is instructive to compare the quantization of the monopole charge with the fluxquantization of a magnetic vortex in a superconductor (section 2.12). In both cases it is the magnetic flux which is limited to a discrete set of values. Furthermore the basic mechanism behind the quantization is also very similar. In both cases the magnetic field interacts with electrically charged particles represented by a Schrodinger wave function ~(t,t) . In the case of a monopole the phase of this wave function can make a quantized jump when we go once around the Dirac string, while in the case of the vortex the phase can make a quantized jump when we go once around the vortex. There is however one essential difference between the two cases: In the case of a monopole we have b'een giving very general arguments to show that the jump of the phase is identical to the magnetic flux. Since we believe that a consistent description of this world must necessarily include quantum mechanics, we therefore conclude that if we should ever succeed to find a magnetic monopole its charge will be quantized. In the case of a magnetic flux string this is no longer so! We can easily construct flux strings where the flux is not quantized. (Observe too that otherwise the Bohm-Aharanov effect would disappear) . It is of course still true that the phase of a wavefunction can only make a quantized jump when we go once around the fluxstring,but this jump need not in general be related to the magnetic flux in the string. When a mag-
503
II
netic fluxstring happens to lie in a superconductor this will have two very special consequences: a) The length of
$
will be constant and nonzero inside the super-
conductor. [The length represents the density of the Cooper-pairsl ~
b) The gauge covariant derivative of essentially the Cooper-currentl
vanishes.
[It represents
This implies the following rela-
tion between the phase ~ and the gauge potential
A:
(9.81)
Conditions a) and b) are characteristic for the so-called oFdered media and it is only through the interaction with an ordered medium that the magnetic flux in a flux string becomes quantized!
SOLUTIONS OF WORKED EXERCISES: No. 9.1.2
(d Fa )F YS +ln ai3 d (FyoF ) 4 i3 yo S Y l(d Fa )FYS+l(d Fa )FYS+lnaSFYo(d F ) 2 S Y 2 S Y 2 S yo
+
+
S~o
S~Y;
Y~o
l(d Fa )Fyo+l(d Fa )FoY+lnaSFYo(3 F 2
0
1 as
2n
1 as
2n
Y
2
[(doFSy)F
yo
Y
0
S yo
2
+(3 y F So)F
[doFi3y+dyFoS+3SFYOlF
oY
+(dSFyo)F
yo
)
1
yo
But according to (9.2 ) we can exchange
with
o No. 9.2.1
a)
A trivial calCulation gives 3R _ y. dY - R
Using this We obtain
3R
=~
3z
R
1 ] __
d [ 1] xFR-Z). 3 [ dX R(R-z) = - R (R-z) 2 I dy R(R-z)
b)
-
y (2R-z) R3 (R-Z) 2
• ,
2[_1_] dZ R(R-z)
=
J;,. R'
We can now calculate the curl! The x- and y-components offer no difficulties, but the z-component is a rather long story. If you split off the term z/R 3 and USe the relation x 2 +y2 = R2 _z 2 _£2 you should, however, obtain the result listed. In the limit £~ the first term reproduces the monopole field, so it is the remaining term Which produces the string. It is preferable to decompose the second term in the following way:
II
504
E2 (2R-z) ~STRING
BE
2E 2 (R-z) + E2 Z gM 2E2 A gM E2 Z A - 4n R 3 (R-z)k - 4n R 3 (R-z)2k
= =
Now, choose a point on the ~ositive z-axis: (?,?,zo)' I~ the limit where E~O we can replace R-z by ~. Thus on the posltlve Z-axlS the two terms behave 2zo like gM 1 1 (E«l) (E«l) 2) 1)
1T
~
n
£T
Thus the first term is approximately constant on the string, while the second term diverges. Consider the two terms as a function of the cylindrical radius p. They have a behaviour somewhat like the behaviour indicated on the figure 183.
p
p
E«l
Fig. 183a
Fig. 183b
Clearly the first term vanishes in the limit where E+O, while the second may very well produce a a-function. To investigate this consider it as a function on the plane perpendicular to the string. We want to calculate the integral:
!:~
-
..3...
E
2
Z
4nJR2R3(R_Z)2dxdy
In the limit E+O the main contribution comes from a region close to the string. But then we can safely perform the SUbstitutions
Using polar coordinates we therefore get
. gM E2 Z 11m - 4n J R2 R 3 (R-z)2 dxdy E~O
The latter integral is actually independent of E as you can easily see by performing the sUbstitution P=Et The integral then becomes a standard integral with the value ~ and We finally get . gM E 3 Z 11m - 4nJR2R3(R-Z)2 dxdy = -gM E~O
But then the limit is a two-dimensional a-function! For a point on the negative zaxis we know that R-z~2tzl and therefore the string terms vanishes in the limit E-+O.
0
505
II
No. 9.8.2 To calculate the integral explicitly we first find a coordinate system in Minkowski space where the coordinate expression for SaS(x) is especially simple. This is done in the following manner: Consider a space-slice. Then we USe the string itself as the positive x 3 -axis. In these special coordinates the position of the monopole is always
(X I ,X 2 ,X 3 )
=
Time
i
intE
(0,0,0)
and the string always occupy the positions:
Xl
= x2 = 0
x3 ~ 0
We say that these coordinates are adapted to the string. We can now evaluate the components of S ,
Fig. 184
~ ~ (x-x~ ( A , I ,2) axY dx e ,A ) EaSyoaIT aIT
SaS(x) = gm I UU As
I
dA dA
2
a parametrization of the sheet we simply choose
XO
i l
= A2 , SaS
Then the expression for
SaS(X)
=0
x2
i 3
= 0,
-AI
reduces to
= -Eas30gmIuo'(x-i(AI,A2))dAldA2
If X3 is negative the sheet can never cross the point X = (xo,xl,x2,X3) integral automatically vanishes. If X3 is positive we get: Iuo~ (x-x(A1
,A2})dA l dA2 =
and the
Iuo(xo-io (AI ,A2))O(XI)O(X2)O(X3_i3 (AI ,A2))dA l dA2 o(X I )O(X2)
Consequently We have shown
SaS(X) = -gmEaS30o(XI)o(x2)0(X3) Now it is trivial to compute the magnetic flux associated with face. If the surface is hit by the string, we can USe (X I ,X2 ) surface in a neighbourhood of the string:
InS =
I-
S through any surto parametrize the
SI2 d Xl dx 2 =-gMIo(x l )o(x 2 )dx l dx 2 =-gM
a
No. 9.4.'
The angular momentum in the direction of then given by: A
roO
j
(where
fa
r[f
~. ~ ~ -M 0 OE (r~ oE)(ror) =47fEoJ-r-30
To evaluate this integral we observe that
A
r
]
d3
fa
x
denotes a unit-vector) is
506
Inserting this relation we get
~
~
I('? .;)
II
I [(f.;) E] d 3x
f .j = - ~4 _ 0_ _ (ij'E)d 3x + ~4 ij. _ 0 _ o nor nor
In the first integral we use that
ij'E =
JLQ3(;_;,) Eo
where
;,
is the position of
the charged particle. In the second integral we use Gauss' theorem to convert it to a "distorted" flux integral: Jij.[('?o:;t) E]d 3X
=
~fcoseSinede~
0
o which thus vanishes automatically. Consequently
I' .j = -K(f 'r') o 0 This shows precisely that the component of j perpendicular to that the component of j along t, is given by -K.
n
;,
vanishes and
No. 9.4.2 Consider a beam of electrically charged particles carrying a specific anergy E . A particle with impact parameter b will be scattered at a certain angle e=e(E;b) . Let us briefly recall the definition of the differential crOss section. For a given scattering angle e we consider the values of the impact parameter b corresponding to e (Remark: There may apriory be several different values of b producing the same
en.
bi -
".\----------r
Fig.
185
Then we consider a small ring between the area
e
and
s+~e
. On the unitsphere it represents
~n = 21TSine~e
The values of b corresponding to this ring will itself form a number of rings from b i to bi+~bi. The total area of these rings is given by ~O
=
E21Tb.~b. 1. 1.
i
The differential cross section
(9.82)
dOte) = dQ
~ is then defined as
beam area = detector area
l:21TMb
21TSine~e
= Ebl~1 d(Cose)
where we sum over all values of b corresponding to the given value of e. We know that a given particle moves On a cone. It is customary to characterize this cone by the angle X, where 1T-X is the angle spanned by the cone. This angle can be expressed in terms of t and K • Considering the triangle shown on fig. 186. we immediately get
507 X Cot 2
2
II
mvb
= lKf = lKf
As we shall see, the scattering angle e depends only on X We can therefore rearrange eq. (9.82) as (9.84) It remains to determine e as a function of X . Here it is profitable to use an adapted coordinate system with the symmetriaxis as the third axis, i.e. j is pointing along the third axis. The position of the particle ;(t) can now be expressed in terms of the radial distance r and the azimutal angle
....
dr
v = dt =
~ ....
drA ~
+ dt Xr
On the other hand we obtain from eq. (9.39) that
jx;
= (rxm;)x; = r2m;
v
(~.;)~ + ~x:t
- ;(~)
and consequently (9.86)
mr By comparing the expressions
(9.85)
and (9.86)
we then deduce
Using eq. (9.38) we can now integrate the above equation for choose
1
= ± ~[~ mv 2
arcsin
~] ro
where + refers to positive time and - refers to negative time. This leads to the following asymptotic values lirrujl(t)
(9.88)
t~co
(Since
has the total increase
1\
_~( --00) :
e
times around the symmetry-axis.) e in terms of X . From eq.
It I
is identical to the angle between
~(+w)
and
508
II
1\
v (-"')
6 1\
1\
Fig. 187
r(-"')
-r(-"')
Thus we obtain cose
-;(+00);(-00) -Cos2 ~[Cos~(+oo)cos~(-OO)+Sin~(+OO)Sin~(-oo)1-Sin2 } 2Cos2
-~- .Sin2r~ • .!L] 2
lmvb
2
1
Using that
~ mvb
= 122+K2 = 11+tan2 X
2
2
=_1_
Cos X 2
we finally obtain the desired formula cose = 2Cos2
f Sin2 [ cos~xJ -
1
2
If we put
!. I; =_2_
Cos X 2
this implies that
d(~~se)
=2Si~os}[-Sin21;+I;Sin~cosI;1
Okay, substituting this into eq. (9.84) mula for the differential crOss section
We have derived the following marvelous for-
.!L I; =_2_ Cos} Having gone that far let us make some comments. First you should compare it with the Rutherford formula for Coulomb scattering
(9.92)
do
lIe 2)2 =1I\2E
1 Sin"S!. 2
Observe that the energy dependence ].s different in the two cases:
(~) monopole ~ i
1 dO) (
Then we consider forward scatterin~, where e is very small. From eq. (9.89) that to lowest order 0=X Furthermore we observe that and
Cos1;;"'O
we see
II
509
In the case of forwards scattering we can therefore use the following approximation
e«
_1_
1
Sin4& 2
which has a structure very similar to the Rutherford formula. For the case of backwards scattering. things are more complicated. On the one hand w~ noW get.cont:ib~t~ons from several values of X. On the other hand ~ b~comes slngular, l.e. lnflnlte at a series of angles en converging towards n . Th~s is due to the term I
Sin2s-~Sin~Cos~
which can no longer be neglected. The denominator becomes zero when
=~
tans
For positive ~ this equation has an infinite number of solutions these solutions corresponds a specific value of e. The angles en are called rainbOW angles.
a
~
'
. To each ~f nfor which ~
=m
d(l
No. 9.5.2 a) b)
~·t
t-r
~.(tX(-iflV-q!)] =
[;x(-inV-qA)].;
-in;- [;xV] -ili.CtxV]-x: to component expressions:
a
=
IT No. 9.5.3 The analysis of the eigenvalue equation (9.65) is most easily performed using the ralslng and lowering operators J± = J l ±iJ2 • Neglecting the normalization for a moment we know that J±P jm
a:
P jm±l
In the Schwinger formalism the raising and lowering operators are given by K
J±
=-~+ a~ ± mc:~: h]
K
=-ll/l-?
[a: ~:5] ±
From this we obtain the following recurrence relations P
jm±l
a:
Il_X2[~ ± mx+~ ax l-xTJ
P.
Jm
These recurrence relations can be further rearranged using the identity P'+fP = e±ff(x)dx In this case
~ [e±ff(X)dx p]
'
510
II
and therefore ff(x)dx
= -~(m+~)ln(l-x)
- ~(m- K)ln(l+x)
The recurrence relations are then rearranged as P.
~l
a:
,
'(
K
'(
-K)t[
-'(
K)
-'(
K)
d (l-x)+;; m+ f;" (1+x)+;; m - f;" P. dx ~
(l-,(J' (l-x)' ±m±fi) (1+,J ±m+r;
]
Ey induction we then deduce
( * )
P.
( ** )
P
,
K)
,
K)
(x)
a:
(1_x)2(n+m+i;" (l+x)2(n+m-i;"
( jm- n x)
a:
(l _x)~(n-m -~) (l+x)Hn-m +i)
Jm+n
This is our main result! From the general theory we know that vanish for sufficiently large n. But this implies that
(l_x)Hm+i)(I+X)Hm-i)p. (x) Jm
and
are polynomials. Eliminating Q.
Jm
Pjm(x)
= Q.
Jm
Pjm+n
and
P.
Jm-n
(x)
we get the following relation between
and
: K
K
Qjm(x) = (I-X)m+il(I+X)m-KQjm(x) But as Q. and Q! are polynomials this is only consistent if are both I~tegers! Jm Using the relations
and
m-J;K
and we see that
m and
K
f;"
are both either half-integers
or integers!
From the recurrence relations (*,**) we can also obtain explicit expressions f£r the functions P. (x) . We know that P .. (x) is killed by the raising operator J+ Therefore we get Jm 1 JJ P .. (x) =
, (.
K)
'(.
K)
C'(l-X)' J +i;" (l+x)' J-i;" JJ . IKI This is a regular normalizable solution provided J~ we then obtain the rest of the functions:
. Using the lowering operator
511
II
chapter 10 SMOOTH MAPS - WINDING NUMBERS 10.1 LOCAL PROPERTIES OF SMOOTH MAPS In this chapter we are going to examine maps from one manifold to another and unfortunately it is going to consist of some rather technical investigations, but there are at least two main reasons for doing this anyhow: First of all maps actually play an important role in physics. When we e.g. discuss space-time symmetries we consider certain maps from space-time into itself, such as rotations or translations, and we then investigate what happens to field configurations during such symmetry transformations.
(This will be explored carefully in the sections
11.4-11.6). It is also worth noticing that field configurations themselves are maps from space-time into a field space. E.g. a complex scalar field, say a charged Klein-Gordon field, is simply a smooth map from space-time into the complex numbers, and properties of such maps may contain non-trivial informations about the system we are considering.
(Examples of such non-trivial properties will be given in section
10.8-10.9 ). The second reason is that it turns out that we can use maps to transfer differential forms from one manifold to another in such a way that the various operations of the exterior calculus, i.e. the exterior derivative, the wedge product etc., are preserved. This gives a much greater flexibility in the computation of various quantities. Okay, as we can not talk us out of it, let
M
m
and
N'n
be dif-
ferentiable manifolds. We want to consider what happens when we construct a map from one manifold to another, explain what we mean by Let
f
f
f : M m ~ N n.
being smooth at a point
We want to
P.
be a continuous map.
Consider a point Po in M and the corresponding point Q = f(P ) O O
i'n
N.
I f we
introduce coordinate systems covering
Po
and
QO' then f is represented by an ordinary
f
Fig. 188
512
II
Euclidean function: (y1, ... ,yn) = f(x1, ... ,xm) Let
U
f- 1 (U) is an open O is continuous. But this shows
Q ' Then
be an open neighbourhood around ~
neighbourhOOd around
because
f
that the Euclidean representative f is well defined in a neighbourm hood of (X 1 , ..• ,x ), where (x01, ... ,xom) are the coordinates of 0 o
PO' Thus i t has meaning to ask if the Euclidean representative f (x 1 , ••• ,x m). [If f is not supposed to be continuous
smooth at
is
o
O
then the Euclidean representative f need not be defined at pOints close to (x 1 , ••• ,x m) and you can no longer wOrk out partial deri-
O
O
vatives!] This motivates the following definition:
Definition 1 m A aontinuous map, f: M ~ N n in
M,
is said to be smooth at a point
i f the aoordinate representative
P
f
is an ordinary smooth m EuaZidean map at the aorresponding aoordinate point (X 1 , •.. ,x ).
o
o
Observe that if just one coordinate representative is smooth, then they are all smooth, because different coordinate systems depend smoothly upon each other. Consider now a smooth map
f.
Then a coordinate representative
will have partial derivatives of arbitrarily high order. The Jacobi matrix
[ayi/axj]
local properties of in
M
is of particular interest to us as it controlls the f. In a neighbourhood of a point P(Xb , ••• ,x m)
o
we have the Taylor expansion
a
i
.
~ AX] + higher order terms
ax]1
Thus when we approximate
f
~n
Xo
a neighbourhood of
P
with a linear
map, this linear map will precisely be generated by the Jacobi matrix. we can make these vague ideas more precise in the following way: Consider the tangent spaces show that
f
T P (M)
and
T f (P) (N ).
We will now
generates, in a canonical fashion, a linear map f
* :
TP
(M )
~ T f (P) (N )
which controls the lOcal properties of
f
and which is represented by
the Jacobi matrix! TO see how this comes about, we proceed in the following way: Let ~P be a tangent vector at curve A(t).
P.
Then
~P
is generated by a smooth
513
II
f
----------------------i.~ t-axis But we can use the smooth map a smooth curve th~ transport of
that
uQ P
N.
A(t)
The transferred curve
into A'
at Q = f(P), which we define to be Q However this only makes sense if we can show
Vp'
is independent of the particular curve vp'
ordinates of with
to transport the curve on
u
generates a tangent vector
generate
f
A' (t) = f(A(t))
Fig. 189
A(t)
we chose to
TO see that this is actually the case, we work out the co-
UQ:
Let A
be given by the parametrization
corresponding to
xi(O).
Then
~P
xi = xi (t) ,
is characterized by the
coordinates dx
i
dt!t=O is characterized by the parametrization The curve A' = f(A) yi = f"i(xi(t)). Using the chain rule, we therefore find the following
...u
coordinates of .
b~
Q
:
d i
=~ dt\t=O
But this clearly shows that the coordinates of coordinates of
~P'
UQ
not on the particular curve
only depend on the A.
Furthermore
we immediately get that the induced map, f* : Tp(M) ~ TQ(N)
is represented by the Jacobi matrix: (10.1 )
[~]\
p
.
Especially it is a linear map! The induced map satisfies a natural composition rule:
II
514
Theorem 1 Let there be given three differentiab~e manifo~ds and two smooth maps: L...!.... M.JL. ,"I. Then (gof)*
(10.2)
=
M,
N
and L
g*of*
Proof: This is an immediate consequence of the chain rule. Let us introduce three coordinate systems: (x 1, ... ,Xl) covering (y 1, ... ,ym) covering
P Q
in L f (P)
and
R
g(Q)
(z1, ... ,zn)
covering
(gof) * is the linear map T p ( Then Jacobi matrix,
L)
in M in ...
TR (
N N)
generated by the
i az ~ ayk axj
i az axj or equivalently,
We conclude this section with a discussion of the behaviour of vector fields. Let there be given a smooth vector field We know that
f
V(x)
on
M
generates a pointwise map
f* : Tp(M) ~ Tf(P)(N) Does this mean that vector field
f*(V)?
f* actually maps the vector field No, not in general!
V
onto another
Let us explain where the
troubles may come from: (a) If f is not surjective, there are pOints on (b)
M
at which
we do not attach any tangent vector. If f is not injective, there exist two points which are mapped into the same point
Q.
P and P 2 1 Consequently the
and V(P 2 ) may very well be map1 ped into two different tangent vectors at Q = f(P ) = f(P 2 )· 1 N, the resulSo when we transport a vector field from to M ting "vector field" will in general be a "multivalued" vector field which is not "globally defined on N". (Consider, for instance, a constant map f which maps the whole manifold M into a single pOint). But even if f is bijective, we may be in trouble. Consider the coordinate expression for the transported vector field: two tangent vectors' V(P )
b i (y1, ... ,yn) =
~. axj
a j (x 1 , .•. ,x n )
515
II
We must demand that it depends smoothly on the coordinates (y1, ... ,yn). If the inverse function f- 1 is smooth too, then everything works out fine and we get
2Lj
(10.3)
ax
If-1 (y)
. a j (f- 1
(y) )
where the right hand side is smooth. But if in trouble! We must therefore demand that
ff
1
is not smooth we are
is not only bijective
but that i t is a diffeomorphism too!
Conclusion:
Only a diffeomorphism generates a well-defined trans-
port of vector fields from one manifold to another.
We proceed to study the local properties of induced map
f.
Let us define the rank of
the dimension of the vectorspace stems around
P
and
f(P)
f
f
f*[T p ( M)].
in terms of the
at a point
P
is is well known that the rank of
equal to the rank of the Jacobi matrix
to be
(Using coordinate syf
is
Df(P).) As in the Euclidean
case we can then define: Definition 2 A smooth map map
f
P
is regular at a point
f
provided the induced
has maximal rank.
We can introduce still a useful concept. Let
f: M
~
N
be a
smooth map and let Q be a point in N. Then we consider the preimage f- 1 (Q) (whic~can be empty if f (M) does not cover Q): Definition 3 *)
Q
We say that in the pre image of critical value.
is a regular value i f
Q.
If
(Remark: Q
Q
f
is not regular,
is regular at all points then it is called a
is also counted as a regular value i f the
preimage is empty.)
*)The above definition differs slightly from the definition used by the mathematicians in the case where Dim M < Dim N. See e.g. Guillemin/Pollack [1974].
516
II
Remember that the tangent spaces have the same dimension as the underlying manifold. We can then distinguish between three cases: Dim M = Dim N , Dim M < Dim N , Dim M > Dim N We will now discuss the three cases in some more detail: (a) Dim M = Dim N As the first case we consider the especially important one, where M and N has the same dimension n. are vectorspaces of the same dimension Then Tp( M) and T Q { N) n, and f is regular at a point P exactly when f* maps T P ( M) isomorphically onto T Q ( N) • This means that the Jacobimatrix Df(P) is a regular square matrix, i.e. it has non-zero determinant. But then the inverse function theorem from analysis tells us that f actually maps an open neighbourhood U of P onto an open neighbourhood V of stricts to a diffeomorphism ,f: U ly nice locally!
Q ~
v.
and furthermore that f reThus f behaves extreme-
f
Fig. 190 -1
The inverse map f IV (f*)
-1
: TQ(
then generates the reciprocal map: M)
~ Tp( N)
where (10.4)
This is a trivial consequence of the composition theoEem (10.2) since -1
the composite map of flU and flU reduces to the identity map. We can also consider a regular value Q in N. If f- 1 (QO) O is non-empty we can find Po such that f(PO) = Q and by definition O f is regular at PO' It therefore maps an open neighbourhood U of Po onto an open neighbourhood V of Q This argument can be O strengthened: Suppose f- 1 (QO) is finite and put 1
=
{P ,···,P k } 1 Then there exist disjoint open neighbourhoods f- (QO)
U i
of
Pi
and open
II
517 neighbourhoods
Vi
of
Q
so that
f
restricts to diffeomorphism
But here we can safely replace V1 ""'V k by their intersection V ~ V. which will be an open neighbourhood of Q • We O i=1 l. 1 can then shrink U to U1 U n f- w) and f will now restrict i i to diffeomorphisms, f:Ui~Vi
f: U!
l.
~
V
•
Thus we have shown the following useful Lemma:
Lemma 1 Let f:Mn~Nn that
f-l(Q)
be a smooth map and let Q be a regular value such
is finite,
say
f-l(Q) = {P1, ... ,P k }. Then there exists a single open neighbourhood V of Q such that decomposes into disjoint neighbourhoods
U1"",U k of
P1"",P k
and
such that the map f restricts to diffeomorphisms: f:Ui~V
,i=l, ... ,k
Fig. 191 Remark: In the above lemma it is essential that It is in g~ false when f-1(Q) is infinite.
f- 1 (Q)
is finite.
n m (bl DimM < DimN : A smooth map f: M ~ N (m < n) which is everywhere regular is called an immersion . Let f be such an immersion. In analogy with the discussion of Euclidean manifolds we expect f(M) to be a submanifold of N. Locally everything works out all right and we can transfer coordinate systems from M to f( M). But globally we may be in trouble because f need not be injective so that f( M) can have self intersections. (See figure 192 ). Consequently we must demand that f is injective. But that is not enough. We still have to worry about the topology since the inverse map f- 1 : f( M) ~ M need not be continuous. We eliminate this by demanding that f is a homeomorphism, i.e. f is injective and both f and f- 1 are continuous. Then it should come as no great surprise to you that the following theorem holds:
518
II
Q
N
M
Fig. 193
Fig. 192
Theorem 2 Let (1) (2)
then
f(M)
f: ~ ~ ~ (m < n) be f is an immersion f is a homeomorphism,
a smooth map such that
is a submanifold of N and
f:
M ~ f(M)
a diffeomorphism.
A smooth map which satisfies both the above properties is called an embedding and we say that M is embedded in N. m
(c) DimM > DimN: A smooth map f: M N n (m > n) which is everywhere regular is called a submersion. Now let f be a smooth map and let Q be a point in N We consider the pre image f- 1 (QO)' cf. O fig. 193, wh~ch consists of all solutions to the equation f,1 (x 1, .•• ,x
i. e.
m)
fn(x 1, .•• ,x m )
We have previously discussed how to construct Euclidean manifolds using equations of constraints (See section 6.3 ). Using those techniques we can generaliz;e theorem 2 , section 6.3 to the following theorem:
Theorem 3 Let f:!.f'",~ (m > n) be a Smooth map. If Q is a regular vaZue in N, then either (1) f-~ (Q) is empty or (2) f- (Q) is an (n-m)-dimensional submanifold of M. We recapitulate the preceding discussion in the following scheme (See also figure 192 and 193 ). m
n N
If
is a regular map then f it is called an
Such a map can be used to construct submanifolds in
m < n
Immersion
N
of dimension m
m > n
Submersion
M
of dimension
f: M
~
m-n
519 In the preceding
~iscussion
II
we introduced the notion of regular
points and regular values. They were important in the characterization of the behavior of smooth maps. But to use the theorems just obtained it will be necessary to control the existence of regular points and values. For a specific map we can of course compute the Jacobi matrix and check its rank and thereby determine the actual positions of the regular pOints and the regular values. But interestingly enough it turns out that it is possible to make general statements which holds for any smooth map. Consider for simplicity ordinary functions
f: R
~
R.
Then a regular
pOint is a point where
fl (x)
*
O. It
is easy to construct smooth functions which are everywhere regular, e.g. fix)
= x.
N
It is equally easy to con-
struct smooth fUnctions which are nowhere regular, e.g. the constant function
f(x)
=
o.
~~----------
M
Thus we cannot hope
____~~x Fig. 194
for general statements about regular
points. But consider now regular values. Even in the worst case of a constant function there is only a single critical value, so the critical values seems to be very exceptional. This turns out to be a general feature of a smooth map. Consider a smooth map,f: ~~Nn. The points in M can be divided into two types: (a) Critical pOints Similarly the pOints in
(b) Regular pOints
N
can be divided into 3 types:
(a) Image of a critical point
(b) Images of only regular points (c) Not an image of any point
A famous theorem of Sard now states that the set of criticaZ points is mapped into a zero-set of vaZues is a zero-set in
every pOint in
N
N.
N,
i.e.
the set of criticaZ
Equivalently we can say that almost
is a regular value.
(For a complete discussion of
zero sets on manifolds you should consult e.g. Spivac [1970] or Guillemin & Pollack [1974]). Remark: By abuse of notation we will call a smooth map regular even if it does possess critical points as long as the critical points are all isolated.
E.g. the function indicated on fig.
194 will thus, slightly
incorrect, be referred to as a regular function.
II
520
10.2 PULL BACKS OF CO-TENSORS Let
f:
M
m~
Nn
be a smooth map. We have seen that we run into
difficulties when we try to transport vector fields from one manifold to another. We will now investigate what happens when we try to transfer other objects. It is instructive to consider maps first: Suppose
R
~
M
is a smooth map into
ly push it forward to a smooth map
,1
curve on
,.
into
M.
represents a
M
But smooth curves gene-
basic mechanism behind the transport of tangent vectors as explained in section
10.1 . Suppose then that
smooth fUnction on ~
N. This 1s
rate tangent vectors and that is the
,. ,/ fo
R
<pof: M
Then we can clear-
M
~N
A smooth map
,-
,. ,.
R
connected with the transport of vectors.
f
M
fo
R
<pof ,-
, ,,
on
,. ,.
,
N M.
~
R
is a
This can be pulled back to a smooth function This can now be used to transport covectors.
,~ f,
COnsider a covector
ill
on N •
can find a smooth function which generates
W,
Le.
W
Then we
N
=
d
R I
and
this smooth fUnction is pulled back to
M
f
•
the smooth function N
we can show that
<pof: d(<po f)
Ni~R
If
is indepen-
dent of the particular function sen we can therefore pull back
w
to
d(<pof) •
determine the coordinates of the pulled-back covector denoted
which will be
*
f w.
The above considerations suggest that it is natural to push forward tangent vectors and to pull back covectors , but it even turns out that pull-backs have nicer global properties than vector transports. This we will now investigate in some detail.
521
II
We start by reinvestigating the pull-back of covectors: be a co-vector field on vector field on on
M.
M
N.
which will be denoted
Then a tangent vector in
tangent vector in
w
Let
We want to construct an associated co-
T f (P) ( N).
T p (M)
f *w.
Consider a point
p
is transported to a unique
This motivates the following definition:
Definition 4 (10.5)
------------------+.
Fig. 195
Clearly
* T P (M ) f,,-,:
~
R
is a linear map since
f*: T p (M )
R ~
Tf(p) (N) is line:r but then we have constructed a globally defined covector field f w on M. It remains to be shown that it is smooth. is characterized by To do that we introduce coordinates. Then f *w the component functions (10.6)
bi(x)
=
-+ cf *wle,>
•
=
-+ cwlf*e i>
=
a.
J
av j . k =
(y(x»~. ~
ax
a.(y(x» J
lY.!. ax~
But the y-coordinates depend smoothly on the x-coordinates since f is a smooth map, and the coefficients a.(y) are smooth functions, J
This shows that the w is a smooth covector field on N. right hand side depends smoothly on x. So the miracle has happened. We never get in trouble when we try
since
to pull back covector
fields!
