Lecture Notes in Physics Edited by J. Ehlers, MLinchen, K. Hepp, Z0rich, and H. A. Weidenm~Jller, Heidelberg Managing Editor: W. Beiglb6ck, Heidelberg
6 I
G. A. Goldin, R. Hermann, B. Kostant, L. Michel, C. C. Moore, L. O'Raifeartaigh, W. Rehl, D. H. Sharp, I. T. Todorov
Group Representations in Mathematics and Physics Battelle Seattle 1969 Rencontres Edited by V. Bargmann, Princeton University II
Springer-Verlag Berlin.Heidelberg • New York 1970
PREFACE
The Battelle Seattle Summer Rencontres in Mathematics and Physics have been established in order "to provide a channel of communication between mathematicians and physicists".
Each year a topic of common interest was chosen which--it was
hoped--would lead to fruitful discussions. The 1969 Rencontres dealt with "Group Representations Physics".
In the view of most participants,
expectation.
I believe,
in Mathematics and
these Rencontres lived up to
The discussions were lively, mathematicians
and physicists found many
questions of common interest to talk about, and they learned from each other.
On
the mathematical side the emphasis was on the problems of infinite-dimensional
re-
presentations,
a subject which has attracted and continues to attract many workers
in functional analysis.
On the side of physics the applications to quantum theory
of both finite- and infinite-dimensional
representations were considered, with par-
ticular attention to the theory of elementary particles,
including its most recent
branch, "current algebra", which leads beyond the framework of Lie groups and calls for new methods. As it should be, individual contacts and informal gatherings played a major role during these Rencontres. in Mathematics
In addition, there were four series of lectures--two
(by E. M. Stein and by C. C. Moore) and two in Physics (by L. Michel
and by L. O'Raifeartaigh)--and
a number of seminar talks by various participants.
The present volume contains three of the four lecture series below) and five seminar talks, written up by their authors.
(see the Note
The reader is expected
to be familiar with the main facts of the theory of group representations. other hand, to help the mathematician,
On the
Michel and O'Raifeartaigh have carefully
Note. Due to the pressure of other work Professor Stein was, Unfortunately, unable to prepare his lecture notes for publication. His course of lectures was entitled "Special Topics in the Representation Theory of Semi-Simple Groups" and dealt primarily with the analytic continuation of representations. Professor Stein has kindly supplied the following list of his publications on this subject. (a)
E. M. Stein, "Analytic Continuation of Group Representations", Advances in Academic Press, New York and London.
Mathematics, 4 (2), 172-207 (April, 1970). (b)
A. W. Knapp and E. M. Stein, "Singular Integrals and the Principal Series I. and II.", I. ~oc. Nat. Acad. Sci. U.S.A., 63 (2), 281-284 (June, 1969); II. Ibid., 66 (i), 13-17 (May, 1970).
(c)
A. W. Knapp and E. M. Stein, "The Existence of Complementary Series". To appear in Problems in Analys~s, Symposium in Honor of Solomon Bochner. Princeton University Press.
IV
stated the quantum theoretical postulates and described the experimental evidence on which the group theorectical treatment of quantum physics is based. The contributions to this volume range from systematic accounts of extensive fields to reports on current research on very specific questions. papers by Moore and by Kostant may be called "purely mathematical".
0nly the
All other
papers--although highly mathematical in content--are concerned with, or at least motivated by physical problems.
A few remarks on the interrelation between these
more physical papers may help some readers.
The papers by Michel and O'Raifeartaigh
give a very comprehensive account of the application of group theory to the most varied problems of quantum physics.
While the two papers are, in the main, comple-
mentary to each other they overlap in some measure, but wherever they do, the discussion of the questions they both treat is sufficiently different to be highly interesting and illuminating.
The contributions by Goldin-Sharp and by Hermann are
concerned with current algebra.
An introduction to the concepts involved here may
be found in the last chapter of Michel's paper and, in greater detail, in the last chapter of O'Raifeartaigh's paper.
Lastly, the papers by Todorov and by RHhl are
independent of each other and of the remaining papers in this volume. On behalf of all participants I take great pleasure in thanking the officers and the staff of the Battelle Memorial Institute, who did so much to provide an ideal setting for these Rencontres. Special thanks are due to the technical typing staff at the Battelle Seattle Research Center, for their painstaking efforts in preparing this volume for publication.
April, 1970
V. Bargmann
TABLE OF CONTENTS
CALVIN C . M O O R E Restrictions of Unitary Representations to Subgroups and Ergodic Theory: Group Extensions and Group Cohomology . . . . . . . . . LOUIS MICHEL - Applications of Group Theory to Quantum Physics Algebraic Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
L. O'RAIFEARTAIGH - Unitary Representations of Lie Groups in Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
144
BERTRAM KOSTANT - On Certain Unitary Representations Which Arise From a Quantization Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . .
237
I. T. TODOROV - Derivation and Solution of an Infinite-Component Wave Equation for the Relativistic Coulomb Problem . . . . . . . . . . . . . . .
254
W. RUHL - Tensor Operators
279
for the Group
SL(2,C)
. . . . . . . . . . . . .
G. A. GOLDIN and D. H. SHARP - Lie Algebras of L o c a l Currents and Their Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ROBERT HERMANN
-
List of Attendees
Infinite Dimensional Lie Algebras and Current Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
300 . •
312 339
RESTRICTIONS OF UNITARY REPRESENTATIONS TO SUBGROUPS AND ERGODIC THEORY: GROUP EXTENSIONS AND GROUP COHOMOLOGY by Calvin C. Moore*
PREFACE
These notes are divided into two rather distinct parts, the first of which concerns the restriction of unitary representations of a group to one of its subgroups, and the connection of this with ergodic theory, while the second part concerns group extensions and the connection of this with unitary ray representations. Some background concerning representation theory is assumed and the reader should consult relevant portions of Mackey's notes [33], and survey article [34], Dixmier's book [9], and Chapter I of [3].
The square brackets refer to the common bibliography
for both Part I and Part II at the end.
PART I.
RESTRICTIONS OF UNITARY REPRESENTATIONS TO SUBGROUPS AND ERGODIC THEORY
i.
INTRODUCTION
This first part concerns the general question of what happens when one takes a unitary representation of a locally compact group and restricts it to a subgroup
H
of
G.
G, say an irreducible one,
One source of interest in this problem is
ergodic theory as we will indicate below, but we believe a thorough study of this type of question will shed much light not only on representation theory as such but will produce much useful information concerning the structure of locally compact groups and their subgroups.
Gelfand and Fomin [13] were perhaps the first to realize
the relevance of this kind of problem concerning unitary representations for ergodic theory.
They showed how one could study geodesic flows on surfaces of constant nega-
tive curvature by looking at unitary representations of the group
SL2(R ) .
This
approach was extended by Parasyuk [43], Mautner [35], Green [i], Auslander and Green [2], and the author [39].
Part of these notes are an exposition of some of the
results in [39] without proofs, and the reader is referred to this paper for further details.
We shall also discuss some related results which will appear shortly.
Chapter I of [3] contains an exposition of some aspects of the theory of unitary representations which we shall use as a general source both for Part I and Part II.
* Department of Mathematics, University of California, Berkeley, California
94720.
We shall suppose for the moment is a Lie group with Lie algebra of
G
and let
X
G
of
~(x(t))
is a one-parameter
operator
where
exp
denotes
such that
he a continuous
Then
~(exp(tX))
operators.
= exp(itA)
operators.
where
G
and given
X 6 G, the various
arbitrary
representation.
is limited
dimensional, by
of
A.
If
G
~(t) = exp(itA).
the second
is to determine
for the unitary
operator
the operator
defines
G
tion of
H
of
G
and restricts
and assumes
and
X, and for an
of
to
A
on the possi-
R, so that
G
is one-
a unitary representation
G.
concerning
of A,
Our results will con-
differently; of
H.
G
suppose
and suppose
We assume
that that
that this representation
G, and then ask what conclusions
can be drawn from this fact. of
for a given group
is semi-simple.
subgroup
is a representation
to a representation
A
representation
of these opera-
type of the operator G
the real line
A
case, namely w h e n
is a one-parameter
= exp(itA)
can be extended
is understood
or that one can put limitations
This same problem can be viewed slightly
z(exp(tX))
self adjoint
exp
In this case we can extract no information
the opposite
X E G
G, and so
= iA, but we shall not
and properties
and in fact the same situation holds for any vector group
H = {exp(tX)},
into
that hold for fixed
is for instance
then any self adjoint
cern exactly
G
One might hope to be able to say that the spectrum of
to a very few possibilities
ble eigenvalues
R
possibilities
The object is to get results
is a one-parameter
in this Rencontres.
The problem that concerns us specifically
A = idz(X).
G
It has an infinitesimal
d~(X)
the definition
that
unitary representation
(usually unbounded)
We write
in other lectures
our approach
= exp(tX)
the exponential map of
concerning
tors since this is discussed
x(t)
there is a unique
in the usual way for unbounded enter into more discussion
~
G.
group of unitary
or in other words, A
Let
be an element of
subgroup
generator,
G.
in order to illustrate
Clearly whether
concerning
one starts with a
H, or whether one starts with a representa-
that it extends
to
G
comes
in physics
for if
to exactly
the same thing.
It
is only a matter of emphasis. Such problems
are relevant
group of a quantum mechanical of
G
on the Hilbert
system,
space associated
rather one has a ray representation doesn't matter,
in
G
one has associated to the system.
of
In any case,
tum, angular momentum,
and so on.
conclude about the spectrum
level this
form part of the infinitesimal
H.
question
generators
algebraic
We have spoken about the restrictions subgroups
A = -ida(X)
interpretations
to be treated for various
such as energy,
momen-
to ask what one can
type of these operators,
tion of a larger group of some specified
one-parameter
a unitary representation
the operators
It is an obvious
or unitary
symmetry
(This is not quite true, but
G; at our present heuristic
have in many cases natural physical
fact that they
is some postulated
and in fact can be gotten around by well known methods
in Part II of these lectures.) X
G
based solely on the
of a unitary
representa-
structure.
of representations
of a group
One can raise the same kinds of questions
G
to
for larger
subgroups
of
G.
Some of the theorems below make sense in this generality
and we
shall state them in that form.
2.
STATEMENT
We turn now to the statement a special
following
H
Let
G
has property
E
(in
v
~(h)v = v
h £ H, we have
for every
~
of
of the representation
have only the obvious
for ergodic
theory.
If
H
of the Lie algebra,
generator
A = -ida(X)
presentation
~
of
~(g)v = v G
of
G
every
then the condition
essentially
0
we say that
such that
g 6 G
~(h)V c V
and every
v ~ V.
tion of any representation presentations
~
the condition
having any eigenvalues.
asserting
its closure
H
subspace
G
to
E
sional representation
of
those subgroups
the re-
that a subgroup may
if for every representation space of the re~(g)v = v
forbids
H
E
and
WM
is a one-parameter
H
E
subgroup
A = ida(X)
from
is motivated
by ergodic
theory
are satisfied,
or
(E) WM
to consider of a group
one may infer
or weakly mixing (in
closed subgroups. G
(WM).
G) if and only if It is
can never have proper-
~
of
G
induced by the trivial one-dimen-
Furthermore,
if
G
is abelian,
no proper closed sub-
Our goal is at least in some cases
to characterize
above, we see that we are going to have to assume
non-commutative,
subre-
generator
are ergodic
has property
for
the restric-
of a given group which have one or bohh of these properties,
from the two examples sufficiently
K. E.
WM
if satisfied,
these conditions
the representation
to an
the infinitesimal
of course
of the Hilbert
If again
does, and so it would suffice
group can have property
V
property
the infinitesimal
that certain group actions
be examining
forbids unless
to have any finite dimensional
ones.
The terminology
H
singles
corresponding
h ~ H, we must have
H
easy to see that a proper compact subgroup ty
stronger
has property
forbids
We note that a subgroup
for the
vectors.
for every
of
In other words,
This definition
{exp(tX)}
as an eigenvalue,
This condition,
and as we shall see later, whenever results
H
other than the obvious
of a Lie group,
vectors.
group
and every finite dimensional
presentation
G,
H, the condition says that we
is a one-parameter
has G-invariant
More precisely,
of
such that
and are looking for invariant vectors
We want also to single out a slightly have.
g 6 G.
the we
may or may not have which will be quite relevant
from having
G
for every
the G-invariant
H
a subgroup; z
as
as dis-
to introduce
H
space of this representation
to the subgroup
ones, namely
out a property which a subgroup
compact group and
concerns
generators
G) if for every representation
in the Hilbert
if we have a representation
the first of which
for infinitesimal
the theorem it is convenient
be a locally
and for every vector
restriction
eigenvalues
In order to formulate
definition.
shall say that
element
of our results,
case the study of possible
cussed above.
OF RESULTS
and that
H
is sufficiently
non-compact.
that
and G
is
#
Semi-simple Lie groups are certainly one of the most important classes of groups, and they are in a sense as non-commutative as possible.
The main result
below will characterize those subgroups of such a group which have properties and
WM.
E
This result is contained in [39] and we refer the reader to this paper for
a more detailed discussion. If
G
is a semi-simple Lie group, let
G/Z, where
Z
is the center of
~ni=l G*i
G.
G*
be its adjoint group, that is,
It is well known that
G*
is the product
of a finite number of simple Lie groups, each having center reduced to the
identity element. bra of
G
This is the global version of the decomposition of the Lie alge-
into a sum of simple ideals.
Let
Pi
denote the projection of
G
onto
G~. We shall say that a subgroup H of G is ~otally non-compact if Pi(H) has l non-compact closure in G~ for each i. Intuitively this says that H sticks out 1 non-compactly in the adjoint group of each simple factor of G. If G is simple with finite center, the condition is simply that the closure of
H
be non-compact.
Theorem i
For a semi-simple group
G
and a subgroup
H
of
G, the following are
equivalent (a)
H
is totally non-compact.
(b)
H
has property
E.
(c)
H
has property
WM.
Thus for a totally non-compact subgroup
H
of
G
and any representation
of
G
which has no G-invariant vectors, we can conclude that the restriction of
to
H
has no invariant vectors, or for that matter no finite dimensional invariant
subspaces.
Even if
H
only partially satisfies the non-compactness conditions we
can still extract information. of
H
into the first factor
For instance if GI
H c G I × G2
any finite dimensional subspace for
H
is left fixed pointwise by
lows by a detailed analysis of the proof in [39]. the case of a one-parameter subgroup operator
A = -ida(X)
and if the projection
is totally non-compact, then one can conclude that GI.
This fol-
If we specialize the theorem to
H = {exp(tX)}, X E G, we can conclude that the
has no eigenvalues provided
H
is totally non-compact.
This result overlaps with the O'Raifeartaigh theorem [42], and gives a stronger conclusion under much stronger hypotheses.
To check that a one-parameter
group is totally non-compact is in any given situation, a rather routine matter. The result above contains as special cases the results of Gelfand and Fomin, Parasyuk and Mautner mentioned above. For one-parameter groups we can in fact get much more information concerning the infinitesimal generator
A = -ida(X)
pact one-parameter subgroup. small number of possibilities.
when
X E G
The unitary type of
A
generates s totally non-comis in fact limited to a rather
~$e introduce the Hilbert space
H+(n)
which is to
consist of all square integrable functions on the interval
(0,~), with Lebesgue
measure, taking values in a standard n-dimensional Hilbert space an integer or
+~.
(-~,0), and on
Let
H-(n)
(-~,~), and let
and
H(n)
H n.
Here
n
is
denote the similar spaces of functions on
M!(ni)(M(n))
denote the unbounded self adjoint
+
operator on
H-(n)(H(n))
We note that
Hi(n ±)
which is multiplication by the scalar function
is a non-negative
(respectively a non-positive)
f(x) = x.
operator.
Theorem 2
Let
G
be semi-simple and let
parameter subgroup, and and let
A = -ida(X).
~
be a representation of
Then if we write
positive and negative parts of equivalent to
H = {exp(tX)}
be a totally non-compact one-
G
with no G-invariant Vectors,
A = A+ + Ani
A, there exists
where
A+
and A ±
such that
Ais
are the
~ unitarily
M±(n±).
The proof of this is contained in [39] and the reader is referred to that paper for the details. such that its powers
We also note that if w e have a single element g
n
g
of
G
form a totally non-compact subgroup, we can obtain an en-
tirely analogous result for the unitary type of the operator
~(g)
(see [39]).
The result above for one-parameter subgroups is best possible in that all + choices of n and n occur, and if one considers a subgroup for which the hypothesis fails, then one can find a representation for which the conclusion fails. fact, for
G = SL2(R)
In
and for a one-parameter subgroup generated by a nilpotent + n and n- occur. The situation for irreducible
matrix all possible choices of
representations of SL2(R) is quite interesting; for the principal and complemen+ + tary series, n = n- = i, and for one discrete series, n = i, and n- = 0, while + n = 0, and n- = i holds for the other discrete series. The one representation of the principal series which is not irreducible decomposes into two irreducible sum± n .
mands which behave like discrete series for
For higher dimensional semi-simple groups, the situation becomes a bit simpler.
~iore precisely,
isomorphic to ~.
if we exclude any group
G
which has a simple factor locally + SL2(R) , the only possible choices for n and n- are either 0 or
This fact is implicit in the argument contained in [39].
A = -ida(X)
alence, one of three types, M+(~), M-(~) replacing
Thus in this case
for a totally non-compact one-parameter subgroup is up to unitary equiv-
~
or
M(~).
Since changing
by its contragradient representation will interchange
X M+
into and
-X
or
M-, we
really have only two distinct cases, which we can classify as one sided spectrum or two sided spectrum. W e can raise the question of when every totally non-compact one-parameter group has two sided spectrum for every representation. in progress has shown that this is true for
SLn(R)
for
S. Scull in a dissertation n ~ 3.
B. Kostant has
proved that this is also true whenever the Weyl group of the maximal compact subgroup
6
K
of
G
contains
the element
not every element of
SPn(R)
-i.
In this connection we should also remark that
has one sided spectrum as w e shall see in a later sec-
tion of these notes. Finally we should like to indicate one application tum physics.
Let
P
denote the Poincar~
group a larger symmetry semi-simple, physical
group
and that
information
~
G
GI
X
~
GI
invariant vectors
A = -ida(X)
tent the representations
of
of mass
P
G
containing
P
i and 2 and the remarks
is a subG
To extract
can look like.
P
is
it, and Theorems
Indeed
cannot be of interest.
corresponding
The spectrum of
1
and let us assume that
A
Now
to translation controls
that can occur since in an irreducible
m, the energy operator has spectrum Theorems
on a Hilbert space. and decompose
since such vectors
is the energy. of
G P
to qaun-
P
We shall assume that
about what this decomposition
be the element of the Lie algebra of
time so that
of
to
of these results
and let us assume that
origin.
is a representation
be the largest normal subgroup
there are no let
of u n ~ w n
one would restrict
and 2 above supply information let
group,
in
to some exrepresentation
[m,~).
following
them give us the following
result.
Theorem 3
Under the above hypotheses, M-(~)
or
M(~),
Proof. subgroup
as defined
spectrum of
X
isomorphic
multiplicities
is unitarily
in Theorem
is totally non-compact to
SL2(R)
it should be noted,
to verify
either
to
M+(~),
in
that the one-parameter
GI, and that
so the result follows. is obvious
G1
has no fac-
(The result concerning
on other grounds
once one has the
A.)
This result says that in the decomposition sentations
equivalent
2.
It is a simple algebraic matter
generated by
tors locally
A
corresponding
to arbitrarily
For a survey of the representations
of
of
~
on
P, we must find repre-
small mass or zero mass or imaginary mass. P, see the article of 0'Raifeartaigh
in this
volume. As we have said before, we shall not enter into the details Theorems
1 and 2.
There is, however,
one important
of the proofs
fact, Lemma 4.2 of
the proof of Theorem i, for which we now have an alternate argument. says that a one-parameter property
E.
T. Sherman.
The argument The alternate
of the universal
argument
group of methods
[39] used in This lemma SL2(R)
SL2(k), where
covering
groups of this group defined of the covering
k
is a p-adic field,
in [40]. groups of
and
the
and also for the
To carry this out however, SL2(R)
has
due to
is global in nature and has in addition
that it works
analysis
covering
in [39] is based on infinitesimal
property
the algebraic
for
subgroup
of
SL2(k)
one needs
Contained
in
[40].
We will carry out the proof for
redefines
the meaning
over word for word
of the symbols
x(t)
=(~
w(t)
= x(t)y(-t-l)x(t)
We let
~(x(t))v =
of
= v
y(t)be and
SL2(R)
for all
(z(y(-t-l))v,v)
Itl ÷ ~.
h(t) t
and v
as ÷ v
eter group
h(t),
t -I.
s # 0
Since
g E SL2(R ) .
3.
where
that
h(t)
is a
is a unitary re-
APPLICATIONS
= i.
We let
÷ (v,v) = 1
as
is orthogonal
= a(t) It follows
(this is Mautner's
This completes
to
v,
we see that immediately
that
Since the one-paramlemma
[35]), it follows
the proof of the lemma in question.
TO ERGODIC THEORY
to the applications
of representations
little direct connection with ergodic
s(t)
~(h(s))v = v.
theory and defer for a later section a treatment the restriction
and we define
lim(~(w(t))v,v)
(~(h(t))u,v)
~(h(t))u ÷ v.
E
We shall now turn our attention
cerning
~
is
(~(w(t))v,v)
(~(h(t))u,v)
and then that
t > 0, has property for
x(t),
that
shows that
= a(t)v + s(t)
Is(t) l2 = i.
Itl ÷ ~, or equivalently,
~(g)v = v
godic
Let us suppose
Itl ÷ ~, we see that
~(h(t))u
under consideration
is a vector of unit length such that
(u,u) = i, and note that
for any
subgroup
It may be verified
An easy calculation
la(t) l2 +
Is(t) l ÷ i
that
t.
so that
~(h(s-lt))u
that if one
the argument will carry
of the matrix
= w(t)w(-l).
and that
W e then write
and note that
the transpose
and if we let
u = z(w(-l))-iv
that we will introduce,
that the one-parameter
diagonal matrix with entries presentation
with the understanding
to the general case.
W e may clearly assume ~).
SL2(R)
to er-
of some more questions
to subgroups.
theory whereas
of these results
con-
These final results have
Theorems
i and 2 have a very
direct connection. Let us first introduce theory. called
Let
M
the setting
be a Borel space,
the Borel sets,
and let
G
that is, a set equipped with o-field be a locally
of the second axiom of countability. transformation
in which we are going to study ergodic
We shall @uppose
group so that we have specrified a map
f(g,m)
= g • m
of
onto itself and such that the function
M
this bijective into itself. f(g,m)
such that for fixed
map,
G
g E G, the map
of
G x M
f
of
should consult Chapter
I of
acts on
in the sense M
as a
G x M ÷ M, written
m + g • m
M
is a hijective map to each
g
in
G,
if
G
is given the o-field
is given the product Borel f-l(E)
Thus not only is varies
M
group if the map
where
G x M
is a Borel function
[3] for further
of subsets,
separable
into the group of all such maps of
into
M, but this Borel automorphism
G
of
that associates
by the open sets ancL where
A function
domain for every Borel set in the range. morphism
G
that f
is a Borel transformation
is a Borel map from
of sets generated structure.
is a homomorphism
We say that
= g • m
compact group,
is a Borel set in the m ÷ g • m
"smoothly" with
details.
g.
a Borel autoThe reader
8
For any Borel set
E
of
the transform of
E
by
g • ~
g
using the formula
of
D
G-invariant r, mny
by
g.
If
M, g • E = {g • m l m E E} ~
(or simply invariant)
is a measure on
if
M, we define the transform
g • ~(E) = ~(g-I g • ~ = ~
. E).
for all
We say that
g E G.
~
is
Unfortunately,
interesting measures arising in practice fail to have this property, but pos-
sess instead the weaker property of quasi-invariance. iant if
~
and
concretely,
on
M
g • ~
are equivalent
~(E) = 0 = ~(g • E)
If G
is a Borel set called
G
A measure
~
in the sense of absolute continuity,
for all
is the additive group of the integers,
is equivalent to
~
then
~
If
or more
g E G. then specifying an action of
is the same thing as specifying the Borel automorphism
sponding to the group element one.
is quasi-invar-
h • D = ~
is invariant
h
or
M
for some measure or if
correh •
(or respectively quasi-invariant).
The
case of a single measure preserving automorphism of a finite measure space is the classical setting for ergodic of
G
consists
theory (see
[15]).
If
in giving a one-parameter group
subject to the joint measurability
condition.
G
h(t)
is the real line, an action of Borel automorphisms
of
M
This condition is readily verifiable
in cases of interest and indeed in general it is a condition that permits us to w o r k with actions of groups If
M
that are not discrete groups.
is a compact manifold and if
X
is a
C
vector field on
M, then
the usual existence theorems for ordinary differential equations provide us with a one-parameter such that
group
h(t)
(t,m) * h(t)(m)
of diffeomorphisms is a
of
M
such that
h(t)
map and hence certainly Borel.
C~
= X(h(t))
and
Such a flow
may or may not leave invariant a measure, but if for instance
X
type, then Liouville's
Since this subject
theorem provides
an invariant measure.
is of H a m i l t o n i a n
is discussed in Kostant's article in this volume we w i l l not go into more details here (see also
[4]).
Before proceeding in our general context of which we have seen several examples above, we must impose a regularity condition on more precisely, we shall assume that
M
M
of a technical nature;
is an analytic Borel space.
referred to [3] for further exposition concerning
this condition;
a condition that is satisfied in all reasonable examples. quasi-invariant measure on
M
for
G.
respect to the measure g • E = E
for all
In other words, complements. M = M 1U measure.
M2
~, or that
g ~ G,
equivalent
One says that ~
in any case it is
Suppose that
~
is a
(In fact this is not just a property of
but rather a property of the set of all measures of a quasi-invariant measure class.)
The reader is
G
to
~,
~, so we may speak
acts ergodically on
M, with
is an ergodic measure if whenever we have
and some Borel set
E, then
~(E) = 0
or
~(M - E) = 0.
the only invariant Borel sets under the action are null sets or their
It also says that the action is indecomposable where
M1
and
M2
are disjoint
in that we cannot write
invariant Borel sets of positive
A rather natural modification whenever
~(g • E A E )
~(M - E) = 0.
Here
that is, the points ~(g • E A E) = 0
= 0
for all
g • E A E
of this definition
g E G and some Borel set
denotes
the symmetric
is clearly more restrictive
a non-trivial
result of Mackey
Suppose now that
~
of a null set.
D(E) = 0
For a general
or
of the two sets,
A set with that
The second definition
and if
G
is countable
locally compact group it is
(see [32]) that the two conditions
is a finite invariant measure
are equivalent.
for an action of the
integers.
This as we have seen is specified
formation,
and if the action is ergodic one has the Birkhoff
For
that
and one would be asserting
than the first,
they can easily be seen to be equivalent.
in assuming
E, then
difference
in one but not the other of the two sets.
might be called almost-invariant,
any such set is a null set or the complement of ergodicity
consists
by a single measure
preserving
ergodic
trans-
theorem
[15].
f E LI(~) ,
i
n
lim~-~--~ n-~o for almost all line.)
x.
u(x)
(ui(x)) = [
fd~
J
(There is a similar
If we interpret
ics where
Ei=0f
this formula
statement
for ergodic actions
in its classical
is the evolution of a state
side is the time average of a function
f
x
context of statistical
is to hold for almost all initial
difficult
to see that the validity
The question of equality
then the left-hand
(a dynamical variable)
and the right-hand
statistical
mechanics,
problem of showing
and the ergodic
that certain actions
Not only for this reason, tions in ergodic of actions available
are ergodic
(see
but for many others,
was initiated
to ergodicity.
has a long history the question
conditions
We have remarked
by Gelfand and Fomin for geodesic
theory and unitary representations
before flows,
are closely
to the
one of the fundamental
ques-
for an action or a class
Our object here is to review a general method
section.
in
[24]).
for doing this by means of unitary representations
sults of the previous
ergodic
theorem just reduces
of these two
In fact it is not
is equivalent
and phase averages
theory is to supply sufficient
to be ergodic.
The equality
states of the system.
of such a formula
of time averages
mechan-
after unit time,
side is the phase average of the same dynamical variable. averages
of the real
one has
and to apply the re-
that the present
technique
although
the observation
connected
goes back to
that
Koopman. We shall assume now that we are dealing with actions space
M
erality of
G
with a finite invariant measure that
define
(~(g)f)(x)
is a unitary that
~(M) = i.
on the Hilbert
~
= f(g-i
operator,
L2~M,~). . x).
More precisely
unitary representation
one can detect ergodicity of invariant
for
It is easy to verify
and using joint measurability
is a continuous
non-existence
~, and we may assume without
One can then define an associated
space
of a group
of
f E L2, and
that for each
loss of gen-
g E G, we g E G, ~(g)
one can show
The key observation
of an action merely by looking at sets is equivalent~othe
on a
unitary representation
of the action, G.
G
non-existence
~.
is that
Heuristically,
the
of invarimnt measurable
10
functions, existence
and since the space has finite measure of square summable
invariant
this is equivalent
to the non-
functions.
Lemma 3.1
The action of g E G
implies
that
G
on
f
M
is ergodic
if and only if
is a constant almost everywhere,
~(g)f = f
for all
and hence a constant
in
L2(M,P). There is a somewhat namely weak mixing. G
on
M.
We notice that
that the product measure M
is weakly mixing
trivially
implies
stronger notion than ergodicity which is often useful,
For this one must define the Cartesian G
acts on
~ x ~
M x M
[15] if the action of
ergodicity,
g • (m,n) = (g • m,g
~
• n)
One says that the action of
G
on
M x M
and if it is satisfied,
action is weakly mixing and ergodic. the unitary representation
by
is invariant.
square of an action of
is ergodic.
on
This condition
every Cartesian power of the
Weak mixing can also be detected
(see [15] and
and G
by looking at
[39]).
Lemma 3.2
The action of sional subspace
V
G
of
on
M
L2(M,~)
is weakly mixing
if and only if any finite dimen-
invariant under
~
consists
of constant
functions
(and hence is one-dimensional). These lemmas serve to motivate
the definitions
Section 2 above since these definitions sional subspaces R
of representations
or the integers
mixing.
of measurable
sets
of a group.
Z, an action may possess
To motivate
Finally,
and
if
G
the yet stronger
this we note that the ergodic A
of properties
concern invariant vectors
E
and
WM
in
and finite dimenis the real line
property
theorem implies
of strong
that for any pair
B lim p(h(t)A N B) = ~(A)p(B) t~
in the sense of Cesaro limits. t ~ Z.
~
[15].
is the action defined
One can find a sufficient
for this to be the case;
projection valued measure group
h(t)
for
t ~ R
One says that the action is strongly mixing if the above limit exists
usual sense tion
Here
T) corresponding
P(E) = 0
if
E
P
to
condition
indeed by Stone's
~.
We shall say that null set.
theorem
P
in the
in terms of the representa[25], there exists a
on the Borel sets of the dual of
is a Lebesgue
or
G
is absolutely
(R
or the circle
continuous
if
See [15] for the following.
Lemma 3.3
If
G
is as above,
the subrepresentation
of
~
then an action of on the orthogonal
G
is strongly mixing provided
complement
of the constant
that
function
11
has an absolutely continuous projection valued measure. It is clear that strong mixing implies weak mixing and that Theorem 2 of Section 2 is exactly the sort of result that will enable us to establish that certain actions are strongly mixing. More specifically the theorems from Section 2 will be applied in the following manner.
Suppose that we have an action of a group
invariant measure, and suppose that on
M
that
preserving H
~
G
G
H
on
M
in the action of a larger group
G.
H
on
and
which acts
on
M.
WM
H
of
In
may be "embedded"
We shall assume that the larger group
be ergodic by some method, and pick a subgroup E
H
H
M
H
G
acts
also necessarily
Equivalently we can start with an action of
Properties
with a finite
determined by the fact
ergodically, and then ask if we can conclude from this that
ergodieally.
M
coincides with the given action of
other words, we are assuming that the given action of
acts ergodically.
on
is a larger group containing
such that the action of
is a subgroup of
H
G
on
G, and ask if
M H
known to also acts
from Section 2 are immediately relevant to
this situation.
Theorem 4
Suppose that acts ergodically on
H M
has property
E
(respectively
WM) in
with a finite invariant measure.
Then
G, and that H
G
is also ergodic
(respectively weakly mixing).
Proof.
We consider the representation
an invariant function for G, and by ergodicity of
H
then by property
G, f
is a constant.
same argument we can conclude weak mixing if
~
of
G
on
L2(M,~).
If
f
is
E, it is an invariant function for Hence H
H
acts ergodically; by the
has property
WM.
The following follows immediately using Theorem i.
Corollary. and if
H
If
G
is semi-simple and acts ergodically on
is totally non-compact in If
H
is a subgroup of
G
G, then
H
M
as above,
is ergodic and weakly mixing.
isomorphic to the real line (or the integers)
we can also obtain results concerning strong mixing using Theorem 2.
Theorem 5
Let
G
be semi-simple and let
H c G
phic to the real line (or the integers). then the action of
H
If
be totally non-compact, and isomorG
acts ergodically on
M
as above,
is strongly mixing.
Although the hypothesis of the above theorems, as far as
H
is concerned,
may seem rather special, their interest lies in the observation that this hypothesis
12
of embedability satisfied theory.
in a larger group
for some classical
G
flows,
is satisfied that seemingly
One way that an action of a group
is transitive; g E G
that is if
such that
identify
m
the coset space
hypotheses
on
and
g • m = n.
6
with
M, one can conclude
of the details
given concretely by iant measure means
that
is immediately
that
M
is
G/F
for a fixed
(See
that
G
With our
and that
[30] and
M
such a measure,
can be (see
m, one can
gF + g • m.
The action of
Our assumption
also possesses F
G/F.
is if it
M, there is an element
by means of the map
of this reasoning.)
on what
have no connection with group
that this is a Borel isomorphism
g • (hF) = ghF.
severe restrictions
Indeed it is
can be seen to be ergodic
F = {glg " m = m}
M
closed so that we may as well assume exposition
G
are given points of
If w e put
G/F
in many cases.
on
G/F
G-invar-
and this places A transitive
seen to be ergodic by the first of the two definitions
is
is then
has a finite
[7] for instance).
F
[3] for an
rather action
of ergodicity
above. One of the simplest bedable
in a transitive
= {(Zl,'.',Zn) , z i numbers
rationally
tion.
and let
independent
[15].
action of
Leon Green
act on
since
by
Let
be an n-torus.
t • (Zl,...,Zn)
however,
T n = Rn/z n
[i] has established
[13] observed
that geodesic
Tn We pick real
= (exp(itan)Zl,'--,
if and only if the
a.l
are
that this action can be embedded where
a beautiful
Zn
is a lattice.
the same for horocycle
certain higher dimensional
semi-simple
(see also
of constant
actions of the group
flow,
manifolds
groups,
generalization
the same ideas
flows on surfaces
in transitive
on the same idea and the following cerning
Tn
We observe,
Rn
which again uses exactly
are embedded naturally observed
R
Izil = i}
The
proof of the result quoted above is based in its essence on this observa-
nil manifolds Fomin
action is that of a rotation on a torus.
This action is well known to be ergodic
in a transitive classical
of an action of the real line which is em-
complex numbers with
al,,.. , a n
exp(itan)Zn).
examples
and Mautner [35].
Gelfand
negative
G = SL2(R);
The proofs
of ergodicity
immediately
curvature [43]
flow on
are all based
all such results
from Theorems
to and
Parasyuk
did the same for geodesic
general result subsumes
and follows
of this result [2]).
con-
4 and 5.
Theorem 6
Let G/F
G
be a semi-simple
Lie group and let
has a finite invariant measure,
(or the integers),
then the action is strongly mixing.
tries
(see
SLn(Z), [8]).
is ergodic
F
of subgroups
the subgroup
of
F
and weakly mixing.
satisfying
SLn(R)
be a subgroup
such that
be totally non-compact.
H
instance
G/F
H
natural action of
Many examples
on
and let
If
this condition
consisting
of matrices
H
Then the
is the line
are known;
for
with integral
en-
43
We shall close this brief discussion a duality author
theorem of sorts which was noticed
(see [39]).
F
operate on
independently
since we let
H\G = {Hglg ~ G}
for any pair of closed subgroups
by
F
H
act on
H
of
H
G/F.
y • Hg = Hgy -I.
and
theory with
by G. W. Mackey and the
Above we were dealing with two subgroups
there was an assumed asymmetry let
of one aspect of ergodic
and
F
of
G
but
We can just as well
The following
fact holds
G.
Lemma 3.4
The action of H\G
H
on
G/F
is ergodic if and only if the action of
on
is ergodic. In general a coset space
G/F
or
H\G
has no invariant measure,
less a finite one, but it always has a unique quasi-invariant and it is with respect
to this measure
As an application
of this,
group of matrices with integral sisting of matrices and
F
H
is totally non-compact
principle
says that
and is in fact
Rn
F
and let
is ergodic on
Since
H\G.
is measure
linear action.
theoretically
F = SLn(Z),
H denote the subgroup
(i,0,...,0).
n ~ 2.
and let
Then H
G/F
[28],
The space
H\G
the sub-
of
G
G/F,
the duality
is easy to identify and the action of
Since a single point is a Lebesgue
the same as
Rn
con-
has finite volume
is ergodic on
minus the origin with Lebesgue measure,
this space is the natural set, HIG
G = SLn(R )
entries,
if
class
class that the above lemma applies.
let
with first column
measure
much
G
on
null
and we have the following
result.
Theorem
7
The linear action of
F = SL (Z) on R n with Lebesgue measure is ergodic. n M o r e o v e g the same is true for any F such that SLn(R)/F has a finite invariant measure. This easily stated result does not appear to be amenable
to any direct
approach.
4.
In Section
2 we studied
ting a representation decomposes.
These results
questions
certain aspects
of a semi-simple
nection was discussed additional
MORE ON RESTRICTIONS
are of immediate
in Section concerning
Theorem 2 for instance
concerned
3.
of the general problem of restric-
Lie group to a subgroup interest
in ergodic
and examining
theory and the con-
In this final section we Want to discuss
the restriction the restriction
how it
of representations of representations
some
to subgroups. to one-parameter
subgroups,
and it is natural
of representations groups.
to raise similar questions
to more general subgroups
From the general version
know that any unitary the Borel subsets function
representation
(see [33]) this projection
We are interested
H, particularly
of Stone's
of the dual group
general
of
H
H.
Together with an appropriate
leads to a projection
that
P
is absolutely
if and only if G
and if
~
H
determines
H
of
on
lutely continuous.
It is natural
G.
one may immediately
corresponding
tion
G = KAN.
rately. groups,
satisfied when
that
Any abelian
G
subgroup
A, and a part from
H = A.
We will say P(E) = 0 if
results
of
~
that if
subgroup
of
G
then the projecto
H
is abso-
is not an issue since the
is a compact
subgroup
valued measure
concerning
for
P(E) = 0
of when other abelian sub-
of any group
for a vector sub-
eigenvalues
of any operator
algebra of the Lie algebra of
is simple,
and we write
H, a
the Iwasawa decomposi-
is in some vague sense made up of a part from N, and we shall consider
the three cases sepa-
We have already noted that the question posed is trivial of
T. Sherman
the answer is affirmative,
Theorem
H
about the projection
and we turn to subgroups
sider the case
if
P
in physics.
Let us now suppose
K, a part from
H.
vectors,
with the restriction
to an element of the enveloping
of some interest
on
one-parameter
with no G-invariant
on
multiplicity
Theorem 2 above says in particular
to raise the question
conclude
valued measure
to Haar measure
We note that non-compactness
above is trivially
If one knows such results
question
G
associated
groups have this property.
H [25], we
the representation.
to Haar measure)
is equivalent
is a totally non-compact
tion valued measure
group,
P
is a Haar null set. and if
is a representation
conclusion
(with respect
E c H, and that
E
is semi-simple
continuous
abelian sub-
class of this projection valued measure
and in particular we would like to compare it to Haar measure
any Haar null set
the restriction
theorem for abelian groups
valued measure
in the equivalence
concerning
A.
Since
A
for compact
sub-
is abelian, we may as well con-
[48] has observed
at least in a special
case that
and the same holds in general one can easily see.
8
If
G
is simple and if
G-invariant
vectors,
tion of
to
~
A
~
is a unitary representation
then the projection valued measure
is equivalent
to Haar measure on
of
G
with no
associated with the restric-
A.
Moreover, the multiplicity
is uniform. We now turn to the consideration at present 2n x 2n
is a counterexample.
matrices
preserving
of subgroups
We consider
a non-degenerate
of
N, and here all we have
the symplectic
group
SPn(R)
skew bilinear
form.
Since the funda-
mental group of this group is the integers,
there is a unique double
G
of
group
N
for this group contains
SPn(R).
(This is Weil's metaplectic a normal
[51].)
abelian subgroup
V
of real
covering group
It is easy to verify isomorphic
that
to the vector
45
space of real of
V
n x n
with a group
symmetric matrices. T
0
[51] has constructed a representation tion (see also Shale [46]).
V
with
V
of rank one. P
~
V = V
Thus, if
n x n
matrices
(that
and all diagonal entries are one). of
G
~
to
V
Weil
called the metaplectic representa-
quite easily.
by means of the bilinear form
In fact if one identi-
~(a,b) = Tr(ab), this projection
is concentrated on the set of positive definite matrices n > i, this is a Haar null set, and gives an example where
is singular with respect to Haar measure.
one-parameter subgroups of non noticed for
is the semi-direct product
One may compute the projection valued measure on
associated with the restriction of
valued measure on
N
isomorphic to all strictly triangular
is, all entries above the diagonal are
fies
In fact
SL2(R)
V
This example also shows that many
have one sided spectrum and hence that the phenome-
persists in higher dimensions.
There is another point worth noticing about this situation; let denote the kth tensor power of the representation
~
with itself.
~(k)
Then it is quite
easy to check that the projection valued measure associated with the restriction of ~(k)
to
V
is concentrated on the set of positive definite matrices of rank equal
to the minimum of
k
and
lutely continuous spectrum. representation of
G
n.
Thus
~(n)
is the first tensor power to have abso-
Since it is virtually obvious that any discrete series
has a projection valued measure which is absolutely continu-
ous, any connection between discrete series and the tensor powers of to the situation for
analogous
n = i, is likely to involve large tensor powers.
We notice that the condition above for abelian groups tion valued measure on the dual group lutely continuous,
~
H
H
that the projec-
associated to a representation be abso-
can be rephrased so as to make sense for any subgroup
H.
The
condition can be readily seen to be equivalent to the condition that the representation of
H
should be unitarily equivalent to a subrepresentation of the direct sum
of the regular representation with itself infinitely many times. for any
H, and we shall say following
This makes sense
[33], that a representation of
H
fying this condition is quasi-contained in the regular representation. happens,
satis(If as often
the regular representation is equivalent to the infinite direct sum of it-
self, the condition is simply that the given representation is a subrepresentation of the regular representation.) We do not have any general theorems concerning this situation, but there is one case of special interest. G = KAN, we let knows that subgroups of
B
B
If
G
be the centralizer of
is semi-simple with Iwasawa decomposition A
in
K, and define
B = MAN.
One
is a group [6], and in some sense it is one of the most important G.
representations and
M
The principal series representations of of
G
G
consist simply of the
induced by the finite dimensional representations
plays a key role in the structure of
is simply the subgroup of triangular matrices.
G.
When
G
is
of
B
SLn(R), then
B
[6],
~6
Our interest here will be the study of restrictions of
G
to
B
something
the regular addition
in the spirit indicated
about the structure representation
to its obvious
and for this we shall need to know
of the regular representation
of
B
is rather striking.
one-dimensional
ducible representations, simple
above,
all infinite
dimensional
is the discrete
sentation ducible
of
B
irreducible
G = SLn(R )
B
has in
~i' i = i, 2, 3, 4. ~
representations,
for
n ~ 3
as a continuous
dimensional
is the discrete
finitely often. forthcoming
[33], say
representation
of
B
A
is the ThUS
with in fact only a
or do we find the same phenomenon
On the other hand for
infinite
presentation
for
decomposes
summands.
G = SL2(R) ,
It is easy to see
G = SL2(C).
Is this an accident A calculation
When
exactly four other irre-
types entering into the decomposition.
that the same is true for
groups?
B.
often, ~ = ~(~i @ ~2 @ 73 @ ~4 )"
~. each taken infinitely l direct sum of irreducible
finite number of distinct
of
It is known that
representations,
calculation will show that the regular
direct sum of the
of representations
reveals
that the regular repre-
direct integral
G = SPn(R),
representations
of
for other semi-simple
and contains
B
4n
such that the regular re-
direct sum of these representations,
We shall now state a general
no irre-
one can find exactly
each taken in-
criterion which will appear in a
paper.
As usual we consider W = MI/M
is a finite group,
morphisms
of
the normalizer
MI
of
A
A, and hence also on the Lie algebra of
not contain the automorphism the Lie algebra of
A).
in
K, and note that
called the Weyl group, which acts as a group of auto-
a ÷ a -I
of
A
A.
The group
(or equivalently
If it does we shall say that
-i
W
the map
belongs
may or may Y ÷ -Y
to
of
W.
Theorem 9
If
G
is semi-simple,
either decomposes
as a discrete
then the regular
representation
direct sum of irreducible
(with a finite or countable number of inequivalent continuous
direct integral with no irreducible
occurs if and only if If irreducible "discrete
-i
belongs
representations
series",
square integrable
G
since
B
of
series,
of the r e g u l a r
B
which
and sufficient
are in "general posiare
Finally we note condition
that is, there are irreducible representation.
as
since they
case that the matrix entries
is not unimodular.
[21] has given a necessary
have a discrete
which are summands
representations
B
as a
number of
of the regular representation
as in the semi-simple on
B
B
W.
which are summands
functions
that Harish Chandra group
or it decomposes
but we prefer to call them generic representations
One may not conclude
of
The first possibility
-i E W, we may think of the finite or countably infinite
are in a sense those irreducible tion".
to
summands)
summands.
of its subgroup
representations
that the
representations
It is interesting
to note that
47
his condition hence if
G
implies
that
-i E W
has a discrete
(and in fact is almost equivalent
series
so does
to it) and
B.
Once we have this kind of control over at least some of the representations of
B, it is natural
tion
~
of
G
to
to raise the question of what the restriction B
looks like.
This technique
G = SL2, and Stein in his lectures with
B
replaced
is exceedingly
in this volume uses a similar
which has no G-invariant regular representation
series
of
quasi-contained
above question
symmetric
and
N
in general, ~
matrices.
representation
restriction
of
know is false.
V
of
~ B
of a simple group quasi-contained
G
in the
of
H.
a normal
covering
subgroup
If the restriction of
B, it follows
is quasi-contained
G
of
of
~
to
is our friend
from
the We
space of
is quasi-contained
in
above that its further
in the regular representation
Again we do not know what the best possible
H
to the
to the vector
B
by the comment
case follows
the answer
SPn(R) , n ~ 2.
isomorphic
con-
to any subgroup
The general
Unfortunately
and the counterexample
of the double
formula.
that any representation
G, has its restriction
representation
contains
the regular
to
from the observation
induced representations.
is negative
n × n
except
This is true for all the series of representations
in the regular
representation
B = MAN
for
[20] which are used to obtain the Plancherel
representation
known facts concerning
have
B.
this follows
tained in the regular
metaplectic
vectors has its restriction
by Harish-Chandra
For discrete
fruitful technique
by an even larger subgroup.
One may ask if it is true that any representation
constructed
of a representa-
of
theorems
V
which we
are in the gen-
eral case.
PART II.
5.
GROUP EXTENSIONS
STRUCTURE
AND GROUP COHOMOLOGY
OF LOCALLY
COMPACT GROUPS
In this second part we shall take up a rather representations, structure
and indeed here the major
of locally
compact groups.
sions comes from the phenomenon ever,
to treat these questions
problem,
different
considerations
The motivation
of ray or projective
will concern more the
for the study of group extenrepresentations
properly, we feel it is better
and then come back to the original
questions
aspect of group
of groups;
to first widen
using the general
how-
the
techniques
which we shall develop. We shall suppose and that
G
operates
More precisely,
p(g),
that A
G
and
A
g ~ G, the map
and further
are topological
is a topological
we are given a continuous
such that for fixed denote by
on
that
groups with
transformation
map of
G × A ÷ A written
a ÷ g • a is an automorphism p
is a homomorphism
of
A
abelian,
group of automorphisms.
G
of
(g,a) ÷ g • a A, which we
into the group of
18
automorphisms that
p(g)
of
A.
The hypothesis
is an automorphism
"smoothly"
with
g.
logical G-module
If
G
and
A
satisfy
A
is a G-module
space,
includes
G A
on
A
and if
A
If
~(g)az(g) -I 6 A
Va ~ A, then A.
~
operators
A
defines
is a G-module.
Finally,
as a G-module.
Such modules will be called of a given
first of all an exact sequence
A
is
unitary represen-
that of
(g • a) = ~(g)(a) G
on a Hilbert space such that
for each
is any topological
G
is a topo-
g 6 G
an
A, equipped with the strong operator
topology,
A group extension
if
varies
If for instance
on this Hilbert
fines
A
A
is a continuous
then one may verify
that
p(g)
not only
[37].
g • a = ~(g)a~(g) -I
It may be verified
assures
A, but that
is again a representation
is some group of unitary
automorphism
of
~
group
g • a
of examples.
and if
in the usual sense,
as a G-module.
of
the above, we say that
a wide variety
space with its norm topology,
tation of defines
of the topological
or simply that
This definition a Hilbert
of joint continuity
group
g • a = a
trivial topological
by a given topological
de-
G-modules.
G-module will be
of groups i I÷A÷E÷G+I
where
i
is an injection
and where
the kernel of
is continuous, that
~ is
A
into
E, and
is exactly
E, and that
and open.
This means
but as topological E/i(A)
and
G
first of these having the quotient
phism of
A
of
E
above
A
with
which depends
this automorphism a G-module.
uity.
i.
topology. G
A
and i(A)
only on
A.
g, and not on the choice of a ÷ g • a
for defining The reader
G-modules
is referred
One of the simplest sion of the circle the real line
T
group.
examples
Z
G
i
topology the
g E G
assump-
and if
g'
is an a u t o m o r g'.
We demand
by
of
that
A
as
A, then by the
G x A ÷ A
following
is one of the main moti-
the condition
of joint contin-
[38] for more details.
of a topological
by the integers
of
of the map
as we did by imposing
groups,
in the ~efinition
This observation
to [37] and
G,
and we assume
has the relative
We note that if
the joint continuity
onto
we impose an algebraic
We note that whenever we have an extension a G-module,
E
can be identified
a ÷ i-l(g'i(a)(g') -I)
be the given automorphism
becomes
i(A)
as topological
Finally,
on
of
We assume not only that
onto its range,
that
groups where
~(g') = g, then
from the axioms for a topological vations
is a surjection
may be identified
tion to take account of the action of is an element
~
the range of
but also that it is a homeomorphism
continuous
not only as groups, from
of ~
group extension
is the exten-
(viewed as a trivial module),
defined
by
R, namely 1 ÷ Z ÷ R ÷ T + 1 .
Another
example which is of more significance
follows.
Let
operators
on
viewed
H H
be a Hilbert
space,
and let
with the strong operator
as scalar multiples
of the identity
particularly U(H)
topology. operator
in quantum physics
is as
be the group of all unitary Then the circle group is a normal subgroup,
T and
19
U(H)/T = PU(H)
is called the projective
unitary group.
Then
i + r ÷ U(H) ÷ PU(H) + i is an extension of
PU(H)
set of one-dimensional on
P(H)
classic
by
r ÷ a • r
of
being trivial module.
Indeed if
H, then any
defines
a • r
= l(Ul,U2) l
We observe
a = tb
naturally
and if
G
chanics
say that we have a homomorphism
If
H
P(H)
that the projective
morphic
with
in this heuristic
assume that let
space associated
of
discussion).
then if
p
G
of
corresponding
Thus
A
trans-
the func-
to
a
U(H)/T = PU(H)
and is iso-
system, into
system,
the axioms of quantum me-
PU(H)
(except for those sym-
but this will not change anything
[5].
is the projection
is precisely what is
For the moment let us
from
U(H)
onto
PU(H), we
and then
a group extension
a subgroup
a.
by picking unit vectors
Such a homomorphism
I÷T+Gr becomes
under
with a quantum mechanical
or ray unitary representation
G c PU(H);
G r = p-l(G),
r i ~ P(H)
t ~ T.
metries which we would want to be anti-unitary, essential
is the
transformations.
is a symmetry group of this physical
known as a projective
r 6 P(H)
onto itself which preserve for
transformations
to a group of projective
is the Hilbert
P(H)
a collineation
[5]) says that except for anti-unitary
which is defined
agree if and only if
a 6 U(H)
is the transform of
(see [52] or
these are the only maps of
f(rl,r2)
u. 6 r.. I l b ~ U(H)
T, T
where
theorem of Wigner
formations, tion
by
subspaces
of
G
by
+G÷I
T.
Even when we do not want to identify
PU(H), we Shall see that we can still construct
G
as
a group extension
I÷T÷E+G+I where
E
has a homomorphism
The fact that projective of a suitable
into
U(H),
or in other words,
representations
group extension
can be viewed
a unitary representation.
as ordinary
is a well known and fundamental
representations
fact.
In these notes we want to present a brief outline of a systematic group extensions mately related
and more generally
to the initial
and to the references
G. W. Mackey
[29],
sions.