Clearly the above analysis may be ex-
tended to arbitrary co-tensor fields:
522
II
Theorem 4 Let
Mm ~
f:
Nn
be a smooth map. Then
f * : T (O,k) ( N) f*
i. e.
~T(O,k)(M)
pulls back a smooth
tensor field on
M.
T
If
generates a map
f
c"tensor field on is a smooth
to a smooth co-
N
cotensor field on
N,
then
the pull-back is given by,
*-+
(10.7)
-+
-+
-+
f T (V 1 , •.. ,vk) ; T (f*v1 ,··· ,f,..vk) which can be written out in components as follows, (f *T) .
(10.8)
.
= T.
(x)
1.1···l.k
Observe that if
.
~ i
(y (x) )
ax
J 1 •· ·Jk
and
M =N
f
1
is the identical map, then we
recover the usual transformation formula for the exchange of coordinates on a manifold! The pull-back has several simple properties: Theorem 5 (a)
f
is linear, i. e. * * * f (T+s) = f (Tl + f (S)
(10.9) (b)
f
*
f * (T®S) ;
(10.10)
(c)
f
* (T)
® f
H
*
* (S)
If we have three manifolds L
* (AT)
f
commutes with the tensor product, i. e.
L
M
.5...
(gof)*
(10.11)
N
L, M,
N
and two smooth maps:
then
= f*og*
Proof: Only the composition rule is worth consideration. It can be derived directly from the composition rule for vector transports (10.2): * ~ ~ (gof) T(v1'.·· ,vk)
def -+
-to
T[,(gof) *v1'···' (gof) *vk1
T[(g*(f*v 1 ),···,g*(f*v k )]
. . . , ..• ,v :t f * (g *T)[V k1 1
=
*
-+
-+
(g T)[f*v 1 ,···,f*v k l
If you prefer, it can of course also be derived from the component expression for the pull-back using the chain-rule. We shall esepcially use the pull-back in two cases: a) Metrics: If
g
is a metric on
degenerate cotensor to a cotensor
field
N,
then
g
is a symmetric, non-
field of rank 2. Consequently we can pull it back f *g
of rank 2 on
M.
It will obviously be symmetric, but it need not be non-degenerate!
523
(If e.g.
f
is constant,then
needs not be a metric on
f
*g
II
*
vanishes everywhere). Thus
f g
M.
Exercise 10.2.1 Problem: Let N be a Riemannian manifold with the positive-definite metric g. Show that f*g is a metric on M if and only if f* is everywhere injective. Exercise 10.2.2 Problem:
(a)
(b)
Let M and N be manifolds of the same dimension. Suppose N is equipped with a Minkowski metric and ~et f: M N be a smooth regular map. Show that f g is a Minkowski metric on M 4 Let N be the Minkowski space R with the usual metric and let M be the real line R As the smooth map we consider f(t)
=
(t;t;O;O) f *g
Show that f* is everywhere injective, but that vanishes identically. b) Differential forms:
If
T
is a differential form on
T is by definition a skew-symmetric cotensor we can pull it back to a cotensor metric, i.e.
f*T
N,
then
field. Consequently
field, f *T, which will be skew-sym-
is a differential form too!
The pull-back of a dif-
ferential form is particularly simple. Let us introduce coordinates on M
and
N.
(7.8)
Then we may decompose T as 1 i1 ik T = k! T. . (y) dy l\ ••• l\ dy ~1"
'~k
From the transformation rUle (theorem 4) we get that
f *T is characte-
rized by the components: j1
T . . . . . (y(x) ) ~ i J1 Jk ax 1
jk ~ i k ax
Le. it can be decomposed as j
(10.12)
f *T =
1.k."
But then you see that
T
. J 1 ••. Jk
f*T
( Y (x »
i dx 1
~ i
i l\ ••• l\
dx
k
ax 1
is obtained from
T
by performing the
substitution
~
dx j axj so as usual the formalism produces the simplest possible answer. (10.13)
dyi
Before we proceed with the investigation of differential forms, we will briefly discuss how one can extend the transport of tangent vectors to a transport of arbitrary tensors. One must then restrict to diffearorphisns.
524
II
Definition 5 A diffeomorphism, f:
vectors from
N
to
Mn ~
N n,
generates a pull-back of tangent
M,
f*:TQ(N)
... T
-1
f
(M) (Q)
is a tangent vector at
If ( 10.14)
f
*;;Q =
Q
in
N
is given by
then
(f -1 ) * ;;Q '
and in coordinates it is characterized by the components, ... i [f* Ql
( 10.15)
But once we can pull-back both tangent vectors and covectors we can clearly push forward tensors of arbitrary type. Definition 6
A diffeomorphism ,f: of type
(p,q)
from
M
M
to
n
~
generates a transport of tensors
N:
T f(P) If
T
is a tensor of type
(p,q) ( N)
(p,q)
at the point
p
then
f*T
is
given by
.... * * *.... *--jo (f*T )(w ' ••• ,w; Y1""'v) = T(f w , ••. ,f W; f v , •.. ,f v ) q 1 ... 1 ~ q ...1 ~ and in coordinates it is characteri~ed bY,the components i i ~ ~ k 1 q ~m 1·· kp (f-1(y»~ .•~ k •• k L R.1 •• iq J Jq (10.17) --jo
.aL2. ax
1 ,
ax
p
ay
1
ay
Observe that in this way smooth tensor fields are transported into smooth tensor fields. Of course we could equally well pull them back using the inverse map. Theorem 5 can now be generalized as follows Theorem 6 Suppose f: M n N n is a diffeomorphism inducing the tensor transport f.: T (p,q) ( M) ~ T (p,q) ( N ). Then
(a) (10.18 ) (b) (10.19 ) (c)
f*
is linear, i. e.
f* (T+S) = f * (T) + f*(S) f* (AT) Af* (Tl f* commutes with the tensor product, i. e. f*(T~) = f * (T) GI f* (S) f* commutes with contractions, i. e.
525 (10.20 ) (d)
C C f*( T ) = [f*T1 If we have 3 manifolds
L ,
and 2 diffeomorphisms ,
L ~
(10.21 )
n
II
M
n
,
N
n
of the same dimension
M ~ N ,
then
(gof)* = g*of*
Proof: There is no need to go through all the details. It suffices to obf: M
serve that when
N
is a diffeomorphism we can use
transport a coordinate system from is a coordinate system on on
N.
to
focj>:U
if
f
cj>: U
to
N,
i.e.
N
is a coordinate system
~
M
Using such associated coordinate systems the coordinate ex-
pression for f*T
M then
M
f
reduces to the identity map
yi
=
xi
and
T
and
get identical components.
In the rest of this section we restrict our considerations to differential of the exterior calculus the pull-back of differential
fo~.Because
forms is a very powerful tool. Consider a smooth map
f: M
N.
It
generates pull-backs:
f*:J\k(M)~J\k(N)
k=0,1,2, •••
which we know are linear (theorem 5). We will investigate that part of the exterior calculus which only depends on the manifold structure and not ,e.g., on a metric. Thus, for the moment, we are concerned only with the wedge product, the exterior derivative and the integral of differential forms. Theorem 7 The pull-back commutes with the wedge product and the exterior derivative, i.e.
* (T"S)
(10.22)
f
(10.23)
f*(dT )
(f * T)
"
(f * S)
d(f*T)
Proof: To avoid drowning in indices we work i t only out in the case of one-forms. * * * * * * (a) f (A"Bl = f (A®B-B®A) = f (A) ® f (B)-f (B)®f (Al so the result is a trivial consequence of theorem 6b.
f
* (Al"f * (B)
526 (b)
[f * (d(A)lkR.
=
(NB!)
=
II
~~= (~_ aAi)~~=~~_aAi~
[dAl ij axk ax'!'
ay~
a~ axk axR.
a ~_ a * a * -R. (Ai k) - ---l<£f Ali - axR.[f Alk ax ax ax"
j
_a_(A } . Y':) axk Jaxi
axk axR. axR. axk
= [d(f*A )lkR.
Observe that there is a subtle point involved in these rearrangements: If you actually carry out the differentiations involved, you will pick up two extra terms involving the double derivatives ~ k 1 and ~ 1 k ax ax ax ax but since partial derivatives commute they cancel automatically. U Mathematicians have a very efficient way of representing computation rules like those of theorem 7. Consider especially the rule concerning the exterior derivative
( 10.23)
f
*od
= do f
*
When you pull back a differential form and take the exterior derivative it does not matter in which order you perform the operations. This is clearly an example of a commutative law as you know it from elementary number theory. Now consider the following diagram where we have represented the various maps involved in (10.23) as arrows. If you take an f* element in the upper right corner you can map it along the arrows and evidently there are two possible ways of mapping it down to the lower left corner~where it can arrive either as d (r'T) or as f*(dT), but according to (10.23) they are identical so the end result is independent of which f* route you actually follow in the diagram. This'is expressed by saying that the diagram is commutative.
..
..
As another example consider the composition rule expressed in theorem 6d. Here we have 3 maps involved corresponding to the following diagram and the computation f~ M , g * rule in theorem 6d simply states that this L - -.........---N diagram is commutative. (gof)* When we corne to integration of differential forms we must be some-
what more careful. We want to find out what happens to an expression
like
J rl
rank
k
when we pull it back. Here and
T is a differential form of
is a k-dirnensional orientable regular domain in
N
527
II
We must therefore also investigate what happens to n when we pull that back: In general f- 1 (n) will not be a k-dimensional orientable regular domain of
M
for the following reaSOns: It need not be
compact, i t need not be a submanifold, and it need not be k-dimensional!
To see this consider the following example 1: Let f (t,x)
Examp Ze
Then
f: M
M
= R 2, t 2 _ x2
R
N
and put
is a smooth map although i t is not regular at
N
(0,0).
The set n [-1,1] is a 1-dimensional regular domain in N, and f- 1 (n) is the "strip" between the hyperbolas as shown on the figure.
N +l
f
)
a
n
-1 Fig.
It is not compact, and it is not one-dimensional.
196
0
Consequently we must proceed a little differently. We must try to
n'
n'
in
M
cally onto the regular domain
n.
Consider the following example:
R
and consider the smooth map
find a regular domain
ExampZe
2
Let f(x)
M =
N
such that
f
maps
diffeomorphi-
=
_x_ 1+x2 which has the following graph. Then
n=
[-~,~]
is a 1-dimensional re-
gular domain in N • The pre image f- 1 (n) = R is not a regular domain, but
n'
[-1,1]
c::
M
is a
1-dimensional regular domain in
M
which is mapped diffeomorphi-
Fig. 197
II
528 cally onto f
n
by
f.
(In one dimension this essentially means that
is monotonic when we restrict
f
to
n').
Exercise 10.2.3 Problem: Reexamine example 1 and try to find a suitable l-dimensional regular domain n' which is mapped diffeomorphically onto n = [-1 ;1].
Okay, let
n
be a given regular domain such that we can find
which is mapped diffeomorphically onto
n.
n'
Then we can compare the
integrals
Jn' f*T
and
Theorem 8 Let
T be a differential form on
N
k-dimensional orientable regular domain in
n in
of rank
k.
n
Let
such that
M,
f
be a maps
diffeomorphically onto a k-dimensional orientable regular domain N.
Then if
f
preserves the orientation
if
f
reserves the orientation
(10.24)
Proof (outline): The proof will proceed in two steps. Step 1: Here we assume that
intn
is a simple subrnanifold. Then we
may cover it with a single coordinate system the positive orientation:
~: U
bij., intn c: M;i.e.x i
(~,U)
which generates
= ~i(A1, ... ,),k)
u ,.'" N
Fig. 198
AS
f
maps
intn
diffeomorphically onto
generate a coordinate system on f
,: U
int f(n):
~, int f(n)
c: N
int f(n)
we can use
f
to
529 In these coordinates the restriction of
II f
to
intn
is represented
by the Euclidean function
lli=Ai;
i = 1 , •.. ,k
i.e. the identity map. If f preserves the orientation, then (fo~,U), 1 i.e. the (11 , ..• ,llk)_ coordinates generates a positive orientation on reserves the orientation, then (fo~,u) generates a nef (n). If f gative orientation and this costs a sign when we evaluate the integral
using these coordinates. Okay, working out the two integrals in these especially adapted coordinate systems we get for the case where
f
actually preserves
the orientation:
J
*
ax
i
1
(f T). . (x('\)) - 1 U ~1"'~k . a,\
J1 (y (x))E.Y...-
ax
~
1 . J1 T. . (y(x('\))) EY1 U J1··· J k . aA J1 T. . (y(ll)) ~ 1 u J l' •• Jk all
J J
ax
i. J d'\ 1 ••• d,\k
. a,\k.
Jk ~1 ~~ 1 ax. i k . a,\ Jk ~ d,\1 .•• d,\k j
~k
aA dll 1 ••• dll k =
all
JTf(n)
StfrP 2: Hfrrfr intn is an arbitrary manifold. Wfr can thfrn COvfrr it with a vositivfr orifrntfrd atlas intn c
U
iEI
~.(U.) J.
J.
Using a partition of unity Wfr now cut up T in small pifrcfrs so that frach littlfr pifrcfr is frfffrctivfrly conCfrntratfrd in onfr of thfr simplfr manifolds ~(U.). From stfrP 1 thfr linfrarity of f* and thfr linfrarity of thfr intfrgfal Wfr now dfrdUCfr thfr dfrsirfrd formula for thfr gfrnfrral casfr.1
If we restrict ourselves to diffeomorphisms we can especially push differential forms forward. The statements of theorem 7 and 8 now carryover trivially for the transport of differential forms using diffeomorphisms, i.e. the transport commutes with the wedge product and the exterior derivative and i t preserves integrals.
II
530
10.3
ISOMETRIES AND CONFORMAL MAPS In the preceding section we considered pull-backs of differential
forms. This time we will concentrate on metrical aspects of smooth
(M n'(/1)
maps. Suppose
and
(N n'(/2)
of the same dimension and with metrics
are differentiable manifolds (/1
and
(/2'
If
f: M .... N
is a smooth regular map we have seen that we can pull back f *(/2
new metric manifold f
M
on
(/2
to a
We have now two metrics on the same initial
M and we can then investigate various metrical aspects of
by comparing
(/1
f * (/2 •
and
Consider e.g. a smooth curve smooth curve
f (rJ
on
N •
r
on
M.
It is transferred to a
((.)~
Now suppose we want to compare the lenghts of corresponding arcs. Then we get
N
A A2 . B dP)dA { Arclength of P P 2 on M}= ~A J 2/(/1 dA dA 1 A1 df(p) _ dP and using that we furthermore get f*dA A } - f A21r,(/2 (~, df(P) df(p) ~) dA .-{ ArC~ength of f (P 1 ) f (P 2) on N A2 1 [Arclength of. P P dp f*dA)dA = fA /f*(/2(~r, ~r)dA = on M rela~ive1t~ = 1 lthe metric! f*(/2 .
------ar- -
J(/2(f*~r,
(~r,
L11 ~~~~(
----
l
J
But then we see that we only have to consider the initial manifold
M
where we can compute the arclengths of
metrics
(/1
and
f *(/2
P7P2
relative td the two
The simplest possible behavior of a smooth map and
f *(/2
\\.
f
is
whe·~.
(/1
\
coincide.
\
\\ Definition? A smooth regular map, f:
M
n ~ N n , is called a local isometry i f
it preserves the metric, i.e.
* f(/2= (/1 It is called a globaL isometry i f it is a diffeomorphism too. Observe that isometries preserve arclengths. Isometries are however very special maps so it is usefull to discuss a broader class of
531
II
maps where we still have some controll of what is going on. Definition 8 n
n
~ N , is called a conformal!!!EE. i f it rescales the metric, i.e. there is a strictly positive scalar field Q2(x) on M such that M
A smooth regular map, f:
* f!12
n 2 (x)!11
Remember that by abuse
of notation we call a smooth map regular
even in the presence of critical points as long as they are all isolated. Let us first give some general remarks about isometries and conformal maps to acquaintain you with these new concepts. Suppose
M
a pcint in map
f*
As we have seen a smooth map
from the tangent space
Tp(M)
f
P
is
generates a linear Tf(p) (N).
to the tangent space
Consider first Riemannian manifolds where the tangent spaces are Euclidean. When
f
is an isometry then
*
... -+
so that
-+-+
-+-+
!11 (u,v) = f !12 (u,v) = !12 (f*u,f*v) preserves the inner product between two tangent vectors,
f*
i.e. an isometry preserves the length of a vector and the angle between vectors. When
f
is a conformal map then
2 -+-+ * -+... -+-+ n (p)!1 1 (u,v) = f !12(u,v) = !12(f*u,f*v) so this time the inner product between two tangent venctors is rescaled.
Evidently the lengths are no longer preserved, but angles are since Cos
(ii,v)
=
and this is
rescaling. Thus a conformal map
preserves the angles between arbitrary tangent vectors. But the converse also holds, i.e. any angle preserving map is a conformal map. In fact we can even show Lemma 2 A regular map, f: M - N,
is conformal i f and only i f it preser-
Ves the right angles.
Proof: Clearly it suffices to show that if two Euclidean metrics determine the same right angles, then they are propcrtional. Let
a
be a
II
532 fixed vector. Any vector posant parallel to
~
sant orthogonal to
~.
b
now has a unique decomposition in a com-
and a compoObserve that
this decomposition is independent of the choice of metric because they detennine the same right angles. It is well-
known that the parallel compos ant is given by
...a
Fig. 200
We thus get g' (b,~)
Thus the ratio
g' g
is independent of
b.
g (b,~)
(['
(~,~)
g (~,~)
(b,i::) (b,i::)
Since the metrics g and g' are symmetric coten-
sors it must also be independent of proportional.
g' (b,~)
Le.
g' (~,~)
0
....
a, i.e. the metrics g and g' are
Then we consider manifolds with Minkowski metrics. Now angles are no longer well-defined objects, but observe that the null-cone structure is preserved by a conformal map, i.e. time-like vectors are mapped into time-like vectors, null-vectors into null-vectors and spacelike vectors into space-like vectors. This time we can then show Lemma 3 A regular map, f: M
N,
is conformal i f and only i f it preser-
ves the light-cone structure. Proof: Clearly it suffices to show that if two Minkowski metrics generate the same light-cone structure, they are proportional. Let time-like vector and
~
a space-like vector. Then the line
intersects the light-cone in two different pOints corresponding to the two values and
A2 • On the other hand solves the equation
A1
and
A1 A2
~
a+
be a A~
533
o = g (~+Ab, -:;'+Ab) = g
II
(-:;.,-:;.) + 2Ag (-:;',b) + A 2g (t,b)
......
(Observe that this actually has two roots, since positive and
g (b,b)
,., "1
.
"2
and But here 1.1 conclude that II
g I (c,c)/g
constant k. g I (u,~)
=
1.2
II (~,~) -+ -+ g (b,b)
are independent of the choice of metric so we
1(-:;',-:;') (b,b)
(c,c)
is independent of
...c.
Then we finally get O;{g I (u+~, 'll+~) - g I ('ll,'ll) - g k
= ~g
is strictly
is strictly negative). We then get
g'
Le.
g (a,a)
......
-+-+
-+-+
Let us put it equal to a I
(~,~)}
(u+v, u+v) - g (u,u) - g (~,~)}
= kl
('ll,~)
Exercise 10.3.1 Introduction: Let RP+q be the pseudo-cartesian space equipped with the metric
A dilatation is a map represented by
D
y
An inversion is a map represenr,ed by
I
Y
A translation is a map represented by T Problem:
y
i i i
AX
i
~
<xl x';> xi + a i
(a) Show that a dilatation is a conformal map with the conformal factor
n 2 (x) =
1.
2
.
(b) Show that an inversion is a conformal map with the con-
formal factor
1
2
---
(x) = <xlx>2 (Strictly speaking we must restrict ourselves to the manifold M = R1 {xl<xlx> ,. O} since the inversion breaks down on the "cone" <xix> = 0).
n
.>+,
(c) Show that the. transformation
C = ITr is given by i x1+a 1 <xlx> y 1+2
n2 (x)
=
(1 + The transformation C transformation. (Hint: conformal maps f and g formal factor given by
2
534
II
Wor'ked exercise 10. 3 . 2 Problem: Show that the stereographic projection from the sphere onto the plane rr:S2'{N}~ R2, is an orientation reversing conformal map with the conformal factor
=~ 2
n2(8,(j))
Sin
where
e
2
is the polar angle.
Okay, by now you should feel comfortable apout isometries and conformal maps. We proceed to investigate various concepts which can be derived from the metric. Let Mn be a manifold with metric
Then we have previously in-
g
troduced an equivalence relation between tensors of various types.
(See
section 6.9) In coordinates this corresponds to the raising and lowering of indices using the components of the metric tensor. Thus the cotensor .
'k
T~j = g~ Tki etc. Now when we use a diffeomorphism to transport tensors from one manifold to another we should be careful. Suppose f:(Mn,gl) ~ (N n ,g2) Tij
is equivalent to the mixed tensor with components
is a diffeomorphism and that
T
then there is no reason why sors on
N
and
f.T
and
g2)
(with respect to
•
T'
M
are equivalent tensors on
f*T'
should be equivalent ten-
But we know that
f*
commutes
with tensor products and contractions. If for instance
we therefore get i
(f*T') j Consequently we conclude
Lemma 4 Suppose f:(Mn,glJ + (N n ,g2J is a diffeomorphism. If T and T' are equivalent tensors on M with respect to gl, then f*T and f*T'
are equivalent on
N
Thus we see that unless
with respect to f
f*gl
is an isometry it will not respect the
equivalence relations induced by the initial metrics M and N.
gl
and
g2
on
This observation is of vital importance in physics. Consider a scalar field
~
(M,g)
and let
be the Euclidian space
R3
(with the
standard metric) representing the physical space. The energy density is then given by
H
=
~ai~
ai
~
But here we have used the equivalence relation induced by the From the field itself we can only construct the
covector
d~
metric~
with the
535 components
di~'
So when we use the contravariant
II
di~
components
it is implicitly understood in this expression that we have used the
metric components to raise the index. Thus it would be more correct to write out the energy-density as
= ~gijd.~d.~ l. J
H
This is a very common situation in physics: Indices are contracted using the metric. Now observe that the metric is a fixed geometrical quantity. It is physically measurable, e.g. the arclength of a curve in a physical space can be measured with great accuracy in the laboratory. We are not free to exchange this metric if we want to compare the predictions of the theory with the experimental results. Suppose now that we have been given a diffeomorphism of space into itself f : R3 ~ R3 • Then we can investigate the transformed field configuration
~'
=
f*~
.
E.g. we can compare the energy densities of the original and the transformed field configurations at corresponding points and But they are only identical if
f
is an isometry since the metric is
fixed. This distinguishes the isometries from a physical point of view: They leave various physical quantities invariant, i.e. they act as symmetry transformations.
We have previously discussed time-like geodesics on a manifold with Minkowski metric.
(See sec. 6.7). We will now extend the concept of a
geodesic. Motivated by the discussion in section 6.7 we define Definition 9 A geodesic on a manifold paramatrized by
x~
d2x~
(6.49)
The parameter
M
= X~(A) ~ +
with a Minkowski metric
g
is a curve
which satisfies the geodesic equation
r~
dx~ dx 8 = 0
~i3 dA
dA
A involved in the geodesic equation is called an
~
fine parameter.
Observe that the affine parameter is only determined up to an affine parameter shift
536
A = as+b
II
(aiO)
i.e. a geodesic satisfies the same equation (6.49) with respect to the new parameter
s
.
Consider a point at
• Then
P
around
P.
vp
~p,
P
To see this we introduce coordinates
In these coordinates P is represented by coordinates ai A geodesic through P .with tangent vector
~P
and
-+
there is exactZy one geodesic which passes through
and has tangent vector
x~
M and a non-trivial tangent vector
in
P
by
then has to satisfy the geodesic equation d2xi
-a-v-
+
r
i
dx
jk
j
dxk
"d."I""d."I"
0
with the boundary conditions and But this second order differential equation has a unique solution by the well-known uniqueness and existence theorem for ordinary differential equations. If we perform an affine parameter shift
A
=
as+b
(afO)
the tangent vector placed by
a~p
-+
vp
is re-
so we have ac-
tually shown that to each point P
and each direction at
P
correspond exactly one geodesic. Worked exercise 10.3.3 Problem: (a) Let r be a curve on M parametrized by xCJ. = x(l(s). Show that is a geodesic ir and only if it satisfies an equation of the rorm
(10.25)
A(s)
a::
(b) Show that the corresponding arrine parameter (10.26)
A
=
J:
r
A is given by
expU:2A(SI)dSl ]dS 2
(up to a linear change or the parameter)
The affine parameter is not only characterized by the simplest possible form of the geodesic equation. It is also distinguished by the following property
537
II
Theorem 9 Let vector
A
be an affine par·ameter on a geodesic. Then the tangent
~ = ~~
has constant length, i. e.
(1(v(>.) ,V(>.))
is constant
along the curve.
Proof:
r
d dx el dx i3 , dALgai3CDl CDlJ
dx ll dX" , CDl d>' J
= 2g ai3
Here we can exchange
due to the identity
a dx [ By ldx" gaBCDl ~g dllg"y CDl
=
a dX [ B ]dX" gaSCDl r II" CDl
Thus we get a a 2 i3 d [ dx dXS] dX [d x d'\ gai3CDl CDl = 2g aB CDl CfP" +
r
S
ll dx dX"] II" CDl CDl
and this last expression vanishes automatically for a geodesic.
0
So the affine parameter is a natural parameter on a geodesic! We can now divide the geodesics on (a)
Time-~ike
time-like so that (b)
M into three classes:
geodesics, where all the tangent vectors are
(The affine parameter can then be normalized dx el dxS gaSCDl CD\ = -1 ).
Nul~-geodesics,
where all the tangent vectors are null-.
vectors. (c)
Spaae-~ike
geodesics, where all the tangent vectors are
space-like. so that
(The affine parameter can then be normalized dxel dx S gaSCD\ CD\ = +1 )
We can then show Theorem 10 (1)
Isometries map geodesics into geodesics and preserve the affine parameter.
(2)
(This is valid fo-::' Riemannian manifolds too).
Conformal maps preserve null-geodesics. If parameter on the null geodesic affine parameter on
f(r),
r,
where
then
n2
,\
is an affine
fn2('\)dA
is an
is the conformal factor.
Proof: The first proposition is almost trivial since isometries preserve arc lengths and since geodesics are in general characterized as extrernizing the arclength. Special care should however be paid to nullgeodesics, but here the result will follow from the second proposition.
538
II
To prove (2) it suffices to consider a single manifold consider a rescaling of the metric
g
= g.
n2 g
+
Now if
M and to r
is a
null-geodesic then it especially satisfies the geodesic equation
d~~
(6.49)
W
+
ga~
+
r~
dx~ dx~
~e dI" dI"
=a
Under a rescaling r~~e
the Christoffel field
r~~e
g~e
= r~~e
=
n2(x)ga~
given by (6.47) is changed into
+ o~~aelnn + o~~a~lnn - g~ea~lnn
Thus in the rescaled metric we get
d2x~ -~ dx~ dx il _[d~~ ~ dx~ dX B ] dx~ d ~. dx~ dx il ~+r ~~dI"dI"- "(IT2+r ~ildI"dI" +2dI"d"lnn-[a lnnlg~~dI"dI" Here the first term vanishes because
r
is a geodesic, and the third
term vanishes because the tangent vectors are null-vectors. Thus we get
But according to exercise 10.3.3 this shows that
r
is a null-geodesic
with respect to the rescaled metric. However it gets the new affine parameter given by
~(,,)
=
J"exp o
[Js 2dds (lnn 2 )ds 0
l
]dS 2
o
=
-------
In many applications it is preferable to control the set of all possible global isometries: Consider a manifold
M with metric
and suppose
(
= (
[¢~I
=
id*g
g This follows
= g
By the same argument it follows that if
M ~ M, is a global isometry i f and only i f
it is on the form,
y~
(10.27)
where
A
(A~~)
=
A~~x~+ba
is a Lorentz matrix.
539
II
Proof: The proposition is intuitively clear since an isometry preserves geodesics, i.e. it maps straight lines into straight lines, and thus it must be an affine map of the form
yrJ. = ACl. xi3+brJ. S with respect to an inertial set of coordinates. From this the rest follows easily: ~*g
The pulled back metric
is characterized by the components
Le. Consequently the metric is preserved if and only if
n i.e. if and only if in section
6 . 6)
A
= 'A+~
'A
is a Lorentz matrix.
(Compare the discussion
.0
The transformation (10.27) should be compared with the Poincare transformation (ch. 6
theorem 3) which has exactly the same form al-
though a different meaning! The Poincare transformation (ch. 6
theo-
rem 3) was a coordinate transformation: The points are fixed while their inertial coordinates were exchanged. The new transformation (10.27), on the contrary, moves
the
points
while keeping their iner-
tial coordinates fixed. It is called an active Poincare transformation, We have thus shown that the isometry group of Minkowski space time is the Poincare group. If we start out with the Euclidean space
n R
equipped with the
usual Cartesian metric, then the isometry group is called the EucZidean group of motions. In Cartesian coordinates an isometry will then be represented by a linear map, (10.28) where
A
y, = AX, +b, is an orthogonal matrix, i,e.
A+'A
Exercise 10.3.4 Problem: Consider the unit sphere 52 in R3 . (a) Show that an orthogonal transformation in R3 = AX, where AEO(3) 1 generates a global isometry on 52 . (b) Show that the geodesics on 52 are the great circles. (c) Let r be a geodesic and ~(r) its image under a global isometry ~. Show that the common points of r and ~ (r) must be fixpoints of ~ •
Y
(d)
10.4
540 II Show that a global isometry on 52 belongs to either of the following 3 types: (1) The identity map. (2) A reflection in a plane through Origo. (3) A rotation around a line through Origo.
THE CONFORMAL GROUP We will now study conformal transformations in the Minkowski
space in more detail.
The basic "defect" of conformal transformations
in Minkowski space, as compared with isometries, is that they do not act as linear transformations.
Nevertheless one can generate them
frc:rn linear transfonnations in a higher dirnens ional space.
The linear repre-
sentation of the group of conformal transformations will be the main topic of this section. It turns out to be instructive to consider the general case of a q RP x R equipped with the standard inner pro-
pseudo-Cartesian space duct:
Naively we should only consider conformal transformations, which are smooth everywhere.
This turns out to be too restrictive.