A large part of the contents
These results
be equivalent
G
to [37] and
in particular
[38] for
the pioneering
this point of view concerning
work of
group exten-
of these notes will be the subject of a forth-
and we will try to summarize
extend and generalize
One of the most important of a given group
We refer the reader
cited there,
[31] who originated
coming paper of the author, volved.
of a theory of group cohomology which is inti-
problem.
more details
theory of
problems
by a given topological
if there is a commutative
the major new points
those in [38] and is to classify G-module
A.
in-
[39].
the set of all extensions Two extensions
are said to
diagram
I÷A÷E+G÷I +
+
+
1 ÷ A ÷ E'÷ G ÷ 1 of continuous
maps where the end vertical maps are the identity maps and where the
20
middle vertical is an isomorphism of topological groups. is not sufficient to assume that
E
and
E'
It should be noted that it
are isomorphic as topological groups
to have equivalent extensions, but rather there must be a particular isomorphism which respects the data of a group extension. set of equivalence classes of extensions of
One of the first facts is that the G
by
A
forms a group
Ext(G,A)
by
means of the Baer product (ef. [16]), and at least in many cases this group is given as a two-dimensional cohomology group the other cohomology groups
H2(G,A).
It turns out to be useful to study
Hn(G,A), both to gain a better understanding of exten-
sions, and also to have at hand general methods of computing
Ext(G,A)
in many
specific situations.
6.
G-MODULES
After this introduction we shall now proceed to some of the details. will henceforth assume that
G
second axiom of countability.
We
is locally compact and separable in the sense of the (Local compactness seems to be essential for this
treatment, although we hope in the future to be able to dispense with it; separability is an assumption of a more technical nature used to avoid certain pathologies.)
We shall also assume that
able by some complete metric.
A
is separable, metrizable and moreover metriz-
Following Bourbaki, one might call such groups polon-
ais, and we denote the family of all such groups by
P.
Since we will always be
dealing with G-modules, we consider all polonais G-modules which we denote by We note that
P(G)
contains all separable locally compact G-modules
tensions were studied in the case of locally compact
A
A.
P(G).
Group ex-
in [38] and [39], and one
of the key points in the present treatment is that we now enlarge the category of modules to
P(G).
In addition to including many important and interesting examples
which were excluded before, we also achieve more technical versatility in that the larger category will contain cohomologically trivial modules, will enable us to define induced modules in a natural way, and will allow us to construct resolutions without going outside the category. If
A,B E P(G)
a G-homomorphism
f
of
A
into
intertwining homomorphism, that is, one satisfying that
P
(resp.
B
is simply a continuous
f(g • a) = g • f(a).
We note
P(G)) is closed under the operations of countable Cartesian prod-
ucts, closed subgroups (closed submodul~s), and quotient groups (quotient modules). In addition, if we have a sequence of elements of i i ÷ A p ÷A
P
~ A" ÷ l
which is exact in the sense of Section 5, then one can show that if
A"
and
A"
are in
P.
The same is clearly true for
phisms in the sequence above are G-equivariant. and it is only assumed that
i
and
~
P(G)
Morever, if
A E P
if and only
if the homomor-
A, AP, A"
are in
are continuous, it follows by classical
P,
21
closed graph theorems
(cf. [3]) that
i
is a homeomorphism and that
~
is open,
and hence that the sequence is exact in the sense of Section 5. In addition
P
is closed under the following construction which might be
described as a sort of direct integral. portance to us. bra of
(M,~)
[0,i]
Let
be a o-finite measure space such that the measure alge-
is separable
[14].
This means that we may as well assume that
with Lebesgue measure together with a countable number of atoms.
A E P, and define A
(M,u)
This construction will be of paramount im-
U(M,A)
U(M,A)
i (the identity in
M
to
A) almost everywhere.
is then an equivalence class of measurable functions, all of
which are equal to each other almost everywhere. f-l(0)
is
Now let
to be the group of all measurable functions from
modulo the group of functions equal to
An element of
M
is measurable in
M
for every open
0
(A function in
A.)
f
is measurable if
It is clear that
U(M,A)
is a group under pointwise multiplication. We topologize cisely let
Pl
U(M,A)
by the topology of convergence in measure; more pre-
be a bounded metric on
finite measure on
M
equivalent to
~.
A, which always exists, and let We define a metric on
~
U(M,A)
be a
by
P(f'g) = I Pl(f(x)'g(x))d~(x) which is always finite since
~
is finite and
Pl
is bounded.
Lemma 6.1
U(M,A)
with
p
Pl
and
of the choice of If namely let
A = T H
f E U(M,T).
be Then
U(M,T)
P, and the topology is independent
~.
is the circle group,
U(M,T)
has a natural interpretation;
L2(M) , the space of square integrable functions on f
defines a unitary operator on
(U(f)h)(x) = f(x)h(x). so
as defined above is in
Clearly
U(f) = U(g)
H
by multiplication by
if and only if
may be viewed as a group of unitary operators on
that the topology on ogy when we view
U(M,T)
U(M,T)
M, and let
f = g H.
in
F,
U(M,T)
and
It may be verified
introduced above is exactly the strong operator topol-
as operators.
If
A = R
is the real line, U(M,R)
is a
topological vector space; in fact a Frechet space, although it is not locally convex. Finally,
if
formula
(g • F)(m) = g • (F(m)).
copies of
A ~ P(G), we can define an action of
A.
If
M
G
on
U(M,A)
by means of the
This may be thought of as a direct integral of
is an atomic measure space, the construction does give the
Cartesian product of copies of
A.
In analogy with direct integrals of representa-
tions
[33] one might hope to find a reasonable definition of a measurable map of
into
P(G), and then define a direct integral where the "fiber"
is allowed to vary instead of remaining constant as above.
A(m)
over
m ~ M
Since we have found no
use for this kind of construction as yet, we shall not proceed any further.
M
22
The group
U(M,A)
the most important
for
for our purposes
and then intuitively
m2
into
has many interesting
A.
properties,
is a "law of exponents".
a function of two variables
as a function of one variable variable
A E P
(say
on
M
We let
into
A
M = M I × M2,
can be thought of
ml) into the space of functions
Such a correspondence
and one of
of the second
holds exactly and indeed follows
from
a version of the Fubini theorem.
Lemma 6.2
There is a canonical topological
A most important M = G
any
(separable)
will be denoted by
A E P.
I(A)
U(M I x M2,A )
onto
U(MI,U(M2,A))
A E P(G)
by the map
f
G
so that
defined by
of
I(A)
as
also operates
(f(a))(s)
on
is when
In this case
is itself a G-module
act by translation:
G
U(M,A)
group with Haar measure.
and we note that
In fact we simply let
If in addition I(A)
of
special case of the construction
is a locally compact
U(M,A)
isomorphism
groups.
(g • F)(s)
for
= F(g-ls).
A
we can embed
A
into
f
is an equivariant
= s -I • a.
Lemma 6.3
If phism of
A E P, then
A
I(A)
onto a closed
is in
submodule
P(G). of
Moreover,
I(A)
isomor-
so that
1 ÷ A ÷ I(A) ÷ U(A) ÷ i is exact where
U(A)
is the quotient module.
It is clear that with coefficients cohomologically
in
A.
trivial
I(A)
is in some sense the regular
In the case of a finite group, in that
Hn(G,A)
= 0
for
n ~ i
can also be proved in the present
context will be of vital
above would then assert
A
module,
that any
[45].
we have defined
importance.
If
have regarded functions, submodule
elements
of
as a submodule
= s-lf(g)
is understood~.
I(A)
of
I(A)
as functions
duced representations defer these details
and hence is in
of
G, and if
I(A).
for almost all pairs
We have engaged
but this poses no problem.
such as inducing to our forthcoming
is
6.3
trivial algebra.
Finally,
once
A 6 P(H), we
More precisely, (s,g)
in
instead of equivalence
All of the expected
I~(A)
where
classes of is
properties
in stages hold in our context, paper.
H × G
in the usual abuse of notation and
It is easy to show that P(G).
G
it is but a short step to the notion of
is a closed subgroup of
the induced module,
I~(A) = {fI f E I(A), f(gs) Haar measure
H
Lemma
from homological
is almost never locally compact.
the regular representation,
induced representations. I~(A),
I(A)
of
I(A)
The fact that this
in a cohomologically
and this fact will allow us to use many techniques
It should also be noted that
define
may be embedded
representation
it is known that
a
closed of in-
but we shall
23
7.
Having discussed a more explicit
GROUP EXTENSIONS
the G-modules
discussion
of group extensions
analogy with the case of discrete Hn(G,A),
n ~ 0
dimensions; g E G}.
A E P(G).
namely
For
morphisms
for
H0(G,A)
groups
of
G
into
A
and group eohomology.
[ii], we shall introduce
In complete
cohomology
These groups have simple interpretations
= A G, the G-fixed points
n = i, and a trivial G-module,
classes of continuous Ext(G,A),
which will enter into our theory, we turn to
in
HI(G,A)
groups
in low
A = {alg • a = a
for all
will be the continuous
homo-
(while for a general module we will have equivalence
crossed homomorphisms).
the group of topological By way of introduction
of how one may parameterize
For
n = 2, H2(G,A)
group extensions
to the cohomology
the group extensions
of
G
by
will be
A.
we shall begin with a discussion of
G
by
A
using cocycles.
Let
I+A÷E÷G÷I be a given extension. direct product
of
G
the direct product
The identity and
of
G
element of the group
Ext(G,A)
A, and in the special case when and
A.
This extension
A
is the semi-
is a trivial G-module,
is characterized
that one may find a continuous
homomorphism
f
(~of)(g)
the following
is to compare a general
by
A
= g.
The idea behind
to the semi-direct
such that
~of
homomorphism. groups,
product.
of
It is natural
is the identity map, and compute
G
by the property
it would not be sensible
the defect of groups,
f
of
G
such that
into
it is simply a fact of life that such a continuous
E
f
extension
of
G
f
to
E
of
G
from being a
but with topological
to choose any arbitrary map
map
E
to consider a map
Since we are dealing not with abstract
want to look for a continuous
back into
f.
satisfying
Ideally one would the above,
however,
map does not always exist.
Indeed
even in the case I+Z+R÷Z÷I such a map does not exist, a general base
G
extension would and fiber
topological
space.
A
and in general imply that
E
the existence viewed
sidered
= s
for
s E G,
(see [9]).
such as those continuous
in a neighborhood Borel map
f
G x G A
f
of
G
Other choices
into
E
element of
as
him we ob-
satisfying
of an appropriate
at the identity element of
of the identity
E = G × A
this, and following
G
G, but we believe
map may be con-
or those continuous that the choice of a
leads to a theory which is in general more satisfactory.
Once we have selected sometimes
fiber bundle with
would be a trivial bundle and so in particular
Mackey has shown how to resolve
serve that one may always find a Borel map (nof)(s)
of such a continuous map for
as a principal
called)
we note that
into the subgroup
and we notice
i(A)
such a Borel function a(g,h) of
E.
f
= f(g)f(h)f(gh) -I
(or cross section as it is is a Borel function from
We view it as a function
that it is a Borel function,
from
and as a consequence
G x G
into
of the associative
24
law in
E
satisfied the "cocycle identity", a(s,t)a(st,r)
for all G x G
s, t, r, G x G x G. into
A
= (s • a(t,r))a(st,r)
We denote by
Z2(G,A)
the group of all Borel maps of
satisfying this identity, and call such functions 2-cocycles.
The
group structure is understood to be multiplication of such functions pointwise. have associated now to each element of
Ext(G,A)
an element of
depends on the selection of a Borel cross section f
by any other Borel cross section
f
only by multiplication by a 2-cocycle of the form Borel function
b
of
G
into
note the group of such by extension in
Ext(G,A)
Z2(G,A)/B2(G,A) noted by
A.
of
f', the cocycle
into
for some
B2(G,A), and notice the very important fact that to each
we can associate a unique element of the quotient group This quotient group is deG
Ext(G,A)
into
H2(G,A)
may be verified to be a homomorphism If
A
is locally compact
[30] has shown that this map is surjective as well.
Ext(G,A)
A E P(G).
Hn(G,A).
A £ P(G), we define a complex of groups
is the set of all Borel functions from 6
We are able to show (see
This construction gives a param-
in terms of a cohomology group and also motivates the in-
troduction of the general cohomology groups
a coboundary operator
with coefficients in
A.
below) that this is also true for any
If
If we replace
(s • b(t))b(s)b(st) -I
of groups, and moreover may be seen to be injective.
eterization of
E.
changes, but it changes
H2(G,A), the two dimensional cohomology group of
The map of
Z2(G,A), but this
We call such functions 2-coboundaries, and de-
which is independent of any choices.
the topological G-module
Mackey
G
a
We
from
n
Gx.-.xG
Cn(G,A)
into
(n
Cn(G,A), n ~ 0, where factors) into
cn+I(G,A)
Cn(G,A)
A, and we define
by the classical formula
[ll], (~nf)(sl,-..,Sn+l)
= s I • f(s2,s3,..-,sn+l)
f(sls2,s3,...,sn+l)
-
... ± f(sl,''',SnSn+l )
¥ f(Sl,''',S n) where we are writing if
f
define and
A
additively.
The verification that
is a Borel function is routine Zn(G,A)
Hn(G,A)
H2(G,A)
to be the kernel of
~ n
to be the quotient group
and
Thus
B0 = 0
Borel}.
and
is a Borel function ~n+16n = 0.
to be the range of
We
~n-l,
n = 2, this gives the group
A, and
For 60
n = 0, a function of
is given by
~0(a)(s)
Z 0 = H0(G,A) = A G, the G-fixed points in
A.
For
Z 1 = HI(G,A) = {flf(st) = f(s) + f(t),
By a classical theorem of Banach, every such Borel homomorphism is auto-
matically continuous phisms of
For
as defined above, so everything is compatible.
n = i, and a trivial G-module, B 1 = 0, and f
Bn(G,A)
Zn/B n.
zero variables is by convention an element of = s • a - a.
~n f
[38], as is also the formula
G
into
homomorphisms of
(cf. A.
G
[3]), so
If
into
G A
HI(G,A)
acts on
is the group of continuous homomor-
A, B 1 # 0, and
or functions satisfying
Z1
consists of Borel crossed
f(st) = s • f(t) + f(s).
Such
25
a function classes
is by the same theorem of Banach continuous
of continuous
of a coset of
Elements tations
[31].
tive unitary
of
Let
of
= {f(s) = s • a - a Z2(G,T)
p
be a continuous
f
of
PU(H)
H
and hence an element
A
consists of
where a class consists
a E A}.
in the study Of unitary ray represenof
G
into
PU(H),
as defined previously.
back into
of
HI(G,A)
into
homomorphism
space
f(p(s))f(p(t))f(p(st)) -I = a(s,t)
Z2(G,T),
G
for some
arise naturally
group of a Hilbert
Borel cross section and then
crossed homomorphisms
BI(G,A)
so
U(H)
by general
theorems
H2(G,T).
It is clear that the element of unitary represnetation
H
of
"induces"
standing
p
[31].
Thus an analysis
of when a ray representation
Even if the element Mackey's
of
H2(G,T)
H2(G,T)
z
H2(G,T) of
G
on
is crucial for an under-
"is" in fact an honest unitary representation.
is non zero we can still construct
theorem a group extension
as above,
can be seen to define an element of
is zero if and only if we may find a continuous which
the projec-
We can find a
of
G
by
according
to
T
1 ÷ T ÷ E ÷ G ÷ i and one may verify
that
which is of the form ray representation
of
E
possesses
t ÷ t • 1 G.
on
COHOMOLOGY
of cohomology
First of all we selected a particular
perhaps
to give us what we wanted somewhat artificial
by the observation definition
We defined
that ray repre-
ad hoc.
(Borel functions)
and moreover
which
we selected a
6 . One's doubts are further compounded n of Section 6 suggest a somewhat different
to be all Borel functions
the possibility
from
of replacing
group of equivalence
classes modulo null functions
(Haar measure)
A, and we denote this group by
into
or
Hn(G,A).
Cn(G,A)
but one is led to consider
comment
is grantedly very
class of functions
of
H
of a group extension.
Hn(G,A)
in low dimensions,
definition
on
the given projective
GROUPS
groups
that the constructions
of the groups
our earlier
as ordinary representations
8.
happened
and which "induces"
This makes explicit
sentations may be interpreted
The introduction
an "honest" unitary representation T
Gx---xG = G n" into
Cn(G,A)
of measurable c__n(G,A).
to verify
that 8 as above is a well defined map from n that we get cohomology groups H n(G,A) = z_n(G,A)/B_n(G,A)
by
U(Gn,A),
functions
A,
the
from
Gn
It is not difficult
C n to -where
C n+1, and hence -Z_n is the kernel
of
6 and B n is the range of 6n_ I . The cocycles in dimension zero consist of n the kernel of 60 or the elements a of A such that s • a = a for almost all s
in
G.
and hence
It is not hard to see that this implies H0(G,A)
= A G.
If
A
sion one are exactly the equivalence that
f(st)
= f(s) + f(t)
that
is a trivial G-module, classes
s • a = a
of functions
for almost all pairs
s
and
for all s 6 G,
then the cocycles f t.
from
G
Similarly,
to
in dimenA
such
in dimension
26
two we look at functions which satisfy the cocycle identity above for almost all triples
(s,t,r).
A result of Mackey in [36] suggests that such an approach is not
as outlandish as it first appears. Motivated by the above, together with the possibility of a wide variety of other choices of cohomology groups we ask if we can somehow find a set of reasonable axioms which any cohomology theory should in principle satisfy, and then prove that there is up to isomorphism only one way of satisfying these axioms. that this is the case, and moreover that the groups cocycles do satisfy these axioms.
Hn
and
Hn
We shall show
defined above by
We then will know not only that these two defini-
tions of cohomology groups agree, but also that any other attempt to define cohomology groups satisfying the axioms below must necessarily lead to the same groups. (a) each
Our first axiom is of a general algebraic nature.
A ~ P(G), G
fixed, and for each
n ~ 0, an abelian group denoted by
such that these are "functors of cohomological type". that for any G-homomorphism of
Hn(G,A)
into
(gf)n = gnfn i
Hn(G,B)
when
g
f
of
A
We assume given for
into
Hn(G,A)
More precisely, we assume
B, we have induced homomorphisms
fn
such that the law of composition is s a ~ s f i e d :
is a G-homomorphism of
denotes the identity homomorphism of
A
B
into
into
C.
Moreove~
in = 1
where
A, and we assume that for any
short exact sequence i ÷ A" ÷ A ÷ A ~ ÷ i in
P(G), we have natural coboundary operators
~ : Hn(G,A ") ÷ Hn+I(G,A ~)
such that
n
the infinite long sequence 0 ÷ H0(G,A ") ÷ H0(G,A) ÷ H0(G,A ") ÷ HI(G,A ") ÷ ... ÷ Hn(G,A) ÷ Hn(G,A ") ÷ H n+I(G,A ~) ÷ H n+I(G,A) ÷ ... is exact (see [38] and [45]). (b)
The second axiom demands
(c)
The third axiom is a vanishing axiom which is motivated by the cohomol-
ogy of finite groups; namely we demand polonais group, where
I(A)
H0(G,A) = A G
Hn(G,I(A))
for any
= (0)
for
A ~ P(G).
n ~ i, and every
is the "regular representation" as defined in Section 6.
Axiom (c) is of course the really crucial one; it asserts that certain modules are cohomologically trivial and although there is a great deal of motivation for it from the cohomology of abstract groups, it does represent a definite choice. One could conceivably select some other class of modules and assume them to be cohomologically trivial, and this would lead to a unicity theorem for some possibly different cohomology theory.
Our defense here is that the groups defined by co-
chains above do satisfy this vanishing axiom, and that the groups
I(A)
do seem to
play a natural role in analysis and group representations. The following unicity theorem follows immediately from Lermna 6.3 and standard methods of homological algebra.
2?
Theorem i0
If A ~ P(G)
H~(G,A), i = i, 2
for a fixed
G
are two assignments of cohomology groups defined for
which satisfy Axioms (a), (b), and (c) above, there are
canonical isomorphisms of
H~(G,A)
onto
H~(G,A)
for all
n
and all
A.
I
One of our major results is that the groups
Hn(G,A)
and
H n(G,A)
defined
above by Borel cochains, and equivalence classes of measurable cochains do satisfy these axioms.
Theorem ii
The groups
Hn(G,A)
hence are isomorphic. in
and
Hn(G,A)
satisfy Axioms (a), (b), and (c) and
More precisely, the map which attaches to each Borel cochain
Cn(G,A), its equivalence class in
Cn(G,A),
induces this isomorphism on co-
homology. The verification of Axioms (a) and (b) is routine in both cases (see [38]); however, the verification of Axiom (c) is non-trivial.
In fact for
n = i, this
verification is for all intents and purposes equivalent to Mackey's general version of the Stone-von Neumann theorem in [26].
A close examination of Mackey's argument
in [26] reveals that what is essentially being proved is that tually one wants to replace
T
by a unitary group
this would lead us into non-abelian cohomology details however are the same as when
H
U(H)
HI(G,I(T)) = 0.
(Ac-
on a Hilbert space, and
(see [45]~
is one-dimensional
The essential analytic so that
U(H) = T.)
Theorem ii is proved first for
n = i, and then the general case is reduced to this
case by an induction argument.
The argument follows in spirit the argument for ab-
stract groups where in fact the result is trivial; however, there are non-trivial analytical complications concerning null sets in our case. In view of Theorem ii we shall henceforth use the notations Hn(G,A)
interchangeably;
Hn(G,A)
and
our choice of notation will serve to emphasize that we are
interested in a particular facet of these groups which may be evident from one of the definitions, but not the other.
We note in particular that such results as the
above are not approachable if one stays within the category of locally compact Gmodules, and that essential use is made of non-locally compact modules. We have remarked before that we have a natural notion of induced modules which gives us for each subgroup of
G.
A ~ P(H), a module
I~(A) E P(G)
where
H
is a closed
A very useful tool for finite groups is Shapiro's lemma [45] which
relates the cohomology of
A
with that of the induced module.
Theorem 12 There are canonical isomorphisms and all
n.
Hn(H,A ) _~ Hn(G,IH(A) )r_
for all
A ~ P(H)
28
The proof is obtained by noting that both sides of the above as functors on P(H)
satisfy Axioms
that for
(a), (b), and (c), and then one applies Theorem i0.
H = (e), this is simply the vanishing theorem.
We note
Also we note for
n = i,
that this theorem is essentially Mackey's imprimitivity theorem [27].
9.
ADDITIONAL PROPERTIES
We shall now discuss some additional properties of these cohomology groups, and in particular nail down the connection with group extensions. have already seen that
H0(G,A) = A G
tinuous crossed homomorphisms of a trivial G-module, into
A.
Ext(G,A) A
and that
G
into
A
n = i, HI(G,A)
For
n = 0, we
is the group of con-
modulo principal ones, and if
A
is
it is simply the group of all continuous homomorphisms of
G
In Section 7 we constructed an injective homomorphism of the group of equivalence classes of topological group extensions into
H2(G,A).
locally compact, Maekey has shown that this map is onto, but his argument
does not extend since it makes essential use of the Haar measure on
A.
For
[30]
We have an
alternate argument which works in general and which we outline below. If
a ~ H2(G,A), we embed
image of the class
a
in
A
H2(G,I(A))
into
I(A)
by Lemma 6.3 and let
under the map given in Axiom (a).
H2(G,I(A)) = 0, a p = 0, and so there is clearly an extension of sponding to
a', namely the semi-direct product
extension of
G
by
A
corresponding to
I(A) • G.
a ~ H2(G,A)
I(A) • G
G
by
be the
Since I(A)
corre-
We wish to construct an
and on general principles we
would expect this extension, if it exists, as a subgroup of we pick a cocycle in the class
ap
I(A) • G.
In fact if
a, we can immediately construct a subgroup
and then prove that it has all the required properties.
Ep
of
(This particular
construction is virtually forced on us, again by general principles.)
Thus
Ext(G,A) = N2(G,A). The higher cohomology groups have as yet no direct interpretation, however, we certainly do expect
H3(G,A)
to contain obstructions to the construction of non-
abelian extensions as in [3], Chapter IV. When the cohomology groups are constructed via equivalence classes of measurable cochains, another interesting and significant property emerges. since
cn(G,A)
is a polonais group, and since it may be readily checked that the
coboundary operators hence in
P.
Namely,
Thus
~n
are continuous,
H__n(G,A)
it follows that
is the quotient of a group in
z_n(G,A) P
is closed and
by a subgroup and,
hence when given the quotient topology, is itself a topological group.
a priori reason for
B_n(G,A)
There is no
to be closed, and it is an unpleasant fact of life
that it is not always closed so that
Hg(G,A )
may not even be Hausdorff.
The
closure of the identity element in such a group is a closed subgroup, and upon dividing by it, we obtain a Hausdorff group which in the case of
H_n(G,A)
is simply
29 Zn(G,A)
divided by the closure of
polonais, and
Hn(G,A)
Bn(G,A).
This quotient group will again be
will satisfy all the axioms of a polonais group except with
"metric" replaced by "pseudo-metric".
Thus
Hn(G,A)
is in a class of groups one
might reluctantly call pseudo-polonais. In any case, the fact that
Hn(G,A)
and hence
Hn(G,A)
more or less reasonable topology will be quite important for us. strengthen Axiom (a) above and prove that the groups
Hn(G,A)
have a natural and In fact we can
are functors of co-
homological type taking values in the category of topological groups. n = i, and if of
G
into
pact sets. topology. H2(G,A)
A A
is a trivial
G
module, HI(G,A)
has a natural Hausdorff topology, namely that of Convergence on com-
It may be verified that the topology on
HI(G,A)
coincides with this
In [39] a great deal of effort was devoted to constructing a topology for for various
G
and
A
by rather
ad hoc methods.
that this topology coincides with the one above on exists.
Moreover, if
being continuous homomorphisms
H2(G,A)
It is not hard to show whenever the former
Details of this will appear in our subsequent paper. One reason for seeking a topology on
Hn(G,A)
(aside from the esthetic one
of expecting a topological object when one starts with topological data) is so that one can hope to make sense out of the spectral sequence for the cohomology of a group extension (cf. [19]).
If
H
is a closed normal subgroup of
with finite groups leads us to hope for a spectral sequence
E p'q
G, the analogy converging to
r
H*(G,A)
with
= HP(G/H,Hq(H,A)) E P'q 2
gin to make sense, we must have topology.
(see [19]).
We observe that for this to be-
Hq(H,A) E P(G/H), and in particular it must have a
We can show that there is always a spectral sequence of this type, and
moreover that if
Hq(H,A)
the expected formula.
happens to be Hausdorff then the
E~ 'q
term is given by
The existence of such a spectral sequence is quite important
since it is an almost indispensable tool in making all but the simplest calculations of our cohomology groups. the case when
A
The reader is referred to [38] and [39] for examples in
is locally compact.
We shall close this section with one final result concerning direct integrals of G-modules.
Recall from Section 6 that if
A ~ P(G), the group
a natural structure as G-module which we called the direct integral.
U(X,A)
had
Since
Cartesian products are a special case of this, and since cohomology commutes with products, we may ask if the same is true for integrals and we have the following result.
Theorem 13
If
Hn(G,A)
is Hausdorff, we have an isomorphism of topological groups Hn(G,U(X,A))
= U(X,Hn(G,A))
.
The content of this result is that a cocycle with values in a direct integral module
U(X,A)
may be represented as a direct integral of cocycles.
If
n = 1
30 with trivial action, the side condition is satisfied and since one cocycles are homomorphisms, this result essentially gives us a new proof of the existence of direct integral decompositions of unitary representations.
i0.