In that
case the only non-trivial conformal transformations are the dilations. We will therefore admit conformal maps which break down at a suitable subset.
The standard example of such a conformal transformation is
the inversion: (10.29) (cf. exercise 10.3.1). origin:
<xix>
= O.
It breaks down at the null-cone through the 2 y2 = xit furthermore
From the relation
follows that the closer a pOint is at the null-cone through the origin, the farther away the image is.
In some sense the inversion therefore
maps the entire null-cone through the origin to infinity.
We would
therefore like to enlarge the pseudo-Cartesian space by adding a "nullcone at infinity".
In the enlarged space the inversion will then be
a diffeomorphism, which exchanges the null-cone through the origin with the null-cone at infinity.
541
II
To get an idea of what that actually means let us take a quick look at the Euclidean space Rn. In this case the "null-cone" is just the origin itself. We must therefore add a single point to Rn, so that the inversion can exchange thi~ point with the origin. For this purpose we consider the unit sphere Sn in Rn 1 Using a stereographic projection from the north pole we can then map Rn onto Sn - {N}. N
ints from infinit
p-axis
s
Fig. 203
P2= Cote
n Thus we can 'replace R by Sn - {N}. The north pole now represents the point at infinity. Furthermore the stereographic projection is a conformal map (cf. exercise 10.3.2), i.e. the metric on the sphere and the transferred metric from Rn are conformally related. From the point of view of investigating conformal transformations we can therefore equally as well work on Sn. Now suppose we identify the Euclidean space Rn with Sn - {N}. What does the inversion look like on the sphere? As indicated on figure 203 it is easy to see that the inversion corresponds to a reflection in the equator. On the sphere the inversion is thus a nice diffeomorphism, which exchanges the south pole (i.e. the origin) with the north pole (i.e. the point at infinity). Remark: The sphere Sn is a compact manifold. By adding a point at infinity we have therefore compactified the plane Rn. In the mathematical oriented literature, the sphere is therefore called the one-point compactification of the Euclidean space.
Motivated by this example we now return to the pseudo-Cartesian RP x Rq . First we enlarge the pseudo-Cartesian space by adding
space
two extra dimensions:
One time like, labelled
u, and one spacelike,
labelled v. In this way we obtain the pseudo-Cartesian space R1 + p x Rq + 1 in which a typical point will be denoted w = (Uix1, ... ,xp;y1, ... ,yq;v) =
(UiZCliV)
The goal is to embed the pseudo-Cartesian space RP x Rq as a suit1 able subset of R + p x Rq + 1 (similar to the embedding of the Euclidn ean space R as a sphere Sn inside Rn + 1 ). This will be done in a tricky manner! In the enlarged space R1 + p x Rq + 1 we introduce the null-cone through the origin:
K.
In the first step we then embed
542
RP
x
Rq
II
K.
isometrically as a subset of
K
section of the null-cone
RP
~
the bijective map
~ (z)
By construction
(
M to be the interu-v = 1, and consider
and the hyperplane q x R ~ M constructed in the following way
Z ;
1)
2
is an isometry.
~
Define
~
To see this we observe that
actually generates a global cordinate
system on
M
The basic frame
vectors are given by
M re-
Consequently the metric coefficients of the induced metric on duces to __ ab ++ ab ++ gab = <eale b > = (+)z Z + <ealeb>(!)z Z = <ealeb> Since the embedding is an isometry we can simply identify
RP
x
Rq
with this particular section M, which henceforth will be denoted q M(R P x R ). The section M(R P x Rq ) will ultimately be replaced by another section of the null-cone
K, but before we proceed with the construc-
tion we must take a closer look at the conformal structure of the nullcone
K
A
K is a null vector
generator of ~
=
It generates a line R,+
w
(u,i",v) ~~
K,
on
= \(u,t,v)
(\(+00
called a characteristic line. A given characteristic line !~ will intersect the section M(R P x Rq ) at most once, and there are characteristic lines which do not intersect M(R P x Rq ) at all. They correspond to the lines which are parallel to the hyperplane
~ = (u,t,v), where
they are generated by null-vectors are precisely the null-vectors where
t2
u-v = 1, i.e. u=v.
But these
= O.
Thus there is a one-toone correspondence between characteristic lines missing M(R P x Rq ) and points on the null-cone through the origin in the original pseudoRP x Rq (cf. fig. 204 a). Consequently the exception-
Cartesian space
al lines represents points on the null-cone at infinity! Consider now two local sections N2 on the null-cone K. N1 and Suppose furthermore that the characteristic lines intersect N1 and
N2
at most once.
Then we have a natural map 11
:
N1
~
N2
543
II
Fig. 204b
Fig. 204a
obtained by projection along the eharacteristic lines (fig. 204b). The basic observation is the following one: Lemma 5 The projection along characteristic lines,
IT:
Nl
~
N , 2
is a
conformal map. Proof: . d uce new coord · ·In R'+p x Rq+' lnates , . We nee d two radial It is preferable to lntro variables r"r Z and p+q homogeneous variables a' , ... ,eP,.p , ••. .pq: Z
r,
rZ Z
z Z u + (x') +
v
z
+ (y' )
z +
...
...
+ (xP ) + (yq)
z
,a P
xp/u
.p' = y'/v, ••. ,¢q
yq/v
8'
Z
x'/u, ••.
(As usual we have troubles with the homogeneous coordinates, which break down when u=O or v=O. Since we are only interested in a local result, we will simply assume that u=O and v=O does not intersect N, or N ). The null-cone K is then Z characterized by the equation If we put the common value equal to r we can introduce the following intrinsic coordinates, (r,e', .•. ,eP,.p', ••• ,.pq) on the null-cone K. A characteristic line is then given by a fixed set of the homogeneous coordinates. Furthermore the local sections can be parametrized as follows: r = f
(e ' , .•. , eP , ¢ 1 , •• • ,.pq)
r = g(e' , •••
,a P ,¢' , .•• ,¢q)
We can therefore use (e', ••• ,eP ,¢' , ••. ,.pq) as intrinsic coordinates on N, and HZ. With this choice of coordinates the projection map IT is simply represented by the identity map!
544
II
We must now determine the various mrtrics inyolved in the game. Consider first q the complete pseudo-Cartesian space R +P x R + with the canonical metric de
2
= -
du
2
1 2 p 2 1 2 q 2 + dv2 - (dx) - ••• - (dx) + (dy) + ••• + (dy )
In the new coordinates this is reexpressed as
K, where
The null-cone ds
2
IK
=
dS~
HZ
dS~
1
2
.
now gets the induced metric: .
kl
and
N1
f2(8,<j»[ =
= r 2 = r,
r [ - e ij (6)d8~d8J + d
Finally the two sections N1
r
C
ij
k
I
(<j»d<j> d<j> ]
are equipped with the induced metrics
N2
j (e)de i d6 + dkl(<j>ld<j>kd<j>l]
g2(6,8)[ - C (e)d8 i d8 ij
j
+ dkl(<j»d<j>kd<j>l]
Consequently
i.e.
Exercise 10.4.1 q RP x R as the interIntroduction: In the above discussion we have embedded 1 using the bijective section M1 of the null-cone K and the hyperplane u-v map
of the null-
Show that the projection along the characteristic lines, 1T:
M1 '" M2
'
corresponds to an inversion in ' RP x Rq •
Having obtained this len'ma we can now apply it to project q M(R P x R ) into a suitable section of K (analogous to the sphere in the Euclidean case). This new section should then include the "nullcone at infinity", i.e. each characteristic line should intersect it exactly once. Unfortunately we run into a slight technical problem: In general it is not possible to find a single section, which is intersected exactly once by each characteristic line. We shall therefore adopt the following strategy:
Let r and 1 Cartesian space
r
545 II denote the radial variables in the complete pseudo2 1 p R + x Rq + 1 :
2 =u2 12 + ••• + (x p2 + (x) )
r 1
Denote by N the intersection of the null-cone the hypersphere
Clearly straint
N is a submanifold defined by the following equations of conr
1
~
Topologically N is therefore a product of the two unit spheres sP Rq + 1 • Consequently N is a hyper-torus: in R1+ p and sq in
N
sP
x
sq
The hyper-torus sP x sq is a nice section on K , but each characteristic line will actually intersect it twice in antipodal points (see fig. 205). Consequently each point in the originaZ pseudo-Cartesian space
sP
x
RP
x
Rq
is represented by a pair of antipodaZ points on
sq.
Fig. 205
q If follows from lemma 5 that the projection map, ~: M(R P x R ) ~ sP x sq, is a conformal map. From the pOint of view of investigating conformal transformations we can therefore equally as well work on sP x sq, except that we must restrict ourselves to transformations which maps pairs of antipodal points into pairs of antipodal points. q In coordinates the projection from sP x sq to RP x R is given by ~(u,z]l,v)
u-v
II
546 consequently the pOints where
u
and
v
coincide are "sent to infin-
ity". But according to our previous analysis these points are in a one-to-one correspondence with the null-cone in RP x Rq . Thus sP x sq is obtained from RP x Rq by adding a "cone at infinity". Because sP x sq is a compact subset of R1 + p x Rq + 1 it is often q referred to as the confopmal compactification of RP x R • q Consider once more the Euclidean space R . In this case the conformal compactq ification reduces to soxS , but the zepa-dimensional sphepe SO only consists of two points, u = ±l. In this case (and only in this case!) the conformal compactification therefore breaks up into two disconnected components {+l} x sq and {-1} x sq Thus we need not double count the points in the conformal compactification because we can cimply throwaway the component {-l} x sq ! But then it is superfluous to enlarge the space with f time-like coordinate, i.e. we simply enlarge Rq to the q Euclidean space R +, and use the unit sphere sq in this enlarged space as the conformal compactification. It is now easy to show that the point at infinity corresponds to the north pole and that the projection along the characteristic lines corresponds to the stereographic projection (cf. fig. 206). Thus we have a nice simplified picture in the Euclidean case.
Conformal compactification of the line.
Fig. 206
RP
Using the conformal compactification of the pseudo-Cartesian space q it is easy to construct conformal transformations on RP x Rq • x R
Consider the matrix group nal matrices operating on
O(1+p; q+1) consisting of pseudo-orthogo1 R + p x Rq + 1 . Each matrix in O(1+Piq+1)
generates a conformal transformation on
sP
x
sq
in the following way:
Notice first that a pseudo-orthogonal matrix S preserves the inner product in R1 +p x Rq + 1 • Especially it maps the null-cone into itself.
Consequently it maps the hyper-torus
caZZy onto a new subset,
S[Sp
x
sql, of
K.
sP
x
sq
K
isometpi-
To get back to the hyper-
torus we then project along the characteristic lines!
In this way we
547 II have constructed a mapping of the hypertorus into itself which we denote by
1[(S']:
Alternatively we can describe 1[[8] in the following way: Each pair sP x sq generates a unique charac{p , -p}, on of antipodal pOints, teristic line
S
formation line
l'.S(P) •
and ~p
x
sq.
.9- p on the null-cone K The pseudo-orthogonal transmaps this characteristic line into another characteristic
The image of
P
is then the intersection between
According to lemma 5 the combined transformation
.9- S (P)
1[[S]
is now a conformal transformation of hyper-torus into itself. From the construction follows immediately some basic properties of 1[[S].
First it maps a pair of antipodal pOints into a pair of anti-
podal pOints, i.e. it can also be considered a conformal transformation RP x Rq . Next the assignment of a conformal transformation to
on
each pseudo-orthogonal transformation constitutes a representation of the pseudo-orthogonal group, i.e. = = 1[[ S2S1]
-
= 1[[ S2]
Notice, however, that the correspondence between pseudo-orthogonal transformations in O(1+p;q+1) and conformal transformations in q q RP x R is not one-to-one. This is because a point in RP x R responds to a pair of antipodal points on
sP
x
sq.
cor-
A pseudo-ortho-
gonal matrix which interchanges antipodal points will therefore generate the identity. matrix:
-
T.
There is precisely one such pseudo-orthogonal
It follows that each conformal transformation is in fact
generated by a pair of pseudo-orthogonal matrices
{S, -S}.
This is
the price we have to pay when we want to represent conformal transformations by matrices! The conformal transformations generated from
O(1+p,q+1) evidently
constitute a group, known as the conformaZ group and denoted by In anaZogy with the conformaL compactifioation of now represent each conformaL transformation in "antipodaL" matrices in
RP
C(p,qJ
x
Rq
C(p,q).
we can
by a pair of
O(l+p,q+l).
Let us investigate the structure of the conformal group a little closer.
548
II
Suppose that the pseudo-orthogonal transformation S actually preserves the additional coordinates (u,v), i.e. that it is on the form
o 5 o
(10.30a)
s E O(p,q)
Then the corresponding transformation on ZCl
~
~ '"
(u; (u-v) zCl ;v)
RPxR q
RP
Rq
x
reduces to
(u; (u-v) s (3z (3 IV) CI
sP x sq
sP x sq
Consequently the conformal group
C.(p,g)
contains the group of
origin preserving isometries 0 (p,g). In the remaining investigation it suffices to consider pseudo-ortho-
gonal matrices of the form
* S
[:
I
*
:]
They will be divided into four general types
+ --2-
(10.30b)
s
C].!
(I)
-2-'-
(10.30c)
Ii (A)
-sinhA
o
c\}
1 + --2-
c
- -r-
coshA
1 -
2 -2-
I
-ev
-
c\}
].!
;
-2-
I
1
c\l
1
-
c"
-2-
Each of these types constitute a representation of a particular subgroup of e(p,g), i.e.
The first type, (10.29)
The second type, (10.31)
The third type, (10.32)
S(I), represents the inversion ].! z
~
z].!
T(c].!), represents the group Of translations z]'!
~
z].!+ c].!
D(A), represents the group of diZatations z].! ~ eAz ll
549
Finally the fourth type, formal translations
II
C(c~), represents the group of special con-
(10.33)
Let uS check the last of these statements just to illustrate the principle: A point z~ in RP x Rq is represented by the point [u,iu-v)z~;v] on the hypertorus sP x sq. By the pseudo-orthogonal transformation C(c~) this is mapped into the point [u ' ; z'~;v'] where
u'
z,1l = (u-v)z~ + c ll (u+v) v'
RP
In
x
=
Rq this corresponds to the point z,~ ~,
(u-v)z~ + c~(u+v) (u-v) + 2
u-v
+ 2
u-v
[u; (u-v)z~; v]
It remains to determine the factor (u+v)/(u-v). Since on the hyper-torus, it follows that u2 + (u_v)2 x 2 = v 2 + (u_v)2y2
is a point
From this relation we especially get u 2 _ v2 = (u-vt{y2-x 2 ) i.e.
u2 _v 2
----2
(u-v) Inserting this the induced mapping from z~ ~
u+v
=-;:;::v RP x Rq into itself finally reduces to
z~ + c~
We can also summarize the above findings in the following way:
RP
Pseudo-orthogonal transformations in D1+P x Dq+l:
conformal transformations in
Transformations preserving both and v
Pseu~o-or~ho&onal ~ransformations: ,"S" E O(p,q) • z ~ S SZ
u
zCl
x
zCl + c Ci
Transformations preserving
u-v
Translations:
Transformations preserving
u+v
Special conformal transformations: CI zCl + cCl
Transformations in the u-v-plane
Dilatations:
Reflection
Inversion
in the u-z-hyper-plane
z CI _
e Az (l z~
zCI ~
q
R
550
II
By now you should feel comfortable about the general structure of the conformal group
C(p,q).
It has the following simple characteri-
zation:
Lemma 6 On a pseudo-Cartesian space
RP
x
Rq
the conformal group,
C(p,qJ,
is the smaZZest group containing the isometry group and the inversion. Proof: First we make the trivial observation that we can generate special conformal transformations using only translations and the inversion. This is due to the identity
(cf. exercise 10.3.1). dilatations.
It remains to show that we can also generate
But that follows from the identity
a 1 = _c = a = a = _c a D(1+
(The verification is rather tedious but straight forward.
Perhaps the
quickest way to verify it is to show that _c a
a
a
_c a
1
[T(1+
RP
x
Rq).O
We conclude this section with a few remarks about whether or not C(p.q) in fact contains all possible conformal transformations. The answer is affirmative if the dimension of RP x Rq is greater than two, though we will not prove it here. The two-dimensional case is an exceptional case because any holomorphic map of the complex plane into itself is in fact conformal. This will be discussed in section 10.6. Never-the-Iess
C(2) still depotes the conformal group constructed a-
bove. Notice that
C(2)
is essentially the Lorentz group
Exercise 10.4.2 Problem a) Show that the conformal group C(2) of the complex plane into itself:
az + S
w = yz + 0
and
w =
~ 0
yz +
0(1.3)!
consists of Mobious transformations (ao
- Sy
1)
II
551 b) In the following we represent the Mobius transformation
w(z)
(az + S)/(yz + 8)
=
by the 2x2-complex matrix:
rLya ~]
Show that the matrix group SL(2,C) (consisting of 2 x 2 complex matrices with determinant 1) constitute a prepresentation of C (2) (consisting o of orientation preserving Mobius transformations).
10.5 THE DUAL MAP It still remains to investigate that part of the exterior calculus which also depends on a metric (and orientation), i.e., the dual map and its related concepts, like the inner product between differential forms, the co-differential
5
~.
and the Laplace-Beltrami operator
The dual map plays a special role, because it is connected to the
M be a manifold of dimension
dimension of the manifold. Let a dual map
Then
*
*
a differential form of rank manifolds with dual maps
Okay, suppose
(m-k)
*1
*2
and
and let
*1(f*T)
map. Then we would like to compare
N
M and
can only have the same rank if
and
M
+
and n N
f*(*2T)
N
with
k
into
are
be a smooth but they
are of the same dimension! M
and
N
have
n.
N
To begin with we let with a metric
M
m
f
For this reason we will assume in what follows that the same dimension
m
converts a differential form of rank
be an n-dimensional orientable manifold
We will of course assume that
g
M is an n-dimensio-
nal orientable manifold, but a priori we do not assign any metric to
M.
f: M
We will let
back the metric
~
N
be a smooth regular map. Then we can pull
to a metric
g
f*g
on
M
(cf. exercise 10.2.2).
Equipped with this metric we have a Levi-Civita form map
*f
Th~or~m
M.
on
(10.34)
f
and a dual
12
Let th~ regular map sing). Then and
E
f We can then show the following theorem:
commutes
f
Mn ~ Nn
f*E = Ef
be orientation preserving (rever-
(f*E = -E ) f
(anti-commut~s)
with
th~
dual map, i.e.
(10.35)
Proof: It is sufficient to check the case where tion. Let
P
be a point in
maps an open neighbourhood neighbourhood
V
of
Q in
M
U
N
As of
P
f
f
preserved the orienta-
is regular, we know that it
diffeomorphically onto an open
552
II
o
Fig. 207
U
We may safely assume that
is in the range of some positively orien-
(~,O)
ted coordinate system
in these coordinates, then
[fo~,Ol
But then we can use
positively oriented coordinate system covering f
V.
as a
If we express
reduces to the identity map:
yi
f
= xi
.
But that is not all! When we pull back the metric, then the metric components are unchanged
as are the components of any differential form
T
we might like to
pull back. Thus locally the two manifolds are indistinguishable if we identify them using the map
f.
0
We now address ourselves to the more realistic situation where we have been given two orientable manifolds with metrics and If
f: M
dual maps
~
*1
N
is a smooth regular map, then we want to compare the
and
*2
on
pulled back to a metric dual maps generated by
M and
f*g2 f*g2
N.
on
M
~nd
g2
But observe that
g2
is
and the relationship between the are completely controlled by
theorem 12. We are thus really left with the problem of comparing the two metrics
gl
and
f*g2 • This reduces the problem to the following:
We consider just a single manifold
M
and see what happens with
the dual map (the co-differential, the inner product etc.) when we exchange one metric with another. Of course everything turns out to be simplest when we do not exchange the metric at all! When we work with isometries we can therefore immediately take over the results obtained in theorem 12. Combining this with theorem 7 and 8 we then get
553
II
Theorem 13 Let
f : M
(a)
~
N
be an orientation preserving local isometry. Then
f
commutes with the dual map, i.e.
f
commutes with the co-differential and the Laplace-Beltrami
(10.35)
*l [f*T] = 1*[*2T].
(b)
operator, i.e. (10.37 )
5f*T
(c)
If
f
=
f*&T
hf*T = f*hT.
is a global isometry it preserves the inner products
= <Tis>
Almost equally simple is the case when
f
is a conformal map. It
M
will be preferable to consider at first a single manifold
and see
what happens if we perform a rescaling of the metric. Theorem 14 Let
g
and
manifold (a)
g' = Q2(X)g
n M and
E
are conformal ly related:
E'
E' = Qn(x)t
(10.38)
(b)
be conformally related metrics on the
Then
Let
*T (10.39 )
* 'T
Then
k
T be a differential form of rank
and
are conforma l ly related: *'T = Qn-2k(x)*T
Proof: is an
(a)
g' (x)
nxn
=
matrix. Thus we have:
DetC'
=
Det[Q2(x)C]
=
Q,n(x)DetG
from which we immediately get
RlXT (b)
Ea
1
••• a
"'an
When we compute the components of
n
*T we must raise the index.
This involves the contravariant components of the metric, l But they are the components of the reciprocal matrix
gij.
G-
•
Cl"
'ck U
Therefore they transform with the reciprocal factor: g,ij
=
Q-2(x)gij
Using this we immediately get 1
blCl
k ;gr('XT Qn-2
k
E al ••• a
~k
bl ••• b
k
g'
bkc k
•• 'g'
1
(x)-, Ig (x)
k.
E
al"
'an - k b l • "b k
g
blCI
T c •••• c
.
"'g
bkc k
k T
n
554
II
Combining the results of theorem 12 and 14 we then finally obtain:
Theorem 15 Suppose
(Mn,gl)
and
(N n ,g2)
are orientable manifolds. Let
f
be an orientation preserving (reversing) conformal map, so that f*g2
,,2 {xl gl
=
(a)
(10.40)
(b)
(10.41)
Then (f*E2
f*[E21 = "n{X)El f*[*2T1
Observe that if
n
"n-2k(X)*1[f*T1
:r
is even and
=
(f*[*2TJ k
has rank
-"n(X)El) -" n
"2
n-2k
(x)*1[f*T1)
then
ac-
tually commutes (anti-commutes) with the dual map
(10.42)
f*[*2TJ = *df*TJ
{f*[*2T1 = -*df*Tll
10.6 THE SELF-DUALITY EQUATION We have previously introduced the concept of self-dual and antiself-dual forms (see ex. 7.5.3 and ex. 7.6.9). They can only be constructed in spaces of even dimension and they satisfy the equation (10. 4 3) rank T = ~ dim M *T = AT but due to the identity (7.43) A is constrained to very special values i.e.
A = {
±l ±i
if
**T
T
**T -T Observe that in the latter case the self-duality equation only works if
for compZex valued differential forms. For future reference we collect the various possibilities in the following scheme:
,
(10.44)
The self-duality equations
Euclidean metric
Minkowski metric
Dimension
T
2
I-form
*T = tiT
*T = ±T
4
2-form
*T = ±T
*T = ±i T
E:nercise 10.6.1 (a) Let n = dim M be even and let rank T = ~. Show that T can be decomposed uniquely into a self-dual and an anti-self-dual part
Problem:
+
(b)
-
T=T +T. Show that the above decomposition is an orthogonal decomposition, i.e. (T+ I T-)= 0, in the case of a Riemannian metric.
555
II
Next we observe that the self-duality equation has the following important invariance property:
Theorem 16 The self-duality equation is conformally invariant. Proof:
T is a self-dual n rank T = '2 so that a conformal map commutes (anti-commutes) with the dual map. This is an immediate consequence of (10.42). If
or anti-self-dual then
dim M
=n
is even and
'0
Observe that an orientation reversing map actually maps a solution to the self-duality equation
*T
AT
=
into a solution of
*T
= -AT
.
The self-duality equation is by far the most important example of a conformally invariant equation and it comprises many well known partial differential equations in disguise. To get familiar with it we will study in some detail the historically most famous self-duality equation: ILLUSTRATIVE EXAMPLE: THE CAUCHY-RIEMANN EQUATIONS. Here we consider maps from
R2
into itself: , 1.e.
{WI W2
= w2 (x ,y )} = WI(X,y)
Such a map generates a complex valued l-form dw = dWI + idwt , and we will demand this to be anti-self-dual *dw = -idw • If you write out the components you will immediately find that this is equivalent to the following partial differential equations (10. 4 5)
2..:'i.= dY
_~ dX
which are nothing but the famous Cauchy-Riemann equations! Here it is preferable to introduce complex numbers, i.e. to identify in the canonical fashion and consider complex functions i.e.
R2
with
C
w = w( z)
Such a complex function generates the complex-valued l-form dW dWz a?lz 3zd
dw = + (For the notation you should consult exercise 7 .6.9and 7.4.3~ But here dz is antiself-dual and d~ is self-dual, so this is precisely the decomposition of dw into a self-dual and an anti-self-dual ~art. (Compare with exercise 10.6.1) But then dw is anti-self-dual if and only if ~ = 0 which is the complex form of the CauchYZ Riemann equations. To summarize we have thus shown (compare with exercise 7.6.9) :
Lermza 7 A complex function w = 'W( zJ is holomorphic (anti-holomorphic) if and only if it generates an anti-sel/duaZ (self-dual) l-form dw.
556
II
Let us now characterize the holomorphic fUnctions geometrically. Here we have the following well-known characterization:
Theor-em 17 A non-t1'iviaZ compZex funation w = w( z} is hoZomorphia (anti-hoZomo:rphia) if and onZy if it is an o1'ientation pr-eserving (or-ientation r-eversing) aonfoT'mal map. Proof: We know that the Cauchy-Riemanns equation are con formally invariant, i.e. if w is any holomorphic fUnction and f: C ~ C is an orientation preserving conformal map then w( f( z) ) is a holomorphic fUnction too. Applying this to the identity function w( z) = z which is trivially holomorphic, we see that f is holomorphic too. To prove the remaining part of theorem 17 we consider the standard metric on C given by
(10.46)
iT
= cfx!sdx
+ dyedy
= ~(d2PIiz
+ d7M.z)
If f: C ~ C is any smooth map represented by the complex function therefore obtain the following pulled back metric f*g =
f f*g
= f(z)
we
1f (c/.w0!£i + IhQdw)
dW 2 = (ldZI
When
w
+
1 dW 2
dZ I ) g +
d d 3W (az -az)dzxdz + (a;: W W
dW - d:z)dv
is holomorphic (anti-holomorphic) this reduces to dW 2
= 1a;: 1
(f*g =
g
'w
1~zl
2
g)
which shows that in both cases f is a conformal map! (Strictly speaking we should cut out the points where w is stationary, i. e. 3w = o. They correspond to the points where the map is not regular, and for a non~ivial holomorphic function they are isolated). n
One of the reasons for the importance of the Cauchy-Riemann equation to physics, lies in the fact that any solution to the Cauchy-Riemann equations is automatically a harmonic function, i.e. a solution to the Laplace equation in 2 dimensions: (10.47)
d*dq,
=
0
Le.
To see this well-known result in our new formalism we let w =w(z) be a holc:rrorphic function. Then rlw is anti-self-dual and we get d*dw = d[-idw) = -id", = 0 • Thus the real and imaginary parts Wl and W2 are harmonic functions. (The same argument holds for anti-holomorphic functions.) This is a common feature in many applications of the self-duality equations. They are brought into the game to simplify the search for a solution to a second order equation. In this case it is the Laplace equation (10.47)
d*dq,
=
0
557
II
which is trivially solved by solutions to the self-duality equations.
*dcp
(10.48)
±idcp •
=
But the self-duality equations are only first order equations, and are thus easier to handle! Interestingly enough it turns out in many applications that the "relevant" solutions to the second order equations automatically solves the first order equations!
("Relevant" means solutions that sa-
tisfy appropriate boundary conditions, integrability conditions or other conditions imposed by the problem at hand.) Remark: In the example concerning the Laplace equation this is precisely known to be the case, i.e. not only is the real part of a holomorphic function automatically a harmonic function, but all harmonic functions are generated this way. Let us also take a look at this well known result in our new formalism: Suppose cp is a real harmonic function and take a look at the complex valued l-form 2
l1 az d z
It is necessarily anti-self-dual and because
is harmonic it is a closed form
cp
d(2 l1 ~P dZAdz = 0 az dz) = 2 azaz But then it is locally generated by a complex valued function
(* )
df
f,
= 2l1 az dz '
and by lemma 7 this function is holomorphic. It remains to be shown that real part of f . Now observe that since cp is real we have
-
cp
is the
an. -
df = 2#dz
(**) Combining (*)
and
(**) we now get
~d(f+f) = ~dz + ~Z = dcp i. e.
cp
~(f+f)
Exercise 10.6.2 Problem:
(a)
(up to an irrelevant constant).
(The wave equation). Show, using theorem 15 directly, that the Klein-Gordon equation for a massless scalar field cp in (l+l)-dimensional space time,
d*dcp =
(10.49)
(b)
a2
axz-)CP
=
0]
is conformally invariant, i.e. the wave equation is conformally invariant. Suppose cp is a massless scalar field generating a self-dual (antiselfdual) field strength:
*dcp (c)
a2
[(w
0
=
[*dCP = -dcp]
dcp
Show then that it automaticalJy solves the wave equation. Show that the self-duality equations reduce to,
~
= -
~
[~ = ~]
and that the complete solution is on the form,
558
II
= f(x-t)
[
=
f (x+t)
1,
where f is an arbitrary smooth function. Let
(10.50)
)
.
After this long digression on the Cauchy-Riemann equations we conclude with the most important example of a conformally invariant field equation in classical physics: Theorem 18 Maxwell's equations for the electromagnetio field are oonformal invariant in (3+1)-dimen?ional spaoe-time.
Proof: This can be proved directly using theorem 15 (compare with exercise 10.6.2 and 10.6.3 below). It is however instructive to reduce the Maxwell's equations directly to the self-duality equation, although it is a little bit tricky: First we complexify the field strength, cf. exercise 9.1.3 F = F+i*F
(10.51)
In this way we produce an anti-self-dual 2-form *F On the other hand: If
=
F
*F-iF = -iF is an anti-self-dual 2-form then it is neces-
sarily of the form (10.51). Maxwell's equations can now be cast into the following complex form: (10.52 )
*F
-iF
rtF
=0
Here the second equation is invariant under arbitrary smooth transformations, while the first one is invariant under orientation preserving conformal transformations.
(Under an orientation reversing map
mapped into a self-dual form *F' But then
FT
real part of
F'
F
is
i.e.
= iF
is anti-self-dual and has the same real part, so the F'
will still solve Maxwell's equations.)
'II:Bl'oi,se 10.6.3 Probl.em: Show, using theorem 15 directly. that the Maxwell equation for the massless gauge potential. A in (3+1 )-dimensional space time.
rl*dA = 0
is conformal.l.y invariant.
[ (a )J a)J) -~ A
-
a
v (a)J A )J ) = 0 1 •
559
II
ILLUSTRATIVE EXAMPLE: CONFORMAL MAPS ON THE SPHERE In the previous example we have discussed conformal maps of the plane into itself and we have seen that they were given by the non-trivial holomorphic functions, f:C ~ C. This time we will consider conformal maps from the sphere into it'tself, f:S2~ S2. It is worth observing that this example is closely related to the preceeding example, since the stereographic projection from the sphere into the plane is itself an (orientation reversing) conformal map. (See exercise 10.3.2). You might think then that the results obtained in the previous example can be carried over trivially, i.e. the conformal maps from the plane into itself are transferred to the conformal maps from the sphere into itself ueing the conformal stereographic projection. But as we shall see this is not quite true, so let US take a closer look. Suppose we have been given a conformal map, f:C ~ C. We would like to lift it to a map 1:S2 ~S2. But here we can get into trouble, since the lifted map is apriori not well-defined at the north pole (which corresponds to infinity in C).
C
C
Fig. 208 Consequently f can only be lifted ir it has a well-defined limit at infinity, since only in this case we can extend 1 by continuity at the north-pole. Consider then a map from the sphere into itself, g:S2~ S2. We want to project it down to a map g:C ~ C. But here we should be careful about the points which are mapped into the north pole. They give rise to singularities in the projected map In the case where g is conformal, i.e. g holomorphic (anti-holomorphic) this means that g will get poles. Combining these observations you might now guess the answer:
g.
Theorem 19 An orientation preserving (reversing) map g:S2~ S2 only if it is projected down to an algebraic function w(z) = p(z)/Q(z)
(10. 53 ) ~here
w(z)
is conformal if and
p(z)/Q(z)
P,Q aPe arbitrary polynomials.
We now show this in detail: 2 The basic idea is to treat the sphere S as a complex manifold. We already know what a real manifold is (cf. the discussion in sec. 6.1-6.3), and the corresponding definition of a complex manifold is a straight forward ~neralization: We identify R2n with Cn. Let U be an open subset of C and f:U ~ C a smooth function of n complex variabels: f(zl, ... ,zn). We say that f is holomorphic (i.e. complex-analytic) it either of the following two equivalent conditions are fulfilled: 1) i=1., ... ,no 2)
The function f can be expanded in a convergent power series around each point (z~, ••• ,z~) in U:
560 n
f ( z 1 , ... , z )
=
~ E
II
a
Pl",Pn Pl",Pn
(1 1 Pl n n Pn z -z ) •••• (z -z ) 0
0
k
Consider furthermore a mapping ~ from an open subset U of en into e :
~(zl, ... ,zk) = (~~zl, ..• ,zn); ... j(zl, •.. ,zn) We say that the mapping ~ is holomorphic provided each of its components ~l, .•• ,~k are holomorphic fUnctions. Suppose now that M is a real differentiable manifold of dimension 2n. A complex coordinate system on M is a smooth regalar homeomorphism ~ of an open subset U in en onto an open subset ~(U) in M. Furthermore two complex coordinate systems are compatible provided the transition functions are holomorphic maps. With these preparations we finally arrive at the main definition:
Definition 10 A complex manifold is a real differentiable manifold M equipped UJith an atlas of compatible complex coordinate systems. Notice that in the complex case the concept of a smo6th function has been replaced by the concept of a holomorphic funtion. We can now show that the sphere 8 2 is a complex manifold. As complex coordinate systems we use stereographic projections from e into S2. It suffices to consider two such coordinate systems: one corresponding to a stereo graphic projection from the north pole and one corresponding to a stereographic projection from the south pole (cf. fig. 20 9).
-' ' : ---' /4 rm
/'
w-plane
w
w "'-r'"
"
Rew
,
w ~ Tg[~Jexp[-i~J
"
Standard coordinates on the Riemann sphere. z-plane Re z
z ~ Cot[~Jexp[i~J
Fig. 209 The transition function is then given by
~21(Z) = 1.z which clearly is a holomorphic ,function: Equipped with these coovdinate systems the sphere 8 2 is thus a complex manifold known as the Riemann sphere. The complex coordinates generated by stereographic projections from the north pole and the south pole are referred to as standard coordinates. As in the real case we say that a continuous mapping, f: M ~ N, between two complex manifolds is holomorphic, provided it is represented by a holomorphic map when we int~oduce complex coordinates. We can now show, using the same arguments as in the complex plane, that a smooth mapping from thB Ri"mann sphilra int~ i tae Lf i$ confOI"maL if and only if it is holomorphic. All we need is therefore a closer examination of the holomorphic maps from the Riemann sphere into itself. Rema~kably enough we can now classify completely the holomorphic maps from the Riemann sphe~e into itself. This requires some preliminary knowledge about holomorphic functions, which can be found in any elementary textbook on complex analysis:
561
II
Suppose f is a holomorphic function with an isolated singularity at the point z (i.e. f is discontinuous at zo). Then f can be expanded in a Laurent series ar8und z : o 'i: a (z-z )n f(z) n=-co
n
0
Since f is discontinuous at z at least one of the coefficients a corresponding to a negative integer n is non~vanishing. We distinguish between twoncases: Either there exists an integer k such that a ~ 0 while a = 0 when n<-k, or there exist an infinite number of negative in~egers n such ~hat a is non-vanishing. In the first case we say that f has a pole of degree k at z ~ In the second case we say that f has an il30lated singularity at z . 0 Notice that if f has a pole of degree k at ~ then we can rewrite it on the form o
f(Z)=~ (z-z _ r'" o
where g(z) is holomorphic even at the point Zo ( i.e. the singularity can be removed by multiplying f with a polynomial). If, on the other hand, f has an isolated singularity at z , then the famous theorem of Weierstrass states that f maps each punctuated n~ighbourhood of Zo
into a dense subset of c. We will also need Liouville's theorem which states that a bounded holomorphic function defined on the ~hole oomplex plane is necessarily constant. The basic properties of holomorphic maps from the Riemann sphere into itself can now be summarized in the following way:
Lemma
8
non-trivial holomorphic map properties: (a) The preimage of a point is (b) Using standard coordinates w = w(z), which has only a are poles. A
f:S2~S2 is characterized by the foll~ing necessarily finite. it is represented by a holomorphic function, finite number of singularities, all of which
Proof: (a)
(b)
In terms of local complex coordinates the preimage of the point Wo corresponds to the zero-set of the holomorphic function w(z) - w • Consequently the pre image c8nsists of isolated points. But in a compact set, like S2, a subset consisting of isolated points can only be finite. Suppose f is represented by the holomorphic function w = w(z), where we we have introduced standard coordinates by means of a stereographic projection from the north pole. A singularity of w(z) corresponds to a point z where w(z ) = ~, i.e. to a point in the preimage of the north pole. Byo(a) there egn only be a finite number of points in this preimage. Since f is continuous it maps a small neighbourhood of z into a small neighbourhood of the north pole. Due to Weierstrass theor~m it consequently cannot correspond to an essential singularity of w(z).
Using this lemma we can then easily prove the following powerfull theorem:
Theorem 20 Using st~~d coordinates a given holomorphic map tram the Riemann sphere into itself, f:S ~ , is represented by an algebraic jUnotion, i.e. it is on the form - P(z} f{ z } - Q( z}
where P,Q are polynomials.
II
562
Proof: We already know that f is represented by a bolomorphic function w(z) with a finite number of poles. If we denote the poles by z. and the corresponding degrees by k we can remove the singularities by multiplyi~g w(z) with tbe polynomial i Q(z) = n(z-z. )ki i 1 Consider now tbe function p(z) = Q(z)w(z) By construction it is a holomorphic function without poles. Thus it can be expanded in a Taylor series I a zn p(z) n=o n
which is everywhere convergent. If P is non-trivial tbe corresponding function p(~) = Q(~)w(~) = E a z-n z z z n will have a singularity at z = O. According to lemma 8 the function w(i/z) cannot have an essential singularity at z = O. It follows that P(l/z) has a pole of some finite degree k, i.e. an = 0 when n>k. Consequently p(z) itself must be a polynomial
o
10.7 WINDING NUMBERS AS an application of the preceding machinery we now specialize to
the case where
M
same dimension
n
and
N
Let
that almost all points in
are compact orientable manifolds of the f : ~ ~ Nn be a smooth map. Then we know
N
are regular values (Sard's theorem). We
can now show
Lemma 11 Then either Let Q be a regular value in N. or it consists of a finite number of points.
f-I(Q)
is empty
Proof: The proof is somewhat technical and may be skipped witbout loss of continuity. If ~1(Q) is non-empty we consider a point P in it. Then f* maps Tp(M) isomcrpbically onto T (N) and there exists open enighbourhoods U,V around P and Q, so that f ~ps U diffeomorphically onto V (compare the discussion in section 10.1). Especially P is an isolated point in the pre image ~1(Q) . On tbe other band, ~1(Q) is a closed subset of M. As M is compact, r-1(Q) must itself be compact. If f-I(Q) was infinite, it would then contain an accumulation point, which by definition is not isolated. (Eacb neighbourhood of the accumulation point contains other points from the preimage. ) Therefore t-1(Q) can at most be finite. 0
563
II
Let us now introduce coordinates which generate positive orientations on M and N. Let Q: (yl, .•. ,yn) and P : (Xl, .•. ,xn ) be chosen so that Q
=
f(P)
•
lies in the pre image of the regular value
p
Since
f.
maps
V
of
Q.
U
of
Q, ,
i.e. we
Furthermore
f
diffeomorphically onto a neighbourhood
p
If the Jacobiant is positive, then
the orientation and if it is negative then orientation. Since the preimage f-l(Q) number of points, we can now define: Definition
TQ(M)
Det[dyi/dxj]IP' is not zero~
know that the Jacobiant, maps a neighbourhood
isomorphically onto
Tp(M)
f
U
f: U
~
V
~
V
preserves
reverses the
can at most contain a finite
11
The degree of a smooth map
n
f
M
~ Nn
at the regu Zar vaZue
Q
is the integer (10,54)
Deg(fiQ)
E
p.Ef-l(Q)
j Sgn I dyi /dX I I P ,
~
~
Here
sgn[ dy i/ 3xj]
±l
according to whether
ses the orientation. Cons'equently tive number of times
Q
If we let
Deg(fjQ)
=
O.
denote the number of points in
p
f-l(Q)
denote the number of points in
q
Jacobiant, then
preserves or rever-
simply counts the effecis is covered by the map f . I f C l (Q)
empty, it is understood that Jacobiant and
f*
Deg (f j Q)
p
the compact manifold sional example.
and
N
q
need not be constants as
f
Q
with positive with negative ranges through
This is easily seen already from a I-dimen-
E(1;amp tel: Consider the following map, f : S I ~ S I
presented
f-l(Q)
,
where we have re-
by a periodic function:
N
p
q
p-q
1
0
1
2
1
1
3
2
1
2
1
1
1
0
1
211
+ -;------------------~2-11----~
M Fig. 210
564
II
But the following deep theorem holds, which we shall not attempt to prove: Lemma ]0
(Brouwer's lemma)
The degree of a smooth map. values in
f
M
~
N.
is the same for all regular
N.
This common value is then simply denoted the effective number of times The integer
Deg(f)
N
Deg(f)
is covered by
and it measures
M by the map
f.
is therefore referred to as the winding number-,
since i t counts how many times
M
is wound around
N
In the lite-
rature it is also referred to as the Brouwer degree. Lemma If
11 f: M -
N
o.
fails to be surjective. then the winding number is
Proof: If
f-l(Q)
f
fails to be surjective there exist a point is empty. Such a point
point with
Q
Q
so that
is therefore trivially a regular
0
Deg(fiQ) = 0
At this point we take a look at some illuminating examples: Example 2:
As an example in dimension
n=l
we look at
M = N = Sl
If we introduce polar coordinates, we may consider the map
f
.
given
by: f(cp)
Cos ncp [ Sin ncp
i.e.
This is obviously a smooth map with winding number Example J:
In
n=2
dimensions we look at
:
[:::::::] f3(8,CP)
n
M = N = S2.
duce polar coordinates, we may consider the map f(8,cp)
cp' = ncp .
f
If we intro-
given by:
Sin8 Cos nCP] Sin8 Sin ncp i.e.
(8' ,cp' )=(8,ncp).
Cos8
This is clearly a smooth map outside the poles and it maps the first sphere
n
circle
8
N •
times around the second sphere. In fact, it maps each little 8
0
in
M n
times on the corresponding little circle in
565
If
f
II
was a smooth map,
it would therefore have winding number
n.
fortunately
is not
necessarily
f
Un-
smooth~
In
general it is singular at the poles. This cannot be seen from the above coordinate expression since the polar coordinates themselves break down at the northpole and southpole. This is a subtle point, so let us investigate it in some detail. We may introduce smooth coordinates at the northpole, say the standard coordinates
x
and
y
They are related to the polar coordi-
nates in the following way:
If
f
x
SinS Coscp
y
SinS SinCP
Arcsin~
S
i
Arctg
is smooth, then the partial derivatives af ax
will depend smoothly upon (x,y) . sic coordinates of
-+
-+
and
ex
af ay
and
e
For simplicity we compute the extrinthat are tangent vectors to the second
y
sphere. Using that:
~l
dS ax
dCP ax
-1
ax
ay
as
dCP ay
2Y 2Y
acp ax
1
r
l-
acp
as
Coscp CosS
Sincp CosS
Sincp SinS
Coscp SinS
we easily find the following extrinsic components of
and
Cos
r
~x
To be continuous at the northpole pendent of depend
cp,
when we put
explicitly on
cp
n = ±l
For
n
1
of a sphere onto itself. For
n
-1
smooth~
these vectors must be inde-
But the two first components
unless they reduce to
This miracle only happens for above map is not
(S = 0)
S = 0
~y
n CoscpSin ncp] n CoscpCos ncp
Hence if
0
or n
f
Cos 2 cp+Sin 2 cp=1 ±l,
then the
it reduces to the identical map it reduces to the antipodal map
of the sphere onto itself. They are obviously smooth.
o
566 II Exeroise 10.7. 1 Introduction: Let X: [o,~l + [o,~l be a surjective increasing smooth function such that all the derivatives vanishes when S = 0 or S = ~ . (You can e.g. put X(S) ~
XeS)
S ___ 1_
= p,Joe A
with Problem:
t(1T-t) dt
11
1
= fllo
e- t(~-t) dt
(a)
Show that the following map is smooth: g(S,~)
SinX(S) • Cos n~l Sinx(S) • Sin ~
= [
(b)
Fig. 212
CosX( S)
Show that the only critical values are the northpole and the southpole.
The examples above can clearly be extended to produce smooth maps Sn ~ sn
with arbitrary winding numbers.
Consider once more two arbitrary compact orientable manifolds and
of the same dimension
N
If
T
on
M
is a n-forrn on
N
,
n
f :
M~
N
M
be a smooth map.
then we can pull it back to an n-forrn
f*T
We want to compare the integrals: and
JMf*T
Since
and let
f
tical. However, Deg(f)
f NT
need not be a diffeomorphism, they are not necessarily iden,
f
is characterized by an integer, the winding number
which tells you how many times
N are effectively covered M. We can therefore generalize theorem 8, section 10.2 .. The proof is omitted since it requires a greater machinery than we have dev~loped yet. (See e.g. Guillemin/Pollack [1974».
by
Theorem 21
(BrouUJer's theorem) JMf*T = Deg(f)fNT
(10.55)
Remark: Let Q be a regular value in image is finite:
N
with a non-empty preimage. Then the pre-
f -1 (Q) = {Pl"",Pk } According to lemma 1, section 10.1, we can find a single open neighbourhood V of Q, and disjoint open neighbourhoods U of Pi such that f restricts to diffeoi morphisms: i
= l, ••. ,k
With these preparations we consider a differential form T of rank n which vanishes outside V. Then f*T vanishes outside Ul U..• U Un Therefore an application
o~~:;o:em~8'fse;:~o: l~.~il~:1 i=l
Ui
i=l
ax]
Pi
fv T
567 But
JVT
is independent of
J~*T
i
II
Consequently we get
= Deg(f,Q)·JNT = Deg(f)·JNT
where we have used Brouwer's lemma. This shows that the formula works for a differential form with a suitable small support.
Using Brouwer's theorem we can now construct an integral formula for the winding number. Let there be given a Riemannian metric on
N
Then it induces a volume form, the Levi-Civita form: E = IgdxlA •.. Adx n . Because
is compact, i t can be shown to have a finite positive vo-
N
lume: (10.56) f: M
If
~
N
is a smooth map we therefore get: JMf*E
=
Deg[f)J
£
N
Deg[f)·Vol[N)
or (10.57)
Deg[f)
This is the formula we are after! To be able to apply it, we must construct the volume form first. Consider especially the sphere
Sn-l.
Lemma '12 Let
n
(X1, ... ,X )
be the extrinsic cartesian coordinates on the
unit sphere
Then the volume form with respect to the induced metric is given by: a a an _ 1 (10.5B) n -
Proof: The induced metric on rounding Euclidean space vectors at a point +
P
Sn-l n R
is the metric inherited from the surLet
(~l' ..• '~n-l)
be a set of tangent
on the sphere. What we must show then is that
+
n(vl, ... ,v _ l ) is the Euclidean volume or the parallelepiped spanned n+ + n by vI' .•• ' v n - l in R (Compare exercise 10.7.2 below)'. lOr simplicity we prove this only in case P
n
= 3,
i.e. for the tWo-sphere. (x l ,x 2 ,X 3 ) Then the ra1 dial vector ~ = also has coordinates (X ,X 2 ,X 3 ) Let (~,;) be two tangent vectors at P with the coordinates (U l ,U 2 ,u 3 ) and 1 2 3 (v ,V ,v ) . We then get: Let
be a point with the coordinates
OP
568
II
This is the three-dimensional volume of the parallelepiped spanned by +
+
r,u
+
and
v
+
But
.
is a unit
r
+
vector, which is orthogonal to
~.
and
u
Consequently the three-
,
dimensional volume of the parallelepiped is equal to the two-dimen-
~'
sional area of the parappellogram
a
~
spanned by
~. We have
and
Q(~,~)
thus shown that
is the
area of parallellogram spanned by
+
+
u,v
Fig.
o
and we are through.
213
Consider a compact orientable manifold M of dimension n and cp : M ~ Sn be a smooth map which we parametrize as: n cpa = cpa(X) , l: [cpa(x) )2 = 1 Here we have used the extrinsic coo~a£nates of the sphere. The winding let
number of
cp
(10.59)
Deg[cp]
is now given by the formula: 1 1
=
J
cp*n
Vol(Sn) M
=
J
g
n:Vol(Sn) M ao···a n
a a a cp o dcp l A•.• Adcp n
This formula, which is based upon a particular choice of volume form on the sphere, has turned out to be very useful in various
applicat~ons.
Remark: It can be shown that, (10.60) where
r
is the
r-function characterized by the recurrence relation: r(~) '"
(10.61)
liT,
r(l) '" 1 ,
r(x+l)
= xr(x)
We list a few particularly useful cases: (10.62)
Vol[Sl) '" 2~ ; Vol[S2)
4~; Vol[S3)
8 2
3~
Exercise 10.7.2 Problem: Consider the sphere Sn and let us introduce intrinsic coordinates on the sphere (e l , ... ,en). They induce canonical frames in the tangent+sp!ces: ~l""'~n' The induced metric is now characterized by the components: gij = ei·ej and the volume form is given by E
1)
1 n Igde A •• • Ade
Show that
Ii = volume 2)
=
Let -+
+
+
of the parallelepiped in
Tp(S
n) spannedby ~el, ••• ,e+ . n
(vl' ... ,v ) be any set of tangent vectors and show that n -+ n -+-+ = volume of the parallelepiped in Tp(S) spanned out by v" ••• ,v ' n
E(Vl, ... ,vn )
569
II
(Hint: Use the Legendre identity: Det(A .• E.) = Det(A ij )Det(B ij ) J
1
+
where Aij are the cartesian components of components of Bi ).
A.
and
Bij
are the cartesian
l
Exercise 10.7.3 Problem: (a) Consider the unit sphere Sl and introduce the usual polar coordinate ~. Show by explicit calculation that the volume form (10.58) reduces to n = dep. (b) Consider the unit sphere S2 and introduce the usual polar coordinates (S,ep) Show by explicit computation that the volume form (10.58) reduces to n = SinSdSAd
Remark: It is important to observe that the concept of a winding number is only well-defined for smooth maps between aompaot manifolds. If
M
is not compact then the preimage may consist of infinitely many points, so that the local Brouwer degree is not well-defined. But even if it is well-defined it need not be constant on
N.
You might still be
tempted to consider
an integral expression like 1 a o al an (*) fM n.'V 0 IN Ea '" an ¢ A••• Ad. o But here the integral need not converge and even if it converges, it
d.
need not reproduce an integer! Consider for instance the smooth map • : R2 ~ S2
where
•
denotes stereographic projection from the center
(cf. figure 214).
Points at infinity
Fig. 214 It covers the lower hemisphere. Points on the upper hemisphere have Brouwer degree 0 while points on the lower hemisphere have Brouwer degree 1. Similarly the integral intuitively clear since
•
(*)
takes the value
~.
All this is
covers half of the sphere. So you may be ~
tempted to attribute the winding number with me, as long as you recognize that
~
to this map and that is okay is not an integer.
II
570
10.8
THE HEISENBERJ FERR(J>'INlNET As an exemplification of the machinery built up in this chapter
we will now look at a famous model from solid state physics. We have previously studied superconductivity ( see section 2.12 and 8.7). This time we will concentrate on a ferromagnet. A single atom in the ferromagnet may be considered a small magnet with a magnetic moment proportional to the spin. At high temperature the interaction)energy between the local magnets is very small compared with the thenreU energy. As a consequence the direction of the local magnets will be randomly distributed due to
~l
vibrations.
FOr sufficiently low temperature, how-
ever, this interaction between the local magnets becomes dominant and the local magnets tend to lign up. We thus get an ordered state characterized by an order parameter, which we may choose to be the direction of the local spin vector. Thus the order parameter in a ferromagnet is a unit vector. If we introduce Cartesian coordinates it can be represented by a triple of scalar fields
[$1 (X);$2(X) ;$3(X)] subject to the constraint
$a$a
=
1.
CONTINUUM LIMIT $Cx)
H
=p
=
[$IC~);$2(~);$3C~)]
I V$a. v$a
d2x
R2 We are going to study equilibriun configurations in a ferromagnet. Consider a static configuration $(~), where $(~) is a slowly varying spatial function. In the Heisenberg model for ferromagnetism one assumes that the static energy functional is given by (10.63)
corresponding to a coupling "between nearest neighbours". To obtain the field equations for the equilibrium states we must vary the fields $a. But here we must pay due attention to the constraint $a$a
= 1.
The order parameter
$
is not a linear vector field, i.e. we
are not allowed to form superpositions!
571
II
The field equations for the equilibrium configurations are given by
1
(10.64a)
or equivalently
1
(10.64b)
Proof: We incorporate the constraint by the method of Lagrange multipliers. Consider therefore the modified energy functional
~[~a(X);A(X)]
=
~
+
Now all the fields,~a(x) and A(X), must be varied independently of each other. Performing the variations A(X)
~
A(X) +
£~(x)
and
we obtain the "displaced" energy functional
~[£]
=
~[O]
+
£J
+
£2~
+ £2
+
2£
-1>
As usual we then demand
o = d~[£l - J
+
2
<$aIJ&d~a + 2A(X)~a(x»
+
<~(X)I~2(X)
+ <~(X)I~2(X)
- 1>
- 1>
This leads to the Euler-Lagrange equations
As usual the equation of motion for the Lagrange multiplier degenerates to the equation of constraint. The constraint can now be used to eliminate A(X) from the equations of motion. Differentiating the constraint twice we obtain
(*) Consequently
where we have used the equation of motion and the identity (*). Inserting this back in the Euler-Lagrange equations we finally obtain:
(The equivalent version of this equation is obtained by dualizing it!) From now on we restrict ourselves to the case D
=
0
2, which has the
most interesting topological properties. Thus we consider a two-dimensional ferromagnet, where a spin configuration consequently corresponds
572
to a map
$:
R2 ~ S2
This suggests that we introduce a topological quantity, the winding number Q, which tells us how many times the sphere is covered
Q[~] =
(10.65)
'"
.l..Je: il:ad$b"d~C 811 abc'" '" R2
But here we encounter the usual problem, as the integral need not be well-defined because R2 is not a compact manifold.
Furthermore it need
not be an integer even if it is well-defined ( cf. the discussion in the preceeding paragraph ).
We must therefore inwoke a boundary condi-
tion and fortunately we have in this model the natural choice: The only spin aonfigurations that are physiaally relevant are those with a finite energy, i.e.
This leads to the boundary condition (10.66)
lim p +
i.e.
$a
(10.67)
d$a
=
0
00
must be asyptotically constant lim $a(x)
=
$a
P'+OO
0
Exactly which components the constant vector at infinity has we cannot say. All unit vectors are equally likely to be the one Nature chooses. But observe that once we have fixed the unit vector at infinity we can not change it without breaking the condition of finite energy. In what follows we choose the north pole as the asymptotic value of the order parameter: (10.68)
lim $a(X) p +
=
[0;0;1]
00
The boundary condition (10.68) has the important consequence that we can now compactify the base space. As usual we perform a stereographic projection of the plan,e into t..ile sphere, 11:5 2 R2. Notice that the North pole corresponds to the points at infinity. We can now lift a spin configuration in the plane, $a(x), to a spin configuration on the sphere,
~a
= 1I*$a.
We then extend the lifted field ~a
to the north pole by con-
tinuity
cf. the discussion in section 10.6. As a consequence of our boundary conditions we therefore see that a spin aonfiguration with finite energy aan be lifted to a map 52 ~ 52, whiah maps the north poZe into the north pole. This shows that the winding number is well-defined for all
573
II
the permitted configurations. The space E consisting of all the smooth finite energy configurations therefore breaks up into disconnected sectors
En' where each sector is characterized by an integer n:
(10.69)
n
= 8~J
€abc$ad$bAd$C
S'
Remark: As a consequence of (10.42) the energy functional is conformal invariant:
H$2[$l
!J
Consequently we can solve the field equations (10.64) on the sphere, whenever it is advantegous, since any solution to the equations of motion on the sphere automaticallyprojects down to a solution of the equation of motion in the plane.
We proceed to investigate various equilibrium configurations. Let us first determine the vacuum configuration for the Heisenberg Ferromagnet. The classical vacuum is characterized by having vanishing energy density. This implies that d$a vanishes, i.e. $a is constant. Thus the classical vacuum corresponds to a spin configuration where all the local spin vectors point upwards. Clearly it has winding number zero. We then proceed to examine the non-trivial sectors. The fundamental problem is whether we can find ground states for the non-trivial sectors, i.e. whether we can determine a configuration in the sector En which has the lowest possible energy among all the configurations with winding number n. We call such a ground state a spin
~ave.
(Notice that
the energy functional (10.63) only contains a "kinetic" term. Although the model is based upon scalar fields it thus corresponds precisely to the eXceptional two-dimensional case, where Derrick's scalings argument does not apply, cf. the discussion in section 4.7 ). Observe that by defihition a spin wave is a local minimum for the energy functional (10.63) and consequently it represents a solution to the second order differential equation (10.64). It is a remarkable property of the Heisenberg Ferromagnet that we can actually explicitly determine all the spin waves. The first step in the analysis of spin waves consists in a reduction of the second order differential equation for the spin wave to a first order differential equation. This reduction is due to a decomposition of the energy functional
Bogomo~ny
(10.63). Consider the quantity
d$a, i.e. ai~a. It carries two indices: A space-index, i, referring to
574
II
the physical space, and a field index, a, referring to the field space. Let us concentrate on the field index for a moment. Consider the vectors They are tangent vectors on the unit sphere in field space. We may now introduce a duality operation in this tangent space. As usual for a twodimensional vector space it corresponds to a rotation of
~TI.
We denote
it by # and observe that in terms of the Cartesian components for the tangent vector the duality operation is given by (10.72)
#d$a
= ~abc$bd$C
Observe especially that #2= -1, and that # preserves the inner product in the tangent space. Worked exercise 10.B.l Problem: (a) Show that
<~ald~a> = <#d~al#d~a> = <*d~al*d~a> (b) Show that the topological charge, i.e. the winding number, is given by
(10.73)
n[$l = ~<*d~al#d~a>
(10.74)
(c) Show that any field configuration automatically satisfies the differential equation
d#d~a
(10.75)
=
~a(#d~bAd~b)
(Compare this with (10.64b) ! )
From exercise 10.8.1 we now easily obtain the desired Bogomolny decomposition (10.76)
(with II
If
<
I > )
But then we conclude a) The energy in eaah seator En is bounded
belo~
by
(10.77)
b) A aonfiguration
~ith ~inding
number n is a spin wave (i.e. a
ground state for the seator En) i f and only i f it satisfies the first order differential equation: (10.78)
&a {#d$a *dcp _#d$a
(when n is positive) (when n is negative)
(This is kno~n as a double self-duality equation).
575
II
Exercise 10.8.2 Problem: Show, by explicit c0mputation, that the first order equation (10.78) implies the second order equation (10.64). The next step will be to show that every solution to the double self-duality equation (10.78) corresponds to a well-known geometrical object. In our case it turns out that we get the following nice characterization of spin waves: A spin configuration,
$:R2
is a spin ~ave exactly when it
S2,
generates a conformal map from the plane into the sphere. Proof: Let ~1'~2 be the canonical frame vectors generated from the Cartesian coordinates in vectors
01$,02$.
If
$
R2.
Notice that they are lifted to the tangent
is conformal this forces
01$,02$
to be orthogo-
nal vectors of the same length. We leave it to the reader to verify that the converse holds, i.e. that the above property actually characterizes the conformal maps from the plane into the sphere. Now suppose
$
is a configuration with winding number
n
which sol-
ves (10.78). This first order differential equation can be rearranged as follows But from these equations it follows immediately that al$,02~
$
thogonal vectors of the same length. Consequently
are or-
is a conformal map.