EXAMPLES AND APPLICATIONS
We want to conclude with some examples, some computations, and some applications of the general theory above. Suppose that G-module.
G = G1 × G2
and suppose for simplicity that
A
is a trivial
Then either as a consequence of the spectral sequence above, or as a re-
sult of explicit computations (cf. [31]), we may obtain a structure theorem for H2(G,A)
as follows: H2(G,A) = H2(GI,A ) @ H2(G2,A) @ HI(GI,HI(G2,A))
.
The first two terms are easy enough to understand and represent the contributions of the factors
G1
and
G2
to the cohomology of
term representing the interaction of immediately compute
G1
G 2.
H2(Rn,T), H2(Zn,T), and
it is easy to verify that
H2(Tn,T)
lows readily by induction that
H2(Rn,T)
Since
n(n - i)/2, and that
H2(Zn,T)
dimension
n(n - 1)/2, and that
H2(Tn,T) = 0.
V
H__2(Rn,T) and
and torus
S.
by induction on
H2(Zn,T)
R n = Rn-I x R, and so on, it fol-
is isomorphic to a torus
H2(Rn,T)
slightly more involved statement holds for
contains a unique skew symmetric H2(Rn,T)
G.
n(n - 1)/2.
A
A similar but
is a trivial locally compact
H2(G,A) = HI(~I(G),A )
Furthermore
with the group of such
H2(G,A)
where
~I(G)
is the
is Hausdorff in its natural
topology and this topology coincides with the compact open topology on = Hom(~I(G),A)
of
H2(Zn,T).
is a semi-simple Lie group and if
G-module, it is classical [47] that
vergence.
V
again of
Moreover the topology defined above
functions which is a vector group of dimension exactly
usual fundamental group of
S
coincides with the usual topology on the
continuous bilinear function, and one may identify
G
Indeed
The isomorphism can also be implemented quite explicitly
since one may show that each class in
If
n.
by looking at the possi-
is isomorphic to a vector space
dimension
vector
This enables us for instance to
H2(R,T) = H2(Z,T) = H2(T,T) = 0
ble group extensions in these three cases.
on the groups
G, while the final term is a cross-
and
HI(~I(G),A)
which in this simple case is simply the topology of pointwise con-
This result also holds for any trivial G-module in
similar result holds for a much broader class of groups
G
P(G)
and moreover a
if one is willing to
suitably redefine and generalize the notion of the fundamental group
~I(G)
of
G
(see [40]). Using the spectral sequence of the previous section one may compute when
G
H2(G,T)
is a semi-direct product of a semi-simple group and say a vector group.
One
31
may verify
in this case known results
similar kinds of groups.
We refer the reader
Another application cerning non-locally
for the inhomogeneous
of this material,
compact G-modules
be a finite or infinite
to [38] and
dimensional
concerns
Lorentz
and
[39] for more details.
and especially the following
torus where
group,
of our results situation.
con-
Let
A = Tn
n = i, 2, -.., ~, and suppose
that
i ÷ A ÷ E ÷ G ÷ i is a group extension
of
G
by
A
where
E
is locally compact and abelian.
a trivial and well known consequence
of the duality
groups
(that is, represents
that such an extension
Ext(G,A))
and so
E = A @ G
With this result
splits
is a direct
sum of
in mind for Cartesian
to ask if a similar result holds for direct
A
It is
theory of locally compact abelian
and
products
integrals
the identity G
element
as topological
of circles,
of
groups.
it is natural
of the circle group and the
answer is affirmative.
Theorem 14
If
1 ÷ U(X,T) ÷ E ÷ G ÷ i
abelian and
G
locally compact,
as topological
is an extension of
then the extension
that
E
and so Theorem 13 is applicable. is assumed
case when useful
U(X,T)
following
to be abelian,
= T
in settling
topological
by
U(X,T)
is split so that
with
E
E = U(X,T) + G
groups.
The idea of the proof is quite simple; Hausdorff
G
where the result
Moreover
H2(G,T)
After some extra argument
using
is the fact
the problem is thrown back using Theorem 13 to
certain questions
groups.
one may verify that
is known.
concerning
Theorem 14 is found to be quite
the structure
exactly the same technique
of non-locally
compact
allows us to establish
the
result.
Theorem 15
If
H2(G,T)
locally compact
= (0)
and
then any extension
U(X,T)
a trivial G-module
i ÷ U(X,T) ÷ E ÷ G ÷ i
This final result leads to a very useful theorem concerning groups of von Neumann algebras which will have some applications theory.
Suppose
Hilbert phism
that
B
space and that f
of
G
u(g)
in
set down in [22]. B
such that
question we raise is whether a continuous
Neumann algebra of operators
is a locally compact
group.
Thus for each
f(g)(b)
unitary representation
field theory when for instance
B
of
G.
automorphism
in quantum field on a separable
of
B
satisfying
the con-
g E G, we have a unitary
= u(g)b u(g) -I
one can choose
G
We suppose given a homomor-
into the group of inner *-automorphisms
tinuity requirements operator
is a v o n G
with
splits.
the operators This question
for all u(g)
b ~ B.
is relevant
is some algebra of observables
The
so that they form
and
in quantum G
is some
32 symmetry group of the physical system.
If
B
is the algebra of all bounded opera-
tors on Hilbert space, a moment's reflection will show that we are raising exactly the question of when a projective or ray representation of
G
can be converted into
an ordinary representation since the group of *-inner automorphisms of
B
is
PU(H).
It follows from our general discussion of group extensions that we can do this for projective representations if of
G
H2(G,T) = 0, or equivalently if every group extension
by the circle group splits as a product.
The theorem to follow asserts that
the same is true in the general context described above.
Theorem 16
If of a v o n
f
is any homomorphism of
Neumann algebra
described in [22], and if of
G
with
~(g) 6 B
B
G
into the group of *-inner automorphisms
on a separable Hilbert space, continuous in the sense H2(G,T) = 0, then there is a unitary representation
such that
f(g)(b) = ~(g)bz(g) -I
The proof is almost immediate for the map cohomology class center of
a
in
B, such that
H2(G,W) a = 0
where
W
f
for
b 6 B.
immediately gives rise to a
is the group of unitary operators in the
if and only if a representation
~
as described in
the theorem exists.
However, by the structure theory of von Neumann algebras
of the form
and the result follows by Theorem 15.
When
U(X,T)
G = R, Kadison in [22] established a special case of this.
R. Kallman [23] has obtained a far more general result.
W
Recently
For the Poincar~ group,
another case of physical interest, L. Michel has already obtained the above result by rather different methods [36].
is
33
REFERENCES
[i] Auslander, L., et al.
"Flows on Homogeneous Spaces", Annals of Mathematics Studies, No. 53, Princeton (1963).
[2]
Auslander, L., and Green, L. (1966).
[3]
Auslander, L., and Moore, C. C. "Unitary Representations Groups", Mem. Am. Math. Soc., No. 62 (1966).
[4]
Avez, A. "Ergodic Theory of Dynamical Systems", Notes, University of Minnesota (1966).
[5]
Bargmann, V. "On Unitary Ray Representations of Continuous Groups", Ann. Math., 59, 1-46 (1954).
[6]
Bruhat, F. "Sur les Representations Math. France, 8-4, 97-205 (1956).
[7]
Borel, A. "Density Properties for Certain Subgroups of Semi-simple Groups Without Compact Components", Ann. Math., 72, 179-188 (1960).
[8]
Borel, A., and Harish-Chandra. Ann. Math., 75, 485-535 (1962).
[9]
Dixmier, J. "Dual et Quasi-dual d'une Algebre de Banach Involutiv", Trans. Am. Math. Soc., 104, 278-283 (1962).
"G-induced Flows", Am. J. Math., 8-8, 43-60 of Solvable Lie
Induites des Groupes de Lie", Bull. Soc.
"Arithmetic Subgroups of Algebraic Groups",
[10]
Dixmier, J. Les Paris (1964).
[11]
Eilenberg, S., and MacLane, S. Ann. Math., 48, 51-78 (1947).
[12]
Eilenberg, S. "Cohomology Theory in Abstract Groups, II", Ann. Math., 48, 326-341 (1947).
[13]
Gelfand, I., and Fomin, S. "Geodesic Flows on Manifolds of Constant Negative Curvature", Uspehi Mat. Nauk, Z, 118-137 (1952).
[14]
Halmos, P.
[15]
Halmos, P. Lectures on Ergodic Theory, Publications of the Mathematical Society of Japan, No. 3 (1956).
[16]
Hochschild, G. (1951).
"Group Extensions of Lie Groups I", Ann. Math., 54, 96-109
[17]
Hochschild,
The Structure of Lie Groups, Holden Day, San Francisco (1965).
[18]
Hochschild, G., and Mostow, G. D. Math., 6, 367-401 (1962).
[19]
Hochschild, G., and Serre, J. P. "Cohomology of Group Extensions", Trans. Am. Math. Soc., 74, 110-134 (1953).
[20]
Harish-Chandra. "Representations Acad. S c i . , 40, 1076-1077 (1954).
C*
Algebres et Leur Representations, Gauthier-Villars, "Cohomology Theory in Abstract Groups, I",
Measure Theory, Van Nostrand, New York (1950).
G.
"Cohomology of Lie Groups", Illinois J.
of Semi-simple Lie Groups V", Proc. Nat.
34
[21]
Harish-Chandra. "Discrete Series for Semi-simple Lie Groups", Acta Math., 113, 242-318 (1965).
[22]
Kadison, R. V. "Transformations of States in Operator Theory and Dynamics", Topology, ~, 177-198 (1965).
[23]
t" Kallman, R. " Spa_tally Induced Groups of Automorphisms of Certain von Neumann Algebras" (to appear).
[24]
Khinchin, A. Mathematical Foundations of Statistical Mechanics, Dover, New York (1949),
[25]
Loomis, L. An Introduction to Abstract Harmonic Analysis, Van Nostrand, New York (1953).
[26]
Mackey, G. W. 326 (1949).
"A Theorem of Stone and von Neumann", Duke Math. J., 16, 313-
[27]
Mackey, G. W. 593 (1951).
"On Induced Representations of Groups", Am. J. Math., 73, 576-
[28] Mackey, G. W.
"Induced Representations of Locally Compact Groups I", Ann. of
Math., 55, 101-139 (1952). [29]
Mackey, G. W.
Les Ensembles Boreliens et Les Extensions des Groupes", J.
Math. Pures Appl., 36, 171-178 (1957). [30]
Mackey, G. W. "Borel Structures in Groups and Their Duals", Trans. Am. Math. Soc., 85, 134-165 (1957).
[31]
Mackey, G . W . "Unitary Representations of Group Extensions", Acta Math., 99, 265-311 (1958).
[32]
Mackey, G. W.
"Point Realizations of Transformation Groups", Illinois J.
Math., k, 327-335 (1962). [33]
[34]
Mackey, G . W . "The Theory of Group Representations", mimeographed notes, University of Chicago (1955). Mackey, G. W.
"Infinite Dimensional Group Representations", Bull. Amer. Math.
Soc., 69, 628-686 (1963). [35]
Mautner, F. I. "Geodesic Flows on Symmetric Riemann Spaces", Ann. Math., 65, 416-431 (1957).
[36]
Michel, L. "Sur les Extensions Centrales du Groupe de Lorentz Inhomogene Connexe", Nucl. Phys., 57, 356-385 (1964).
[37]
Moore, C. C. "Extensions and Low Dimensional Cohomology Theory of Locally Compact Groups, I", Trans. Am. Math. Soc., 113, 40-63 (1964).
[38]
Moore, C. C. "Extensions and Low Dimensional Cohomology Theory of Locally Compact Groups, II", Trans. Am. Math. Soc., 313, 63-86 (1964).
[39]
Moore, C. C. "Ergodicity of Flows on Homogeneous Spaces", Am. J. Math., 88, 154-178 (1966).
[40]
Moore, C. C.
"Group Extensions of P-adic and Adelic Linear Groups", Inst.
Hautes ~tudes Sci. Publ. Math., (35), 5-74 (1968). [41]
Mostow, G. D.
"Cohomology of Topological Groups and Solvmanifolds", Ann.
Math., 73, 20-48 (1961).
35
[42]
O'Raifeartaigh,
L.
"Mass Differences and Lie Algebras of Finite Order", Phys.
Rev. Lett., 14, 575-577 (1965). [43]
Parasyuk, O.
"Horocycle Flows on Surfaces of Negative Curvature", Uspehi Mat.
Nauk, 8, 125-26 (1953). [44]
Segal, I.
"An Extension of a Theorem of L. O'Raifeartaigh",
J. Functional
Analysis, ~, 1-21 (1967). Cohomologie Galoisienne,
[45]
Serre, J. P.
Berlin, Springer (1964).
[46]
Shale, D. "Linear Isometries of Free Boson Fields", Trans. Am. Math. Soc., 103, 149-167 (1962).
[47]
Shapiro, A. 585 (1949).
[48]
Sherman, T. "A Weight Theory for Unitary Representations", 18, 159-168 (1966).
[49]
Seminaire "Sophus Lie", Paris (1954).
[50]
Weil, W. L'Integration dans les Groupes Topologiques et ses Applications, Hermann, Paris (1940).
[51]
Weil, A. "Sur Certains Groupes D'Operateurs Unitaires", Acta Math., iii, 143211 (1964).
[52]
Wigner, E.
"Group Extensions of Compact Lie Groups", Ann. Math., 50, 581-
Group Theory, Academic Press, New York (1959).
Canad. J. Math.,
APPLICATIONS
OF GROUP T H E O R Y TO Q U A N T U M P H Y S I C S ALGEBRAIC
ASPECTS
by Louis Michel#
T A B L E OF C O N T E N T S
0.
INTRODUCTION
i.
COVARIANCE
IN Q U A N T U M T H E O R Y AND ITS M A T H E M A T I C A L
i.i.
What
i. 2.
Group
i. 3.
G-Vector
Is Q u a n t u m M e c h a n i c s Invariance Spaces
i. 4.
Unitary
i. 5.
More Algebras
I. 6.
M o r e on
of
2.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Groups
SU(n)
47
and P e r m u t a t i o n
Groups
Operators.
S(n) . . . . . . . . .
Pseudo
and its T e n s o r O p e r a t o r s PHYSICS
Group Theory
2. 2.
The C o r r e s p o n d e n c e
2. 3.
Particle
2. 4.
The H y d r o g e n A t o m
2. 5.
The H e l i u m A t o m
of M a s s
P a u l i Principle. Atomic
Shell States
Principle m
. . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
and A t o m i c P h y s i c s
62
. . . . . . . . . . . . . . . . . .
62
. . . . . . . . . . . . . . . . . . .
62
in a S p h e r i c a l l y
Symmetric
Potential
.....
64
. . . . . . . . . . . . . . . . . . . . . . . . . .
71
The E l e c t r o n
Spin
Structure -Peri0dic
Atomic
Spin and E u c l i d e a n
2.10.
Molecules
2.11.
Measurement
. . . . . . . . . . . . . . . .
Table
. . . . . . . . . . . . . .
in a G i v e n S h e l l - Spin O r b i t or G a l i l e a n
Invariance
Coupling
........
. . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . of Spin and S t a t i s t i c s Spectra
3. i.
T h e Set of K n o w n N u c l e i
3. 2.
Isospin
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Scientifiques,
73 74 78 79 81
of N u c l e i by the Study
S T R O N G AND W E A K I N T E R A C T I O N S
Etudes
57
65
2. 8.
des Hautes
54
. . . . . . . . . . . . . . . . . . . . . . . . .
2. 9.
PHYSICS:
51
Roots
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. 7.
43
. . . . . . . . . . . . . . . . . . . . . . . . . . U(n)
2. 6.
43
45
2. I.
# Institut
. . . . . . . . . . . . . . . . . . . . .
and M o r e Tensor
SU(2)
A T O M I C AND M O L E C U L A R
NUCLEAR
. . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
of D i a t o m i c - M o l e c u l e 3.
TOOLS
38
91 - B U R E S - S U R - Y V E T T E
- France.
82 84 84 86
57
3. 3.
4.
. . . . . . . . . . . . . . . . . . . . . . . . .
89
3. 4.
Shell M o d e l . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
The Hadrons . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
3. 6.
The Other Particles
98
RELATIVISTIC
IN-VARIANCE.
and the O t h e r THE D I S C R E T E
4. i.
The P o i n c a r ~
Group
4. 2.
Relativistic
Invariance
SYMMETRIES
C, P, T . . . . . . . .
its A u t o m o r p h i s m s ; and I n t e r n a l
. . . . . . . . . .
Zeeman Theory
i01
.....
i01
Symmetries . . . . . . . . . .
102
Irrep
4. 4.
Particle
States
4. 5.
Particle
Polarization . . . . . . . . . . . . . . . . . . . . . . .
4. 6.
Invariance
4. 7.
How
to O b s e r v e V i o l a t i o n . . . . . . . . . . . . . . . . . . . . .
115
a.
Action
115
b.
Parity Violation
5.3.
105 and Irreps
Under
of
P, T, C
Time R e v e r s a l Violation
T0 . . . . . . . . . . . . . . . . .
on O b s e r v a b l e s
. . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . .
Invariance
, . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . .
SYMMETRIES
SU(3)
of
P × Z 2 ( C ) ; PCT T h e o r e m
c.
a.
2.
TO
CP
THE I N T E R N A L
5.
of
and
Interactions
4. 3.
5. i.
6.
Invariance
3. 5.
4. 8.
5.
u(4)
OF HADRONS . . . . . . . . . . . . . . . . . . . .
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . .
The O c t e t s
. . . . . . . . . . . . . . . . . . . . . . . . . .
108 ii0 113
116 118 119 121 121 121
b.
The M a s s O p e r a t o r . . . . . . . . . . . . . . . . . . . . . . .
122
C.
The First B a r y o n D e c u p l e t . . . . . . . . . . . . . . . . . . .
124
d.
Other
e.
Cross-Sections
Geometry
SU(3)
on the
Electromagnetic
Multiplets and Decays SU(3) - O c t e t
Electromagnetic
b.
Weak
Critical
5. 5.
SU(3)
5. 6.
SU(6),
125
. . . . . . . . . . . . . . . . . . .
125
Interaction
Interaction
Quarks,
in
SU(3) . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
Orbits of a G - I n v a r i a n t
× SU(3)
ACKNOWLEDGMENTS
124
of R e s o n a n c e s . . . . . . . . . . . .
and W e a k I n t e r a c t i o n s
a.
5. 4,
. . . . . . . . . . . . . . . . . . .
F u n c t i o n on a M a n i f o l d
M .....
128 128 129 131
Symmetry . . . . . . . . . . . . . . . . . . . . . .
134
Current Algebra,
138
. . . . . . . . .
Boot-Strap,
Etc . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
143
38
0.
INTRODUCTION
Since you mathematicians and we physicists came here to meet together, there is no need to emphasize that we both believe that the progress of physics requires for its theoretical formulation more and more advanced mathematics.
I
thought fit however to give you the opportunity to read what Dirac wrote on this subject, 38 years ago, as an introduction to the very paper where he predicted the existence of the "antielectron", which we now call positron. Notwithstanding Dirac's prediction, when positrons were observed one year later by Blackett and Occhialini, and by the Joliot-Curies, they were not immediately recognized.
And Anderson who was the first to identify a positron (in cosmic rays)
did not know Dirac's paper.
This illustrates the communication difficulties which
existed and still exist between theoretical and experimental physicists. also expect them between mathematicians and physicists. culties due to my use of English.)
You should
(Not to speak of the diffi-
You and I are here determined to overcome them,
but I beg your patience in advance. Extract from
Proc. Roy. Soc., Ser. A, 130, 60 (1930):
Quantised Singularities in the Electromagnetic Field By P. A. M. DIRAC, F. R. S., St. John's College, Cambridge § i.
Introduction
The steady progress of physics requires for its theoretical formulation a mathematics that gets continually more advanced. This is only natural and to be expected. What, however, was not expected by the scientific workers of the last century was the particular form that the line of advancement of the mathematics would take, namely, it was expected that the mathematics would get more and more complicated, but would rest on a permanent basis of axioms and definitions, while actually the modern physical developments have required a mathematics that continually shifts its foundations and gets more abstract. Non-euclidean geometry and non-commutative algebra, which were at one time considered to be purely fictions of the mind and pastimes for logical thinkers, have now been found to be very necessary for the description of general facts of the physical world. It seems likely that this process of increasing abstraction will continue in the future and that advance in physics is to be associated with a continual modification and generalization of the axioms at the base of the mathematics rather than with a logical development of any one mathematical scheme on a fixed foundation. There are at present fundamental problems in theoretical physics awaiting solution, e.g. , the relativistic formulation of quantum mechanics and the nature of atomic nuclei (to be followed by more difficult ones such as the problem of life), the solution of which problems will presumably require a more drastic revision of our fundamental concepts than any that have gone before. Quite likely these changes will be so great that it will be beyond the power of human intelligence to get the necessary new ideas by direct attempts to formulate the experimental data in mathematical terms. The theoretical worker in the future will therefore have to proceed in a more indirect way. The most powerful method of advance that can be suggested at present is to employ all the resources of pure mathematics in attempts to perfect and generalize the mathematical formalism that forms the existing basis of theoretical physics and after each success in this direction, to try to interpret the new mathematical
39 features in terms of physical entities (by a process like Eddington's Principle of Identification). A recent paper by the author# may possibly be regarded as a small step according to this general scheme of advance. The mathematical formalism at that time involved a serious difficulty through its prediction of negative kinetic energy values for an electron. It was proposed to get over this difficulty, making use of Pauli's Exclusion Principle which does not allow more than one electron in any state, by saying that in the physical world almost all the negative-energy states are already occupied, so that our ordinary electrons of positive energy cannot fall into them. The question then arises as to the physical interpretation of the negative-energy states, which on this view really exist. We should expect the uniformly filled distribution of negativeenergy states to be completely unobservable to us, but an unoccupied one of these states, being something exceptional, would make its presence felt as a kind of hole. It was shown that one of these holes would appear to us as a particle with a positive energy and a positive charge and it was suggested that this particle should be identified with a proton. Subsequent investigations, however, have shown that this particle necessarily has the same mass as an electrontt and also that, if it collides with an electron, the two will have a chance of annihilating one another much too great to be consistent with the known stability of matter.+tt It thus appears that we must abandon the identification of the holes with protons and must find some other interpretation for them. Following Oppenheimer,tttt we can assume that in the world as we know it, all, and not merely nearly all, of the negative-energy states for electrons are occupied. A hole, if there were one, would be a new kind of particle, unknown to experimental physics, having the same mass and opposite charge to an electron. We may call such a particle an antielectron. We should not expect to find any of them in nature, on account of their rapid rate of recombination with electrons, but if they could be produced experimentally in high vacuum they would be quite stable and amenable to observation. An encounter between two hard x-rays (of energy at least half a million volts) could lead to the creation simultaneously of an electron and antielectron, the probability of occurence of this process being of the same order of magnitude as that of the collision of the two y-rays on the assumption that they are spheres of the same size as classical electrons. This probability is negligible, however, with the intensities of y-rays at present available. The protons on the above view are quite unconnected with electrons. Presumably the protons will have their own negative-energy states, all of which normally are occupied, an unoccupied one appearing as an antiproton. Let me just remind you that antiprotons were first observed twenty-four years later (1955). There will be many advanced seminars on the applications of group theory to quantum physics.
So I believe that these lectures must be introductory~ and that I
have to present concepts that will be used by all physicists here.
That will be
Part i.
t Proc. Roy. Soc., Ser. A, 126, 360 (1930). #t H. Weyl, Gruppentheorie und Quantenmechanik,
2nd ed., p. 234 (1931).
tit I. Tamm, Z Physik, 62, 545 (1930); J. R. Oppenheimer, Phys. Rev., 35, 939 (1930); P. Dirac, Proc. Cconb. Philos. Soc., 26, 361 (1930). tttt J. R. Oppenheimer, Phys. Rev.,35,
562 (1930).
40
The ultimate goal of these lectures will be to bring you to the present problems on the subject, mainly in the field of the fundamental particle physics. Then there might be some overlap with Professor O'Raifeartaigh's
lectures, but there
should be no inconvenience to see some aspects of physics from probably two different points of view.
We have to face the f a c t t h a t
fruitful discussions of frontier
problems of physics between mathematicians and physicists are difficult, because these problems often cannot be presented in a formalized language, but only through some physical analogy.
So, obviously, to understand what is the problem, one must
know some physics! In these lectures I will therefore present a quick survey of applications of group theory to atomic, molecular and nuclear physics.
Often, I will even
follow an historical approach.
Indeed, physicist minds are partly conditioned by
the recent history of physics.
But I also hope to use the power of your language,
mathematics,
to convey to you a maximum of physics in a minimum of time.
Of course
I shall have succeeded only if I have also been able to convince you that physics is fascinating~ It is fit to end this introduction by the history of the birth of our subject.
Less than three years after the first paper on quantum mechanics
(W. Heisenherg, Z. Phys., 33, 879 (1925), there appeared the first two papers devoted to the application of group theory to quantum mechanics: •
E. P. Wigner, "Einige Folgerungen aus der Schrodingerschen Theorie fur die Termstrukturen",
•
Z. Phys., 43, 624 (1927).
F. Hund, "Symmetriecharaktere von Termen bei Systemen mit gleichen Partikeln in der Quantenmechanik",
Z. Phys., 43,
788 (1927). Wigner will surely be the most quoted author on our subject.
Let us just say that,
with J. von Neumann, he applied group theory to atomic spectra ("Zur Erkl~rung einiger Eigenshaften der Spektren aus der Quantenmechanik des Drehelektrons I., II., III., Z. Phys., 47, 203; 49, 73; 51, 844 (1928)), and published a self-contained book on this question:
E. P. Wigner, Gruppentheorie und ihre Anwendung auf die
Quantenmechanik der Atomspektren, Vieweg, Braunschweig (1931). It is remarkable that two famous mathematicians, Hermann Weyl and Van der Waerden, also published very early books on our subject: •
H. Weyl, Gruppentheorie und Quantenmechanik,
•
Van der Waerden, Die Gruppentheoretische Methode in der Quanten-
Hirzel, Leipzig
(1928).
meehanik, Springer, Berlin (1932). Then the excellent, but more elementary book, by E. Bauer, In~oduction &
la Th~orie de8 Groupes et 8es Applioation ~ la Physique Quantique, Presses Universitaires de France, Paris (1933), continued a list of books which, today, may have reached several dozen.
41
For the interested
mathematicians
first books, but in their second,
revised
I would still recon~nend the two very and enlarged
edition:
of Groups and Quantum Mechanics,
Methuen,
New York
by J. J. Griffin of Wigner's
(1949);
the translation
London
(1931); Paper Back reprint, book,
and Its Application to the Quantum Mechanics of Atomic Spectra, New York
Dover,
Group ~eory
Academic Press,
(1959).