On the other hand it is not too difficult to show that any conformal map actually solves (10.78). We leave the details as an exercise: Worked exercise 10.8.3 Introduction: Let $:R2~ 82 be a smooth map and introduce the following vectors in field space:
..p,
i.e.
€,
i.e,
Problem: (al Show that if
$ is
,0,$
$xo,$
'-J J
'-
€,
,0,$
'-J J
'-
+
$'0,$ '-
a conformal map then the following holds
Pl,i\ = Ql'Q2 = PI'QI = (b) Show furth!;.rmor!;. that if PI = P2 = 0
P2'Q2
$ is
=
Pl'Q2
=
P2'QI
=
0
a conformal map then either
or
This concludes the proof:
0
As an example we notice that the stereographic projection itself is a conformal map and thus it represents a spin wave with winding number
-1 (it reverses the orientation!) Observe also that the stereogra-
phic projection
can be used to lift any spin configuration ~:R2.... R2 to
576
II
a spin configuration ~:S2~S2. Consequently there is also a one-to-one correspondence between spin waves and conformal maps from the unit sphere into itself which maps the north pole into itself. In the final step we introduce aomplex analysis. In the present ,case it is completely trivial. It is well-known that the sphere is isomorphic to the extended complex plane S2
C
...
* =C
u { co}
and that the ortentation preserving conf<mnal maps are holCllOrphic, cf. the discussion of the Riemann-sphere in section 10.6. But holomorphic maps are necessarily algebraic, i.e. they are on the form w(z)
(10.79a) FUrthermore
w(co)
=
= !:ill Q(z)
with
implies that
P,Q polynomials.
DegP>DegQ. Thus we have finally
explicitly construced all spin waves with a negative winding number.*) Similarly a spin configuration with a positive winding number corresponds to an anti-holomorphic map on the sphere, i.e. it is on the form: (10.79b)
w(z) =
~m
with
P,Q
polynomials, DegP>DegQ
It would also be nice to find a simple formula relating the winding number to the polynomials P and Q. That is easy enough: The map w=P/Q is a smooth map and according to Sard's theorem there are plenty of regular values. Let Wo be a regular value. Then the preimage consists of all solutions to the equation P(z) - woQ(z) =
o.
As DegP>DegQ it has Degp distinct solutions. map preserves the orientation. Thus this means that
$
w
FUrthermore a holomorphic
has winding number
DegP
and
has winding number -DegP.
Remark:
Rather than relating spin waves to conformal maps one can relate them directly to holomorphic (or anti-holomorphic) functions. Using a stereographic projection IT from the unit sphere in field space to the complex plane we can project the order parameter down to a single complex field given by ~2
1.\1
W
= -~l_iji3
+ i-'l'-I.\l_iji3
In the same way "he tangen" vector a,$a on the unit sphere is projected down into the tangent vector diw in the complex pl~e, i.e. d$a ~* dw Since IT is conformal it preserves the right angles. Furthermore it reverses the orientation so that ~* -id.w #d$a ]!.* -idw i.e. 1
Similarly it follows from the linearity of Aa
e: i } }
1f*
~
e:, .d.W 1J J
IT*
i.e.
that
*d$a JJ..* *dw
*) A.A. Belavin and A.l1. POlyakov,"lIetastable states of two-dinensional isotropic ferromagnets", JETP Lett. 22 (1975) 245.
577
II
Putting all this together we therefore see that the double self-duality equation (10.78) is projected down to the ksual self-duality equation for a holomorphic function *dw = -idw
Okay, this concludes our discussion of spin waves, i.e. the ground states for the various sectors. We might still ask if there are other finite energy solutions to the full field equations (10.64) Apriori there could be local minima lying somewhat above the ground state, or there could be "saddle points". But Woo*) has investigated this problem and using complex analysis he has shown that the answer is negative: Any finite energy solution to the seaond order equation (10.64) automatiaally solves the first order equation (10.78) Worked exeraise 10.8.4 Problem: a) Show that a spin configuration represented by the complex valued function w(z) has the energy density 2 {ow H = (l+ww)Z
(10.80)
oZ
b)
c) ( 10.82)
~2W
OZdZ
=
~
oW
ozJ
oZ
zw
illi: ~
and
oZ dZ
(1+WW)o;2~
=
2w
~ ~
Consider the following two complex-valued functions
awaw
fez)
d)
+
Show that the corresponding equations of motion are given by (l+ww)
( 10.81)
aw
oZ-
=
awllw
oZ oZ (l+wwP
and
g(z)
=
az oZ (l+ww)Z
Show that fez) is holomorphic and that g(z) is anti-holomorphic provided w(z) solves the equation of motion Show that if f has no poles, then w itself must either be holomorphic or anti-holomorphic.
In the above exercise we have almost proven that a solution to the full field equation (10.64),which has finite energy, is automatically represented by a holomorphic or an anti-holomorphic function. There is however one possible loop-hole: The quantity
f (z)
=
oW oW oz
az
(l+wwF
might have a pole, or equivalently the energy density might have Consider, e.g., the non-admissible solution w(z)
*) G.
\\00,
=
zls
"Pseudoparticle configurations in two-dimensional ferromagnets",
J. Math. Phys. 18 (1977) 1264
a"pole~
578
II
(It is non-admissible because it is multi-valued!) Neglecting the branch cut for a moment we notice that it produces the energy density
1 H = 2Izl(l+lzl)2
Consequently the energy density has a "pole" at
z=O . But it is still
integrable (as you can easily see if you introduce polar coordinates). Thus we cannot exclude the possibility that
f(z)
might have a pole.
One must then investigate the solution to the first order equation (10.82) very carefully to exclude that possibility too, i.e. near a pole of like
f(z)
w(z) =
we must show that
w(z)
is necessarily multi-valued
z~. FOr further details you should consult the original
paper of Woo. Exeraise 10.8.5 Introduction: Let F be an ordinary function: F:R1R and let w(z) be a complex valued field characterized by the energy density H =
F(wW){~ ~ az az
+
~ az ~} aZ
In analogy with the Heisenberg ferromagnet we will only be interested in static configurations with a finite energy. Problem: a) Show that the equations of moti~n are giv~ El 2 F d_W =-F'w ~ ~ and F ~ =-F'w ~ ~ azaz dZ dZ azaz dZ az b) Show that
fez) = F(ww) ~ az ~ az
is a holomorphic function provided w(z) solves the equations of motion. c) Specialize to a static massless complex Klein-Gordon field in (2+1) space-time dimensions and try to characterize the static solutions with finite energy. (There are none except for the trivial solutions w(z) = constant!) Hint: Show that it corresponds to the case F=l.
10.9
THE EXCEPTIONAL f4-MODEL
As another interesting example we will consider a model in two spacedimensions, which on the one hand is related to the Ginzburg-Landau model for superconductivity, on the other to the Abelian Higg's model (cf. the discussion in the sections 8.7-8.8). It will be based on the static energy functional:
(10.83)
D
579 ,cf.
(B.60) and (B.BB).
II
For the moment we leave the potential unspe-
cified except that we assume that it is gauge invariant, i.e. it is
,
a function of 1~12
and furthermore that it only vanishes at the non-
zero value I~I= ~o' The field equations for a static equilibrium configuration are given by
au
(10.B4a)
-
(10.B4b)
D*D~
2~2~[
- &B
i~[~D~ - ~D~J 2
cf. the worked exercise B.7.3. As usual we consider only finite energy configurations. This leads to the boundary conditions: (10.B5)
lim p -+
=
B
0
D~
lim p +
co
=
lim
0
..,
p +
co
As in the case of ordinary superconductivity we have furthermore flux quantization, i.e. (10.B6)
JR2 B
=
=
n
21f e
where n is related to the jump in the phase of the Higgs' field when we go once around the flux tube. So far nothing is new. We will now try to see if we can find the groundstate configurations for the non-trivial sectors
(where nfO). As
usual the investigation of the ground states will be based upon a Bogomolny decomposition of the static energy functional. We start by guessing a reasonable set of first order differential equations. Guided by the (l+l)-dimensional models and the Heisenberg Ferromagnet we try the ansatz: (10.B7a)
B
(lO.B7b)
*D~
v'2u[~]
E
iD~
= *hU[~l
(i.e. the self-duality equation) .
They lead to the following pair of second order differential equations
-D*D~
(lO.BBa)
=
-iD2~
=
-eB~
=
-eV2U[~1
(where we have used exercise B.7.2).
-5B = -5*{2U[~1
(*)
=-
*dV2U[~1
But from the self-duality equation we get *d~
=
iD~
+
ie~*A
Consequently we get
*d~
=-iD~
-
ie~*A
5BO
II
~*d$ + $*d~
Inserting this into equation (*) we can rearrange this as
au
1
(lO.BBb)
- 5B
=-
v'2U[$)~2 i[cjlDcjl - cjlDcjl)
Compairing (lO.BBa-b) with (lO.B4a-b) we see that the first order differential equations are compatible with the second order differential equations provided the potential satisfies the identity
This restricts U to be a fourth order polynomial of 'the form
(lO.BB)
U[cjl) = §2ClcjlI2_ cjlo)2
with
Icjll < cjlo
Notice that the above potential represents a spacial case or the potential energy density in the abelian Higg's model where we have put e2 (lO.B9) A= 2 The abelian Higg's model based upon the potential energy density (lo.BB) is known as the exceptiona l $"-mode l. In superconductivity it corresponds to the case where (10.90) (K is the Ginzburg-Landau parameter (B.67). The identity (10.90) follows immediately from (lO.B9) when you perform the substitutions A ~ 8
and e ~ q/h ). Okay, in the exceptional cjl"-model we thua have the possibility of reducing the field equations (10.B4) to the first order equations (lO.B7). This suggests the following Bogomolny decomposition: H
~ + ~<*Dcjl+iDcjll*Dcjl+iDcjl> + a rest term
All we must check is that the rest term is a topological term, i.e. that it only depends on the "winding number" n. We leave this as an exercise to the reader: Worked exercise 10.9.1
Problem: (a) Show that the winding number is given by (10.91) n = ~ <*dcj>lidcj» (This should be compared with (10.74)!) 11'1'0
(b) Show that the exceptional cj>"-model possesses the Bogomolny decomposition (10.92)
H
= !\lB+/2U[cjl)EW
+
~1*Dcjl+iD$1I2 : n1lcjl~
We therefore conclude as usual:
(with
II
1\2= < I> )
581 (a)
The energy in the sector
H ~ Inl1f
(10.93)
(b)
E n
II is bounded below by
A configuration with winding number n is a ground state i f and only i f it satisfies the first order differential equation:
(10.87 )
In the above discussion you might feel that the first order equation (10.87a), which preceded the Bogomolny decomposition, was sort of "pulled out of the air". We did try to justify it by referring to our earlier experience with (l+l)-dimensional models, but you may not find that very convincing. We shall therefore present another argument, which allows one to make reasonable guesses in such a situation. The argument is closely related to the reasoning behind Derrick's scalings argument. We know that a pure scalar field theory in two dimensions cannot possess stable non-trivial static solutions. In the above model it is thus the precense of a gauge field which makes it possible to stabilize a static configuration. Let us look at this in some detail: Consider the scaling transformation
D" (i{)
'=
A~ •
It is a conformal transformation woth the conformal factor
QZ(x) =
A2
Let us furthermore introduce the scaled configuration
From (10.40-41)
(or by working out the coordinate expression directly)
we now get
RZ
=
DA CR2)
JU[
H[
+
!
+
A-ZfU[
582
II
A stable configuration and especially a ground state must now satisfy
Thus the condition of stability leads to the "viriral theorem": A stable configuration satisfies the ldentity: (10.88) Thls should be compared with (4.75). In analogy wlth the (l+l)-dimenslonal case we now suggest that a ground state not only satlsfles this ldentlty globally, but that it ln fact satisfles it point-wise, i.e. that B = 2:v'2U[cjl)
E
Thus we recover preclsely the ansatz
(lO.87a).
So all we have got to do is to solve the first order equations (10.87).
Cnfortunately that ls a very hard task, slnce one cannot wri-
te down the explicit solution.
[sing advanced analysls it can be shown
that the most general solution with n flux-quanta depends upon 2n arbltrary parameters correspondlng to the center posltlons of n flux tubes each carrying a unit flux.
(Ior details about the machlnery requi-
red to analyse the general solution you should consult the book of Taubes and Jaffe) • If one speciallzes to spherical symmetric configuratlons, the analysls becomes very slmple. If we make the ansatz n
(10.89)
A(x) = eA(p)dcp
it is easy to see that thls represents a flnite energy conflguration with n flux quanta provided ¢(p) and A(p)
satlsfies the boundary condl-
tlons (10.90)
lim p -+
4i'(p)
0;
0
lim' 4i'(p) p -+
co
1;
lim A(p) 0
p -+
0;
lim A(p) p
-+
1,
co
cf. the dlscussion in section 8.7.
Worked exercise 10.9.2 Problem: Show that the field equations (10.87) reduce to n dA 1 2 2( ~2) &b = ncjl(l-A) ~ (10.91) p dp = 2e cjlo l-cjl ; p~ when we spacialize to the ansatz (10.89)
It follows fran standard theory of first order differential equations that there is preclsely one solution (cjl(p) ,A(p)), which interpolates between (0,0) and the "equilibrium point" (1,1) .
583
II
SOLUTIONS OF WORKED EXERCISES: No. 10.3.2 In polar coordinates the stereographic projection.is.giv~n by P = 2Cot~ ; ¢ = ~ The Jacobi matr1x 1S g1ven by
(;5) • [- ";'~
:1
(Especially we observe that the Jacobiant is negative, i.e. the stereographic projection is orientation reversing.) The metrics gl and g2 are characterized by the components
= G1
=
[1 0e] 0
= G 2
Sin 2
=
[1
p
0
Fig. 215
The pull-backed metric is then characterized by the components
f*g2
["~"
=
s+ -
D21ii2D21
2
1 "2
_1_
A = A(S)
then
o
a
4Cot
28
Sin 4il. 2
-------
No. 10.3.3 If we perform a general parameter shift and so the geodesic equation is transformed into 2
d Xll + rll d s2
as
dx
a
ds
B ll dx = dx ~ ds ds A' (s)
Thus the most general, covariant equation for a geodesic is on the form A( s) --~ A' (s)
with On the other hand any smooth function A(s) =
A(s)
~ = A' (s)
can be written on the form
[lnA'(s))'
with
o
584
II
No. 10.8.1 (a) The first equality is a consequence of the fact that # preserves the inner product in field space, since it corresponds to a rotation. The second equality follows from theorem T chapter 8.
8~J**[d$a]"#d$a
(b)
trr<*d$al#d$a> =
=
(c)
d#d$a '" d[£ b $bd$c] '" £ b d$bAd$c
_~Jd$a"£abc$bd$C
=
~J£abC$ad$bAd$c =n[$]
a cae
Notice that Fa = £ b
a c
d$b"d$C
is the vector in field space obtained by taking
the cross product of the two tangent vectors d$b an~ d$c. Consequently it must be proportional to the radial vector $a and since ~a is a unit vector we
¢
Fa = $a($bFb) = $~b£
abc
d$CAd$d= $a(#d$dAd$d)
nU
No. 10.8.3 + + + (a) Let us write out Pl,P2,Ql and !1 = 02~ - ~'01~ Ql '" 02~ + ~'01~ Then we get
o and
+ + + + ~ ~ Pl·Q2 = P2·Ql = -2g1~ d2I But for a conformal map ~:R2~ we know that 01$,02$ are orthogonal tangentvectors of the same length, so this forces the right hand side to vanish. (b) We start with thx following important remark: If 01$ (or d2~) vanishes, then aLL fOur tangent-vectors Pl,P2,Ql and Ql va-
nish. To see this let 01$
vanish. Then we get
P2 = -$'02$ and Q2 = $'d2$ ~ut+th$Y are o~thogonal vectors so.this forces 02$ to vanish too. Clearly Pl,P2,Ql and Q2 must now all van~lh.+ + + To prove (b) we now observe that Pl,P2,Ql and Q2 are all tangent vectors in the same tangent plane to the unit sphere. As they are mutually orthogonal at least two of them has to vanish. But apparently there are more possibilities than t~ose lilted i~ b). + + Suppose PI and Ql van~shes. Then 32$ = !(Pl+Ql) hal to vanish too and by the above remark they iherefo~e all vanish. Similarly PZ=Q2=O make them all vanish. Suppose then that PI and Q2 vanish. From Q2=0 we get 31$ = $.32$ and when we insert this in the expression for PI we obtain = = d2~ - $K($XdZ$) = 232$ • So again 02~ has to vanish, and therefore they all vanish. Similarly P2=Ql=0 make them all vanish. This leaves with the two possibilities listed in (b).
o PI
0
No. 10.8,-4 '; (a) From section 10.3 we know that the stereographic projection is a conformal , map with the conformal factor \12 S· 4 (a) 1 = i l l 2 = Cl+wW)2 Consequently we get 1dlWl2+ 13 2wI 2 !(I Ol$1 2 + 102$1 2) = 2 (1+wW) 2 Thus the energy density may be re-expressed in terms of complex variables as follows:
585 Ow aw} 2{aW OW +~ az ~ az (l+wW) 2
aiWa? 2(1 +WW)2
B
II
(b) The corresponding Euler-Lagrange equations can be written in complex form as aB aw
=
.1.f3B } + .1. LaB} azla~
azla~
az
az
~~ = fzf}~}
+
~{a~}
la az
'a aZ:
E.g. the first equation reduces to 2(l+wW),w aw ~ + ~ ~ (1+WW)4 az oZ: oz az
aw
= 1-
_1_ QH + J.. _1_ az (1+wWl'az a~ (1+wW1 2 az 2 w - 2(1+ww ~ aW. oW ow ya -)[~+w~l2R (1+wW)2 azaz - 2 (1+wW) [azw+waz1ai + (1+wW azaz alw az dZ (1 +
wW)"
which after a long, but trivial, calculation reduces to a2aw OW (1+wW)a!~ = 2waz az:
o
(c) We must show that f satisfies the Cauchy-Riemann 2 2 1 -)2 a w aw + ow a w _ aw~ 2(1+ww-) M = ( +ww az azaz az (1 + wW)"
mz
az mz
2 2 _ _1_ _ {(1 -) a w '>.-: ow~ (1+ww-) a w _ 2w awa~ } - (1+wW)3 +ww azaz - ~w azaz + az:az azaz Inserting the equations of motion we see that the right hand side vanishes automatically. In the same way we can show that g(z) is anti-holomorphic. (d) If fez) has no poles it is an entire function, i.e. a global holomorphic function. As the energy density vanishes at infinity we conclude that
!~I -~-O · 11m (1+w.W) -
IQRI
· az 0 11m = (1+w.W) '" p"''''' Consequently !f(z)! ... 0 at infinity and therefore f(z) must be bounded. But a bounded entire function is necessarily constant (Liouville's theorem) Therefore fez) must in fact vanish identically. But then the equations of motion (10.81) reduces to aw/a'l. = 0 or aw/dz = o. p ...
co
0
No. 10.9.1 (a)
<*d~lid~>
=
-iJ~Ad~ R2
(b)
=
-iJd[~d~l
=
R2
lim Po"'''''
-iJ~d~
2n1f~~
p=po
Expanding the normS we immediately get
111 B-v'2U[~h:,,2 =
+ l"I
!<sls> +
Consequently the rest term is given by A + B =
D=d-ieA
586
II
Here B1 is the topological term we are after B1 Furthermore B,
= ~<*d~lid~>
= nIT~%
vanishes automatically since B,
= ~ie2<~*AI~A> = -~ie2fl~12AAA
but
AAA
0
Thus we are left with B2 and B3 which we can rearrange as B2+B3 =
-~ef(~d~+~d~)AA = -~eJdl~12AA
Here we can safely replace 1~12 with 1~12_~~ since the constant drops out anyway when we take the exterior derivative. Thus we get B2+B3
= -~eJd(I~12_~%)AA = -~efd[A(I~12-~%)1
+
~efB(I~12_~%)
The first term is converted to a line integral at infinity and it now vanishes sincel~12_~% vanishes at infinity. To summarize the rest term has thus been broken down to
fBV2U[~1 R
J~Be(I~12-~%)
+
+
nTI~~
R
But here the first two terms obviously drops out provided
J2U[~1
~e(~~_1~12)
U[~l
i.e.
(In this way we have actually recovered the
=
~e2(1~12_~ij)2
~'-model!)
U
No. 10.9.2.
Let us first look at the equation (10.84a) B = ~e(~ij_I~12) E On the other hand we get
B =
dA
=
ndA
e dO
ndA
dPA~
epdp
E
By comparison we thus obtain the first field equation. The other one is a bit tricky. First we observe that (cf. exercise 7.5.5) *dp
= -pd~
and
*~
~dp p
Then we get by a trivial calculation D~ = ~oe
i~
•
~ in~. (
dpdp + ~o~e
In l-A)~
The two sides of the self-duality equation therefore reduces to *D~
in~ = l~oeapdp
.. iD~
~i~
- ~o~e
n(l-A)d~
By equating the coeeficients of dp or field equation.
~
we now finally obtain the second
587
II
chapter
II
SYMMETRIES AND CONSERVATION LAWS 11.1 CONSERVATION LAWS We start by investigating the general structure of conservation laws in physics. They are concerned with a vectorfield
J
satisfying the
equation of continuity (see (7.88»:
(FgJll) [ -L~ Fg II
&J = 0
(11.1) The vectorfield
J
=
0] .
n is a don , given by the
is interpreted as a "current" and if
main in a spaceslice, then the flux of
J
through
integral
Q=-f*J,
(11.2) is interpreted as a "charge".
(The minus sign is conventional. Compare
with (8.53). Consider now a tube as shown on fig. 216,
J
and let us assume that the current
vanishes outside the tube.' (The width of the tube can be very large and the fact that the current vanishes outside the tube in practice means that "the current vanishes sufficiently fast at spatial infinity".) We can then show the following lemma:
Lemma 1
xl
(The tube lemma)
A smooth "current"
J
which satisfies the continuity equation
(11.1), wiLL produce the same flux face
n
-fn*J
through any closed hyper sur-
intersected by the tube, or equivalently, the "charge"
contained in a hypersurface wiLL be the same for aLL hypersurfaces interseoted by the tube.
588
II
Proof: Let n 1 and n2 be two hypersurfaces intersected by the tube. We will assume that they are far apart. Together with the wall of the tube they then constitute the boundary of a 4-dimensional regular domain W. Be careful about the orien-
w
W! From Stokes'
tations relative to theorem we now obtain
o (If
= Jw*&J= Jwd*J = Jaw*J = J n2 *J-J n1 *J' n1
and
n2
are not far apart we
introduce a third hypersurface are far apart from transitivity).
n1
and
n3
02
Fig. 217
which
and use
D
Okay, having constructed the machinery we can now apply it to some important cases: Example 1:
CONSERVATION OF ELECTRIC CHARGE.
The electric current is represented by a co-closed I-form. Let us prove that this leads to the conservation of electric charge. To avoid convergence problems we assume that the current is confined to a tube in space-time. Suppose
S
is an inertial frame of reference, and let
Fig. 2l8a
Fig.2l8b
0 1 and O2 be three-dimensional regular domains contained in space slices relative to S such that n 1 surrounds the charges and currents at time
t1
and similarly for
the total charge at time
t1
02
at time
, respectively
t2
{see fig.
2l8~.
(8.53» Q{ttl =
-J n1 *J
Then
t2 , is given by (Cf.
respectively
Q{t2)
=
-J 02 *J
589
II
But according to the tube-lemma they are identical, i.e. the charge measured by an observer in
S
is independent of time.
Actually we can prove something more. Suppose
S1
and
different inertial frames of reference. Then we can let gular domain contained in a space-slice relative to
S1
gular domain contained in a space-slice relative to
S2
As we have seen the electric charge measured in
S1
S2 n1 and
are two be a ren2
a re-
(see fig. 218b).
,respectively
S2 ,
is given by the integral Q1 =
-J n1 *J
respectively
Q2 =
-J n2 *J
But by the tube-lemma they are identical. Thus the electric charge is independent of the observer too. We have now given the promised proof of the first half of Abraham's theorem (section 1.6): Theorem 1 If
J
is a co-cLosed l-form, which vanishes sufficientLy fast at
spatiaZ infinity, then the "charge"
-J n *J
Q =
=
J
JOdx 1dx 2dx 3 xO=t O
is independent of time and the observer. Example 2:
CONSERVATION OF ENERGY AND MOMENTUM.
We start by observing that the energy and momentum densities of a field configuration are represented by a
symmetric tensor of rank 2
(Cf. the discussion in section 1.6). At first thought you may therefore think that the analysis of energy and momentum falls beyond the scope of exterior calculus, which deals only with skew symmetric cotensors. On the other hand, we want to define the total energy and the total momentum of a field configuration. This involves integrals of the form
J xO=tO
TCLo d 3 x
and it would be nice to have an invariant, i.e.
geometricaZ, characte-
rization of these integrals. To do that we "mimic" the discussion of electric charge. We select an inertial frame of reference
S. This inertial frame of reference
is characterized by a set of inertial coordinates rating the basic tangent vectors -+
-+
-+
-+
eo ,e1 ,ez ,e3 These constitute four vector fields, which loosely speaking specifies the direction of the CarteSian axes and the direction of time. By con-
II
590
tracting the cotensor
T
with these four vector fields, we obtain
four new l-forms (11.3)
which in our inertial frame are characterized by the components Po : TasO~ = Toe It follows immediately that Pi
Po
Pi : TaSO~ = TiS represents the energy current, while
represents the momentum current along the xi-axis. Observe, that
although the currents PO,Pl,P2,PS are purely geometrical quantities, they depend strongly on the choice of the inertial frame of reference
,
in question. In terms of the inertial coordinates associated with that the energy momentum tensor
T
5
we know
satisfies the continuity equation
(1. 38)
and we therefore conclude
~ where
(Pa)S
S (p a ) S
= 0
are the contravariant
components of
P
a
' Consequently
,the current
Pa is conserved, which can be expressed in the geometrical form &P a = 0 • Thus PO,Pl,P2,PS all give rise to conserved quantities. If 0 is a space-slice characterized by the equation xO=tO , then the integrals (11.4)
are all time-independent provided the energy and momentum currents vanish sufficiently fast at spatial infinity. Let us interpret these integrals. Using that dxo=O along n we get p = f (p )odx 1 dx 2dx s = a xO=t O a Observing that in an inertial frame _Too=T ly obtain that (11.5) (11.6)
-p
°
P.l.
f
!Toodxldx2dx3 xO=tO
of
x =t
0
Tiodxldx2dxs
° °
final-
the total energy the total momentum along 1 the x -axis
Consequently the total energy and momentum is conserved.
(Do not be
confused by the signs. If we "raise" the index a we get [pO,pi] = [total energy, total momentum along xi-axis] with the correct signs as usual!) We will then investigate what happens when we exchange the inertial frame of reference. Suppose
52
is another inertial frame of reference
591
connected to
81
II
through a Poincare transformation
=
Ar:J. )3+b r:J. S It follows that the new basic tangent vectors yr:J.
~
are connected to the old
AS r:J.
where
e {If
are connected to the
{~
+
through the formula {cf.
is the reciprocal Lorentz
(6.12»
matrix. The new basic tangent
vectors give rise to new energy and momentum currents
P
(ll. 7)
AS
(I) S
r:J.
Consequently the new currents depend linearly upon the old currents. ~1
Let
and
~2
contained in a space-slice relative to relative to
82 ,
~1
be two three-dimensional volumes, such that (see fig. 218b). p (2)
r:J.
=
81
and
02
is
in a space-slice
From the tube-lemma we then get
-J ~2 * (2)P
r:J.
-J * 01
P
(2)r:J.
Using (11.7) this is rearranged as P (1)S
(1l.8)
Formally this shows that the components energy momentum vector transform
(P O ,P l ,P 2 ,P 3 )
AS r:J.
of the total
like a Lorentz co-vector under a
Poincare transformation. However, we can not
interpret
(Po ,PI ,P2 ,P 3 )
as the components of some co-vector for the following reason: It has no foot point, as it is not attached to any specific event in spacetime~
To summarize we have now given the promised proof of the second half of Abrahams theorem (8ection 1.6): Theorem 2
T
Suppose
is a symmetric co-tensor Of rank 2 in Minkowski space,
~STr:J.S=O
that is conserved (i.e.
in an inertiaL frame).
8
Let
be an
inertiaL frame of reference generating the basic vector fieZds
~o, •.. '~3
•
Then the foLlowing holds:
(1)
The l-forms
(2)
The "charges" Pr:J.
Pr:J.
= -JO*Pr:J. =
2
J T °dx l dx dx 3 xO=tO r:J.
are conserved, i.e. independent of time. (3)
If two inertiaL frames of references ted through a Poincare transformation
81
and
82
are connec-
II
592 yc!.
=
AC!. 'xi3+bC!. i3
then the "charge" transforms as a Lorentz-coVariant quantity.
i. e.
A
i3 P (ll i3 c!.
(11.8)
Remark: In the theory of general relativity spacetime is curved and therefore we can· not choose a global inertial frame. The energy-momentum currents then no longer exisi and this lead to difficulties in the interpretation of various quanti ties like "the total energy of a closed system".
11.2 SYMMETRIES AND CONSERVATION LAWS IN QUANTUM MECHANICS Now that we understand the formal aspects of conservation laws we turn our attention to symmetries. To get a feeling for the machinery will investigate symmetry transformations in ordinary quantum
WE
mechanic~
We will also indicate how they lead to conservation laws in quantum mechanics, but these are of a very trivial type involving no geometry, so this aspect is only included for physical reasons. We start by conSidering a single particle moving in a potential V{i)
• The system is completely characterized by its Schrodinger wave
function
~(t,i)
• Let us neglect dynamics for a moment and just con-
sider the wave function as an ordinary complex-valued scalar field defined on the Euclidean space R3 • Suppose now that (fA)A€R is a family of diffeomorphisms. Then we can push forward the wave function
~J,.{i)
(ll.9)
= (fJ,.)*~{i)
But as this operation is linear we may equivalently say that we have constructed a family of linear operators, UJ,.~
(ll.lO) If fUrthermore
fJ,.
{UJ,.)J,.€R' where
(f.J,.)*~
is a family of isometries
the inner product (see theorem 13,
~ion
, then
UJ,.
preserves
10.5):
<&J,.~21&J,.~1> = <{fA)*~2 I (fJ,.)*~l> = <~2 I~!> A
Consequently
{UJ,.JJ,.€R
is a family of unita1y
~J,.~
is normalized too, i.e.
UJ,.~
as a wavefunction, i.e.
Jl6J,.~12 dV UJ,.~
=
operators. Observe that
1 , so we can interpret
represents another possible state
of the system. Such families of unitary operators have been studied intensively both from a physical and a mathematical point of view. Before we go into a further discussion, let us recapitulate a few basic facts from elementary quantum mechanics. Suppose
T
is a Hermi-
593
II
tian operator representing a physical quantity T (e.g. the energy refl2 ... '" presented by H = - 2m~+V(x) or the momentum represented by p=-iflv) A
When we measure this quantity the outcome of the experiment is a number, and the possible numbers we can measure are precisely the eigenvalues of the operator
T. Let
$n
be a complete set of normalized
eigenfunctions with the associated eigenvalues
An'
T$ n = An $ n
If the state of the system is characterized by the wave function
$
we
can decompose it on the above set with Here Jan J 2 is the probability of measuring the number mean vaLue of the physical quantity T is given by <.T> = Jij?r$dV = LJa
(11.11)
n
n
An
and the
J2 A n
Now observe that a Hermitian operator
T
generates a family of uni-
tary operators given by (l1.12) This family is actually a one-parameter group, i.e. it satisfies UA2 + A1 = UA2 UA1 and the Hermitian operator T is called the infinitisemal generator of this one-parameter group. On the other hand, if we have been given a one-parameter group of unitary transformations
U ' then a famous A theorem of Stone guarantees that it is actually on the form
UA
(l1.12)
=exp[-kAT1
for some Hermitian operator
T
(Of course
fl
has only been intro-
duced for convenience. It will not occur in mathematical references!) Thus there is a bijective correspondence between one-parameter groups of unitary operators and physical quantities represented by Hermitian operators. In the preceeding discussion we have seen that a family of isometries generates a family of unitary operators. If we assume that the family of isometries is a one-parameter group, i.e. it satisfies f A2 + A1 = f A2 0fAl then i t will actually generate a one-parameter group of unitary transformations. In the end we will therefore find that isometries in space are linked up with various physical quantities! When
{fA)A€R
is a one-parameter group of isometries we can asso-
ciate a vector field
~
to this group. This vector field is the geo-
metrical equivalent of the infinitisemal generator in physics. Geome-
II
594
trically it is generated by the curves that
fA{P)
pass through
P
at
If we introduce coordinates ,
A=O
fA
i
_
~
fA(P)
at
A=O .
(Observe
thus generating a vector at
will be represented by
and the components of the vector field (ll.13)
A
+
a
P).
yi = f~{xj)
are thus given by
i
5!Y..
a - dA I A=O • We call this vectorfield the character~8tic vector field assooiated with the one-parameter group
~f
isometries.
NB! In what follows we will have to do a lot of differentiations with respect to
A
the suffix IA=O
so in the rest Of this ohapter it will aZways be under-
8
at
A=O. For typographical reasons we will drop
tood that differen tiations a:r-e to be \~arried ou t at
A=O.
Using the characteristic vectorfield we can now construct the infinitisemal generator
T: i
A
!~
d -flAT dACe 1/J)
-
T
The infinitisemal generator
is the first order differential ope-
rator given by (1l.14)
We can also consider veotor partioles. Quantum mechanically they are characterized by a complex-valued vector field complex valued scalar field). But
tt{i)
tt{i)
(rather than a
still has the usual interpre-
tation. Its absolute square measures the probability density of finding •
+
the vector particle at the point (ll.lS)
x
--+i+
-+
P{x),=1/Ji(x)1/J (x)
Again we can use a family of isometries, {fA)AER' to generate a A
...
~
group of unitary transformations, UA1/J = (f )*1/J , w~ere the transformed A wavefunction is characterized by the components 1/J~{i) • We proceed to determine the infinitisemal generator T: iA i d A ~ i d :t i -KI'1/J ;rr{U A1/J) =dA[{fA)*'I'] d
[~j
~]
d A axj 1/J A (f_A(x» i.e. in components we get '"
(11.16)
T1/J
i
=-ifi[(d.1/Ji)a
J
j
-
595
This is a bit
surprising~
II
By construction we know that the quantity
on the right hand side is a
covariant quantity, but the separate
terms are not covariant quantities. However we have a skew-symmetrization involved (in analogy with the exterior
derivative~)
Thus we have
incidently discovered how to construct a new vector field out of two given vector fields, the commutator or the bracket. Exercise 11.2.1 Problem: Let u,'t be two given vector fields on a manifold M characterized by the components ai(i) and bi(i) . Show that (a.bi)a j - (a.ai)b j
(11.17)
J
J
are the components of ~other. vector field, which We denote by [~;;]
USing the notation of exercise 11.2.1 The infinitisemal generator
T
~e
have thus shown:
is the first order differential ope-
rator given by
(11.18) Exercise 11.2.2 PrOblem: Show that the bracket satisfies the following rules: (1) Skew-symmetry: (2) Bi-linearity:
(3) Jacobi-identity: EXerci8e 11.? 3 Problem: Let Tl,T2
[~I~]
= -[~;~] [~;~1+~2] [~!~l] + [~;;2] [~;A~] [~;[~;;ll + [~;[;;~ll + [;;[ii;~]] = 0
A
be first-order differential operators, i.e. of the form
(a)
T1/J =-ift
(for a scalar particlei
(b)
£t =-ift[~;+]
(for a vector particle) is given by the first order differential
(11.19)
Show that the commutator [Tl,T2] operators (a) [Tl,T2]1/J _fi 2
(11.20)
(b)
(for a vector particle)
[T1 ,T2]$ = -1i.2[it;[~1;;:2ll
(for a scalar
partic~)
We are now in a position where we can investigate some symmetry transformations. In quantum mechanics the dynamical evolution is conA
trolled by the Hamiltonian
H , and a unitary transformation
U
is
called a symmetry transformation if it commutes with the Hamiltonian,
i.e.
[U,H]
=
0
This has the well-known consequence: If
1/J
is a solution to the equations of motion, then
transformed stat!!
U1/J.
80
i8 the
596
A
Proof:
Putting
$'=U$
II
we get UH$
But symmetry transformations also lead
o
HU$
to conservation laws. If
is a Hermitian operator representing the physical quantity J*T$dV T
T
then
T
is the mean value of this physical quantity, and we say that
is conserved if
J*T$dV
is constant in time. We can then easily
show If
U
is a one-parameter group of symmetry transformations, then
A
the infinitisemaL generator
Proof: From that
TH=HT
T
is conserved.
UAH=HU
we get by differentiation with respect to A A i.e. the infinitisemal generator commutes with the Hamilto-
nian too. But then we obtain
i~d~I*T$dV
I{-i~:t$)T$dV
=
-fH$T$dV +
+
I*T{ind~$)dV
I~H$dV
o
= I*[T,Hl$dV
o
In our case U is generated by applying the space transformations A fA • Consider the Hamiltonian (l1.2l) We know that
H
fA
2 + = - n2mll+V(x)
commutes with the Laplacian (theorem 13, section 10.5)
which leaves us with
Le. A
UA commutes with
H
i f and onLy i f the potentiaZ
under the transformation group
fA' i.e.
V
is invariant
{fA)*V = V •
11,3 CONSERVATION OF ENERGY, MOMENTUM AND ANGULAR MOMENTUM IN QUANTUM MECHANICS Okay, it is time to look at some simple applications. In
R3
the
isometry group consists of reflections (which are uninteresting for the present purpose), translations and rotations, {Cf.
(10.28».
Translations: In Cartesian coordinates a one-parameter group of trans-
lations is given by (11.22)
597 where
a
II
i
obviously are the components of the characteristic vector field. Since the components a i are constant, we get that the infini-
T
tisemal generator
reduces to the first order differential operator
~
T =-ifla j
(l1.23)
ax]
(This is valid for both scalar particles and vector particles!) For translations along the Cartesian axis we get especially A
Tl
,a
=-lTI ax
a
,a
A
=-iTI ay
T2
T3 = -lTIa-z
i.e. the infinitisemal generator of translations represents the operator of momentum! Furthermore we see that the momentum is conserved if the potential is translation invariant, i.e. if
V
is constant. This is certainly
reasonably enough: In the presence of a non-trivial potential the 'particle will experience forces, and therefore its momentum will change. Rotations: Consider first as an explicit example rotations around the
z-axis. This one-parameter group is generated by the matrix
"-I::;:
-::~:
~l·
i .•.
f,(i,)-~~,
By differentiation we get the characteristic vector field -1
o o
:]
[~] [
-y x
o
which is circulating around the z-axis. Exercise 11. J.1 Problem: Show that the rotation matrix
RA with
can be rearranged as,
Sz
-1
o
o o
o o
so that the characteristic vector field is given by,
~I
=
Sz~1 .
Consider the case of a scalar particle. Then the corresponding infinitisemal generator (ll.14) is the operator of orbital angular momentum around the z-axis (11.24) We get similar results for rotations around the z-axis and the y-axis, so we see that the infinitisemal generators for rotations are the operators of orbital angular momentum.
598
II
In the case of a vector particle things are slightly more complicated. Here the infinitisemal generator is given by the commutator, and we get an extra term
i.e. J
(lL25)
- Yax ~] z =-in[x.1... ~y
+ in
5z
(Compare with exercise 11.3.1). The first term is again the orbLtal angular momentum, but the second term is something new. It can be interpreted as the spinoperator for the vector particle. Observe that the operator
ih 5 can only have the eigenvalues O,±h i.e. the spin proz jection along the z-axis can only take the values O,±l (in mUltiples
of
h). So we see that the vector
of 1 unit.
p~rticle
posses an intrinsic spin
(Compare with the discussion in section 3.6). The operator
~
(11.26)
=
t+§
i.e.
Ji
= Li +
8i
is then interpreted as the operator for the total angular momentum. Observe that the angular momentum is conserved in both cases when the potential is rotationally invariant ( 1. e. when it only depends upon the radial distance
r).
Exercise 11.3.2 Introduction: Let us introduce the skew-symmetric matrices components -
(11.27)
s., ~
characterized by the
j
(Si) k = Eijk
Problem: (a)
Show that
AS.
~
(b) (11.28)
e generates rotations arougd the xi-axis. Show that t~e matrices Si' i=1,2,3 satisfies the commutation rules ~
(c) (11.29)
, [So ,S.] = E" ~
[;;:. ;;;:.]
§k
~
E"
~J k
tK
Show that the operators of angular mom~ntum get the following commutation rules ~
(11.30)
J
~
(d)
~J k
J
Let a. be the corresponding characteristic vector fields. Show that they s~tisfY the commutation rules
~
~
[Ji,J j ] = i~Eijk J k (It should be emphasized that they hold both for scalar particles and for vector particles!)
Notice that it follows from exericse 11.3.2 that the operator of angular momentum for a ticle, (9.50)
scal~r
particle, (ll.24), and for a vector par-
(ll.26), both satisfies Dirac's commutation rule
599
II
Exercise 11.3.3 Problem: Consider a vector particle. Show that the operator of angular momentum commutes with a spin-orbit coupling of the form , 4-4A Vl(r)L'S , 4where L is the operator of orbital angular momentum and S the spin operator.
...
It follows from exercise 11.3.3 that a Hamiltonian of the form A.fI2 ... S (11.31) H = - 2m~ + V{r) + Vl{r)L' is rotationally invariant.
(This is the Hamiltonian for an electron
moving around a proton if we put Coulomb potential and
V{r)
e2 V l (r) = 2m 2 c 2 r
=
3
-e 2 /r
corresponding to the
corresponding to the spin-or-
bit coupling). You can easily show that the orbital angular momentum and the spin are not separately conserved, because they do not commute with the spin-orbit coupling. But according to the above exercise 11.3.3 the total angular momentum will be conserved! The above discussion of the angular momentum can be generalized to more complicated systems. In the above cases the particle interacted with a scalar potential. We can also investigate what happens, when the particle interacts with a gauge potential, i.e. it interacts with the electromagnetic field.
Ear simplicity we consider only the interaction
of the electromagnetic field with a scalar particle. An electro-static field, like the Coulomb field, is generated from the electro-static potential ~(~), which is completely analogous to f ...
the potential V(r) in the above discussion! In that case the operator of angular momentum is therefore given by (ll.24). A monopole field, on the other hand, is generated from a
(singular)
vector potential A{~). In this case we have previously seen that the operator of angular momentum is given by (9.47) The correction term to the orbital angular momentum is determined by the similar correction term in the corresponding classical expression, \ cf. (9.39). The correction term to the orbital angular momentum can be understood in the folis not spherically symmetric. When (fA) is a one-parameter family of rotations we therefore have (fA)~'; A But it generates the spherically symmetric monopole field, B = dA,and consequently we have
~owing way too: The vector potential A entering in the Hamiltonian (9.43)
i.e.
A
and
(fA)~
are gauge equivalent:
600
II
A at A = 0 we therefore conclude (cf. (11.18)):
Differentiating with respect to
( 11.32) -q[ii;it] = dx with X = q. ~~A [A=O We interpret X as the infinitisemal generator of the gauge transformation necessary to compensa~the rotation. It remains to identify X • The characteristic vector field a associated to rotati'ons around the unit vector ft is given by ,
~
+
a = n x r
The corresponding operator for the angular momentum around the n-axis is according to eq. (9.47) given by
ft.~
=
ft.!
+ fi,[-;xqA - Kr]
We can now show Lerruna 2
The infimtisemaL generator of the gauge transformation necessary to compensate for the rotation is identicaZ to the correction term to the orbitaZ anguLar momentwn, i. e. ( 11. 33 ) X = fi· [-;xqA - Kr] Proof: We rewrite the
expressio~
X
for
X as
= -(ftxr).qA -Knr = -q[ii'A
From this we get Using that a.a ~
while
j
+ a.a
i
J
=0
a.A j i gM xk ~ a/ = 41f E ijk ? we can rearrange the above expression as gM i i a·X = -q[aj(a.A ) - Aj(a.a ) + 41f( ~
J
J
E •• a ~Jk
j
k. j + n J a.(1S..)]
Sr
~
r
Here the last two terms cancel automatically When you work them out, and we therefore end up with
o
11.4 SYMMETRIES AND CONSERVATION LAWS IN CLASSICAL FIELD THEORY Symmetries of a classical field theory has far reaching consequences as we have learned from the discussion in sections 3.9 and 3.11. Let us now examine it once more in a slightly more general framework. In what follows we will oniy consider diffeomorphisma of the manifold. Usually we pull back differential forms, but when we restrict ourselves to diffeomorphisms, we can equally well push them forward. Consider a field theory consisting of a field configuration
~a
and the action functional:
Sn[~a] Here
~a
=
JnL[~a]E
can stand for a collection of scalar fields or vector fields
or something more complicated, say a cotensor field. using a diffeomorphism
f
configuration
we can push forward the spacetime region ~a'
n
and the field
601 Definition 1 A diffeomorphism (11.
II
is a symmetry transformation if
f
Sn[
34)
for arbitrary regular domains
n
and arbitrary field oonfigurations
Observe that the one-parameter family (fA)AER generates a tangent vector field on M inacanonical fashion. To each point P we have associated a smooth curve f)..(P) passing through P at A=O , thereby generating the tangent vector df).. (P) + (11.35 )
~
---crx-
ere it is understood that differentiation is carried out with respect to )..""0 !). The corresponding vector field ~ is called the oharaoteristio veotor fieZd associated with the one-parameter family of diffeo( H
morphisms (f)")AER' Suppose
fA CU) f-
A'
'I
11
I
1- -
-
U I
-"- .1
-
f-
~l' Fig. 219
602
Then
fA
II
y~ = fA~(XV) which fo~(XV) = x~ . Observe
is represented by the coordinate map
reduces to the identity map at Ithat dy~/dA
A=O;
i.e.
are the components of the characteristic vector field
~
The space time region n corresponds to the coordinate domain U , and the displaced region to fA(U) . The field configuration ~a is represented by ordinary Euclidean functions ~a(X~) and so is the ordinary Euclidean transport of ~ a , i.e. (fA)*~ a are represented by for A=O functions too which we denote ~ A (x~) , reproducing ~a(X~) A a (The explicit form of ~ a will be calculated later on, when we need it) .
Let us now specialize to a one-parameter family of symmetry transformations. Then we can finally state and prove the celebrated Noethers theorem: Theopem 3
(Noetheps theorem)
If a fieLd theopy with the canonicaL enepgy momentum tensor
~~v
(section 3.2) admits a one-papameter famiLy of symmetpy transformations
(fA)A€R
then the cuprent
(11.36)
is conserved.
Here
aV
ape the components of the chapactepistic vec-
tor field,and ~aAthe components of the displaced field configuration.
(As usual we calculate the derivative at
A=O!)
Proof: The proof is quite technical, though interesting. You may skip it in a first reading. First we notice that the invariance implies
The displaced action can then be eValuated as
Here we run into the technical difficulty that not only the integrand, but also the integration domain depends on A. However,when we calculate the integral using coordinates we are free to use the same ccoPdinate domain U in Rn for all the integrals involved. So let us transform the integral back to the A-independent coordinate domain U , performing the substitution y = fA(x) in the integral. This is legal because fA is a diffeomorphism. We then get Sf
v
«(l)[(fA)*~aJ
J L[q,aA(y(x»;k~~a\y(x»;gai3(y)l,l-g(y(x))I~1
A
ay~ax
U
(It is important to notice that y We can then finally perform the (11.
371
_
dS
0 - ~I
A=O
=
J{[~ dL
U
a
still depends on
~fferentiation and V dcj>a ClL d dX a ~ + a(d cj» dA(~ ~ a Cly dX
A!). obtain:
7aA(y(x»))
dnx
603 +
L[~ (I-g(y(x))
+ I-g(x)
II
~ Ifxl ]}dnx
Here we need some identities which we leave to you as an exercise:
Worked exercise 11.4.1 Problem: Remember that a)
Let
I~I
V
denotes the components of the characteristic vector field.
denote the Jacobiant. Show that
~Iffl =~ (~) = aIJ alJ dA ox dA axlJ
(11 . 38)
b)
a
i.e. it suffices to differentiate the trace of the Jacobi matrix, which in turn gives the divergence of the characteristic vectorfield. Show that
(11.39)
Here it is instructive to divide the terms into two groups. Those containing V and those containing a
o
= fU{A + B}/-g dnx
B separately ~ divergences. Consider first A: aL d~a A=a;p~+~alJ~ a IJ a Using the covariant equations of motion (8.35) A is immediately reduced to
We want to rewrite
A
~d
oL d~a
( 11.46)
A =
----±.T-g
a- (.T-g _ _ oL_.~) IJ a(dlJ~a) dA
Then we turn our attention to the second group: B = - aca o~ ) IJ a
av~a
0 a IJ
V
+ La alJ + L----±- a (.T-g)alJ + a'dL (avgall)a v IJ .T-g IJ gall
[o(~~~a) °v~a + olJvL ]alJa +{L I~g v
a).T-g) +
a!~1l
avgall}aV
But here you recognize an old fellRw in the first paranthesis. It is nothing but the canonical energy momentum tensor ~ IJ (Cf. (3.17)). The second term is due to the fact that we work in a covariant fo¥nallsm where .T-g need not degenerate to a constant. We need still a useful identity:
Worked exercise 11.4.2 Problem: Show that the fOllowing identity holds, provided of motion: ( 11.42)
~a
solves the equations
604
II
Using (11.42) We can finally rearrange the B-term as: (11.43)
--La (;::g¥llaV )
B=
I-g
v
Il
Okay, it was a long tour de force, but now A and B has been rearranged as divergences and we obtain the identity which we have looked forward to find: (11.44) As
0
aL ~ = dS dA = Ju[ 1 a {/-g(~ d" I-g Il lJ"'a
--=
+ aV Tv ll ) ~ I-g dnx O
~
U is arbitrary, this is consistent only if the integrand vanishes.
This is the main theorem! To apply it we must gain some familiarity with the current, especially the last part. Let us try to calculate d<PaA/d"
explicitly. We consider first the case where
<Pa
is a collec-
tion of scalar fields. Then by definition (f,,)*
(x»)
=
=
-1
.
(x»,
(f,,)*<Pa(y(x»
i.e.
"
and the last term vanishes
automatically! We thus conclude:
Theorem 1/
(Noethers theorem for scalar fields)
If a field theory, involving only scalar fields parameter family of symmetry transformations
<Pa' admits a one-
(f"}"tR' then the fol-
lowing current is conserved:
where
o ... aV~ lJ i.e. J = T(a,') v is the canonical energy momentum tensor, and Jil
(11.4S) o
T
=
racteristic vector field associated with
...a
is the cha-
(f"},,ER'
However, in the case of a vector field, i.e.
<Pa=Aa' the last term
no longer vanishes. By definition we get v
f*A (xl Il
ax = --aylJ
A (y v
-1
(x»
<pa"(f,(X»)= (f,l*Aa(y(x» A
A
i.e. ax S -1 =-AS(y (y(x») aya
So in this case we get an extra factor
axS/aya
B aya
=~A
S
(x)
which does depend on
" ! If we perform the differentiation and use Exercise 11.4.1 (b) once more we get
i.e. this time we have a non-vanishing contribution to the current
605
II
which now looks as follows: (11.46)
J
~
=
VO
aT,}
~
-
3L
3(3 A) ~
A"
SOa
as
C<
We will return to this peculiar current in section 11.6 and show JI1 = T~ a v where T~
that it can actually be rewritten on the form
v
v
is the true energy momentum tensor (Cf. the discussion in section 3.2) . In the above discussion we have only considered symmetry transformations generated by space time transformations. We can however easily extend our discussion to internal symmetries as well, where an internal transformation is characterized by only affecting the field components. Suppose we have a one-parameter family of internal transformations,
We say that T
T,: (x) ~ A (x) "a a is a symmetry transformation provided it leaves the ac-
tion invariant, i.e. (11.47) for any space-time region Q. Then we can immediately take over the conclusions obtained in theorem 3: Theorem 5 (Noether's theorem for internal symmetries) If a field theory admits a one-parameter family of internal symmetry transformations, (l1.48)
J
(TA) .. ER' then the current d
is conserved. (As usual we calculate the derivative at A=O).
The standard example of an internal transformation is the phase transformation (x)
~
exp[iA](x)
~(x)
~ exp[-iA]~(x)
where is a complex scalar field. The Noether current (11.48) then reduces to
JI1 _ dL d
. ['" dL
=l.'t'~
11
which is nothing but our old friend (3.67) : We can also generalize the concept of a symmetry transformation. In the preceeding discussion we have assumed that the displaced action is identical with the original action.
From a dynamical pOint of view, how-
ever, two actions will lead to the same dynamics provided they just differ by a boundary term (i.e. their Lagrangian densities differ by a divergence). We may therefore say that a space time transformation is a symmetry transformation in the generalized sence provided it satisfies: (11.49)
606 II If we have a one-parameter family of symmetry transformations in the generalized sence,
(f)
A AER
, we must now be a little careful in the deriva-
tion of the Noether current. This time the displaced action is not constant, so we must first subtract the surface term, which can be rearranged on the form
This produces the following additional term to (11.44):
-J
dS;n dA
dX].1
a].1{dAA[~al}v=g
dnx
n Consequently the Noether current must now include an addtitional term of the form
dX\
-cD:"" [~al To conlude we have therefore shown: The most genepal Noethep cuppent consists of thpee tepms: (aJ A piece,
o JIl = aVT Il
(11.50)
V
1
'
which comes fpom the displacements of the space-time points. (b) A piece, Il _ _
(11.51)
J
2
-
"(
"_L_ d~~ a
II
~
a
) dI-
which comes fpom the displacement of the field. (Notice that it is genepated paptly fpom the space-time symmetpies and paptly fpom the intepnal symmetpies involved.} (c)
A piece,
J~
(11.52)
which comes from the boundapy tepm in the dispZaced Lagpangian.(As usual we calculate the depivative at A=O}.
11.5 ISOMETRIES AS SYM~lETRY TRANSFORMATIONS We can now investigate the consequences of the transformations in the isometry group. Consider Minkowski space equipped with a field configuration
~a.
The dynamics of
~a
is controlled by an action
principle, i.e. we have a Lagrangian density constructed out of the fields and their first order derivatives
L = L[g'~a,d
L[!1,a,d
M, so that the action integral
~L~
itself is a scalar field
is covariant. This guaran-
607
II
tees that the field theory becomes covariant, i.e. the equations of motion will be covariant etc. Now, using a global isometry ~a
pull back the field f*~a
f
we can
and_obtain the transformed configuration
. The important thing to recognize then is that f*L[g'~a,d¢aJ
(11.53)
=
L[g,f*¢a,df*¢aJ
This is not true for an arbitrary diffeomorphism! It only works because f is an isometry (in general we get Consequently we get SQ[f*~aJ
(11.54)
L[f*g, ... J on the right hand side) .
JQ*L[g,f*¢a,df*~aJ
JQ*f*L[g'~Ct,d~aJ
Jnf*(*L[g'~a,d~aJ)
Jf(n)*L[g'¢a,d~aJ
Sf(Q) [~aJ because
f
is an isometry and therefore commutes with the dual map.
We have thus shown
A Poincare transformation is a symmetry transformation for any fieZd theory, where the Lagrangian density is a scaZar fieZd constructed entireZy out of the fieZds and their first order derivatives.
If
~a
is
a soZution to the equations of motion, we can transform it into another soZution by appZying a Poincare transformation. The converse is not true. A specific field theory may very well admit a larger symmetry group than the isometry group. See exercise 11.5.1 below.
Worked exercise 11.5.1 (Compare with exercise 10.6.2 and 10.6.3) Problem: (a) Show that the theory of a massless scalar field in (l+l)-dimensional space time is conformally invariant, i. e. the conformal transformations are symmetry transformations. (b) Show also that the theory of electromagnetism in (3+1)-dimensional space time is conformally invariant. I
Now we are ready
\0
apply the machinery from the preceding section.
All we have to do is to extract some suitable one-parameter families out of the full Poincare group. In what follows we restrict to inertial coordinates.
ExampZe 1: The group of transZa~ions: A one-parameter group of translations is given by (11.55) where
a~
are the components of a fixed four-vector
~. Clearly this
constant vector field is the characteristic vector field associated
II 608 with the one-parameter group. Consequently we get from Noether's theorem that the current
is conserved, i.e. ~
¥v ~]
(The last term in (ll.4G) vanishes because
a~
o = a J~ = aV[a ~
is constant). As
aV
is arbitrary, we conclude that the canonical energy-momentum tensor itself is conserved. As a result we see that invariance under the transZation group impLies conservation of the canonicaZ energy and momentum. Example 2: The group of Lorentz transformations.
A
By definition a Lorentz matrix is a 4x4-matrix (6.32)
ATnA = AnA T = n
i.e.
ATrl
satisfying =
nA- l
This is not a useful characterization when we want to construct oneparameter groups of Lorentz transformations. In analogy with the oneparameter groups of unitary transformations (section 11.3) we will try to construct one-parameter groups of Lorentz matrices on the form exp[Aw]
with some matrix
Here it is useful to observe that ded
w satisfies
exp[w]
w •
is a Lorentz matrix, provi-
=t=
w n -nw i.e. nw is skew-symmetric. If we write the matrix elements of w as w~ and if we raise and lower indices using the metric components, \) (Obthen nw is characterized by the "co-variant" components serve that w~ are not the components of a tensor!) ~~ Whenever nw is a skew-symmetric matrix we can therefore construct a one-parameter group of Lorentz transformations as (11. 56) y, = exp[Aw]x, By differentiation of this we get (11.57)
a, --~ d)'
=
= wX 1
i.e. the characteristic vector field
..,. a
is characterized by the compo-
nents (11.58)
a
V
Exercise 11.5.2 Introduction: Let uS introduce 6 matrices labelled where is skew-symmetric in (po). These matrices are goin~Oto be_chara@~erized by the "co-variant" components n n -n n i.e. n itself is characterized by the components: ~p va ~o vp po
n ,
(npo)~v = 8~pnvo 8~on~p
n
II
609
Problem:
a) b)
Let (ijk) be an even permutation of (123). Show that exp[AQ .. J generates rotatiQns around the k'th axis through an angle A .1J Show that exp[An kJ k = 1,2,3 generates a special Lorentz transformation a18ng the k'th axis with the velocity v given by the relation Cosh A = _1_ Il-v 2 The parameter A is called the rapidity and unlike the velocity v it is additive under composition of special Lorentz transformations in the sawe direction.
J~
Let us co.nsider scalar fields first. We then get fro.m theo.rem 4 that aV~ ~ is co.nserved, i.e. v
o As
W
Vp
~ a J~ ~ ~
W
vp
(0 xp¥v~) ~
is an arbitrary skew-symmetric matrix we co.nclude that
o
(11.59)
~ a [xp~V~_Xv~p~] ~
ConsequentZy invariance under the group of Lorentz transformation implies the conservation of the angutar momentum!
It is co.nvenient to. intro.duce (11.60) Then ~pv~
~pv~
=
xp¥v~_xv~p~
co.mprises the currents o.f the angular mo.mentum density,
but o.bserve that it is net a co.variant expressio.n, net even if we restrict o.urselves to. inertial frames. Fro.m the co.nservatio.n o.f the angular mo.mentum we new deduce
o = Here the
last
a MPV~ ~
=
0 P~v~_o V~P~+xp(a TV~)_Xv(o ~p~) ~
~
~
~
°v~ T
two. terms vanish since
is co.nserved and we end up
with (11.61)
i.e. the cano.nical energy mo.mentum tenser is
b~rn
symmetric. There is
co.nsequently no. need to. repair it. In the case o.f a scalar field the canonical energy mom9ntum tensor is the true energy momentum tensor!
In a geo.metrical language we start o.ut with the inertial frame o.f reference
epo
S . We then asso.ciate to.
is skew-symmetric in
(pO)
S
6 vecto.r fields
. Here"" e po
..,.
e po ' where is characterized by the
co.mpo.nents (11. 62)
i.e.
When we co.ntract them with the energy momentum
tens~r
we o.btain 6 con-
served currents: (11.63)
i.e.
(J
po
)~ = x p ~ 0 ~-x 0 ~ p ~
which are interpreted as the currents o.f angular momentum relative to.
II
610 the
xP-x
o
n
-plane. If
denotes a spaceslice relative to
S
then
the associated charges (11. 64)
J
po
=
-J n *J po
are interpreted as the total angular momentum relative to the
x p -x 0 -
plane.