For The Mathematician
Readers
Physics will be injected venient
Theory
H. Weyl, The
in these notes as needed.
to gather here some information
on physical
However,
it seems con-
constants which might be useful
at any time. We will study quantum phenomena. physics,
the key number
In atomic,
to pass from macroscopic
nuclear,
fundamental
particle
scale is the Avogadro number:
a = 6.0228 × 1023 ~ 6.03 × 1023 It is the number
of atoms in a mass of one gram of hydrogen.
A hydrogen atom is m a d e of one proton (mass
(0.i)
(mass
mp)
and one electron
ms). m
-J~ = 1836.5 m
These two particles
are electrically
+ p , ~ , the absolute value of this
charged,
charge is e =
The most
convenient
1 Faraday
96,600
=
6.03 × 1023
unit systems,
for us, will use
h = (elanck constant) c = (velocity In this system
and a nucleus which A = Z + N topes
number.
= i
contains
Z
protons
and
N
is nearly
A = 3, tritium isotope is
are made of
neutrons;
e.g., for hydrogen
Z = 92; the most abundant
charge and its mass
i 137.04
Atoms of the other elements
N = 0, A = i; A = 2, deuterium;
electric
i
e = (137.04) -1/2 , indeed
the atomic mass number,
For uranium
x (27)-1=
of light)
2 e hc
is a dimensionless
Coulombs
a
Z
Z
electrons
is the atomic number,
Z = i; hydrogen has 3 iso(unstable,
A = 238.
lifetime
12 years).
The neutron has no
equal to that of the proton,
42
m n
--= m So the atom mass is practically
1839.0
concentrated
in its nucleus.
Many more particles will be introduced, different
neutrinos
all with
We remind
the reader
mass is a form of energy. E = me2(l_(v/c)2)-I/2.
zero rest-mass
e.g.,
and zero charge,
that in relativistic
The energy of a particle
The rest energy
the photon
(v = 0)
X, and the four
etc..,
see 3.5 and 3.6.
physics mass is not conserved; of mass
is
m, velocity
We need to choose another unit to complete our unit systems. choice for atomic physics
is the electron mass
momentum
However,
the universal
length
2 m c g h/m c
time
him c 2 = i . 2 8
difference
is the electron volt charge
3.86
=
10 - 1 1 em x 10 - 2 1 s e c .
e
the most conxnon energy unit
It is the energy that a particle with
gains by traversing
an electric field of potential
one volt.
m c s m c P
I
x
(eV).
The conversion with the preceeding
Note that
The best
m . ° Then the other units are:
due to the nature of its measurement,
electric
is
m ge
energy
used for particles
v
mc 2 .
2 2
unit system is
= 0.511 × 106 eV ~'i/2 MeV = 938.256 MeV ~ 1 GeV = 109 eV
1 = ~ ~ (10 -13 cm) × (200 MeV) Before 1932,
electromagnetism. by proportional
only two kinds of interactions
In the static approximation potentials
K/r
where
of the (attractive)
gravitational
between
is independent
two protons
r
were known,
the two interactions
is the distance.
energy to the (repulsive)
gravitation
and
can be described
So the absolute ratio electrostatic
energy
of their distance.
It is Gem 2 P = I= 2 ~ e Thus, gravitation will be completely
137 x 10 - 3 6 175 neglected
in these lectures.+
t The gravitational energy of a system increases roughly as the square of the number N of nucleons while, in neutral matter, the electrostatic energy is roughly proportional to N. So gravitation becomes important only for masses as large as that of asteroids, planets (we know it on the earth!) or stars. It is not a coincidence that most stars have a number of nucleons ~ ~{-3/2 = IOS7 (see for instance E. E. Salpeter, "Dimensionless Ratio and Stellar Structure", in Frontier in Physics, Bethe Festschrift, p. 463, R. Marshak Editor).
43
The binding energy of atoms, molecules, solids, etc., is of electromagnetic origin.
This energy can be released in chemical form, with an order of
magnitude: a × i eV = 23 cal/mole which is ~ 10 -9 to i0 -II the rest-mass energy.
i
If the energy we receive from the
sun came from chemical reactions, the sun would produce it for less than 105 years! There are two other known kinds of interactions:
the nuclear interaction
(see part 3 and 5), stronger than the electromagnetic interaction at distance smaller than 10 -13 cm; the Fermi or "weak interaction" (see 3.6 and 5), which is very short range.
Both interactions are important in stars and nuclear reactions
and can yield an energy up to 10 -3 the rest mass energy.
For the Physicist Reader
All mathematical terms used here are not defined. are known to physicists Salam's lectures in
196~
Of course many of them
(e.g., for the notion of root vectors of Lie algebra, see
High Energy Physics and Elementary Particles, Trieste Seminar
(International Atomic Energy Agency, Vienna (1963)).
in I) come from a modern mathematical terminology.
Some terms (used mainly
They were not absolutely neces-
sary and they are used explicitly as synonyms of other terms generally used by physicists.
Physicists should know the proper mathematical terms of the mathematical
concepts they need:
indeed their students, and even their young children know them
and physicists want to communicate with their students and their children! An excellent and elementary exposition of this modern mathematical language is given in the text
Algebra
by S. Mac Lane and G. Birkhoff, Macmillan, New York,
(1967), particularly Chapter I; note also the list of symbols, p. XVII to XIX.
i.
COVARIANCE IN QUANTUM THEORY AND ITS MATHEMATICAL TOOLS
i.i.
What Is Quantum Mechanics
Less than two years after the first paper (quoted above) of Heisenberg on quantum mechanics, J. von Neumann answered this question in three successive papers in Gottingen Nachrichten, (1927), (pp. i, matische Grundlagen der Quanten Mechanik,
245, 273) expanded in a book: (1930)
Mathe-
(English translation, Princeton
University Press (1955)).#
He later published with G. Birkhoff, "The Logic of Quantum Mechanics", Ann. 37, 935 (1936). This subject is still controversial and lively.
Math.,
of
Two early books on quantum mechanics by physicists are reedited and still very advisable reading:
P. A. M. Dirac, The
Principles of Quantum Mechanics,
Clarendon Press, Oxford Ist ed. (1930) 4th ed. (1958). Quanten Theorie",
Handbuch der Physik,
W. Pauli, "Prinzipien der
i, Springer (1958) ist ed. (1933).t
If you have not read these books it is not too late to do it, but today let us just give a mlni-description of quantum mechanics. a)
To each physical system corresponds a separable complex Hilbert space
~C.
A physical state is represented by a vector
(Normed to 1 for convenience: b)
Each physical observable
<x,x> = I.)
a
(e.g., energy, electric charge, etc.)
is represented by a self adjolnt operator of c)
A
x> 6 ~C.
is the set of possible values of
A
on
~C.
The spectrum
a.
Quantum mechanics does not predict, in general, the value of for the state
a
x>, it gives only its expectation value: <x,Ax> = Tr A P
(i.I) X
where
P
is the Hermitian projector
(Px = Px *)
onto the one dimensional space
X
spanned by
x>.
Note that unit eigen vectors of
Px
(with eigen-value i) differ
only by a scalar phase factor and describe the same state since they yield the same physical predictions.
The projectors
P
are themselves observables.
Indeed
x
TrP P xy
= I<x,y>I 2
is the probability to observe in the state system which is known to be in the state
x> y>
(1.2)
(respectively (respectively
y>) the physical x>).
Part of the art
of the quantum physicist is to code what he sees in nature into vectors of Hilbert space!
This always requires "physical approximations". When we can describe a state by a rank one projector (or a vector up to a
phase) we say that we have a pure state and that we have a complete information on it. More often our information on the state is only partial. case we know only a set of probabilities
c.
in the set of orthogonal pure states
(i.e.,
Pi
(with
In the simplest
Ec. = i) for the system to be
PiPj = ~ijPj) so the expectation
value is E.c. Tr A P. = Tr A R i
i
(1.3)
1
with R = E.c.P. i
Since
I
Tr R = E.c. = 1
I
0 ~ ci, the self adjoint operator
i
R
i
is positive and it is called the
# There are also books on the mathematical foundation of quantum mechanics by mathematicians: G. Mackey, L. Schwartz.
(1.3')
45
density matrix# of the mixture (= not pure) state of the system.
Pure states are
extremal points of the convex domain of states. This leads us to a natural generalization. B
with unit I, generated by the observables.tt
C*-algebra). is
Then a state is a linear functional
VA E B, ~(A*A) ~ O.
One defines a Banach*-algebra
(More specifically it is usually a ~
on
B
which is positive, that
For systems with a finite number of degrees of freedom this
is not an essential generalization.
It becomes so for infinite degrees of freedom
as in quantum field theory and statistical mechanics.
Classical statistical
mechanics can also be put in the same mathematical mould with an abelian algebra.ttt
1.2
Group Invariance
We assume that there is a relativity group considered here.
That is
G
phism between the physics of observables by
K E G
acts on a physical system S
(e.g.,
~
for every physical theory S, and
(its Hilbert space of states
B c L(~)tttt, etc...)
view of G-invariance.
G
and the physics of
can be a rotation).
there is an isomor-
~, its algebra of
~(S), the transform of
S,
This will be called the "active" point of
The "passive" point of view for a transformation group is
simply the isomorphism between the physical description of the same system
S
by
two observers choosing different coordinate frames, G-transforms of each other. For any
g E G, we denote by
P
the transformed of the state gx i
say that
G
P
.
To
xi
is an invariance group is equivalent to saying that all probabilities
of Equation (1.2) are invariant x> E ~,
t
g E G, TrP
P = TrP P gx I gx 2 xI x2
(1.4)
J. von Neumann introduced the density-matrix in 1927 in the papers quoted above.
tt Quite early physicists also considered non-associative algebras formed by the observables and introduced Jordan algebras The first fundamental paper on those algebras is by P. Jordan, J. von Neumann and E. Wigner; "On an Algebraic Generalization of the Quantum Mechanical Formalism", Ann. of Math., 35, 29 (1934). it* I. E. Segal advocated twenty years ago the use of C*-algebra for quantum physics. The fundamental paper showing the benefits from this choice (physical approximation and Feld's s-equivalence; introduction of super-selection rules) is that of R. Haag and D. Kastler, "An Algebraic Approach to Quantum Field Theory", J. Math. Phys., Supplement 848 (1964). Most of the C*-algebra physics is published in the journal: Communications in Mathematical Physics, and is written in a rigorous mathematical style. For statistical mechanics, see D. Ruelle, Statistical Mechanics, Benjamin, New York (1969). Soon there will appear in the collection of C. N. R. S. Colloquia (France) "Rigorous Results on Interacting Systems with Infinite Degrees of Freedom". tttt We denote by
i(~)
the space of linear operators on
~.
46
or
1
12= I<Xl,X2> [2 This means that
G
acts on
(1.4')
~C by isometrics.
Wigner proved in his book (Appendix to Chapter 20) # that either a unitary operator
U(g)
that an antiunitary operator
or an antiunitary operator
V
V(g)
x> ~+ gx> on
~C.
has the characteristic properties
(1.5)
x>, y> E ~, V(~x> + ~y>) = EVx> + ~Vy>
(1.51 )
= <x,y> =
(1.5")
it has an inverse Given an isometry on
~, there is a simple criterion ## for deciding whether it is
realized by a unitary operator U
or
U
or an antiunitamy operator
is defined up to a scalar phase factor.
V
is
We recall
V.
In either case
The product of two antiunitary
operators is a unitary operator. Let U00
V(~)
be the group of unitary and antiunitary operators on
the subgroup of unitary operators.
since it is a subgroup of index two. i.e., no other element than for
g 6 G
group
UI
i 6 G
generate a subgroup
U~)
is an invariant subgroup of
We assume that acts trivially on
E(G)
of
$C and
G ~.
The
U(g)'s
or
V($C) which is an extension of
(phase multiplication of the vectors of
V(~)
acts effectively on
~C, V(g)'s
G
by the
X, leaving invariant the states)
with the action f G --+Aut U I where
Ker f
is the invariant subgroup of index two
transformations and the non-trivial element of -i ~--+ ~ = ~ E U(1). We can also say that
G+
Im f
G+ c G
which acts by unitary
is the complex conjugation
acts by a linear unitary projective representa-
tion and Wigner has coined the word projective "corepresentation" for the action of G
(when
G
is strictly larger than
G+).
Wigner also showed from physical arguments that antiunitary operators are to be used with transformations which reverse the direction of time, this in order that energy be positive: indeed, the time translation t is represented by the iHt iHt e ; if t ÷ -t, i has to go to -i in order that both H and e
operator
be invariant.
#
A more explicit proof of Wigner's theorem has been given by V. Bargmann, J. Math. Uhlhorn, Neumann axiomatics, the equivalent theorem has been proven by G. Emch and C. Piron, J. Math. Phys., i, 469 (1963).
Phys., 5, 862 (1964). See also proofs of slight generalizations by U. Ark~v for Fys~k, 23, 307 (1963). In the framework of Birkhoff and von
## See Bargmann:# for any triplet of vectors x>, y>, z>, <x,y> is invariant under a unitary transformation U and is transformed into its complex conjugate under the antiunitary transformation V.
47
Continuous projective linear unitary representation of finite groups or Lie groups are well known. SO(3,R)
For instance, for the three-dimensional rotation group
these projective representations are in a one to one correspondence with
the "linear irreducible unitary representations" (= irrep lectures) of
SU(2)
the universal covering of
S0(3,R).
through all these This justifies the intro-
duction of spinors in quantum physics. In Part 4 we will study invariance under the relativity groups of nonrelativistic (= Newtonian) mechanics and of special relativity theory. are other invariance groups in physics. acting on
n
But there
For instance the permutation group
identical particles (as the electrons of an atom).
S(n)
In nuclear physics
and fundamental particle physics we shall meet many "approximate invariances". corresponding invariance group is most often a unitary
n × n
matrices, with determinant
1
U(n) for
or
SU(n)
The
group (group of
SU(n)) with
n = i, 2, 3, 4, 6.
We shall have more to say for the word "approximate" symmetry. We will also have to study invariance under a group
G
when
G
is a
symmetry group for a physical system, e.g., the symmetry group (one of the crystallographic group) of a crystal. invariance in physics.
This example raises a fascinating question about group
Surely the interaction between atoms are translation in-
variant (and may be invariant under a larger transformation group).
How is it
possible that atoms aggregate to form a crystal whose lattice is invariant only under a subgroup of a translation group?
When such a phenomenon occurs, i.e., when
a stable state has a lesser symmetry than that of the physical laws we will say that we have a broken symmetry.+ We will continue this Part i by introducing some mathematical tools that we will use quite frequently.
1.3
Let
G
G-Vector Spaces
be a given group.
If you like you can say that we consider a
category whose objects are vector spaces action of
G
on
endomorphisms of
E
(i.e.,
E, with
G D x ~
E
(over a given field
g(x) E L(E), where
L(E)
K) with a linear is the algebra of
xy ~-+ g(x)g(y) = g(xy).
The morphisms of the category are the vector space homomorphisms
E
f~ E ~
compatible with the group action, i.e., they are the commutative diagrams for every x E G, of vector space homomorphisms.
We will call these morphisms G-homomorphisms
of G-vector spaces.
+ This short section on group invariance is too sketchy. Much more should be said of the symmetry of physical laws (e.g., E. P. Wigner, "Symmetry and Conversation Laws", ~oc. Nat. Ac~. Sci., U.S.A.j 51, 956 (1964)) without which symmetries of states, that we have considered, would not last. Of course much more will be said in these lectures.
~8
E
~ (x)
~ E
E'
g' (x)
~ E'
L
Diagram i. Of course we could have also said that we study bimodules
(G- and K-
modules) or even more simply that we are interested in the linear representations of G; and G-homomorphisms morphisms from
E1
are also called "intertwining" operators.
to
E2
form a vector space that we denote
Indeed, it is the subspace of the invariant vectors of intertwining operators for the two representations of Given G-vector spaces,
Note that the GHom(El, E2)G.
Horn(El, E2); they are the G
on
E1
and
E 2.
El, E 2 ... all vector spaces which can be formed
functorially from them are also G-vector spaces, e.g.,
E 1 0 E2, Hom(El, E2) , L(E)
= Hom(E, E), the vector space of the tensor algebra on
E:
E (n) = E 8 E 8 ... 8 E, n
factors,
(and
Given a physical system, let Assume that
3{ is a G-vector space.
T(E) = n@-~--OE(n)
when
E (0) = K), etc. ~C be the Hilbert space of its state vectors.
So is
LOC).
We are then led to the study of
the objects (of the category of G-vector spaces) "above" the physical literature "Tensor operators on
$C".
LOC).
They are called in
(A notable exception is the book
devoted to this subject, Irreduoible Tensoz~al Sets, by V. Fano and G. Racah,
By definition, for physicists, an "E.-tensor
Academic Press, New York (1959)).
l
operator" is a G-morphism (or intertwining operator)from
E.i to
LOC).
If the
representation of
G on E.i is irreducible, then the corresponding G-morphism is called in physics an "irreducible tensor operator". If G acts trivially on E.i then we have "scalar tensor operator". (Just try to remember that tensor operators
on
~C are not operators on
~C!)t
It is time to specify the field of complex numbers since essential in physics.
5£
So often
tor" is a G-homomorphism
T
of self-adjoint operators on the field from IR When
G
to
K.
Generally, of course, it is the field
is a complex Hilbert space. E
is a real vector space and the "E-tensor opera-
of real vector spaces from ~.
However, reality is also
E
to the real vector space
Of course it is always possible later to enlarge
~.
is a Lie group we consider, of course, only continuous differ-
entiable representations so a G-vector space is also a ~-module for the Lie algebra ~_ of L~) ~l on
G.
We denote
G
the vector space of
there is a particular one 5£.
F'
When the representation of
~.
Among the G-tensor operators on
which is also the Lie algebra representation of G
on
~C
is unitary, then
adjoint operators for images which satisfy
# "Scalar" is often used by physicists for "invariant"!
F = iF s
has self-
~9 [F(a),F(b)] When translations,
G
(F(a)F(b) - F(b)F(a)) = iF(a A b)
(1.6)
is respectively the group of rotations, space translations,
etc., F
momentum, energy,
=
...
corresponds respectively to the observables;
time-
angular-momentum,
In the technical sense of i.i, what we called observables
there, are the elements of the image of
F, i.e., for instance, the component of the
angular momentum or of the momentum, in a given direction. clear that G-morphisms on
L~)
But I hope it is by now
are what correspond to the physical concepts with
a tensorial character with respect to a group
G
(other examples:
velocity, mag-
netic moment, electric quadrupole moment, energy-momentum tensor, tensor of inertia, etc. ). Let E
and on
R
3£.
and
U
(unitary) be the representations
By definition of the E-tensor operator Vx 6 E, Vg 6 G,
If
D
and
F' = iF
U(g)T(x)u-l(g)
of
respectively on
= T(R(g)x)
are the corresponding representations D(a) = ~d
G
T
R(e ~a) ~-0 ; iF = ~d
(1.7)
of the Lie algebra
U(e ~a)
-
~=0
then an equivalent definition of the E-tensor operator
T
(1.8)
is
Vx 6 E, Va 6 ~, [F(a),T(x)] = iT(D(a)x)
(1.9)
In a nutshell, I would say that much of the application of group theory to quantum physics consist in the study of the "tensor-operators" (Hilbert) space T1
and
T2 E1
~
of a physical system.
be respectively
E1
and E2-tensor operators on
E 2 9 x @ y q--+ Tl(X) + T2(Y )
E1 0 E2 ~ x 0 y ~
Tl(X)T2(Y)
that we denote respectively
TI ~ T 2
defines a
defines and
on the G-vector
They form a ring~ (and an algebra).
a
Let
~, then
E 1 • E2-tensor operator E1 0 E 2 - t e n s o r o p e r a t o r
T l O T 2.
The latter is generally re-
ducible and can be decomposed into a direct sum of irreducible "tensor operators". I believe that many problems arise which have not been systematically studied by physicists although they work very much with this ring (for fixed and action of
G
on
G,
~).
For instance, if Yx, y 6 G, [T(x),T(y)] = 0.
G
is simple, and
T
is a G-tensor operator and
I believe this implies
dim Hom(G,~) G is infinite.%+
Of course the subalgebra generated by an element is well known; given an A T there is a functorial G-morphism T from the tensor algebra
E-tensor operator
For infinite dimensional ~, the operators T(x) are not bounded so their product is not always well defined. I will forget here this difficulty which has to be faced in quantum mechanics and is considered in O'Raifeartaigh's lectures. %% C. Moore proved it during the Rencontres.
5O
T(E)
on
E
LO0,
to
onical injection of
E
which is moreover an algebra homomorphism. into
If
i
is the can-
T(E)(Im i = E(1)), then the Diagram 2 is commutative.
T-(E)
E
~
~ L('~)
Diagram 2. in the-particular case where i, (see Equation
U(G),
(1.6)) of
~
on
T
is the representation
F
(up to a factor
~, then it appears also in the representation of
the universal enveloping algebra of
T(G)
s
G
> U(G)
~ L(SC)
F
Diagram 3. A F
(in diagram 2) = F • s
A remarkable "scalar tensor operator" is the Casimir# operator. Let
G
be a semi-simple Lie group.
resentation of the Lie algebra
~
D(a)b = a A b,
Let
a ~
on its vector space
D(a)
the adjoint rep-
G
[D(a),D(D)] = D(a A b)
(i.i0)
The symmetrical bilinear Caftan-Killing form
(1.11)
B(a,b) = TrD(a)D(D) is non-degenerate. dual
Gt .
Therefore, it defines a G-isomorphism
This also defines the isomorphism G O G <
iI
O I
GI O G ~
The well known canonical homomorphism operator i on
G
is an invariant
j
.J
i
between
G
and its
i e @ I, (I = identity) ~
Hom(G,G)
is also a G-homomorphism.
G-vector E Hom (G,G) G.
The identity
So
c = (iS @ I) • j(1) is an invariant vector of the Casimir operator on
E @ E c T(E)
with a fixed normalisation and
~.
# Casimir is a physicist with a high position in Philips.
(1.12) A F(e)
is
57 It occurs that neither physicists, nor some mathematicians
(cf., Bourbaki,
Groupes et Alg~bres de Lie, Chapter I §3 No. 7) use this canonical normalization for c.
In the physics literature nowadays,
independant elements of the center of In order to induce physicists
the images by U(~)
F
of a set of algebraically
are called "the Casimir operators".
to use the more canonical point of view ex-
posed here, let us end this section by a very simple theorem proven elaborately in particular
cases in the physics literature.
Theorem If G
G
has no non-trivial one-dimensional representation,
a non-invariant irreducible E-tensor operator on
then
Va E E, tr T(a) = 0.
Indeed, the field (~
dimensional G-vector space, and then
1.4.
for instance)
"trace" E Hom(LOC),~) G
"trace T" = "trace" • T E Hom(E,~) G = 0 Unitary Groups
U(n)
is for
T E Hom(~,L(~)) G,
by our hypothesis.
We have to survey briefly some results on irreps of
integer partitions of
T
is a trivial one-
since
and Permutation Groups
we shall use very much in these lectures.
and if
•, a finite dimensional space,
The irreps of
S(n)#
U(n)
S(n)
and
S(n)
that
can be labeled by
n
~i ~" h~k] [h I ... hi.1 ...
with
kl > h2 "'" hk > 0
and k Zi=leih i = n There is a more picturesque notation of with
n
small squares,
~i
lines of
°k ... hk ]
[h I hI
squares,
which is an ideogram made
~2
lines of
h2
squares, etc.
and called a Young diagram. Example of
VT-[
%
[hil]
%
h I = 9, ~i = i, h 2 = 5, ~2 = 3
•%
• ". <---- dia~ onal
h 3 = 3, a 3 = 2, h 4 = i, a 4 = i n = 9 + (3 x 5) + (2 x 3) + i = 31
L n=
The Young diagrams contain a qualitative information, (vertical)
the diagram the more symmetrical
(antisymmetrical)
31
the more horizontal
are the vectors of the
representations.
t For an exposition of the representations of U(n) and S(n) see Weyl's book, Chapter V. A survey for the needs of physicists has been made by C. Itzykson and M. Nauenberg, Rev. Mod. Phys., 38, 95 (1966).
52
There are only two one-dimensional irreps of [n]
. . . . . .
S(n)
completely symmetrical
: 221"~
n squares
R II [in ] =
",J
n squares
completely antisymmetrical.
Y The representations of
S(n)
are self contragredient.
To each irrep of
S(n)
we
can associate a complementary representation A C~.
with
~i
.k+l-i A = Lj=I ~j' ai = ~k-i+l - ~k-i+2
(1.13)
Its Young diagram is simply obtained by a symmetry through the diagonal. We recall that the tensor product of two irrep contains
[n]
(resp.,
[in]) only if the two irreps are equivalent (resp. complementary) and then it contains
[n]
(resp., [in]) once only. We will also use a shorter symbol
tion of
[ ]%
for a linear unitary representa-
S(n). We call factorial a group representation which is direct sum of equivalent
irreps. Let
~C(I)
be a Hilbert space and
3C(n) = ~ X (1) = X (1) O ... O X (1) 8 ~ (1) By permutation of the factors, tion that we denote representation.
acts linearly on
factors).
~(n)
(1.14)
through a representa-
[ ]3c(n) and which can be decomposed canonically into factorial
We denote
ial representation
S(n)
(n
w(n) ~[ ]~
the subspace of
3C(n)
on which acts the factor-
• [ ]~.
w(n) also denoted ~ ~(i~ ~(n) , also denoted ~ ~(i) , and ~[n]' [ in] are the spaces of completely antisymmetrical and symmetrical rank-n tensors on For instance
~(l).
~(l)
=
and on on
~(n)
Let us assume that dim ~(i) k is finite. Then U(k) acts on n ~(n) by @ U(k). The decomposition of this linear representation of into factorial representations yields the same subspaces
therefore denote by the same symbols
[ IX
-w(n) --[ ]~
the corresponding irreps of
U(k)
One can U(k).
To summarize: for
S(n), [ ]~(n) ~ ~%u%[ ]l
(1.15)
for
U(k), [ ]3c(n) ~ @%sx[ ]X
(1.15)
l
53
where [ ]h
u h = dimension of the irrep of
[ ]h
of
U(k), s I = dimension of the irrep
S(n).
The a b o v e i s a t h e o r e m w h i c h i s a l e i t - m o t i v
of W e y l ' s b o o k q u o t e d i n t h e
introduction and is implicit in the two other books. are l a b e l e d
o n l y by t h o s e p a r t i t i o n s
of
n
When
such that
n > k, irreps of
E~. ~ k , i . e . ,
U(k)
t h e Young
1
diagrams of irreps of
squares;
n = 1,
of
and
U(k)
~
U(k)
have at most
corresponds
n = 0, "."
k
lines, but an arbitrary number
to the k-dimensional
of
U(2)
are labeled
representation.
[hl,h2]
~.
course it
is shorter
a [Ill]
h '1 = h l ~ j~ = ~ p + 2 - j , h ti = h i - %p + l - i
Ee.1 < k , a~ = k - ~ a i ,
to say that
~.J [%i I]
is if
i,j > i.
Of
I 1
t h e Young d i a g r a m of
For
with the integers
X I ~ h 2 ~ 0. The diagram of the contragredient representation of
with
of
(= f u n d a m e n t a l ) r e p r e s e n t a t i o n
to the trivial (= one-dimensional)
example, the representations
n
[Xi,]
i s t h e complement ( u p -
~°
s i d e down) of t h a t
SU(n) group
SU(k)
of
[Xil]
in the rectangle
Representations.
of
U(k)
of
k
lines
of
h1
The restriction of an irrep of
is an irrep of
SU(k).
Irreps of
U(k)
squares.
U(k)
whose Young dia-
grams differ only on the left by a rectangular block of columns of length by restriction equivalent irreps of unambiguously label irreps of inequivalent irreps of
Example.
SU(k)
SU(k).
by Young diagrams.
Moreover, this yields all
[%l,h2]
of
U(2)
SU(2)
obtained by re-
are given by the value of the integer
h I - h 2.
So their Young diagram can be written as a horizontal line of
squares.
For
with
i j = ~(h I - h 2)
is called the spin of the representation.
Casimir operator of
hI - h 2
irreps it is customary to use the symbol D. 3
j
yield
SU(k).
striction of the irreps
where
k
Taking into account this remark, one
The equivalence classes of irrep of
SU(2)
to the sub-
D., 3
that defined by (1.9).
(1.16) 2j + I
as normalized by physicists,
is
is its dimension.
The
j(j + i)I, which is twice
(Indeed, physicists take as Cartan-Killing form
1/2 TrD(a)D(D)). We also recall the well known decomposition Jl + J2 D. @ D. = , ~ 31 32 @J=l Jl-J 21Dj Note that all representations Representations Zk, the cyclic group of
of
SU(2)
are self eontragredient.
of the Adjoint Groups k
(1.17)
SU(n)/Z n.
The center of
elements, so the adjoint group of
SU(k)
is
SU(k)
is
SU(k)/Z k.
54 The representations
of this group are those of
number of squares multiple of
k.
tions are
D. with integral j. 3 even number (= 2j) of squares•
For example,
equivalent
whose Young diagram has a
SU(2)/Z 2 = S0(3).
Their Young diagrams
The adjoint representation space of its Lie algebra,
SU(k)
Its representa-
contain only one line of an
of
SU(n) or its adjoint group is that on the it has dimension n 2 - i and label [2, In-2]; it is
to its contragredient.
Remark For Any Group. linear unitary representation As we saw
S(n)
resentations
and
of
G
S(n)
For any group
G, let
~C(I)
(it may be reducible and
acts on
~C(n) = ~ C (I) •
may be infinite).
~(n) o~[ ]~
Subspaces
are not in general subspaces
be the space of a
dim ~C(I)
of primary rep-
of primary representation
Methods for knowing the nature of the G-representation
of the different
of
G.
~¢(n)
~[ ]h would be interesting
for the physicists
(bosons) and
(fermions).
~(n) [i n]
especially in some case, for %~(n) , W[n ] Here is an example of a result, proven by A. Bohr,
Mat. Fys. Medd. Dan. Vid. Selsk, 2 6 (No. 14), 16. G
is
resentation not contain with spin
S0(3), ~C(I)
D 2.
For any
DI 0
n
is the five-dimensional the representation
in its reduction
of
Hilbert-space S0(3)
on
into direct sum of irreps.
ground state has no spin
1
of the rep-
~(n) = ~ ~C(1) "~[n] (Physically,
state corresponding
does
a nucleus
to collective
excitations.) Of course we also can add that if an irrep of once, then it acts either on 1¢(n) "'[n] or on ~C(n) [in ]
1.5.
More Algebras
and More Tensor Operators•
An algebra on the vector space algebra is symmetrical, Hom(E V E, E), (resp. , metrical, or
(respectively, Hom(E A E,E)).
antisymmetrical
E
G
appears
on
Pseudo Roots of
is an element of
antisymmetrical)
X (n)
SU(n)
Hom(E @ E, E).
The
if it is an element of
Similarly we can define a co-algebra,
co-algebra as an element of
only
sym-
Horn(E, E • E), Hom(E,E V E)
Hom(E, E A E). If
E
is the space of a linear representation
Hom(E 8 E,E) G, resp., of automorphisms When
contains G
for the space
are algebras, resp.,
on
E
of the adjoint representation,
for all simple compact Lie groups and the corresponding the Lie algebra itself.
co-algebras,
compact Lie group, a necessary
is that the representation G
G, elements
of whose group
G.
is a semi-simple
dim Hom(E @ E, E) G > 0 example,
Hom(E,E 8 E) G
of
condition for
has a null weight.
For
dim Hom(G A G,G) G = I antisymmetrical
algebra is
55
In
V
we shall see two examples
on a real irrep space
E
of
the adjoint representation or
i.
G = SU(3)
ture of elementary
algebras uniquely
algebra of particles,
SU(n),
n > 2.
Let us give some properties
since it has been used in the physics
after its introduction
For
dim Hom(G V G,G) G = 0
of a simple compact Lie algebra SU(n),
defined
dim Hom(E V E,E) G = i.
x SU(3), with
It has the latter value for the
of this symmetrical
of symmetrical
by Gell-Mann.
litera-
What follows
is
extracted
from a preprint written in collaboration with L. A. Radicati. 2 G be the n - 1 real vector space of the n × n traceless hermin tian matrices x. The action of u 6 SU(n) on G (vector space of the Lie n algebra) is x u~+ uxu-i = uxu*. The euclidean scalar product Let
1 (x,y) = ~ is invariant
(= i/n
the Cartan-Killing
trace xy bilinear
(1.18)
form).
The
SU(n)
Lie algebra law
is i i x A y = - ~(xy - yx) ~ - ~[x,y] and the symmetrical
algebra law ist
I x V y = ~{x,y} Note that for
n = 2
fijk~ dijk
2 - ~(x,y)~
it is trivial:
In the physics normal basis
(1.19)
literature
where
(1.19')
x V y = 0. (mainly for 2 - i)
(ei,e j) = 6ij(i, j = l,...,n for the structure
{x,y} = xy + yx
n = 3) one introduces and uses traditionally
an orthothe notation
constants
e i A e . 3 = Ekfij kek' e i V e j = Ekdijkek Let us use
F(a),
D(a)
for the linear mappings
of
G
(1.2o)
F(a)x = a A x, D(a)x = a V x (the matrices
are
F(ej)ik = fijk' D(ej)ik = dijk)"
With the scalar product symmetric.
D
and
F
of dimension
D(a) = 0.
of an element
m n - i.
the Caftan subalgebra
F(a)
As is well known,
x, i.e.,
the set
When its dimension C
x
of
x.
is antisymmetric
6 Hom(G,L(G)) SU(n)
are tensor-operators
at the end of 1.3, trace the centralizer
(1.18),
is
D(a)
is
so from the theorem
in the Lie algebra
{y,y A x = O} n - 1
and
SU(n),
is a Lie subalgebra
it is abelian and it is called
(All Cartan subalgebras
are transformed
into each
other by the group.) x, x V x,
C is spanned by the n - 1 linearly independent vectors x (x V x) V x = x V (x V x), ((x V x) V x) V x, etc., up to n - 1 factors
and C is also a subalgebra for the law " V ". The roots of SU(n) are solutions x of the equation r n - (r,r)r n-2 = 0. We shall normaliz e them by (r,r) = i. In a
2 * This is not a Jordan algebra. However, one could have started from the n The cordimensional representation realized by the n x n hermitian matrices. responding symmetrical algebra is a Jordan algebra.
56
Cartan algebra
C, there are
is also one), for every eigen space
C
n(n - i)
a 6 C, the spectrum of
and on the orthogonal Spectrum
Define
(for
normalized roots
space
F(a)
rk, (if
has
n - i
r
is a root,
-r
zeros for the
the spectrum is the set
C~
F(a) l = {i(a,rk)} Cm
(1.21)
n > 2) n-
2
7--~
qk = rk V r k = (-rk) V (-rk)
(1.22)
then (i. 22' )
(qk,qk) = i and they are idempotents
of the V-algebra
n - 4 - qk # n ( n - 2)
qk V qk
We will call them "pseudo roots" every
(they are weights
of
(1.23)
SU(n)) for they satisfy for
a 6 C Spectrum
D(a) l ~ C
=
{n -
~tqk
2,
a"
, ) = (a,r k V rk)}
(all the eigen values have at least multiplicity Let us denote by
(1.24)
2).
I 6 Hom(G A G,G) SU(n)" , v 6 Hom(G V
G,G) SU(n)"" the vector
space homomorphisms %(x @ y) = x A y, ~(x @ y) = x V y and consider
their right inverse o ~' = Identity on and
Note that %'
%'
G, v o ~' = Identity in
G
(1.25)
can be defined for any semi-simple Lie algebra.
As we said
~' define co-algebras on G. If T is a G-tensor operator, using the A T of Diagram 2, one can define the G-tensor operators
and
mapping
A T A T = T o %a
and
A T V T = T o i
(1.26)
and by recursion (...(T
T) ~i
where
"
7. 11
is either
|,
A
11
or
" V ".
T)...) 72
T
(1.26')
~k For physicists who need to see coordi-
i
nates, in
SU(3)
octet space
i g e k, , (e i) = Zjk ~3 dijkej 8 ek l' (ei) = Zj,k - ~ fijkej
% The
fijk
and
dij k
are the structure constants
.%
introduced by Gell-Mann.
57
If we set
T(e i) = Ti, then 1
(T A T)i = Ej, k - ~ fijkTjTk , Note of course that we can define operator when
G
T A T, T V T
dean scalar product. (Ker %)~
G
on
Indeed,
and
E
%
E or
i
=
3
Ejk 5 dijkTjTk
for any real irreducible E-tensor
dim Hom(ETE,E) G = 1
is a compact group when
since there is the irrep of
between
(T V T)
(~
is
A
or
V)
is orthogonal and leaves invariant a eucliv
are then surjective and are isomorphisms
(reap., (Ker ~)~
and
E) so we can define their right
inverse. Let us consider the more particular case when the G-morphism
T
is
F
itself (see Equation (1.6)), i.e., the representation (up to the factor i) of the Lie algebra on F V F
3£.
Then
F A F = iF.
When
SU(3)
is used for elementary particles,
is often called the D-coupling operator (see 5.1b).
For
SU(2), in order to
follow the tradition started in elementary school, we denote by
×
the Lie algebra
law (= vector product) [F(a),F(b)] = iF(a × ~) and by
eijk
(1.27)
the structure constants e i × ej = Ekeijke k
So if
A
is a vector operator (with
(1.28)
A(ei) = A i
+ ÷ i E s..,[A. (A × A) i = jkE£.I3~.,A.A,3 K = ~ jk 13K 3'Ak]
Remark.
Given two G-tensor operators AV
and in particular
A x B.
1.6.
For is trivial,
B =A
SU(2)
O B
o
t,
AA
A
and
B =A
@ B
B, we can also define
o ~
This reduces to Equation (1.26) when
More on
SU(2)
the symmetrical algebra
V
on the adjoint representation
G
dim(G 0 G,G) SU[2)" = dim(G A G,G) SU(2)" = i
Ejl ~ Ej2, Ej3,
dim Hom (Ej 1 8 E32'Ej3 )G = A(Jl'J2'J3) = 0 A(Jl,J2,J3) = 1
otherwise.
A = B.
and its Tensor Operators
Much more generally, given any three irrep on
where
(1.29)
if
lJl - J21 ~ J3 ~; Jl + J2
or
1
(1.30)
(triangular relation), 0
This property, under an equivalent formulation, is called the Wisner-
Eekart theorem by physicists, and groups with the property (1.30) have been called simply reducible by Wigner.
58
Let us give here two references that we shall quote often in this section.
A.
Quantum Theory of Angular Momentum - a collection of reprints and original papers edited by L. C. Biedenharn and H. Van Dam, Academic Press, New York (1965).
B. Spectroscopic and Group Theoretical Methods in Physics, Racah Memorial Volume, North Holland, Amsterdam (1968). In
B
p. 131-136, Wigner proves the following theorem for finite groups.
Theorem
Let
G
be a finite group and
H
a subgroup.
The following conditions
are equivalent a)
The restriction to
H
of any irrep of
decomposed into irrep of b)
H;
G
is multiplicity free when
/
The ring of conjugation classes by
H
of elements of
G
is abelian.
Let us explain a) and b) in more detail. a)
Given an irrep of
G
on
is generally reducible.
~, its restriction as a representation of To say that it is multiplicity free means
that in its decomposition into irreps of more than once.
H
H
no such irrep appears
Equivalently, one can say that the commutant of the
representation of
H
(i.e., the set of all bounded elements of
which commute with every operator of the representation of set is an algebra) is an abelian algebra.
[~)
N; this
That last condition can be
used as definition of multiplicity free for any linear representation of any group. b)
Given
a 6 G, the conjugation class of
A = {hah -I, Vh 6 H}. set
A • B = B • A.
compact groups. this theorem:
by
H
is the set
Given two such classes we define as
{ab, a E A, b 6 B}.
classes,
a
A • B
the
Condition b) states that for any pair of It seems easy to extend Wigner's proof to
Examples of pairs of group and subgroup which satisfy S(n)
and
S(n-
i), U(n)
and
U(n-
From the group law one verifies that the direct product
l).t SU(2) × SU(2)
and its diagonal subgroup satisfies b); by Wigner's theorem this implies (1.30). would be interesting to extend, if possible, Wigner's proof to type
I
It
locally
compact groups.tt
t
After the lecture, Professor G. Mackey gave a proof for compact groups, using his theory of induced representations.
tt Wigner proved in 1941 (paper reproduced in reference A., see 1.6) for finite groups another property equivalent to a) and b). Let ~(g) the number of square roots of g in the finite group G, and v(g) the number of elements of G which commute with g. In a finite group ~ £ ( g ) 2 _ ~(g)3 ~ 0. The equality g occurs if and only if G is simply reducible.
59 Another property of SU(2)
SU(2)
that we have mentioned is that any irrep of
is equivalent to its contragredient.
fines an isomorphism
C:
E. ~ E~ 3 3
For any irrep
between the
with the canonical property
CT
of
C.
space of
of D. J
SU(2)
this de-
and its dual
~ C T = (-l)2Jc E3;
Ej for the transposed
E. 3
D. 3
E~ 3 (1.31)
Physicists normalize
C
by
cTc = 1
(1.32)
We are by now ready to give you a glimpse of the calculus developed independently by Wigner and Racah for the necessity of physics in order to exploit completely rotational invariance in atomic physics (and later on in nuclear physics and many other branches of quantum physics).
Many of the numbers observed in atomic
spectra (sp~cing between neighbors in a family of lines, relative intensity of these lines, etc.> turn out to be algebraic functions of the coefficients defined by I
Wigner and ~acah. Since they are so useful, the literature on Wigner and Racah co! efficients is abundant. They have been found to possess unexpected symmetries, there are u~proven conjectures on them.
However, the language of this physics folk-
i
lore seems to be unknown to the mathematically minded ethnographer. I hope there will be much discussion on this subject in this Rencontres. The rest of this section should help to start it. To write Wigner's "three
j" coefficients,
physicists choose a base in
each Hilbert space
E.s carrier of the irrep Dj, for every j. It is formed by J the eigen vectors of a chosen U(1) (Cartan subgroup of SU(2)) ordered in terms of decreasing eigen value
~
(going from
j
to
-j
by integer steps).
It is obvious
that most of their properties are base independent. Consider an element of the one-dimensional vector space (E~31 8 E~32 8 E.j3)G = Hom(Ej I @ E.32,Ej3)G
(1.33)
and denote it o
o
(jlj 2J3) The isomorphism
C
(1.34
and its inverse, defined in (1.31), (1.32), transform
the tensor (1.20) into its following images
and so on.
(313233) ~ (Eli ~ E I. @ E I. )G = Hom(Ej I 8 E. @ E. ,~)G i J2 33 32 33
(1.35)
(jlj?jq) E (Ej 8 E. 8 E. )G = Hom(~,Ej ~9 E. ~9 E. )G ...... 1 32 33 i 32 33
(1.36)
o @ E. @ E. )G = Hom(EjI,Ej 2 @ E. )O (jlJ2J3) E (Eli 32 33 33
(1.37)
60 o
Equations (1.35), (1.36) show that
o
o
(jlJ2J3) (resp.,
(jolJ2J3)) belong to
a one-dimensional representation of the permutation group of the three factor spaces labeled by
31' J2' J3"
Computation shows that the representation is
,,,, (symmetrical)
if
Jl + J2 + J3
is even
~ (antisymmetrical)
if
Jl + J2 + J3
is odd
(1.38)
The composition of the two homomorphisms o
E.
0 E.
31
32
o
o
(JlJ2J 3)
0 E.
(JlJgJq)
~. ~
....
o~
@ E. 0 E. 32 33
> E.
33
31
(1.39) o
is an element of
Hom(E.31 O E.32 0 Ej3 , E.31 0 E.32 0 E.33)G that we denote
o
o
(jlJ2J3)
q1 2 3) Wigner proved (see reference A and Equation (24.18b) of his book quoted in the Introduction) f
where
dp(g)
o
o
o
D. (g) 0 Dj2(g) 0 Dj (g)d~(g) = (jljvjq)(jljgj~) SU(2) 31 3 . . . . . . . . .
is the invariant
SU(2)
measure of mass
(1.40)
[ dp(g) = i. J SU(2)
This also defines for ~ou, up to a sign, which element of the onedimensional vector space o
o
(E1 8 E 2 0 E3)G
has been chosen by physicists for
o
(31j 2j 3 )" Of course tensors can have their indices contracted (notation stance
oo c
XOO
(abx)(cpq)
is the composed homomorphism ooc
(abo)
E
x); for in-
0 I
0 Eb o E 0 E a p q
ooo
0 I
P
q~ E ~ E
c
P
oE
(cpq)~ q
Wigner's notation is v@ry handy! Note that equation (1.40) yields I
where
Xj
x x x jlJ2J3 SU(2)XjI(g)xj2(g)xj3(g)dD(g ) = (jlJ2J3)(x x x ) = A(jl,J2,J3)
is the character of D.. 3 Approximate expression, asymptotic expression, for large
j's
(1.41)
exist for
the components of those tensors in the basis described above (see ~he thorough bibliography of reference A (see 1.6)). found a 72 elements group of symmetry
Regge (paper reproduced in A (see 1.6)) has ~ Aut(S(3) x S(3))
for the set of components
61
~i~2~3 (jlJ2J3)
o
of
o
o
(jlJ2J3).T
In 1941 Racah and Wigner (both papers reproduced in A (see 1.6)) introduced a "six j" symbol (numerical function of six irrep of recoupling coefficient.
Consider the sequence of
Ee
SU(2)-homomorphisms
(ofa) If 8 ~aoo) coo ~ Ef 8 Ea ~ Ef ~ Eb 8 E c
Ed @ Ec Since
SU(2)), also known as
It is canonical.
c
(o~e~ a=o)~ E e
(1.42)
E
is the carrier of an irrep, this SU(2)-homomorphism must be a e multiple of the identity operator on E . e Its trace defines (up to a sign that I do not guarantee here) the six j's symbol l abc I
- -b+c-d+e+f x xx xx = (-i) (efxa) (~bc) (fbd) (ace)
(1.43)
i defJ Wigner has shown that for given dices
c, f.
def,
labc I" is an orthogonal matrix with ina, b, d , e, 'def
He also proved the relation (in his book, Chapter 24)
= ;If Xa(r)Xb(s)Xc(t)Xd(st-l)Xe(tr-l)Xf(rs-l)d~(r)d~(s)d~(t)
Asymptotically its value is a rapidly oscillating function of some variables, but an average over some range of one argument yields, when
a, b, c, d, e, f
form the
edges of a tetrahadron, the asymptotic value
abcI2 + (24~V) -I def2 where
V
is the volume of the tetrahedron. G. Ponzano and T. Regge (in reference B (see 1.6), first paper) have con-
jectured precise asymptotic formulae for
{abcl whether or not the valu~ of the def arguments can be the length of the edges of a tetrahedron. Furthermore, Regge has found (paper reproduced in A (see 1.6)) the largest
linear group acting on the Z-module generated by the symbols a/2, b/2, c/2, d/2, ~abc e/2, f/2 and having idef I as invariant. It is the group S(3) × S(4) which includes the permutation group of the columns. T Most of these symmetries appear naturally (see Bargmann's paper in A (see 1.6)); for the others see G. Flamand, Ann. Inst. H. P o i n c ~ , ~, 353 (1967).
62
Neatest and very symmetrical expressions for the
(jlJ2J3)
and
{
JlJ2J.3~ symbols can be found in the paper of V. Bargmann (last paper reproduced J4J536~ in book A (see 1.6)) who uses Hilbert spaces of analytic functions as spaces of SU-2
irreps.
2.
ATOMIC AND MOLECULAR PHYSICS
2.1.
Group Theory and Atomic Physics
The application of group theory to atomic physics is essentially of this ÷
type; only the even part ÷
part
÷
÷
-~
f+(r) = i/2(f(r) + f(-r) ÷
÷
÷
-->
f+(rl,r2) -- i/2(f(rl,r2) + f(r2,rl))
to the integral over the whole space
(respectively, the symmetric
÷
of
If(~)d3~r
f(~)
(resp.,
(resp.,
÷
f(rl,r2)) contributes
Ifkrl,r2)a "÷ ÷'13+ rla"3÷" r2).
This
is the explanation of two empirically known facts (before 1926), the Laporte selection rule for atomic spectra and the partition of the helium spectrum into two independent subsets (attributed to ortho and parahelium).
Of course, these examples
are the simplest because they are based on invariance under a two element group (Z2).
We will have to use invariance under
invariance under subgroups of
2.2.
SO(3)
SO(3), S(n)
and
U(2)
for atoms and
for molecules.
The Correspondence Principle
We had a general description of quantum mechanics, but now we have to, know how to study a given physical system.
There does not seem to exist an axiomatic
formulation of the question, so here again, physics is still an art!
However, when
the system under consideration, with a finite number of degrees of freedom can be described by classical Hamiltonian mechanics, the "correspondence principle" tells physicists how to treat it quantum-wise. Let
h(Pk,qe)
be the classical Hemiltonian and dPk
~
dt ..... -k the Hamiltonian equations.
~h ~ = - ~qk'
~h = ~P£
The corresponding observables
pz, Q£
in quantum
mechanics form the abstract algebra with unit PkQz - QzPk = [Pk'Q£ ] = - i ~ 6 k ~ [Pk,Pz] = 0 = [Qk,Qz] where
2~h
is Planck's constant.
(2.1)
63
In the cases we shall study, function of the where
Ht
and
p's H ~t
and
htt
h
is a sum
h = h t + h It
is a function of the
q's.
Then
are the same functions respectively of the
where
ht
is a
H = H t + Hte P's
and the
Q's.
There is yet no synthetic formulation of quantum mechanics as there is for classical mechanics by symplectic manifolds Souriau).
(see, however, work in progress by Kostant,
We also know that the relations between the classical and quantum treat-
ment of the same problem are not simple (see e.g., Van Hove's work in 1951 comparing the two automorphism groups). The Hamiltonian operator is the generator of the group of time translations [H,Qk] = i~Qk , [H,P~] = i ~
.
(2.2)
A representation of the algebra defined by (2.1) and (2.2) was obtained, independently of Heisenberg's work by Schrodinger, using the concept of de Broglie's waves. Indeed, the algebra (2.1) is realized by self-adjoint operators of is the Hilbert space of square integrable functions
~(qi ).
Qk P = qk ~, P£P = ~i ~q% ~ P The
~
are also functions of the time
(t)
where
Then
.
(2.3)
and the SchrDdinger equation is
H = i~ 9-~
.
(2.4)
This representation raises some analysis problems. von Neumann's theorem (J. von Neumann:
LOC)
On the other hand,
"Die Eindeutigkeit der Schr~dingerschen
Operatoren", Math. An~le~,
104, 5]0 (1937)) tells us that all irreducible represeniP k iQ~ tations of the algebra defined by Equation (2.1) are equivalent when e , e are realized by unitary operators.t Quantum mechanics was also discovered by Dirac who gave the neatest for-
mulation of the "correspondence principle".##
In classical Hamiltonian mechanics
one has also a Lie algebra, that of the Poisson brackets functions of the
p's
and the
(P.B).
Let
f, g
be two
q's,
P.B.(f,g) = E% ~f ~g ~q~ ~q~
~f ~$ ~q% ~P~
(2.5)
For systems with an infinite number of degree of freedom, as they appear in statistical mechanics and field theory, this is far from true. Infinities of irreducible representations of (2.2) have been given first by Friedrichs, Van Hove, Ggrding and Wightman, Segal and several other physicists and mathematicians. An excellent thin book on the subject is by A. Guichardet, Alg~bres d'Observables Associ~es aux Relations de Commutation, Armand Colin, Paris, (1969). (See also G. Mackey, Duke Math. J., 16, 313 (1949)). ## Historically, meaning.
the expression "correspondence principle" had a more restricted
•6 4
The Lie algebra of the corresponding quantum observable is [F,G] = i~ As you surely know
quantum observable of l~]2~dqk , where
~
P.B.(f,g)
(2.5 t)
is a solution of Schr~dinger
Equation (2.4), is the density of ~robability to find the system at the coordinate {qk }.
This of course appeals very much to physicists.
like just as well to work with the abstract algebra.
As mathematicians you will As a short, but fundamental
illustration of the use of that algebra, let us prove the Heisenberg uncertainty relations. Let x>
A, B
be the self adjoint operators corresponding to the observables
a, b.
If
<xAx>
is the expectation value of
is a given state of the physical system we study, we have seen that "a"
for
x>
and the mean square dispersion of
probability is given by (Aa) x = l<x,(A - <x,Ax>)2x>l I/2 = I<x,12x>l I/2 = II~II
(2.6)
= A - I<xAx>
(2.6 t)
where
By Schwarz' inequality (Aa)x(Ab) x = If
A
and
B
l li~ll
i< l l'~xll ~ ll ~ 71
•
x'[A'B]x>l
satisfy the same canonical relations as the
P's
and
(2.7) Q's
we do
obtain i (Aa)x(Ah)x ~ ~ ~
2.3
Particle of Mass
Let
V(r)
m
(2.8)
in a Spherically Symmetric Potential
be a spherical symmetric potential, where
r
denotes
I~I.
The Hamiltonian of the particle is 1 ~2 H = ~m + V(r)
which is invariant under the orthogonal group H, ~2, V(r)
are "scalar operators";
~, R
0(3).
and
(2.9)
,
Using the vocabulary of 1.3,
R x ~ = ~ ÷
vector operators.
(So we put an arrow on .them!)
the three-dimensional vector space
E3
÷ ÷ ,Q(D)] = i~
where the Cartan-Killing form
~
a, b,
etc., are vectors of
of the adjoint representation of
should write the canonical commutation relations E~(~)
If,
are (polar and axial) ÷
0 (3), w e
(2.1)
÷ ÷ ÷ ÷ ~1 B(a,b) = i~a,bX
,
has been defined in (i.ii).#
# See also the Appendix on commutation relations at the end of 2.
(2.10)
65
From (2.10) and the definition by the correspondence angular momentum operator
(see end of 1.5), ÷
÷
÷
principle of the
~ = R x ~, we obtain
÷
[L(a),L(b)]
= i~(a
A b)
,
(2.11)
which confirms
that the vector-operator
representing
representation
(up to
Lie algebra on the Hilbert space of our prob-
i) of the
0(3)
lem. Some physicists write prised if in all physics ÷
Pj, L k
are used for
~ • ~, ~ • ~
~(
~) , ÷e(n). ÷
is the
But do not be sur-
text books an orthonormal basis of vectors notations ÷
÷
÷
÷
corresponding
motion generate the algebra
Qi'
÷
Q(ei), P(ej), L(ek),
The operators
for
the angular momentum
etc.
to the observables which are constants
{H} ~ , the commutant of
one deduces from (2.10) and the definition
H.