Worked exercise 11.5.3 Introduction:
Problem:
a)
(11.65)
b) (11.66 )
inertial frames of references SI and S2 and the assOangular momenta JI , J2 given by the above formula inertial coordinat~~ (xa)Poand (ya) associated with are assumed related through the Poincare transformation ya = Aa )l + ba B Show that the basic vector fields and ~2 are related through po the formula -+1 .... lJ v\,l +2 e ejl\.,A pA 0 + [b P~20 - bo ~2] po p Show that the total angular momentum measured in Sl and S2 are related through .. v J2 JI v'W + [b p2 - b p2] po P0 0 p IN A pA 0 p2 where is the total energy momentum measured in S2 jl Consider two ciated total (11.64). The SI and S2
Then we turn our attention to vector fields. Here the conserved current is given by (see
(~1.46)
Jjl
and (11.58))
=w
(p&Vjl _ aL AV) vp x a(a A ) jl P
which leads to the conservation of the slightly more complicated angular momentum (11. 67)
~PVjl _ [ pOVjl_ VOPjl] xT xT +
ra(aA)A aL p L
jl V
_
vl
aL a(aA)A jl P -
Here the first term represents the conventional orbital angular momentum of the field, but what is the origin of the last term? From a physical viewpoint the main difference between a scalar field and a vector field is that a scalar field carries no spin, while a vector field carries spin 1, (compare the discussion in Section 3.6). This is reflected in the fact that the scalar field carries no space-time index, while the vector field carries one. This again leads to the different behaviours under symmetry transformations. formation property of
Aa
It is the nontrivial trans-
which on the one hand allows us to construct
eigenfunctions for the rotation operator with non-trivial eigenvalues, while it on the other hand generates the last term in (11.67). Now, for a field carrying spin we expect the total angular momentum to be composed of a spin part and a contribution from the orbital angular momentum: J
L+S
611
II
and it is only the complete angular momentum which is conserved, not the separate terms. Okay, this suggests then that the last term is interpreted as the spin density.
11.6 THE TRUE ENERGY-MOMENTUM TENSOR FOR VECTOR FIELDS We have just seen that the conserved current (11.46) for a vector field has a complicated structure, involving not only the canonical energy momentum tensor, but also a spin-contribution. We also know that the canonical energy momentum tensor for a vector field fails to be symmetric, and has to be repaired (cf. the discussion in sections 3.5 and 3.7). We will now extract the true energy momentum tensor from the expression Jil
(11.46)
=
a
v¥ v il _
dL
a (alJA a )
A a as S a
The main clue is the observation that the expression (3.17) for the canonical energy-momentum tensor is not covariant, i.e. are not the components of a tensor.
o jJ Tv
(Remember we are working in arbi-
trary coordinates on an arbitrary manifold!). The trouble lies in the factor
avAa' where we have differentiated through a vector field
which immediately destroys the covariance (compare the discussion in section 7.1. For a vector field it is only the field strength FaS = aaAs-asAa which is covariant.) However by construction the complete expression for the current JjJ is covariant, and it would be nice to have this covariance explicitly built into the formula (11.46). This is where the last term comes into the game. It can be rearranged as follows:
Wopked exepcise 11.6.1 Problem:
Prove that
---~--- are the components of a skew-symmetric tensor. a (ajJAa)
Using the covariant equations of motion (8.35) we now rearrange the
II
612
Inserting this we can finally decompose the current into two parts (11.69)
jl
Jjl
J (2)
[-0£
-l-aa[~o(~LA
A av aAjl v
r-g
jl a
V )AVa ]
which are manifestly covariant. From exercise 11.6.1 we learn that
O£
(11. 70)
v
a (a A ) (Ava ) jl Ct
are the contravariant components of a 2-form
S
and the last part is
therefore given by (11. 71)
J
(2 )
from which it immediately follows that it is conserved. Furthermore it does not contribute to the total charge since
= -fn*
Q (2)
J (2)
=-f n*5S = fan*s
where we have used Gauss' theorem. Here the last integral vanishes provided the vector field dies off sufficiently fast at infinity. But then we can simply split off the second part. If we furthermore introduce the abbreviation jl
(11. 72)
Tv
we can now
=
-0£ -aAAv jl
3L jl 0(0 A )FVCt + 8 v L
jl a
express the first component as
J
(11.73)
=
T(~i·)
(1)
We want to show that
TjlV
defined by (11.72) is the true energy mo-
mentum tensor. Consequently we must show that it is symmetric, conserved and produces the same energy and momentum as does
¥jlV
•
(Cf. the dis-
cussion in section 3.2). Since
J
and
J
are conserved it follows that
J
is conserved
too. From translat±66al invariance we therefore get thAe) TjlV served. Furthermore TjlV and ~jlV produce
J
ffi~
same total energy and momentum. Finally i t fol-
lows from Lorentz invariance that
TjlV
produces the following conser-
ved angular momentum
i.e.
o= Consequently
TjlV
is con-
does not contribute to the total charge, so
PVjl
0 jl M
=
is a symmetric tensor.
T VP - T Pv
613
II
Exercise 11.6.2 Problem: a) Consider the theory of electromagnetism based on the gauge potential A . Show that (11.72) rep~oduces the true energy momentum tensor (1.41). U b) Consider the theory of a massive vector field W . Show that (11.72) reproduces the energy momentum tensor given by e~ercise 3.7.1).
Using the true energy momentum tensor for vector fields we can now reformulate Noether's theorem in the case of vector fields:
Theorem 6 (Noether's theorem for vector fields) If a field theory based upon a vector field Au admits a one-parameter family of symmetry transformations (fA}AtR > then the following current is conserved: (l1.73)
JU (1)
=
aVT U v
i.e.
J (1)
where T given by (11.72) is the true energy momentum tensor and where ~ is the characteristic vector field associated with the one-parameter family. Exercise 11.6.3 Problem: Consider a mass-less scalar field in (l+l)-dimensional space time respectively the electromagnetic field in (3+1)-dimensional spacetime. (a) Show that the conserved current and charge associated with the invariance under dilatations are given by (11.74)
(D
=1
or 3)
(b) Show that the conserved currents and charges associated with the invariance under special conformal transformations are given by (11.75) (c) Show that the dilatational charge transforms as a scalar under Lorentz transformations and that the special conformal charges transform as the components of a Lorentz vector under Lorentz transformations. How do they transform under general Poincare transformations?
Exercise 11.6.4 Problem: (a) Show that the energy momentum tensor of a conformal invariant field theory is trace-less, TU = 0 . (Hint: Consider the conservation law u corresponding to dilatational invariance.) (b) Consider a field theory possessing an energy-momentum tensor which is trace-less, TU = 0 . Show that the charges (11.74) and (11.75) are u conserved.
II
614
11.7 ENERGY-MOMENTUM CONSERVATION AS ACONSEQUENCE OF COVARIANCE. In this section we will finally
take up the problem of how to construct
the true energy momentum tensor from a more general pOint of view. We saw in section 3.11 how gauge invariance could be used to construct a general expression for the electromagnetic current. In that case we have a Lagrangian depending on a charged field
~
coupled to an exter-
nal gauge potential Ajl . Performing a variation of the external field, Ajl
~
Ajl + EOAjl , the corresponding variation in the action will be li-
near in
OAjl' oS
dS(E) dE [10=0
and we then showed that the coefficient in front of
8A
jl
actually was
identical to the current: dL
(3.76)
aAjl
Furthermore the gauge invariance of the action allowed us to prove that this current would be conserved as a consequence of the dynamics of the charged field
~
.
In the present case we are considering an arbitrary field configura~a.
tion
To construct the Lagrangian, which has to be a scalar field,
we furthermore need the metric
g. We then construct an action fUnc-
tional of the general form
S = JL[g,$a,d$al~dxlA .•• Adxn (Observe that the metric is involved in the construction of the volumeform too!) The basic idea is to treat the metric as an extermal field and see what happens when we perform a variation in the metric: gjlV
~
gjlv+EOgjlV . The corresponding variation of the action will be li-
near in
ogjlV
and can therefore be written on the form oS dS(E) - dE: 110=0
(11.76) The coefficient
Tjlv(x)
~JTjlV(X)Og
jlV
(x) Fgdnx
is a symmetric tensor and as we shall see it
reduces to the true energy momentum tensor 'in the now well known cases of scalar fields and vector fields. To obtain an explicit expn,ssion for
TjlV
in terms of the Lagrangian we need some useful identLties in-
volving the derivatives of the metric.
Worked exeroise 11.7.1 Problem: Deduce the following identities (11.77)
a)
djl[lnDetM(x)] where
M(x)
= tr[M-l(X)djlM(X)] is an arbitrary matrix function.
II
(11.78)
b)
(11.79)
c)
(11.80)
d)
Using (11.79) we now get OS = _ _d_ ILFgdnx
dEjE=O dL f[ agjlV
+
~gjlVL]og
I1v
Fgdnx
from which we read off: TjlV
(11.81)
=
2~ + gjlV L ag jlV
As an example we take a look at the electromagnetic field. Here the Lagrangian density is given by (cf. (3.36)) L
Note that
L
= -!F F PO = 4 po
-!F F g pa oS 4 po as g
is gauge invariant, so
TjlV
defined by (11.81) will auto-
matically become a symmetric gauge invariant tensor. Using (11.80) we find aL ag jlV
= -!F
F
4 po as
[_gI1PgvagOS_gPagI10gVS]
=-!Fjl F av 2 a
Inserting this into (11.81) we then end up with TjlV = _ Fjl F av _! g jlV F F Po a 4 po and
tha~
is precisely the true energy-momentum tensor for the electromag-
netic field (cf. 1.41). Motivated by this example we will call the tensor defined by eq. (11.81) the metl-ic energy-momentum tensor. Exercise 11.7.2 Introduction: We consider a theory consisting of a single scalarfield ~ and assume that the Lagrangian density is constructed as a function of the two scalar fields
~
and
~
=
gaSaa~aS~
=
(d~jd~)
i.e. L = L(~;gaSaa~aS~) Problem: Show that the metric energy-momentum tensor (11.81) coinsides with the canonical energy momentum tensor (3.17) [Hint: Show that
II 616 Exercise 11. 7. J Introduction: We consider a theory consisting of a single vectorfield A and assume that the Lagrangian density is constructed as a function o¥ the two scalar fields,
e where
= ga8A A = (AlA) and a 8 FaS is the usual field strength,
FaS
°aAS-OSAa'
L = L(ga SA A 'gaPgSoF
F ) a 8' a8 po . Problem: Show that the metric energy-momentum tensor (11.81) coincides with the true energy-momentum tensor (11.72). i.e.
Worked exercise 11.7.4 Introduction: We consider a system consisting of a single relativistic particle and assume that the Lagrangian density is constructed as follows I dxU dxS L =-m ,
J
(11.82) Problem:
1 • --=--0 (x-x(A))dL
I-g
where Xa(A) is a parametrization of the particles world line. a) Show the La ran ian densit (11.82) leads to the conventional action: dxU dx
f
S =-m as(x(A))dI"dI"dA. b) Show that the metric energy-momentum tensor (11.81) coinsides with the conventional energy-momentum tensor (cf. the discussion in section 1.6): a 8 1 °IlV ' T = m dT dT --=--o"(x-x(T))dT .
Jdx dx
I-g
As shown by the exercises 11.7.2-4 the metric energy-momentum tensor coincides with the true energy-momentum tensor for a system consisting of particles, scalar fields and vector fields. Interestingly enough we can now give a general proof for the conservation of the metric energymomentum tensor based upon the covariance of the action. The proof is somewhat similar to the proof of Noether's theorem in section 11.4. Especially it is somewhat technical and you may skip it in a first reading: Theorem The metric energy-momentum (]1.81)
T IlV
= 2~
~nsor
+ gllVL
og\.lv
will be conserved for any field theory based upon a covariant Lagrangian( i.e.
the Lagrangian is a scaZarfield constructed from the metric,
the fields and their first order derivatives). Furthermore the conservation of the energy-momentum tensor can be expressed in a covariant way , as
(11.83)
o
617
II
Proof: From covariance we get
* * * * Sn[~~l = JfA(n)L[fAg;fA~a;dfA~alfAE where fA is an arbitrary one-parameter family of diffeomorphisms. For the present purpose we choose the diffeomorphisms flo. so that they map the interior of n into itself, and such that flo. reduces to the identity on the boundary. Observe that this especially implies that the characteristic vectorfield ! vanishes On the boundary. Introducing coordinates we now get
Sn[~~] = Jlf[g~)y);~~(y);~~(y)]l-i(Y)dny (Observe that the coordinate domain this time is independent of A from the beginning). We assume now that ~ (x) is a solution to the equations of motion. This has the important consequence, ~hat ~A does not contribute to the variation of the action integral. (Observe that the figld this time really is varied. In the case of a scalarfield we get for instance: ~A(y) = ~ (f~l(y))). When we differentiate we therefore need only take into considera~ion the~new variations coming from gA IN
We must then
x = f~l(y) . Differentiating exercise 11. 4.1b) A
at
with
10.=0
we get (using
dg)lV ~ = -( a a )g dA jl av
Inserting this we find
o =
r [Tjlv;.:ga Va jl Ju
+ rB
TjlVas/-gldny jlV Performing a partial integration on the first term we now end up with
o = JU[/~gav(l-gTSV) As the variation
as
+ rSjlVTjlV]aS/-gdny
is arbitrary this is consistent only if
-1-0 (l_gT SV ) + r S TjlV r-g \) jlV
(11.83)
=0
Finally we specialize to an inertial frame. Then the identity (11.83) reduces to
av TSv = 0 which is the standard expression (1.38) for the conservation of the energy-momentum tensor.
n
From theorem 8 we learn too that the covariant expression for the divergence of a symmetric rank 2 tensor is given by
II
618
-1-a (!=gTBV)
(*)
+
I-g v
rB T~v ~v
This should be compared to the covariant expression for the divergence of a skewsymmetric rank 2 tensor: (**)
(l_gF Bv )
-1-a I-g
V
In fact (*) is valid for a skewsymmetric tensor too, since in t~at case t~e co~tribu tion from the last term vanishes due to the symmetry of the Chrlsto~fe~ fleld ln the indices (~v) . Thus (*) is valid for arbitrary rank 2 ~ensors: Th~S lllustrates a general principle: The covariant expression for the partlal derlvatlves.of a tensor will contain the Christoffel field, but once we restrict to skewsymmetrlc tenso:s the contributions from the Christoffel field vanishes automatically and we end up wlth the usual expressions from the exterior algebra.
u.g
SCALE INVARIANCE IN CLASSICAL FIELD THEORIES
In the final sections
we will show how conformal transformations can be
combined in a natural way with an internal transformation.
The com-
bined transformation then acts as a symmetry transformation for a large and interesting class of classical field theories. As a preparation we consider scale transformations (dilatations) n-dimensional Minkowski space.
in
Using inertial coordinates such as a
scale transformation is given by
(11.84) Under a scale transformation a scalar field transforms as
The displaced action is therefore given by
Changing variables, 5 D (Q)
ya
AXU, this is rearranged as
[ ] Jn D*q, =
1
L[q,(X);1" -
a
ax~
nn
q,(x}] A d
x
Consequently a scale transformation is a symmetry transformation provided:
(l1.8S) The standard Lagrangian of a scalar field is on the form L = -
l(a )J q,} (a~q,) -
U(q,}
619
II U(~)
Usually the potential energy density
only consists of mass term,
lm2~2, but in general it can be an arbitrary positive function.
If
the theory is scale invariant, it follows from the kinetic term that n
must be 2.
Furthermore the potential energy denSity must vanish.
Thus we are left with the trivial example of a massless free scalar field in (l+l)-dimensions! Let US now combine the scale transformation with the internal transformation,
Here
d
is a real scalar, called the scaling dimension of the field.
This transformation is very similar to the phase transformation in the theory of complex scalar fields.
The displaced action is then changed
to SD(I1) fA
-d
D*~l
= JI1Lf).. -d ~(x)
-d-1 n n ; A a\l~(x} lA d x
Consequently the combined transformation is a symmetry transformation provided (11 .86)
=
Lf~ (x);
a\l (xl
1
If the theory is scale invariant it follows from the kinetic term that (11. 87)
n-2
d
--r
(Scaling dimension of a scalar field)
This number is then referred to as the scaling dimension of a scalar
field.
The potential energy term must furthermore satisfy the homo-
geneity property:
2-n
AnU[ ;..--2-- ~l= U[~l
Substituting
~
_
1 -
and
x = A
n
'2
we therefore see that
U
must be
on the form: 2n n U(x) = U(1)x - 2
(11 .88)
n 2. 3
E.g. in 4 dimensions this means that a term like
is admissible.
Notice, however, that a mass term is never admissible!
The same analysis can be applied to a vector field placed field is given by
A • IJ
The dis-
II
620 This leads to the displaced action:
A scale transformation is thus a symmetry transformation provided (11.89) If the theory is scale invariant it follows from the kinetic term, 1 F F~v that 4" ~v ' d = n-4 (11.90) (Scaling dimension of a vector field) . -2-
-
This number is then referred to as the sea Zing dimension of the vector
fieZd.
Furthermore a potential energy term must satisfy 2-n )..nU [)..-2-- A 1 = U[A 1 ~
~
This is the same condition as before, i.e. in 4 dimensions this means that a term like, U[A 1 = . ~. (A A~) ~
1fT
2
~
is admissible, while e.g. a mass term is never admissible. Notice that with the above choices of scaling dimensions, the conditions (11.86) and (11.89) actually coincide! ~a
If we in general let
denote a collection of scalar fields and/or vector fields, the
preceding arguments thus show that a theory is scale invariant provided the Lagrangian satisfies the homogeneity property 2-n
)..n
(11. 91 )
n
L [ )..-2- ~.
)..
a' This leads to the following rule:
2" d~ C< 1
A theory in Minkowski space is scaZe invariant provided each term in the Lagrangian scaZes Zike )..-n. The scaZing of a term is found by scaZing each fieZd Zike f~-W/2, and each derivative Zike A- 1 . As an important application of this rule we consider the case of a complex scalar field interacting with the Maxwell field through the rule of minimal coupling.
In this case the standard kinetic term is
modified to
The first term presents no problem, but the two last terms scale like
)..(4-3~/2
and
)..4-2n.
iour in
n=4
dimensions.
Thus they only have the correct scaling behavWe therefore conclude:
621 Minimal coupling is only scale invariant in 4 dimensions.
II
In standard physics literature the above criterium for scale invariance is often stated in a slightly different way using dimensional analysis of the individual terms in the Lagrangian. In units where ~ = c = 1 the action S itself is dimensionless. [In general S and ~ has the same dimension so that S/~ is dimensionless. Cf. Feynman's rule (2.20)]. Any physical quantity T will now have a dimension, denoted [T] , corresponding to a power of a length. E.g., the volume element dllx has dimension n. Each term in the Lagrangian density consequently has dimension -no Consider first the kinetic term - !3,,$alJ $ or - ~ F FIJV ~ 4 IJV Since a derivative corresponds to a division by a length, it reduces the dimension by 1. Consequently - n
i.e. the dimension of
= 2 ([]
(and similarly of
- 1)
AIJ)
is given by 2-n
[$] =~
(11.92)
= -2-
2
Notice that the dimension of the field coincides with the scaling behavior of the field (cf. (11.91». Except for the kinetic term all other terms will now contain coupling constants, which mayor may not carry a dimension. Consider e.g. a mass term !m2 $2 In that case we get 2-n -n = 2([m] + [<1>]) = 2([m] + -2-)
i.e.
[m] = -1.
Consequently the mass parameter m has dimension -1 in any number of space time dimensions. Any coupling constant, which carries a dimension therefore generates a "mass-scale", i.e. a suitable power of the coupling constant has the dimension of mass.
As a famous example we consider Newton's constant of gravity. As we have not yet developed a field theory of gravity, we will use Newton's second law,
ix
mM
m--=Gdt 2 r2 as a starting point. Notice that in units where e=l time will have the same dimension as length. Furthermore the mass has dimension -1, so that we get 2
[G] =
[~] 2
+ 2[r] - [M]
-1 + 2 + 1 = 2
dt
The mass scale in gravity is therefore given by (11. 93)
G-!
(/¥
= 5.5 10- 8
kg)
which is known as the Planck-mass. On the contrary a dimensionless coupling constant can not in any way be related to a mass: It simply generates a pure number. Consider, e.g. the coupling constant in electromagnetism. In four dimensions, A has dimension -1. It follows from consideration of the covariant derivative, IJ d - ieA ,that e is dimensionless. Reintroducing ~ and c, the pure number itlJrepres~nts, is given by (11.94)
2 1 e ftc ~m
which is known as the fine structure constant. Returning to the criterium of scale invariance, we can now reformulate it in the following way:
622
II
A theory is scale invariant precisely when it has the following two properties: 1) It contains no mass terms; 2) AU coupling constants are dimensionless. (By abuse of language a theory with these two properties is often referred to as a massless theory). Given a scale invariant theory there exists a corresponding conservation law.
This can be found from Noether's theorem, where the dis-
placed configuration now includes the action of the internal transformation. Consequently a general field configuration as follows
where
d
is the scaling dimension.
$a(x)
transforms
We thus get an additional contri-
bution to the conserved current coming from the factor
e
-Ad
in front.
Since the characteristic vector field associated with the one-parameter group of dilatations is given by
aV the conserved current reduces to (11.95 ) This is a nice except for one little detail.
Unlike the case of
ordinary space-time symmetries, the current is no longer on the form J~
= aVTv~'
This is very analogous to what happened in the case of
Tector fields, where we also obtained an additional spin contribution and consequently had to repair the canonical energy momentum tensor (cf.
(11.46».
This suggests that even the true energy-momentum tensor
needs to be repaired in prder to absorb the additional term in (11.95). Actually it can only be repaired in the case of scalar fields. This is due to a different structure in the two cases. case the additional term is simply a gradient: (11.96 )
- d$
3L
~
= d$3 ~ ~
~
In the vector case we similarly get (11.97 ) (cf. exercise 11.6.1).
_ dA
3L
a 3(3~Aa)
But this is not a gradient!
In the scalar
623
II
To see this let us temporarily assume it was a gradient, A F].lCl. ~ a].lX[A 1 CI. CI. In that case the functional Tn[ACl. 1= la].lACl.F].lCl.dnX where a ary term].l
is an arbitrary constant vector, can actually be converted to a pure bound-
Tn[A 1 = CI. Consequently the functional Tn[ACI.] they vanish on the boundary. If we
n 1 faX[A ]n].ld - x an].l CI. is independent of variations in ACI. perform such a variation, ACI. ~ ACI. + EBCI.
as long as
[A (a].lACI._a].lACI.) + E{B (a].lACI._aCl.A].l) + A (a].lBCI._aCl.B].l)} + E2B {a].lBCI._aCl.B].l}ldnx, CI. CI. CI. CI. We therefore obtain
o=
dEdiT = a].l6 E=O Performing a partial integration this is rearranged as o= a aCl.A].l + g].lCl. a AV]B dnx ].l v CI. This is only consistent provided ACI. satisfies the identity aCl.A].l = g].lCl. a AV v Taking the trace you immediately see that aCl. AV must vanish! This contradicts our assumption that (*) was supposed to hold for an arbitrary configuration!
f[-
As we shall see later, this difference between the scalar case and the vector case also has the consequence that scale invariant scalar theories are actually conformal invariant whereas scale invariant vector theories are not (except in the trivial four-dimensional case). In the following we shall therefore confine ourselves to the scalar case. As usual we modify the energy-momentum tensor by adding a term satisfying the requirements (3.18). tion is a bit tricky.
This time, however, the construc-
It is based upon the properties of a particular
tensor constructed directly from the metric: (11.98)
gCl.~y6
d~f.
gCl.yg~6
-
gCl.6g~y
Notice the following important symmetry properties (11.99a)
g~Cl.y6
gCl.~OY
(11.99b)
gCl.~y6
gy6C1.~
=-
gCl.~Yo
[.'.w-'''''''' i. ,h. pair of indices
first and last
symmetry between the first and the last pair of indices
Using this we now construct the following correction term (11.100)
1
624
e~V
From (11.99b) it follows that
II
~
is symmetric in
and
v •
Furthermore (11.99a) implies that
a~()~v
= 4(~-!/)
apr
aAa~gA~Pv>21
0
Finally we get by explicit computation that eOo = _ e~v
which shows that
a a i ",2
(n-2)
4 (n-1 )
i
e io
'I'
does not contribute to the total energy and
momentum. Thus (11.100) is an admissible correction term! Using the tensor gA~PV we can also construct a conserved current: (11.101) (Notice that the bracket is skew-symmetric in
A
and
~).
This
expression can be rearranged as J~
6
(n-2)
4lIl=1T
Xv
X 6~v _
[a a gA~PV",2] A P
(n-2) 4
v
+
'I'
(n-2)
4lIl=1T
[g
a (gA~PV",2)]
Av P
'I'
a~>2
On the other hand scale invariance led to the following conserved current (cf.
(11.95-96)).
J~
-
- Xv
T~v
(n-2) ~~",2 + ~ " 'I'
By adding these two conserved currents we thus precisely cancel the unwanted term:
J~
Notice that since
j~
is conserved, we get
a (x T~V) = T ~ + xva~ T~v ~ v ~ But here the last term vanishes trivially because 0
=
a
~
=
T~v
is conserved.
Consequently the modified energy-momentum tensor is traceless! Let us summarize the main outcome of the investigation: Any scale invariant scalar theory in ordinary Minkowski space pos-
sesses a conserved, symmetric, traceless energy-momentum tensor (11.102)
T~v
= T~v
+
(n-2) [ 4(n-1)
g
~v
[]
known as the conformal energy-momentum tensor. current (11.103)
J~
is known as the scale current.
=
x T~v
v
The associated conserved
625
II
11.9. CONFORMAL TRANSFORMATIONS AS SYMMETRY TRANSFORMATIONS
Now that the group of scale transformations is under control we return to the problem of conformal invariance in a classical field theory The main difficulty is that we have to combine the conformal transformation with an internal transformation.
Let us first look at this in-
ternal transformation in some detail:
M is an arbitrary manifold equipped with a metric g and M is a suitable diffeomorphism. We would like to assign
suppose that
f:M
~
a scale to the transformation
f.
If
g
and
f*g
happened to be pro-
portional, the constant of proportionality would define the square of the scale, but in general there need not be any simple connection between
g
and
form instead.
f~g
Therefore we turn our attention to the volume
Since
£
and
f.£
both are n-forms, they are necessar-
ily proportional: (11.104)
= Jf(x)E
f*c
Consequently we can always attach a volume expansion transformation
f.
Notice that if
f
Jf(x)
to the
preserves the orientation, then
according to (10.34)
f*t reduces to Ef (where ' f is the volume form generated from the pull-backed metric f*g). This justifies the name of J f (x). The scale " associated with a scale transfonnat1.on can now in -the general case be replaced by the scale
= [J f
"f(x)
(X)]1/n
Let us look at a few illustrative examples. then
f*g
=g
according to whether f
f
is an isometry,
and consequently J
larly, if
If
f
f
(x)
= ±1
preserves or reverses the orientation.
is a conformal transformation, then
f*g
Simi-
= Q2(X)g
and
consequently (11.105) according to whether
f
preserves or reverses the orientation (cf.
(10.40) ) • We can also work out a coordinate expression for ( 1 0 • 8 ) we get
Jf(x).
From
626 [f* tl
(x)
Ig(£(x»
E.
.
J 1· .. I n
1 ••. n
II
~ ax 1
I~I ax
fg(£(x) )
is the Jacobiant of the transformation
where
Consequently
f.
the volume expansion is given by: f,g(f(x)), I~I
(11.106)
ax
rIglX)I
Notice that in ordinary Minkowski space
Jf(x)
Jacobiant when we use inertial coordinates. expansion has been denoted by capital
simply reduces to the
(That is why the volume
j I).
From the coordinate expression follows an important group property. Suppose
f 1 , f2
are successive transformations z
= f 2 (y)
Y
= f1
(x)
Then Jig (f 2 (f 1 (x))) I J f of (x) 2 1
(11.107)
~
I~!I
v'lg(f 2 (f 1 (X)))1 v'lg(fl(X»1
~ I~~I~I
v'lg(f (x)) I 1 J
f2
(f (x)) • J (x) 1 f1
We are now in a position where we can generalize the internal transformation associated with a scale transformation.
Let
~a
denote a
general field configuration consisting either of a set of scalar fields or a vector field.
Under a diffeomorphism
f
it will then not only be
pushed forward but also multiplied with an additional scale. bined operation will be denoted by [Tf¢a 1 (x)
(11.108) As usual
d
T
The com-
and it is given by
f ; din J -1 (x) • [f*~a] (x)
=
f
denotes the scaling dimension of the field
case of a scale transformation, we know that
Jf(x)
~a.
= nn(x) =
In the An.
Consequently (11.108) reproduces the conventional transformation in the case of ordinary scale transformations.
Notice too that when
f
is an
orientation preserving isometry, it just reduces to an ordinary spacetime transformation. The operation (11.109)
Tf
satisfies the important group property
627
II
To check this we notice first that
(cf.
(10.17)).
Then we apply the right hand side of (11.109) to a
field configuration [T f
2
~a
[T f
Jd~~ f2
(x) [ (f 2 ) * (T f <Pa) 1 (x) 1
Jd/n(x)[(f) -1 2 f2
* (Jd/n(x) -1 f1
• [(f )*"'N 1 )1(x) 'I'~ 1
Using the group properties (11.107) and (10.21) this reduces to
(If you think this proof is too abstract, you should try to work it out in coordinates!).
The essential content of the group property
(11.109) is that whenever
M
then the operation
Tf
f
belongs to a group of transformations on constitutes a representation of this group.
T f acts linearly upon the field ~a' For a given field theory we can now investigate if a space-time
Notice also that transformation
f
is a symmetry transformation.
Using the now famil-
iar techniques we calculate the displaced action:
substituting
y
= fIx)
this is rearranged as
Using (11.106) this is finally rearranged as (10.110) Sf (Il) [Tf<Pa 1
Notice that apart from a scaling of the Lagrangian the following substitutions have been performed
628
II
(11.111)
a[jl~~ • [f*
(11.112)
(11.113)
'\tv (x) ...
gj.lV
(f
(x) )
(f (x) )
The first substitution, associated with the potential term U[<Pa.) , is quite harmless. E.g., in four dimensions the potential term U[
= Q-n(x)
f
Consequently the substitutions (11.111) and (11.112) reduce to
and
Consequently the kinetic term is replaced by Qn{_,g
j.lV
(f(x»
ax P ax(J d d d d --- [Q- ap
dyj.l
629 But since
f
II
is a conformal map we know that P axcr g].lV (f (x)) ax = Q-2(x) g Pcr(x) ay].l ayV
Thus the displaced kinetic term reduces to _ 1
[n-d~ ~
nn-2gPcr
~ "
"Opo/
+
n-d][ -d -d <pap" Q acr
Inserting the scaling dimension (11.87), this finally boils down to
The first term is precisely what we are after, but the last term is not a manifest pure divergence.