Hence,
of
the equation that
of
÷
÷
E E 3, [L(a),H] = 0 means both that the Hamiltonian
or symbolically
[L,H] = 0
is invariant under rotations
(2.12)
and that the angular
momentum is a constant of motion. ~2
The Casimir operator (with the physicists' normalization) of 0(3) is 3 2 = Ei=IL i. As is well known, its values for irreducible representations of SU(2)
are
j(j + i)~ 2
representation.
where
2j
is an integer
Only integer values of
state vector is an eigenvector that the corresponding
of
~2
Consider
two particles
positive integer).
where
r =
is
of mass
with
r =
tot.
is the dimension of the S0(3)
irrep,
when the
j(j + I)~ 2, we say shortly
jh.
The Hydrogen Atom
ml, m 2
The total Hamiltonian
h
2j + i
appear in the
with eigenvalue
angular momentum
2.4.
~ 0; and j
+2 Pl -2m I
electric charge
Ze, -e
for this system of two particles +2 P2
(Z
is a
is
Ze 2
+ - -
,
2m 2
(2.13)
r
- r I.
Introduce the center of mass ÷r 0 = (mlr÷ I + m2r÷ 2)(m I + m2 ) -i ÷
and
r
÷
as new variables
conjugate variables.
instead of
rI
and
÷
÷
r2; let
P0
(2.14) ÷
and
p
the
Then +2 PO
htot.
2(m I + m2)
+
*2 ~--Ze2" h +h (2m r ) = cm
(2.15)
where m = mlm2(m I + m2 )-I
(2.15 I)
66
The motion of the center of mass is described by
while h corresponds to the cm So quantum-wise, we have to study the spectrum of
internal energy of the system.
~2 2m
H
Ze 2 R
h
'
(2.16)
for obtaining the energy of the hydrogen atom levels.
The first quantum study of
the hydrogen atom was made by Pauli, Z. Phys. , 36, 336 (1926) before SchrSdinger's ÷
equation was published.
Pauli did study the abstract algebra generated by
and Equations (2.1), (2.2), and (2.15). stant of motion.
The angular momentum
~ = ~ x ~
÷
R, P, H is a con-
Another constant of motion is the Runge-Lenz vector I = y(Ll + x ~ - ~ x ~) + ~% ~
with
X = mZe 2
(2.17)
Note that ~(LI ÷ x ~ . P ×. I) . (R . ~)~.
~(~2) . . i ~.
P(P
R)
(~2)R + i ~
(2.18)
so we can check that [~,H] = 0, [~,H] = 0 We recall that dices are equal.
(2.19)
.123) sij k = sign of the permutation ~ijk
or
0
if two in-
From now on we will use the Einstein summation convention, i.e.,
summation of repeated indices is implied, and we find (2.20)
[Li,Lj] = i~gijkLk, [Li,Aj] = i~E...~13K~
(2.21)
[Ai,Ak] = -i~2mHSijkL k
~.;=;.I=o
(2.22)
~2 _ 2mH(~2 + ~2) = (Ze2m)21
(2.23)
Let us just consider the bound states of the hydrogen atom. spectrum of
H < 0.
Let
P_
be the projector on the bound states. ÷
write
X- -- XP .
From (2.19) when
X
P XP Furthermore,
-2mH-
They correspond to the
= XP
is
~(~)
or
For any
X
÷
A(b),
= P X = X-
is an inversible positive operator.
positive square root of its inverse and define
Let
(-2mH-) -I/2
K~ = A-(2mH-)-I/21"
be the
Then Equations
(2.20 i) to (2.23) read ILZ i i IL: i i [~ ,~ Lj] = ~ ~ijkLk , [~ ,~ Kj] = % S i j k <
(2.20 S )
I K ~" 1 i [~ ,~ Kj] = ~ gijkLk
(2.21')
L i~--~K-2 +
" K
= K
•
= 0
Z e 2m 2 (-2mH) -i ~-2 = (---~--)
(2.22) (2.23)
67
We last define ~(+1 so the previous
= ~i
~(-1
+ ~i
~(-)
(2.24)
equations read
c (2)
= i iS( )
~(+)2 = ~(_)2 = !)(2((Z-~2 m H ) - 1 4
The spectrum of this operator is integer.
j(j + i) = (n 2 - 1)/4
°0
(2.2s)
- i)
(2.26)
with
2j + 1 = n, positive
So the energy spectrum of the bound states of the hydrogen (n positve integer,
(Ze2) 2 m -~2n 2
~n =
-Z2(e2. 2 m e 2 n2 -~c) 2
atom is
-i (Z~)2mc 2) 2n 2
(2.27)
where 2 e ~c in rationalized fundamental
units of charge,
~
1 137.03...
(2.28)
is the fine structure constant,
a dimensionless
constant of physics.
Some Physical Comments.
The ratio binding
energy/electron
rest-mass
energy
is the number gn
(Z~) 2
mc 2
2n 2
The value of every physical observable we can compute will appear as the product of a pure number and the quantity of same physical dimensions built with the constants 2 e, ~, m, c. Example: length ~/mc = 3.86 x i0 -II cm; energy mc = .51 x 106 eV; time
h/mc 2 = 1.28
x 10 - 2 1 see.
The pure number is a function of
the value of the observable in the unit system ~ = m = c = 1 2 is the value of e in this system. For instance = Z ~ ~
~
only.
It is
that we will use, and
~ (i 10-8 cm)-i = (~ i ~ngs tr~m)-i
We have studied not only the bound state of the hydrogen atom + p e (rap = 1836 me) (the nucleus can also be a deuteron ~ 2m ), but also that of ++ P positronium e e (m I = m 2) , munomium ~ e ( m = 207 me), ~-atom, H-atom, ionized Helium ion
H+e, etc.
More On The Group Aspect.
The states of energy
~
are eigen states of n
~(+)2
and
~(_)2
and they form the space
~
of the irrep n
(j,j)
of
S0(4); n
has dimension (2j + 1) 2 = n 2
(2.29)
68
The Lie algebra of the physical rotation
(~)
is the diagonal of
@ SU(2) = S0(4), so the representation of the rotation group in irrep
(J'J) = (n -2 i' n" -2 i)
of
SO(4)
SU(2)
~n' space of the
reduces to
2j (j,j)Iso(3) = @E=0DE
(2.30)
i.e. , = 0, l,...,n - i
(2.30')
n I
,
~
,
,
:
~-
J
,
:
J
:
3
. . . .
2
(z~__/) 2 m
0
-i 0 1
~=
0
i
2
3
S
p
d
f
FIGURE 2. i.
-2 -i 0 1 2
-3 -2 -i 0 1 2 3
-4 ... etc.
LOWEST STATES, IN A LINEAR ENERGY SCALE OF THE HYDROGEN ATOM There is an infinite number of states with energy between -s and 0 because of the dependence in r -I of the potential for r ÷ =. ÷ L~The eigenvectors of the abelian algebra genera÷ ÷ ted by H, L , L(e 3) form an orthogonal basis for the Hilbert space of bound states. A complete set of labels for them is the quantum numbers n, ~, m; n = positive integer, ~ and m integers 0 ~ g ~ n - i, -g ~ m ~ g, corresponding to the eigenvalues
Z(~)2/2n 2, ~(g + i), m
of
(mc2)-iH, ~-2~2, ~-l~L(e3 )÷
Note that the trivial representation appears once only for each
n, and
from the Frobenius reciprocity theorem w e know that
@2j=0 (j 'j ) = UD0 the induced representation of
S0(4)
by the trivial representation of
(2.31) S0(3).
In
69
other words, space
P_~=
L2
S0(4)/S0(3).
(functions on
$3) since the sphere
(This was exploited by V. Fok, Z.
Physik, 98, 145 (1935), see
From Mackey's theory of induced representations, of the representation of
R 4 D S0(4) = E 4
duced by the trivial representation of tor
# 0
of
P_~
irrep of
R 4.
(euclidean group in four dimensions)
This is an irreducible representation of
considered above).
S0(4,1)
speak of eigenvectors of
H H
in-
E 4.
We can also con-
(obtained by deformation of the
However, the physical meaning of the generators
(representing the elements of Lie algebra) of The spectrum of
(2.31) is also the content
R 4 [] SO(3), the stabilizer of any chosen vec-
as the space of an irrep of E4
is the homogeneous
Z. Physik, 86, 21 (1933).)
also L. Hulthen,
sider
S3
on
E4
or
SO(4,1) is not very trans~rent!
P+~, (positive energy) is
(+0,~).
One cannot
for the positive energy = unbounded states of a proton
and an electron and one has to study their scattering.
However, since
~K+ = ~ +(2mHP+)-i/2
HP+ = P+H
i/~ ÷LP+
is a positive operator one can define and and ÷+ i/h K , which generate a S0(3,1) Lie algebra as was noted and exploited by V. Bargmann,
Z. Physik, 99, 576 (1936). Thus,
dimensional unitary) irreps of an irrep of the inhomogeneous
P+~
is a direct integral of (infinite
SO(3,1), the Lorentz group. S0(3,1)
irrep of spin zero and fixed mass
It is also the space of
that we call the Poincar~ group (it is an
m > 0).
What we observe mainly in atoms are their emission or absorption of photons of frequency = ~l~(~nl - ~n2 )
(2.32)
So the wave length is
nI which is
4 ~ ~
103
n2
times the "size" of the atom.
All large enough frequencies of (2.32) were already seen in the spectrum of atomic hydrogen.
In fact, there is a "fine structure" which corresponds to a
relative splitting in the energy of the states with 2 ~ 1/2 10 -4.
~ # 0
of the order of
The theory even predicts the intensity of the spontaneous emission of photons.
Since its wave length is large compared to the atomic size, the light wave
is a dipole emission# and the probability for spontaneous emission of a photon from
# Electromagnetic waves, predicted by Maxwell's equations, were produced by Hertz with an oscillating electric dipole. An example of such a dipole would be a charge -e rotating around a charge +e with a given frequency ~. That light was an electromagnetic wave was a Maxwell hypothesis and Selenyi, by clever experiments, verified in 1913 that light emitted by atoms was a dipole radiation. (Higher multipoles occur for more complicated charge distributions for which the
7o
state
x>
to state
y>(E x > E ) Y Xxy
(Note t h a t
<xeRx>
=
is 4 ~(E x
Ey)3 -
÷ 2 I<xeRy> I
(2.33)
i s the e x p e c t a t i o n v a l u e of the e l e c t r i c
as we shall see, it vanishes.)
d i p o l e of a s t a t e
The intensity of the emitted light by
N
and,
atoms in X
state
x>
is 4 - E )4e2ZTrP P~P R. ixy = Nx ~(Ex Y k XK y g
(2.34)
The W i g n e r - E c k h a r t t h e o r e m ( s e e b e g i n n i n g o f 1 . 6 ) p r e d i c t s tors, the matrix elements between two g i v e n eigenstates Example.
For
x>, y>
eigenstates of
that for all vector
x>, y>
of
oper~
~2, are eolinear.
H
-i~ + + m <xPy> = <x[H,R]y> = (Ex - Ey)<xRy>
~2
Consider from now on states which are eigenstates of
(2.35)
(eigenvalue
~(E + i)).
Note that <xLy> = 0
if
~ # ~ x y
,
(2.36)
while for a general vector operator <xRy> = 0
if
~x + ~y = 0
or
(2.37)
l~x - ~y I>I
However, we should have taken into account the parity operation H(r) = -r The corresponding
H
operator satisfies
~2 = I, H ~ but, since
~
=
-~, ~
= -~
(2.38)
is an axial vector (2.39)
Eigenvectors
of
~2
have a well defined parity
(the eigenvalue of
~).
Looking at
their realization by spherical harmonics, one finds HE 2 = (-i)~I 2 Thus, when
x, y
(2.40)
have a well defined angular momentum <xRy> = -<x,HH~y> = -(-i)
Y<xHy>
SO
<xRy> = 0
if
Z + Z = even x y
which is Laporte selection rule we spoke of in 2.1. and (2.41) is equivalent to:
(2.41)
The set of both equations
no electric dipole transition:
<xRy> + = 0 <=~
-
(2.37)
%1
# i.
Zi~iqi~ = 0; see work of Mie, Poincar~, Rayleigh, etc., on multidipole moment pole expansion - it's applied group theory!) The trouble was: classically atoms should always radiate and use up their energy fast. Quadrupole radiation in atoms can be observed in exceptional cases (rare-earth, atoms in interstellar vacuum). In a radiation field, electromagnetic emission of photons can be induced and become intense: laser!
71 How Does This Theory Compare With Experiments? very poor.
It is both very good and
Within an accuracy of 10 -4 the agreement for the values of the binding
energy is perfect.
The value predicted by the present theory of quantum electro-
dynamics gives a correction in
(Ze)4/n 3
10 -4 ) so that levels with different
~
(i.e., a relative correction of and same
n
(Ze)2/n
have a small difference in
energy, t What is very bad is the counting of the number of levels.
This can be -+
seen by putting the hydrogen atom in a constant electromagnetic field (F, electric ->
and
B, magnetic).
Then one must add to H . _ _3e _~. em 2
H e
~+
~ . ~
(2.42) ->
The effect of
F
levels of the same
(Stark effect) is well reproduced, but not that of E
should split into
2% + 1
do split, but in an even number of levels! !
levels separated by
B.
Indeed,
eB/2mc.
They
This is due to the electron spin which
we have not yet taken into account (see 2.6).
One should also take into account the
proton spin with effects 2 < e m m ep
~-31> m ~(Z~)3mem~ I)
2.5.
The Helium Atom
It has a nucleus of charge
Ze = 2e
(mass ~ 4mp) and 2 electrons.
After
separation of the center of mass motion, the Hamiltonian for the internal energy is 2 e H = HI + H 2 +-RI2 where
(2.43)
H i = P~/2m - Ze2/Ri , the hydrogen Hamiltonian and the operator
ponds to
711 the relative distance of the two electrons. r12 = Ir 2 - _ neglect the term in eZ/Rl2 (this is a better than 10% approximation) is solved.
We will consider only bound states.
the bound states of hydrogen atom. ~(i) @ ~(i)
with Hamiltonian
the binding energy is
Let
~(i)
RI2
corres-
If we our
problem
be the Hilbert space of
Our simplified helium atom has Hilbert space
H0 @ I + I @ H0
- (Z~)2(I/n~~ + i/n$)/2
where
H0
is that of hydrogen.
So
i.e., the sum of the binding energies
for the two electrons. We assume here that the term
e2/R12
is a perturbation in the technical
sense (see Kato's book for mathematical rigor).
This term breaks the
variance, so the electron levels with different
~
and same
n
S0(4)
in-
no longer have the
J" The difference between the two levels n = 2, % = i and 2 predicted by the theory of quantum electrodynamics is essentially Z4~52 -3, i.e., ~ 103 megacycles and the agreement with experiment is of the order of i0 -I megacycles ~i0 -15 mc2/~, Quantum electrodynamies is not yet well defined for the mathematicians! Refined predictions for positronium, muonium, etc., are also very precisely verified.
72
same energy.
(As we shall see later in 2.6, for a given
n, E
increases with
Z.)
What is left is angular momentum and parity conservation ÷
[L,H]
and the indistinguishability
= 0,
[H,H]
(2.44)
= 0
of the two electrons (2.45)
[SI2,H] = 0 where
S12
is the operator permuting the two electrons 2 = = S12 I, SI2(A O B)SI2 B O A E L~(1) The decomposition of the tensor product
of symmetrical and antisymmetrical
(2.451 )
O ~(i))
~(i) ~ ~(i)
into the direct sum
tensor spaces
3£(I) O~C (I) = X (I) V ~C(I) e~C (I) A ~ C (I)
(2.46)
that we also wrote ~(2)
I (2.46)
= ~[2] @ ~ [ i 2] give the decomposition into eigenspaces of gen atom. - Y 0 x)
Which of the two states
S12.
Let
x, y
be states of the hydro-
x V y = i/¢~(x @ y + y O x) or x A y = i/¢~(x @ y
yield the smallest expectation value for the positive operator
It is obviously
e2/R12 ?
x A y because the two-electron wave function vanishes when
is very large (while that of
x V y
has generally a maximum when
e2/R12
RI2 = 0).
symmetry character yields a new selection rule for the dipole radiation;
This
the matrix
element of the transition operator is proportional to <~I~ I + ~21~s> ÷
Since
-~
RI + R2
= E (
g2
(2.47)
l
= i)
is symmetrical,
P
and
~
must have the same symmetry character
since ÷
÷
I
÷
->
J
I
-~
÷
<~,(R I + R2)P > = <~,SI2(R I + R2)S12~ > = se <~(R I + R2)~ e>
(2.47 ~)
As we announced in 2.1, this shows that the helium levels are to be divided in two sets according to their symmetry characters, and eledtric dipole transitions occur only within each set.
Let me remind you that helium got its name because it was
observed in the sun before being observed on earth. posed of two spectra, one for orthohelium
Its spectrum appears to be com-
(e = +i), one for parahelium
This was a complete mystery before quantum mechanics.
(e = -i).
The explanation was given by
Heisenberg in 1926, "(Uber die ~pektra von Atomsystem mit zwei Elektronen", 39, 499 (1926)). the type
x @ x, as for instance the lowest level (n = I, ~ = 0
It is observed that corresponding rays (e.g., transitions V (n e,ge)
and
Z. Phys~k,
It also explained that the orthohelium has more levels; those of
(i,0) A (n,~) ÷ (i,0) A (n a,gl)
with
for each electron).
(i,0) V (n,~) ÷ (i,0)
n I ~ i,£ S ~ 0) of parahelium
are about three times more intense than those of orthohelium.
To explain it, the
electron spin will have to be taken into account (see also 2.9).
73 2.6.
Pauli Principle.
The Electron Spin
We want to pass now to the case of n-electron atoms.
The internal energy
Hamiltonian is t
~n)
Of course trons.
%
2 e E R.. l~i<j~n 13
n = Ei=IH i +
with
H.
P~l
Ze 2
2m
R.
l
is invariant under the permutation g r o u p
H (n)
S(n)
It is also the case for the electric dipole operator
observables.
(2.48)
l of the
e(EiRi)
n
elec-
and for all
Identical particles cannot be distinguished from each other and every
prediction of the theory must be invariant under When we consider states of
Z
(or
S(n).
n) distinsuishable particles, we con-
sidered (with success for the helium atom) the Hilbert space tensor product of the $C
for each particle. i
S(n)
acts on
Consider again
~C(n) = ~ $C[I)" "
for
~C~n)
by the representation
s ÷ S(s).
of operators.
which represents a state.
{S(s)}'
S(n)
Vs, S(S)PxS(S) = Px
This requires that the vector
x>
(completely symmetrical)
or to
symmetrical).
The othe r spaces
of the other factorial representations
(completely antiof
are excluded as space of physical states. We have used both
use of
$C[ ]~
~C[in] = ~ C (l)
for any vector
belonss either to
~[n] = ~$C(I)
S(n)
of
of the set
As we saw the rank one projectors wich represent
physical state are observables of the theory so x
identical particles.
Invariance under
all observables requires that they are in the commutant {S(s),s 6 S(n)}
n
%
X[n](n > 2)
$C[2 ]
and
~C[12]
for the helium atom.
for atoms does not represent nature.
However, the
Indeed, the ground state
of any atom would have all electrons with the same binding energy (of the order of (Z~)2/2).
Experimentally,
trum for Z-large enough).
only two electrons have this binding energy (X-ray specThe necessary energy (called ionization energy) for re-
moving a first, a second, a
kth ..., the crease irregularly from a fraction of e 2
vectors of some other
~(n)
do appearX
Zth to
electrons of any neutral atom inZ(e) 2/2 . Moreover, as we shall see,
The solution to this puzzle is that
~C[I)" "
is not the Hilbert space of the bound states of one electron in a constant potential. The electron has another degree of freedom~ the spin and the Hilbert space of its states has to be changed into a new X(1) = /(i) @ K( I )
where
i (1)
is the
L2(R3,t)
dimensional Hilbert space.
previously called
3C( I )
and
K(1)
is a two-
Pauli was the first in 1924 to introduce the spin as an
intrinsic angular momentum and magnetic moment for the nuclei, but it was Goudsmit and Uhlenbeck who introduced in 1925 the spin as an intrinsic angular momentum for the electron.
h/2
This explained the number of energy levels which appear in the
Zeeman effect, but it did not explain the magnitude of their splitting. magnetic moment produced by an electric charge
e
Indeed, the
moving with an angular momentum
74
j
(Ampere's law!) is ~ /2-~c
where
e~/2mc
is the Bohr magneton.
g = i, but for the spin
(2.49)
For the orbital momentum, g(l~/~l
lJ/hl = 1/2, g
solved by Thomas in 1925.
gJ/
appeared to be
2.
integral)
This was a mystery
It is a relativistic effect.
It is an experimental fact that we have to use Fermi Statistics for electrons, i.e., the Hilbert space of electronic states of an n-electron atom is
• [in] (= ~ ( i ) n C where
~(i)
(2.50)
is the (new) one electron Hilbert space defined in (2.48).
Since length
)
. ,,(n) @ [(n) wh£L[ ]h _[ ]~.
dim K (I) = 2, the Young diagram of
h I ~ h 2 a 0.
Of course
c [ ]h
has only two lines of
h I + h 2 = n; we will show that
hI - h2
is the
chemical valence. The diagram of It has two columns and
hI - h2
[ ]h
in
h I ~ h 2 ~ 0.
lines of length
l(n) -[ ]h
is the one symmetric through the diagonal.
In other words, it has
i.
h2
lines of length
2
That means that it cannot be completely symmetri-
cal in more than two electrons, i.e., there can be only two electrons at most in each orbital state; then two electrons must have "different spin states", or more exactly, their spin-state has to be antisymmetrical. discovered by Pauli
(Z. Physik,
FIGURE 2.2.
YOUNG DIAGRAM OF THE ORBITAL PART OF n-ELECTRON STATE:
2.7.
This is the Pauli principle,
31, 765 (1925)).
l(n) ~[ ]~
HERE
n = 21.
Atomic Shell Structure - Periodic Table
We can now clearly describe the lowest state of an n-electron atom. orbital part is a vector of
~(n)
The
obtained by filling all the lowest energy states,
putting only two electrons in each electronic orbital state.
Of course, in atoms
75
with
n > i
appear.
electrons,#
the two specific properties of the
The number of bounded states is finite, and the
exists, i.e., states with the same
n
and different
%
I/R
S0(4)
potential disdegeneracy no longer
have different energy.
The
observed order of increasing energy for the states is given in Table 2.1.
TABLE 2.1.
THE ELECTRON STATES ARE LISTED IN ORDER OF INCREASING ENERGY
n
i
2
2
3
3
4
3
4
5
4
5
6
5
4
6
7
...
%
0
0
i
0
i
0
2
i
0
2
i
0
2
3
i
0
...
Spectroscopist Notation
is
2s
2p
3s
3p
4s
3d
4p
5s
4d
5p
6s
5d
4f
6p
7s ...
2(2% + i)
2
2
6
2
6
2
I0
6
2
i0
6
2
i0
14
6
2
Total
2
4
i0
12
18
20
30
36
38
48
54
56
66
80
86
88 ...
Note that for a given
n, the energy increases with
predicted by computation and can be qualitatively understood. charge tial
Ze
and
k
electrons.
%.
...
This is of course
Consider a nucleus of
At infinity an electron will feel the coulomb poten-
(Z - k)e/r, but when it gets near, the probability of finding some of the
electrons at the distance
r ~ n(Z~) -I ~(mc) -I
is not negligible,
electron feels a potential above the Coulomb potential. ence if it is in a than in a
~ = 0
(k + i) th
It feels less the differ-
state (more concentrated inside the sphere of radius
r)
% = i state, ~ ~ 2 state, etc., where the concentration of probability is
more and more on the surface of the sphere. (% = i)
the
k
Note also from Table 2.1 that "p shells"
are filled by the 10th, 18th, 36th, 54th, 86th electron and this is just the
atomic number of the "rare gas" elements, very inactive chemically, Neon, Argon, Krypton, Xenon, and Radon.
We can even give the Mendeleev's periodic table, in
terms of electron states, if we know the order of increasing binding energy of the states
(n,%).
Using the spectroscopic notation %=0
1 23
4 5 6
s p d f g h i This order is, from Table 2.1, (n,%): 6s, 5d ~ 4f, etc.
The sign
~
(++) sign indicates
i
is, 2s, 2p, 3s, 3p, 4s ~ 3d, 4p, 5s ~ 4d, 5p,
indicates roughly the same energy so that the two
shells are filled simultaneously. +
(2.51)
Indeed, for the element of
% = 2
(d-shell) the
(or 2) more electrons in the d-shell, taken from the
# Because of the electrostatic repulsion among the electrons.
s
76
shell.
We can now construct a periodic Table.
TABLE 2.2.
%=if2(2%+i) = 6
~=0 2 electron s state I 2
n=l
~
n=2
Li
n=4 n=5 n=6
PERIODIC TABLE (UP TO ELEMENT 56)
~=3~2(2~+I)=14
%=212(2%+i) = io
P
.#
1
2
3
4
5
6
i!i C6 N7 Si
O8
F9
Ne I0
1
lI~ Ca 20
2
3
4
5
6
7
8
q-
9
i0
+
Sc Ti V Cr Mn Fe Co Ni Cu Zn 21 22 23 24 25 26 27 28 29 ~30
Ge As Se Br Kr 32 33 34 35 36
14 rare earths 58 to 71 I
~9 Zr ~a 40 %" 41 ~42 ~c 43 ~u 44 ~h 45 ~46 ~ 48
Rb Sr In Sn Sb Te I Xe La~Hf 37 38 49 50 51 52 53 54 57[72 Cs Ba 55 56
In d-shell, + (or ++) means (n + l),s) state).
1
(or
[ ~--|
2) more electrons in
n-d
(coming from the
The atomic state of an atom is labeled by the filled states, e.g., Oxygen: (is)2(2s)2(2p) 4, i.e., 8 electrons.
In general, the electrons fill up all states of
lower energy and fill incompletely the last "shell"; for example, in the case of Oxygen, we can add 2 more electrons in the
2p
state.
The question which arises is
which state is the atom ground state for an incomplete shell?
We can label this
state by a Young diagram (let us do it for the first elements). TABLE 2.3.
Z
=
Name
Y~iag of orbital
YOUNG DIAGRAM OF FUNDAMENTAL STATES
I
2
3
4
5
6
7
H
He
Li
Be
B
C
N
is F7
is~
1(z)
Y-diag of spin sta~e
(z)
[]
B
is 2s
is[] 2s
EF
2s 2p
FFP
2s 2p
2s I 2pl~
FFF
??