We therefore rearrange the displaced
kinetic energy density as follows: (11.115) It follows that a conformal transformation with the conformal factor Q2 (x) is a symmetry transformation precisely when 1
(11.116a)
I-g
a].I (Fg
a].llnQ)
= 2-n a].I InQ -2~
a].llnQ
Le. (11 . 116b)
[] InQ = 2;n (dlnnldlnQ)
The latter condition is far from automatically fulfilled, but it tHrns out to be fulfilled for all conformal transformations in the ordinary flat Minkowski space.
Since the conformal group is generated
from isometries and the inversion, it suffices to check that the inversion is a symmetry transformation.
In that case we know from exercise
10.3.1 that the inversion has the conformal factor
,,(x) Using that that
a].llnQ
= -2x- 2 x].l
(11.116) is satisfied.
and
Fg
= x- 2 = 1
it is now trivial to verify
Notice, however, that the above argument
only applies in ordinary "flat" Minkowski space.
In a curved space
time the theory of a massless scalar field is not conformal invariant, at least not when you use the standard Lagrangian! Worked Exercise 11.9.1 Problem: a) Consider a conformal transformation with the conformal factor g2(x). Show that the kinetic term of a vector field is displaced as follows: (11.117)
630 II Let nx(x) be the conformal factor associated with the one-parameter group of special conformal transformations xJl + AbJl<xlx> 2 + 2A
b)
Show that (11.118)
and
(aVlnn ) A
=
0
IA=O
(Hint: Use exercise 10.3.1). Show that the kinetic term associated with a vector field is not conformal invariant when n~4. (Hint: Show that the correction term cannot be a pure divergence when we apply an infinitesimal special conformal transformation).
c)
Exeraise 11.9.2 IntrOduction: Problem:
In this exercise we restrict to the four-dimensional ordinary Minkowski space. Show that minimal coupling is conformal invariant, i.e. show that the gauge invariant kinetic term - a a - !(aa - ieAa)~(a + ieA )~ is conformal invariant.
We can summarize the above findings
(including the results obtained
in exercise 11.9.1 and 11.9.2) in the following way:
In the four-dimensional Minkowski spaae any saale invariant theory based upon saalar fields and veator fields is aonformal invariant. In an arbitrary dimension different from four it is only the saale invariant theories based upon saalar fields, whiah are aonformal invariant. This explains among other things why we could not construct a conformal energy-momentum tensor in the case of vector fields course in the trivial four-dimensional case). ing one:
(except of
The reason is the follow-
Suppose a theory is scale invariant and allows the construcllv such that
tion of a conserved symmetric energy-momentum tensor
T
the associated scale current is given by
SV = x Il
Tllv
Notice especially that the conservation of the scale current implies llv is traceless. that From the vanishing trace it now follows tri-
T
vially that the four currents
[SJ'J =
Sp
given by
« x I x >gllP
-
2x x
P Il
are conserved as well (cf. exercise 11.6.4).
)T llv But these are preCisely
the currents aSSOCiated with the special conformal transformations!
631
II
In a massless scalar theory this is as expected, since such a theory is not only scale invariant but actually conformal invariant.
However, in
the case of a massless vector theory we do not expect any conserved currents associated with the special conformal transformations since such a theory is not conformal invariant! It is instructive to derive the conserved current (11.119) directly from Noether's theorem. Then we have to be careful since the displaced action differ from the initial action by a divergence term (cf. (11.115)): n-2f 1 ( r- 2 ~ ) v_g r- dn x ~ ~a~ v_g~ a Inn
A
This must consequently be subtracted from the displaced action. The Noether current associated with a one-parameter group of conformal transformations is therefore given by: 0 aL d~A J~ = a"T ~ + ~ dA (f),(x)) v ~ IA=O The displaced field configuration is given by
~A(fA (x)) = Jd~~( fA (x) )[ (fA )*
this the Noether current reduces to
(11. 120) which generalizes (11.95). In the case of special conformal transformations, where the one-parameter group is given by xV+AbV<xlx> yV = 1+ 2A
a V = dy = b~<xlx> - 2x~
= (1
0A
is given by
+ 2A
Consequently dQ A = _ 2
~2 1
- 2x
2
v
<
x>
As in the case of scale transformations we thus get an additional term, which we must somehow absorb in the canonical energy-momentum tensor. In analogy with (11.101) we therefore consider the current
J~e=dn-21la,a {[b <xix> n- A P \l
- 2x
V
Since
gA~pv[b <x\j> _ 2x
632
J~ = e
[b <xix> - 2x
II
~
+
3(n-2)ap[(b~xP
2(n-1}
_
x~bP)~21
By adding these two conserved currents we almost cancel the unwanted term:
j~ = J~ + J~ e
= [b \! <xix>
- 2x
v
2Tr1=TT
P
Furthermore the additional term constitutes itself a current which is trivially conserved. We can therefore throw it away, whereby we finally recover (11.119)!
SOLUTIONS OF WORKED EXERCISES No. 11.4.1 (a) The determinant is given by a.
I~I
=E:
.&2
a.l · . ·a.n axl
Using Leibniz' rule we therefore get
d~ 1;51 If we put
A=O
M:~:l):~;
this reduces to
~
I a a.1
d
dA!A=O
(b)
= €a.l ... a.n
We start with the identity
a.
\!
h-1L
=
o\!
aya. axil
Il
Differentiating this with respect to
~axV) a a. dA\aya. For
A=O
~
A we get
v
ax ~Ai:) = 0 +
aya. dA\aill
this reduces to
i.e.
~.axV) dA!A=o\ayll Remark: When trix:
~2L) + dAIA=o'axJl
A is close to
= 0 0
the Jacobi-matrix is close to the identity ma-
II
633
(~] If we work out the determinant and drop all higher order terms the off-diagonal contributions automatically cancel, and we get
I~I_- 1+£ 11+···+£ n n + hlgher . lax order terms S
0
and this is where the trace in (a) comes from! No. 11.4.2
.....La (I-::gfi I_g)l
)l)
v
(Fg)L [.....La ,-- v v-g
+
av L]
In the first term we can use the covariant equations of motion (8.35). In the second term we can interchange a and a . The first two terms can then be reduced to aL
v
)l
o
-a L + -~- a g Q' from which the result follows. V "ga.S v a." ______ _ No. 11.5.1 (a) Let f
be a conformal map, and put
~=f*~
• According to th.15,sec.10.5 we have,
f*[*~l = Qn-2(x)*f*[d~1 where we have used that
<~I~>Q
fQ*~A~ = fQf*[*d~A~l ff(Q)*d~Ad~
(b)
Let
f
be a conformal map, and put f*[*dAl
where we have used
= *~,
n=2. We then get,
=
A'=f*A. From th.15,sec.10.7 we now get,
= r,P-4(x)*f*[dAl
= *dA' ,
n=4 • We then get,
fQ*d~'~dA'
= fQf*[*dAAdAl
ff(Q)*dAAdA
=
0
No. 1.1.5.3 (a) We know that
(NE! Observe that we are using indices). Consequently we get +2
epa
+2
=ypea
covariant indices, not the usual contravariant +2
+l
v
B
~l~a
- Yaep = ypesAa - Yaea.Ap
)~IAS_(x AS+b )~IAa. ( xa.Aa.+b p p S a S a a a. p
634 (b)
II
By definition we have -+
and po = T(e po ") and (b) follows immediately from (a). J
No. 11.6.1
The Lagrangian density for a vectorfield depends only on strength F~v = a~~-avA~ • We therefore get
a~~
through the field-
_ _ a_L_ _ _ dL aF"" _ _ aL p 0_ p 0 aL aL dL ala A ) - aF ~ - aF [o~0a. 0a.cV = ~-~=2~ ~ a. po ~ a. pa ~a. a.~ ~a. No. 11. T. 1
(a)
To prove a) we observe that
a~lnDetM(x)/x=xo
d DetM(x) = MetM(xo)/X=X o
a [DetM(x) ] ~ Det!;J(xo ) /x=x o
=-1 = a~Det[M (xo)M(x)]
But M-l(x )M(x) is a matrix function which reduces to I at ing the ar~ument from exercise 11.4.1 we now immediately get
(b)
x=xo' Repeat-
From (6.47) we get ra.
ag _, a.p ~ ~a. - ~g ax~
Introducing the matrices as ra. = 'TrG-la G ~a.
~
and
=-1 G
we can reformulate this
~
which by a) can be converted to
(c)
Let the matrix we get from a)
ga.~
~ga.~dga.~ = ~ dA
depend in an arbitrary fashion on the parameter " .A. ln / g/ = _1_.A. dA dA
;-rgr
;-rgr = _l_.£&[ dga.~ irgT dga.~ dA
From this we conclude that the symmetric matrices
~ga.~/fiT and
must be identical. (d)
Differentiating the identity
a a.~
o =~ agpa
g
~y
+ga.jlo~oY p a
and multiplying with
a a.~
o =~ gpa
+ ga.Pg
ga.~~y
oa. y
we immediately get:
a~ =~ g +ga.poY ag ~y a
pa
g~Y this is converted to ~
D ______ --'
A . Then
0
635
(b)
II
aL
a~V
dx].l dxv = ;m -dT dT f
')
-
1
1-g
6 4 (x-X(T))dT -
~g
].IV
L
from which we conclude aL- + g].Iv L 2ag].lV
= mfdX].I dT
V 1 -dx --=-o4(x-X(T))dT dT ;_g
n U
No. 11.9.1
for a conformal transformation (*)
a (x) ].I a.
~
(11.112)
reduces to
axP{Q-tX)a[f*<Pa.](f(X)) + [f* ](f(x)) aQ-dl. ay].l axP a. axP f
In the case of a vector field we have aa [f,.Aa.lCf(x)) = 2:..A (x) aya. a
(
\
Therefore (*) leads to the following displacement rule: a a a A ~ axP{n'll [ax A 1 + ax A a n- d} ].I a. ay].l P aya. a aya. a p Using that and this can be rearranged as P a A ... ax axa{n-da A + A a Q-d _ Q-dA ax< ~} ].I a. ay].l aya. Pa a P KayA axPaxa A skew-symmetrization in (].Ia.), which on the right hand side corresponds to a skew-symmetrization in (pa), then leads to the following displacement of the field strength: P a F ax { n -~ ].Ia. a ... x --- ---p + (Aaapn- d- Apaan- d ) } ay].l aya. pa The displaced kinetic energy term therefore looks as follows
636 p
II
a A
T
_ !rf(x)gll"(f(x))gUS(f(x))dX ~ ax ax x 4 ayllayUay"ayS
x{Q-~ since
f
pa
+ A a Q-d _ A a Q-d}{Q-~, + A a,Q-d _ A,a Q-d} a p pa AT TA AT
is a conformal map we know that p
a
A
T
gil" (f(x) )gUS(f(x)) ax ax ax ax dyll ayUay"a? Furthermore d = (n-4)/2 (b)
Thus
=
10.3.1
(11.117).
we have 1
(1+2A
from which (e)
gPA (x)gaT (x)
and the displaced kinetic term now reduces to
According to exercise
QA(X)
=
follows immediately.
If the kinetic term is conformal invariant the displaced kinetic term must differ at most by a divergence term. In terms of inertial coordinates we must therefore have
(**) I f we let fA
(b)
represent the group of special conformal transformations we get from
that
Thus we see that FIl"A b must be a divergence (for an arbitrary choise of b"), but that is impossiblella~cording to the general argument g~ven below (11.97).0
637
INDEX OF SUBJECTS, Abelian gauge theory 58 Abelian Higgs' model 445-50 Abraham's theorem 27,589,591-92 Abrikosov vortex 78 Action of a point particle 33 86 Action of a relativistic field Action of a relativistic particle 294 Action of a relativistic string 443 Adapted coordinate system 396 Affine parameter on a geodesic 535 d'Alembertian 20,364 Ampere's law 5 Angular momentum 28 Angular momentum of a charged particle in a monopole field 489-90 Angular momentum of an electromagnetic field created by a charge-monopole pair 487 Angular momentum operators for a charged particle in a 492-93,495 monopole field Angular momentum of a relativistic string 444 Approximative ground state 142 Arc-length 291 Atlas 256 Backlund transformation 150 451 Bag Basic form 333-336 BCS-theory of superconductor 73 Bianchi's permutability theorem 151 Bion 148,210,211 Bloch wave 235 140 Bogomolny decomposition Bogomolny decomposition in a (1+1)-dimensional scalar field theory 141 Bogomolny decomposition in a ferromagnet 573 Bogomolny decompos~tion in the exceptional ~4-model 580 Bohm-Aharonov effect49-53,78-79,502 Bohr-Sommerfeld quantization rule 200,203 Bosonic spectrum of a charged particle in a monopole field Boundary Bound states (in the sine-Gordon model) 148-49 Bracket 595 Breather 147,210-11 Brouwer degree 564 Brouwer's lemma 564 566 Brouwer's theorem
Canonical energy-momentum tensor 89 Canonical frame vectors in the tangent space 270 Canonical frame vectors in the dual space 303 Canonical identification of tangent vectors and covectors 305-06 Canonical identification of mixed tensors 316-17 Canonical quantization of point particles 68-72 Canonical quantization of relativistic fields 93 Canonical system 65 Cartan's lemma 399 Cauchy problem 109-13 Cauchy-Riemann's equations 365,555 Caustic 173 Characteristic line on a null cone 542 Characteristic vector field associated wtih a one-parameter family of diffeomorphisms 594,601 Charge conservation 7,119-21,588-89 Charged field 116-19 Charge rotation 474-75 Christoffel field 295,373 Circularly polarized light 106 Classical approximation of a propagator 44 Classical vacuum 130 Closed domain 400 Closed form 341 Co-closed ~~f8r.. 364 Co-differential 360-62,421-22 Co-differential of a weak form 459 Co-exact form 364 Coherence length 433-34 Coleman's formula for a quadratic path integral 195 Colour field 451 Commutation rules for wedge products 331-32 Commutative diagram 256 Compact subset of an Euclidean 263,397 " space Complex' coordinate system 560 Complex Klein-Gordon field 113-16 Complex manifold 560 Complex valued differential form 365 Composite particles in non-linear field theories 160-64 Confinement 450-51 Conformal compactification of an Euclidean space 541 Conformal compactification of a pseudo-cartesian space 546 Conformal energy-momentum tensor 624
638 Conformal group 547-51 Conformal invariance in field theories 625-32 Conformal invariance of Maxwell's equations 558 Conformal invariance of the self-duality equation 555 Conformal map 531-33 Conformal map on the sphere 559-61 Conjugate momentum of a 65 point particle Conjugate momentum of a relati vistic field 90 Conservation laws 587-92 Conservation laws associated with conformal symmetry 613,631-32 Conservation law, topological 134 Conservation of angular momentum in field theories 609-11 Conservation of angular momentum in quantum mechanics 597-600 Conservation of electric charge 588-89 Conservation of energy and momentum 589-92,614-18 Conservation of momentum 608 in field theories Conservation of momentum in quantum mechanics 596-97 Constraint equation (in gauge theory) 111 Continuity equation 7,21,381 Contraction of differential forms 356,359 Contraction of tensors 310,313,314 Contravariant components of a tangent vector 272,306 Contravariant quantity 274 Cooper current 75 Cooper pair 73 Coordinate line 270 Coordinate system 254 Correspondence principle in 203 quantum mECc):1anics Cotensor " 308 Coulomb field 14,456-57 Covariant components of a " tangent vector 306 Covariant quantity 274 Covector 300-305 Critical value 515 Cross-product 3,351-52,366 Cross section for scattering in a monopole field 506-09 Curl 5,366 Cyclic coordinates 307,383 Cylindrical coordinates 290 Cylinder (as a product manifold)265
Decomposition of differential forms 332 Decomposition of cotensors 311 Degree of a smooth map 563 o-(delta-) function in Euclidean space 18 o-(delta-) function on a manifold 455 Derrick's scaling argument for scalar field theories 158-59 Derrick'S scaling argument for the exceptional $4-theory 581-82 Determinantal relation 196,197-98 Determinant of a differential operator 194 Determinant of a metric 279,348 Diffeomorphism 515 Differentiable manifold 258 Differentiable structure 257,258 Differential 304 Differential form 336 Dilatation 533 Dilute gas approximation 233 Dirac formalism for monopoles 493 Dirac string 18,383-84,476-81 Dirac's action for a system including monopoles 482 Dirac's commutation rule for the angular momentum operators 493 480 Dirac's lemma Dirac's quantization rule for the magnetic charge 487 482 Dirac's theorem Dirac's veto 478 Dispersion relation 84,90 Distribution 458 Divergence 5,366 Double dualisation 354-55 220-222 Double square well Dual map 352-55, 551-54 Dual map in Minkowski space 374 Dual map in R3 367 Dual map of a weak form 460 Dual vector space 301 Dyon 470 Eckhardt potential 162 Edge condition for a relativistic string 444 Einstein-deBroglie rule 48 Electric current in a gauge theory 120 Electric flux 429,430,432 Embedding 518 Energy current 24,88 Energy density 24,87 Energy-momentum tensor 22-28,320-21 Energy-momentum tensor, canonical 88-89 Energy-momentum tensor, conformal 62). Energy-momentum tensor, metric 615 Energy-momentum tensor of a point particle 23 Energy-momentum tensor of electromagnetic field 26,99-100,320
639 Energy-momentum tensor of massive vector field 108 Energy-momentum tensor, true 89,611-13 Energy split in a double well 232,245 Equation of constraint 111 Equation of continuity 7,21,381 Euclidean action 213 Euclidean group of motions 539 280 Euclidean metric Euclidean propagator 212,214-16 Euclidean vacuum 216 Euler-Lagrange equation for a point particle 33,35 Euler-Lagrange equation for a relativistic field 87,91 Euler-Lagrange equation on covariant form 425 Even form 332 Exact form 341 Exceptional ~4-model 578-82 Exterior derivative 336-42 Exterior derivative in Minkowski space 375 Exterior derivative in R3 371 Exterior derivative of a differential form 337 Exterior derivative of a function 302 Exterior derivative of a weak form 459 Extrinsic coordinates of a tangent vector 271 Faraday's law 5 Fermionic spectrum of a charged particle in a monopole field 497 Ferro-magnet 570 Feynman's propagator 42,167-7 2 Feynman's propagator in canonical formalism 71 Feynman's propagator in momentum space 171-72 Feynman's propagator of a free particle 46 Feynman's propagator of a harmonic oscillator 174,181 Feynman's propagator of a timedependent oscillator 188-93 Feynman's formula for the path-integral 182 Feynman's principle of the democratic equality of all histories 43,93 Feynman-Soriau's propagator for the harmonic oscillator181,187 Fictitious forces 296,299 Field quantum in a free field theory 89-90
Field quantum in a non-linear field theory 94,130-31 Fine-structure constant 621 Flux integral 416 Flux-quantization in superconductors 77 -78,502-03 Four-velocity 293 Free field theories 89-90 Free particle wave function 49 Functional derivative 91
0<
Galilei's principle 298 Gauge covariant derivative 60,69 Gauge group (in electromagnetism) 57 Gauge phase factor 50 Gauge potential 7-11,381 Gauge scalar 59 Gauge theory, basic ingredients 61 Gauge theory of electrically charged fields 116-18 Gauge transformation 10,498-99 Gauge transformation of the action 38 Gauge transformation of the first and second kind 118 Gauge transformation of the propagator 56 Gauge transformation of the Schrodinger wave function 55-57 Gauge transformation on covariant form 307,381 58 Gauge vector Gaussian approximation of the path integral 238 Gauss' law in electromagnetism 5 Gauss' theorem in 3-space 6 Gauss' theorem on a manifold 417 Geodesic 296,535-38 Geodesic equation 296,535 Ginzburg-Landau parameter 436 Ginzburg-Landau theory for superconductivity 432 Gradient 5,366 Gravitational forces 299-300 Green's identities 423 Ground state in quantum mechanics 219-23 Ground state for a double well 232 Ground state for a harmonic oscillator 178 &riound state for a non-pertubati ve sector 138-43 Ground state for a periodic potential 235 Group property for the propagator 46 Hadron 445,450 Hamiltonian 65 Hamiltonian density 87 Hamiltonian formalism for a point particle 65-68 Hamiltonian formalism for a relati vistic field 90-9 4
640
Hamilton's equations for a 66 point particle Hamilton's equations for a relativistic field 90,92 Hamilton's principal function 200 364,422 Harmonic form Heat operator 214 Heaviside function 18 Heisenberg's commutation rules 'for a point particle 71 Heisenberg's commutation rules for a relativistic field 93 Heisenberg's model for a ferro570 magnet Heisenberg's uncertainty principle 176 Helicity of a photon 104 Hermite polynomials 179 Hermitian operator 70 Higg's field 445 Hilbert product of differential 420 forms Hodge duality 354 Holomorphic function 365,555-56,560 560 Holomorphic map Holomorphic map on a sphere 561 252 Homeomorphism Honey-comb structure 334-35 Immersion 517 Impact parameter 490 266 Implicit function theorem 282 Induced metric 285 Inertial coordinates Inertial frame of reference 285 Infinitisemal generator for a one-parameter group of unitary transformations 593-95 Infinitisemal propagator 64 Instanton 216 Instanton in quantum mechanics217-19 Instanton, relationship with tunnel effect 223-25 Instanton gas 232 Integral of a differential form 403- 1'1 Integral of a scalar field 413 Integral of a simple form 418-20 Interior of a regular domain 397 Internal symmetry 116,605 Intrinsic coordinates of a tangent vector 272 Intrinsic spin of a charged particle in a monopole field 495 Inversion 533,540 Isometry 530 Isometry group on a manifold 538 Jacobiant Jacobi matrix
344-45 252
k-form Kink Kink sector Klein-Gordon equation Klein-Gordon field Kronecker delta
327 137 138 95 95-98,427-28 315
Lagrangian density 86 Lagrangian density for the electromagnetic field 100 Lagrangian density for the KleinGordon field 95 Lagrangian formalism and the exterior calculus 424-29 Lagrangian formalism for a monopole 482-86 Lagrangian formalism for a point particle 32-38 Lagrangian formalism for a relativistic field 86-90 Lagrangian formalism for a relativistic particle 295 Lagrangian formalism for a relativistic string 442-44 Lagrange multiplier 571 Laplacian in 3-space 5 Laplacian on a manifold 363,422 Laplace's equation in 2-space 556-57 Legendre transformation 202 Leibniz' rule for differential forms 340 Leibniz' rule in R3 372 Levi-Civita form 350 Levi-Civita symbol 348,350 Light cone 281 Light cone coordinates 150 Linearized equations of fluctuations 153 Linearly polarized light 105 line element 293 Line-integral6,415 Liouville's theorem 561 Locally finite family of functions 405 Lorentz force 4 Lorentz force, generalized 474 Lorentz force in quantum mechanics 50,53-55 Lorentz group 287 Lorentz matrix 287 Lorentz transformation 287 Lorenz gauge 10,112
Ma);he~~c charge 469-70,472 Magnetlc current 471-72 Ma,gnEltic flux 11-14,429,432 Magnetic monopole 13-14,318-20,381-84,469-503 Magnetic string in the abelian Higgs' model 448-49 Magnetic string in a superconductor 78,437 -41 Manifold 258 Manifold defined by an equation of constraint 268
641 Massive vector fieldl07-8,112-13,428 Massless field theories 622 Maximal atlas 257 Maxwell's equations 5,376,381,428 Maxwell's equations including monopoles 472 22,428 Maxwell field Mehler's formula 179 Meissner effect 72,435 Metric 276-84 Metric components 277 Metric energy-momentum tensor 615 Metric on a sphere 281-83 Metric on Minkowski space 288-89 Metric, reciprocal components 278 Minimal coupling 69,117,621 Minkowski metric 280 Minkowski space 284-90 Mixed tensor 312 Mobius strip 346 Momentum density 24 Monopole 13-14,318-20,381-84,469-503 Monopole confinement 450 Monopole in the abelian Higgs' model 449-50 Multilinear map 308 Multi-soliton solution in the sine-Gordon model 153 Nambu string 442-45,452-54 Nielsen-Olesen vortex 448,453 Node-theorem for eigenfunctions in the discrete spectrum 157 Noether's theorem 602-06 Noether's theorem for internal symmetries 605 Noether's theorem for scalar fields 604 Noether's theorem for vector 613 fields Non-linear field theories 9 4 ,127-31 Non-pertubative sector 132 Null-geodesic 537 281 Null-vector Odd form 332 One-parameter family of diffeomorphisms 601 One-parameter group of isometries 593-94 One-parameter group of unitary transformations 593 One-point compactification of the Euclidean space 541 Order parameter in a ferromagnet 570 Order parameter in a superconductor 432 Orient ability of a boundary 401 Orient ability of a manifold 345 Orientability of a regular domain 401
Ortochronous Lorentz transformation
288
Paracompact 406 Partial integration 413 Partition of unity 406 Path-cum-trace integral 199 Path integral 42,181-88 Path integrals and determinants 194 Pauli's exclusion principle 73 Penetration length 435 Phase ambiguity in the path integral 180-81 ~'-(phi-to-the-fourth) model 127 Plane wave 135 Planck mass 621 Poincare group 539 Poincare's lemma for the co-differential 360 Poincare's lemma for the exterior derivative 338 Poincare's lemma for weak forms 460 Poincare'S lemma in R3 372 Poincare transformation 285,286,538-39,607 Poisson bracket for point particles 67 Poisson bracket for relativistic fields 92 Polarization vector for the electromagnetic field 100 Primitively harmonic form 365,422 Probability amplitude 39-40 Product manifold 264-65 Proper Lorentz transformation 288 Proper time 293 Pseudo-orthogonal group 546 Pseudo-orthogonal matrix 546 Pseudo-Riemannian manifold 280 Pseudo-tensor 350,355 Pseudo-vector 3 Pull-back of cotensors 522 Pull-back of covectors 521 Pull-back of differential forms 523 Pull-back of equivalent tensors 534 Pull-back of integralS 528 Pull-back of metrics 522-23 Quadratic Lagrangian 45,194-198 Quantization of the magnetic charge 500-02 Quark 450 Quark confinement 451 Quark potential 452 Quasi-static solution in a non143 linear field theory Rainbow angle Rapidity Rank of a cotensor Rank of a mixed tensor Regge trajectory Regular domain
509 609 308 312 445 397
642 Regular domain Regularity of a smooth map Regular value Relativistic string Restriction of a differential form Riemannian manifold Riemann sphere
460 515 515 442 408 280 560
Sard's theorem 519 Scalar field 274-75 Scalar product between differential forms 357-59 Scale associated with a diffeomorphism 625 Scale current 624 Scale invariance in a relati-
vistic field theory 618-24 Scale invariance of minimal coupling 620-21 Scale transformation in a relativistic field theory 158-59,581-82,618 Scaling dimension of a scalar field 619 Scaling dimension of a vector field 620 Schrodinger e~uation 32 Schrodinger e~uation in the canonical formalism 68 Schrodinger e~uation in the path-integral formalism 62-64 Schrodinge~ wave function 41,55-61 Schwinger formalism for monopoles 492 Self-dual form 355 Self-duality e~uation 554-58 Self-duality e~uation in a ferromagnet 574 Self-duality e~uation in the excepcional ~'-model 579 Simple form 342 Simple manifold 258 Sine-Gordon e~uation 128 Sine-Gordon model 128-29 Singular differential form 454 Singular gauge transformation 18,384 Skew-symmetrization 328-29 Slit experiment 39-40,49-55 Smooth map between manifolds 512 Smooth map in an Euclidean space 252 Solenoid 15-19,52,78-79,306-07 Solitary wave 135 Soliton 144 Soliton-soliton scattering 145-47 Soliton versus instanton 217 Soriau's formula for the propagator of a harmonic oscillator 181 Space like geodesic 537 281 Space like vector
Space slice 431 Special conformal transformation 533 Special relativity, basic assumptions286 Spectrum of the angular momentum in the Dirac formalism 497 Spectrum of the angular momentum in the Schwinger formalism 496-97 Spectrum of the Eckhardt potential 163 Spectrum of the Schrodinger operator 157 Sphere (as a manifold) 259-64,281-83 Spherical symmetry in Minkowski space 289-90 Spinning string 444-45 Spin of a photon 103 Spin of a vector field 610-11 Spin of a vector particle 598 Spin wave in a ferromagnet 573 Standard coordinates on a manifold defined by an e~uation of constraint 267 Standard coordinates on a sphere 261 Standard Lorentz transformation 288 Static solution in a non-linear field theory 138-40 Stationary phase approximation 205 Stationary state 71 Stokes' theorem in 3-space 6 Stokes' theorem on a manifold 411 Stratification 334 Submanifold 395 Submersion 518 Superconductivity 72-79,432-441 Symmetry transformation in field theory 601 Symmetry transformation in quantum mechanics 595-96 -Tangent space Tangent vector Tensor Tensor field Tensor product Tidal force Time-dependent oscillator Time-evolution operator Time-like curve Time-like geodesic Time-like vector Topological charge Topological conservation law Topological current Torus (as a product manifold) Transition function Transport of tangent vectors Transport of tensors Travelling wave True energy-momentum tensor Tube lemma Unitary group Unitary matrix
272 269-76 312 314 310,314 299 188 169 291 537 281 133-35 134 134 265 255 512-15 ~24-25
136 89,611-13 587 58 58
643
Unitary transformation Unit tensor field
498 315
130 Vacuum, classical 132 Vacuum sector 440 Vacuum texture 275 Vector field Vector particle 59 4 -95 270 Velocity vector Virial theorem in (1+1)-dimensional space-time 159 Virial theorem in the exceptional <j>4-model 582 625 Volume expansion Volume form 343-44,351 Volume form on a sphere 567 414 Volume of a regular domain Vortex-string in a superconductor 78,441
Wave equation 557 Weak coupling approximation 199 Weak k-form 458 Wedge product in 3-space 6 Wedge product of differential forms 330 Weierstrass' theorem 561 Wick rotation 212 Winding number 564 WKB-approximation 204-209 World-line 284 World-sheet 442 Wronskian 241-42 Yukawa-potential Zero-mode in a non-linear field theory Zero mode in quantum mechanics
97 158 238