Now w e can generalize what we say for helium. energy state
(i.e.,
configuration
is the most antisymmetrical
sional Hilbert
electrons
in the coordinates
for the spin coordinates
in Table 2.3 for the 2p-electrons
is not possible
k
in the same
k ~ 2(2~ + i)), the lowest energy state of this
so it is the most symmetrical is illustrated
Given
to have a completely space,
for
antisymmetric
so w e can give the successive
TABLE
2.4.
nb of electrons k = 1 filled shells
np-SHELL
2
k
electron
(k = 2, k = 3, ...),
(k = 2, k = 3, etc.). f = i, 2, 3.
tensor
on a
When
k = 4, it
2~ + 1 = 3
atomic states
of
p
This
dimen-
shells.
STATES
3
4
5
6
,a...
spin statefs~ll~
valence
I
i (or 3)
2(or 4)
3
2
i
O
n = 2
B
C
N
O
F
Ne
n = 3
AI
Si
P
S
CI
A
n = 4
Ga
Ge
As
Se
Br
Kr
n = 5
In
Sn
Sb
Te
I
Xe
The ionization
energy
increasing with
(energy necessary
k, as w e expect,
tron state, because Although
to extract one electron from the p-shell)
along a given p-shell,
it is the first one not completely the energy of ns-states
excited state of atoms with that is, an ns-electron
k = i
or
2
is lower is
goes to an np-state.
is
except for the four elec-
antisymmetric.
than that of the np-state,
k = i; (ns)(np)2; This increases
a low
k = 2; (ns)(np) 3,
the valence by two and
gives more bounded molecules. There would be a lot more to say, even from group theory, Mendeleev
Table.
For instance did you notice that the ferromagnetic
Co, Fe and also M n in alloys shall stop here.
about the
are together
in an incomplete
3-d shell,
elements
(Ni,
etc.?
But we
78
2.8.
Atomic States in a Given Shell - Spin Orbit Coupling
The Pauli principle, expression of the Fermi statistics, simplifies the study of atoms.
Instead of studying an n-electron system, within a good approxi-
mation we can (for not too highly excited states) consider
k
electrons outside a
closed shell which has angular momentum zero, electric charge distribution depending on the electron wave function).
(n - k)e
(with a
This closed shell can be
considered as a spherical potential and the Hilbert space of states for trons in an
n - £
k
elec-
shell is k ~(k) = A (~2%+i @ K2)
to a good approximation, metric tensor,
x A y A z ...
Example. = (~) of
for
~(k)
an electron state is a (k
(I) = d i m ~
using a complete set of observable
TABLE 2.5.
Spectrocopist notation
THE
@ K 2 = 6; and
dim~C (2) = 15
Each decomposable tensor
That is what the spectroscopist does
~2, ~2, ~2 = (~ + ~)2, Jz' that is the orbital
~, the spin angular momentum
and its projection on axis
order decomposable antisym-
(complete shell again).
can be given a name or a label.
angular momentum
kth
factors).
% = i, k = ~ d i m ~
k = 6, d i m ~ (6) = 1
(2.52)
s, the total angular momentum
J3"
IS(np)2-STATES IN ORDER OF INCREASING ENERGY
3P 0
3P I
3P2
iS
ID2
0
L and space symmetry
i ~
i ~
I ~
2 [-~
0 ~I]
S and spin symmetry
I b J-J
I t~--I
I
O
0
J = total ang. morn. O 2J+l=nb of states I
>
~-
-2
~-]
1 3
~-~
+
-I
Remark on Table 2.5. states) are below the
j,
2 5
+
i
~
2 5 O
Note that the space-antisymmetric
states as we already emphasized.
+
~
0 i =15 0
F-l H states
(P-
For the symmetric
states, the S-state which feels more the repulsion than the D-states, is above them. Why do the P-states appear in order of increasing effect that we can explain in th~ following way.
J?
This is the small spin-orbit
?9
The orbital state of angular momentum (e~/2mc)~,
L~
while the spin state of angular momentum
g(e~/2mc)~
with
g = 2
(see (2.49)).
produces ~h
a magnetic moment
produces
a magnetic moment
The interaction between the two magnetic
moments has for matrix element# e2c_~_)2 ~ • - 2 "mc < R3 For a state
lj,%,s>
(2.53)
>
the expectation value of
~ • ~
is easy to compute from
~2 = (~ + ~)2 = 12 + 21 • g + ~2 And the expectation value for state
lj,~,S,jz>
(2.54)
(when
Jz
is the eigenvalue of
Jz ) is ->
= ~(j(j i + i) - ~(% + i) - s(s + i))
(2.55)
where I% - s I ~ j ~ % + s~, j + s This explains
the value of
integers
m 0
(2.56)
in Table 2.5.
We have seen that for hydrogen
the
~ (Z~/n) 3, so the expecta-
tion value of the spin orbit term is
S> (Z~)2 Z~2 ~ -2n 2 h
.
This is an order of magnitude. trum
[(is)2(2s)2(2p)6](3s)
largest splitting,
~
n
n
Z~ 2 ~
10 -4
n
In the sodium atom (alcaline = hydrogen (fundamental
like) spec-
state) and the
[ ](3p)
state has the
i.e., nearly 10 -3 , so the very bright
3p - 3s
(yellow) ha-line
is a doublet.
2.9.
In Chapter
Spin and Euclidean or Galilean Invariance
2, Sections
2.7 and 2.8, we have mainly used spin as a new
degree of freedom for the electron. values
This new degree of freedom has two discrete
(often called "up" and "down" in the physics manual)
Hilbert space
K2, of complex valued functions
dimension
The atom Hamiltonian
2.
freedom,
i.e. , it is of the form
Equation
(2.50).
operators
The permutations
of the form
have well defined
I @ S(s)
c [ ]~.
# This is a short for
(2.48) is independent
of the
illustration
applied to
@%(L (n)
[Ix
~ K (n)
of
[]~"
electron spins are represented by
which commute with
The simplest
~ ~ ~
n
set, has
of this spin degree of
acting on the space
H @ I
so the corresponding
defined in a 2-element
~. %(k) +i[
.
]%
H.
So eigenstates
is the helium atom
@
K(k)
[ ]~"
of
H
can
(n = 2).
80
C [ ]X = [2] =
The set of states with =
~
was called orthohelium.
~
Since
dim K (2) = 3
and
dim
K (2) = i, helium states
m (which are tensor products
x A y
or
C = [12] [ ]k
was called parahelium, with
8
x V y
of different
(x # y
hydrogen states)
have the statistical weight (for instance in the Boltzman distribution of thermodynamic equilibrium) of 3 for parahelium and i for orthohelium; this explains that spectral lines of the former are three times more intense than the corresponding lines of the latter. Using the considerations of Chapter i, Section 1.4, on the relations between the unitary groups and the permutation groups, we could also consider the action of the group U(2) on the two-dimensional space K~I)." Its action ~(n) c on .,[ ]~ is through the factorial representation [ ]I' and this action is on ,(n)]~ @ ..[ ~(n)]~ ~(n) C c[ Hamiltonian
of the type
c) = I @ (~Ds) ; it commutes with the I @ (@D[ ]I
H @ I.
These two points of view are formally equivalent for the classification of quantum states, but the
SU(2) c U(2)
Euclidean or Galilean invariance.
Let
has a deeper meaning. G
It is related to the
be either group, and
covering, i.e., there is a surjective homomorphism
G % G.
G
its universal
(As we have seen in
Chapter i, Section 1.2, and will see again in Chapter 4, it is an extension of the relativity group which acts through a linear representation on the Hilbert space of states.
See also O'Raifeartaigh.)
~ SU(2).
product, where action of
There is also a surjective homomorphism
(In the Euclidean case for instance
G
SU(2) on
G = R 3 [] SU(2); D = semi-direct
is the covering of the rotation group.)
~(i); the one particle-state
This gives us the
+ ÷ ~(x,t;o) 6 L2(x,t) @ KO
transfoms
into (U(g)~)(~t,o) = p(~(~)-i • (x,t); ~(~)-i • o) Often physicists prefer to write equivalently integrable function K . o
P
of
x,t
~(I)
(2.57)
as the Hilbert space of square
with value in the two-dimensional Hilbert space
Then (2.56) reads --
÷
(u(g)~)(x,t) --
l
E=I,2DI/2(~0(g))oj~j- (~(~)-i • (x,t)÷)
To s u m m a r i z e , t h e s p i n i s r e l a t e d
to (essentially
(2.58)
the rotation
part of)
Euclidean (and a portion of the larger Galilean) invariance; and it is an intrinsic angular momentum for the electron.
We will study it in 4.5.
The value
g = 2
for
the corresponding electron intrinsic magnetic moment is, however, a relativistic effect (see Figure 2.5). Conservation of angular momentum implies only that + spin angular momentum) be a constant of motion.
In atoms
rately conserved to a good approximation only because Equation (2.48).
H
~ = ~ + S ~
and
~
(orbital are sepa-
is spin-independent (see
81 2.10.
Molecules
The interaction which binds neutral molecule (= coulomb)
atomic nuclei and
(or a charged molecular
interaction.
n
ion) is essentially
Instead of treating directly a
uses the Born-Oppenheimer fixed.
N
electrons
into a
the electrostatic
N + n
body problem,
one
approximation where the (heavy) nuclei are considered
Take for example the Hamiltonian
II for the two protons,
i = i, 2
of the hydrogen molecule
for the two electrons,
i ,+2 ÷2 1 ,÷2 ÷2 H = -~-~tPl + Pll) + ~-~tPl + p2 ) - e2( .
(subscript A = I,
rA, i = I÷r A - ~il , etc.)
~
)
E A=I,II rA,i i=1,2
2 2 e e + ~- + r12
(2.59)
÷
where
d = Ir I - rii I
a parameter
the distance between
in the Born-Oppenheimer
nuclei will be neglected).
When
the two hydrogen nuclei is considered
approximation
d
where
L~ 2) (r,r) ÷
x, y
However,
and the two-dimensional
are hydrogen states.
in this basis, for
diagonal when
d
other nucleus,
so
Hlh When
2Ipl.
- IPl ÷ ~. d
are
state
operator
HIh
s = s + s . x y
is not exactly
d
is
for benzene),
etc.
the two eigen-
When
d ÷ 0, e2/d
e2/d - IPl < 0, and a value of
spins form an antisymmetrical
x @ x, and
state.
Hydro-
(covalent bond) with electrons
in order to form a closed X1 - X2
qualitatively
of
Z. Phys~k, 44, 455 (1927)).
H
"spin" shell, as was
is the "valence"
and quantitatively
H 2 , H2S , H2Se are of the form
(the repulsion of the two
H~C H7
Hence,
and they have an energy difference
for which
type of liaison
XI - X2
(W. Heitler and F. London,
C2H 4 is flat
c ± IPI.
The number of atoms which can be bound to an atomic (spin)
Quantum mechanics has explained
H2S , H20 (= 108°).
are
The ground state is of the type
empirically before 1920, and
why the molecules
(2.60)
i/2(x V y)
is minimum.
form a similar
Xll I I I II X [ I [ I
cp (~¢)
and so does e2/d - IPI"
the two-electron
shell of atoms.
2 discovered
and
IPI ÷ 0
e2/d - IPl
gen (or alcaline)
=
and its eigenvalues
i/2(x A y)
d ÷ ~,
from Fermi statistics
unfilled
Consider first the space h = (x 0 y) ~ (y @ x)
These two states have the same energy
But there is a domain for
for which
~(2),
is finite because each electron feels also the attraction of the
(since it is Hermitian),
of
spaces. subspace
h, the Hamiltonian
Hih states of
(and the kinetic energy of the
is very large, a state of (2.59) is in
the tensor product of two hydrogen atom-Hilbert dependence
as
H --
of the atom.
the covalent bond
It explains, 0~
for instance,
with an angle
> 90 ° H atoms makes the angle increasing from 90 ° for H2Se ,
It explains why NH 3 is a trihedron and CH 4 a tetrahedron, /H
(d
and
~
electrons).
It explains mesomery
why
(e.g.,
--C~H Group theory is so useful for explaining molecular
spectral
We
82
have to skip this subject for now and simply refer to an elementary but elegant textbook, ~ a n ~
Chemistry, by Eyring, Walter, and Kimball, Wiley, New York (1944).
The symmetry group of a molecule is a subgroup of dimensional orthogonal group. group
G
is known.
0(3), the three-
When its shape is known experimentally,
Let us refer to Wigner's paper (GDttingen,
its symmetry
(March 1930),
p. 133) on the characteristic elastic vibration modes of molecule (given by the equivalent classes of
G), as examples of the application of group theory.
studied CH 4 (whose group
G
is
S(4)) as an illustration.
Wigner
The H. A. Jahn, E.
Teller theorem (~oc. Roy. Soc. Ser. A, 161, 220 (1932)) proves that the electron orbital state of "non-straight" molecules cannot transform as an irrep of dimension
> i.
(The irrep has dimension
2
G
of
for molecules whose atoms are on a
straight line.) We will study here only one very important example.
2.11.
Measurement of Spin and Statistics of Nuclei by the Study of Diatomic-Molecule Spectra
The Hamiltonian
H
of a diatomic molecule can be divided into
H = Helectronic + Hvibratio n + Hrotation + H ~ where, to a good approximation, H ~ tronic states of the molecule;
can be neglected.
two nuclei) which minimizes the energy. two nuclei are identical, a fraction of
2
Helectronic
each such state yields a distance
0(2) × Z 2.
gives the elecd
The invariance group is
(between the
0(2)
or if the
Binding energy for such states are typically
(few electron volts).
H
vibration is essentially the harmonic
oscillator Hamiltonian for small oscillations around the equilibrium position fixed by the distance d. 2 to a , and the H to
%(~ + i), ~
The equidistant spacing of the vibration level is small compared rotation yields for each
integer
d
also rotation energies proportional
m 0, and small compared to the vibrational energies
tional bands; in spectrum).
is the symmetry of the molecular state for the permutation group nuclei? its
S(2)
The symmetry depends only on the spin state of the nuclei,
SU(2)
(rota-
If the two nuclei of the molecule are identical, which of these two
(each of spin
j)
irrep is Ds, 0 ~ s ~ 2j;
[-I--] s = 2j, 2j - 2, 2j - 4 .... S
and the rotational state of the system, Since
H
2j - i, 2j [7-]
for
3, ... ~
even,
H
for
~
odd.
is independent of the nuclear spin (to a very good approximatio~
the symmetry character of the nuclear spin state is a constant of motion (with often a lifetime of weeks) and is, as for Nelium, called ortho or para. "statistics", motion.
Because of
the symmetry character of the rotational state is also a constant of
So the rotational spectrum of the molecule divides into two independent
83
sets of transitions - those between even
%, and those between odd
tions occur in both states as quadrupolar energy
~ (~ + 2)(~ + ~ -
The number of
(2j + l)(2j + 2)/2 = (j + l)(2j + i).
states is
(2j + i)(2j)/2 = j(2j + i). (rotational)
lines is
The transi-
~ + 2 + ~, with a (radio-wave) photon
~(~ + i) = 4~ + 6.
states is
sity of spectral
~.
~
nuclear spin
The number of
nuclear spin
~
So if for the molecule the relative inten(2j integer
~ 0), j/j + i
% even/% odd, h r the nuclear spin is j, the statistics of the nuclei is I H x ~I/I~-~ x HI ~L_a ;.Jj ) I = Fermi; if it is j/j + i for ~ odd/% even, the nuclear spin is j the
Ia mentally,
for
xper -
=
only Fermi statistics is found for half odd integral
tron) and Bose statistics for integral
j.
j
(as for the elec-
We will summarize this important experi-
mental fact by statistics = (-i) 2j For instance when only even
~
(2.61)
rotational states exist, we conclude that
j = 0,
and the statistics has to be Bose. Historically, 598 (19309 was
the first nuclear spin measured
(F. Rasetti,
(in 1929) that of N14 (nitrogen molecule N-N).
Z. Physik, 6__!1,
Rasetti found
j = 1
and Bose statistics. But it was then believed that the universe was made of protons + p , electrons e , and photons y, (the only particles then known, and that the nucleus N14 of charge 7e, contained 14 protons and 7 electrons,
thus, half integral
spin and Fermi statistics were expected. This measurement started a crisis in physics. Appendix.
On Commutation Relations
Professor Bargmann pointed out to me that I have spoken of the invariance group of the commutation relations only in the context of rotational invariance Equation
(2.10)).
Surely it is worth mentioning
the general case:
(see
consider the
relations [Pi,Qj ] = ih6ij~ (i,j = 1
to
n).
Let
operator notations
(2.61)
a = (al...an) , b = (bl...b n) E Rn; we can use the tensor
P(a) = EiaiPi, Q(b) = EjbjQj.
Equation (2.61) defines a
dimensional Lie algebra which is a central non-abelian extension R1
(center of
on
R 2n
where
=
Rn
~
~).
~
of
2n + 1
R 2n
by
This extension is defined by the antisymmetrical bilinear form
Rn
o(a @ b, a e • b e ) = a • b t - b • a I
a • b = E.a.b..
The symplectic group
Sp(n)
(2.62)
which leaves this form invari-
i i i
ant is a group of automorphism of
~.
The corresponding simply connected group G has, up to an equivalence, a unique unitary irrep (von Neumann's theorem. Its Schr~dinger realization as operators on the space L 2 of functions of n variables: x = (x I ....x n) is U a = e iP(a) with
(Uaf)(x) = f(x + a); V b = e iQ(b)
a E En, b E E'n dual of the rotation group
SO(3)
En.
with
(Vbf)(x) = ei~b'Xf(x).
In the case of Equation
(2.10)
n = 3.
Here
Furthermore,
leaves invariant the symmetrical linear form
and w e used the corresponding identification of
E3
and its dual.
x,
@
on
E3
8#
3.
N U C L E A R PHYSICS:
3.1.
The n u c l e i have
protons
charge and
N
unstable
m
p
+ e.
excited
and the n u c l e u s
the f o l l o w i n g
which
may transform
B+-decay
÷
it b y
(Z,N);
~-decay*:
spontaneously
two p a r t i c l e s
A = Z + N nucleon~
The g r o u n d state
itself m a y be
into a n o t h e r u u c l e u s
seconds
+ ~(v = a n t i n e u t r i n o ) ;
(Z - I,N + I) + e + + ~
T
can vary
(Z,N) ÷
(Z,N) ÷
which
b y one of
(Z + I , N - i) + e
competes w i t h
(which r e q u i r e s
+
e--capture
less energy).
from 10 -3 sec to 1020 years.
(Z - 2,N - 2) +
(2,2)
for
A > 140
nuclei,
T
from
to 1020 years.
A n d two m u c h rarer
types:
c)
neutron
(Z,N) +
d)
spontaneous
emission: fission
into
Let us call n u c l e i 274 stable n u c l e i
(Z,N - i) + n
rare,
two s m a l l e r n u c l e i
r < few seconds,
(Z,N) ÷
(ZI,NI) +
if they. have a h a l f life of d e c a y
stable
(Z2,N2). r > 1020 years.
are known.
even
N
even
165
A even
Z
even
N
odd
55
Z
odd
N
even
50
A odd Z
odd
N
The m u c h in Figure
odd
greater
3.1 w h i c h
and for a g i v e n
N
have definitely
This is also
distribution
(Z = N = 1,3,5,7)
abundance gives
of
to n o t e
even,
N
even n u c l e i
of stable n u c l e i
from Figure
m o r e isotopes
is
Pb208,
of stable n u c l e i
and
Z
the n u m b e r
true, but less strikingly,
stable n u c l e u s
N - Z m 0
4
is s t r i k i n g l y
for g i v e n
Z
illus-
(isotopes)
(isotones).
It is w o r t h w h i l e (N = 82)
heaviest
These
it contains
are unstable.
(Z - I,N + i) + ~
The m e a n life
trated
n.
--
+ e
(Z,N) ÷
(Z,N) + e
Z
and n e u t r o n s
types of decay.
B--decay n ÷ p
b)
and d e n o t e
states,
+
a)
p
= 1 8 3 6 . 1 0 m = 938.25 MeV, m = 939.55 MeV. The pro~ton has e n Both have spin 1/2. We d e f i n e a n u c l e u s by its n u m b e r Z of
of neutrons,
Nuclei have bound
The Set of K n o w n N u c l e i
are m a d e of p r o t o n s
similar masses
electric
S T R O N G AND W E A K I N T E R A C T I O N S
N - Z
for
N - Z = -i
for the p r o t o n
N - Z = 0
for 13 nuclei;
N - Z = 1
for 16 nuclei;
N - Z
increases
of
Z
increasing (Z = i)
Z = 20, Z = 50,
(and also
Another
and
function
and
for
than their e v e n - n e i g h b o r s .
N = 20, N = 50
Z = 82, N = 126.
in f u n c t i o n
is a slowly
3.1 that n u c l e i
(or isotones)
N
striking
is that w i t h of
N = 28). feature
two e x c e p t i o n s
A = N + Z:
H e 3 ( Z = 2)
#
# What was (2,2).
first
called
with
A
an m - p a r t i c l e
on the a v e r a g e
(N - Z) N 6.10 -3 A 5/3.
has b e e n i d e n t i f i e d
with
The
in the
a H e l i u m nucleus:
85
50
20 71
Number of isotones
82
vN
6 5 4 3 2 i 0 I i0
I 20
I 30
I 40
I 50
I 70
I 80
I 90
I i00
! Ii0
i 120
." N 130
50
20 i0
] 60
Number of isotopes
~Z
9 8
7 6
even z
5 4 3 2 odd Z i 0 I
I
I
I
I
I
I
I
I
i0
20
30
40
50
60
70
80
90
FIGURE 3.1.
NUMBER
vN
AND
~Z
OF STABLE NUCLEI
(Z,N)
Note that there are no stable nuclei for Z N = 19, 21, 35, 39, 45, 61, 89, 115, 123, > A = N + Z = 5, 8, 147, m 208. The heaviest 126 82 Pb208 with Z = 82, N = 126. The most Figure 3.1 is that
vZ
and
vN
I
I
I
i00
ii0
120
AS FUNCTIONS OF = 43, 61, > 83, for 126 and none for stable nucleus is striking feature of
are mainly i, sometimes 2 or 0
for odd Z or odd N. Their value is more irregular for even Z or even N, there are relative maxima for Z = 20 = N, Z = 50 = N, N = 82 and also N = 28.
Z
>
Z
AND
N
86
Nuclear
forces are the most intense
208 nucleons
together#.
in nature,
however,
they do not bind more than
The reason is that they have short range and also that nu-
cleons obey Fermi statistics
which,
for condensed
states,
yield effects
similar
to
repulsion.
lifetime
More than one thousand different unstable nuclei
are known.
T > .i
are found in nature,
(age of u n i v e r s ~ and their decay products
are still produced have Z-even,
in nature
N-even.
(C14).
There exists
All other are man made; more a semi-empirical
formula
and
nuclei
giving
in function of
A = Z + N B(Z,N)
= Zm
p
+ Nm
n
- m(Z,N)
= U A - U Z(Z
~
c
- I)A -I/3
- U A 2/3 - U -(Z -- N) 2 + U (-l)Z + (-I)N A -3/4 s t A p 2 the values of the constants U
U
= 14.0 MeV,
corresponds
sponds
U
to the Coulomb
N
c
=
are in
.61 M e V ,
A I/3" .
to pairing
U
s
among
= 14.0 MeV,
Z
The term
effects
protons U
MeV,
Ut
p
=
favors
34 M e V
The term
U
corre-
c
in a sphere of
to a surface effect which
a minimum
for
IZ - N I
while
As we saw, nuclei with even
than those with odd Z
U
equally distributed
in like nucleons.
is that all known
even, N
Z
and/or odd
N.
Z A
even nuclei have zero spin
at rest).
The distribution
of nuclear spin for odd
A
nuclei
is discussed
in
3.4.
3.2.
Isospin
As soon as the neutron was discovered language
U t = 84.2
corresponds
s
forces;
are more stable and more numerous
(= angular momentum
(3.1)
MeV
average binding energy by nucleon.
repulsion
to
rule without exception
Section
U
a short range for nuclear
corresponds
P and
U
to maximal
radius proportional suggests
some
than half of those
(Weizs~cker's)
the binding energy of the lowest state of (stable or unstable) Z, N
Those with a
for the study of nuclei.
particles,
the nucleons,
Neutrons
(1932),
and protons
Heisenberg
created
are considered
which have five degrees of freedom:
3 continuous
(~)~ a two valued one, o, for the spin and a new one that Heisenberg the fifth degree of freedom,
# The existence
T, and which distinguishes
neutrons
a formal
as the same in space
simply
called
and protons
of neutron stars with a radius of i0 km to i00 km and containing
N 1057 neutrons has been postulated. These stars seem to be observed now as "pulsars". They are indeed gigantic nuclei, but the binding energy is due both to nuclear and gravitational forces.
S?
(Z. Phys.j 77, 1 (1932)); since, like the spin, it is two-valued,
it is now called
isospin.# This Heisenberg
convention has revealed itself more than useful.
it was quickly established and neutrons:
reasonable
that nuclear forces did not distinguish between protons
their differences
small mass difference)
(different electric charge and magnetic moment,
are attributed mainly to electromagnetic
approximation
Indeed
to neglect
effects and it is a
them.
If we denote the Hilbert space of our nucleon states by
~(i) that of a number
A
of nucleons
~(A)
where
P[1A]
(3.2)
is
(i) = ~ [ i A] = e[l A]
is the projector on
@ ((L 2 ~ K ) (A) % [1]
~(i)
.
0
A convenient
~(A) c) "T[%]
(3.3)
approximation
for the study
[1 A]
of a nucleus of
A
nucleons
an average potential potential,
÷
= L2(x,t ) @ Ko @ KT
is to replace the sum of 2-particle
(= sum of 1-particle Hamiltonians)
which is still attractive.
allows us to draw qualitative
interactions,it
by
plus a residual 2-particle
Then the analogy with the study of atomsttt
conclusion.
Using the same type of argument as in
Section 2.6 for atoms, but here with the opposite sign, we know that for the ground state
[%]
in Equation
antisymmetrical
(3.3) should be as syrmnetrical as possible,
as it can be with the restriction
so
[%]c
that it has only two lines.
is as This
implies that the two lines are as nearly equal as possible %i ~ 0; [%]c = [%1,%210 ~ %1 - %2 = I If the nucleus has
Z
pletely symmetrical
in
protons and sup{Z,N}
N
neutrons
particles,
minimal;
(Z + N = A)
%1 + %2 = A
(3.4)
its states are com-
so
(3.5)
%1 ~ sup{Z,N} and
IZ - NI ~ %1 - %2
'
(3.6)
Called isotopic spin since 1936, the name isobaric spin would have been more proper. Anyway it has been shortened into isospin by the natural evolution of language. tt
In fact physicists are more sophisticated: when a sum of 2-particle interaction does not yield a good enough approximation, one adds also the sum of all kparticle (2 < k ~ A) interactions, mainly for k = A (collective effects).
#t% There is still a difference. Atoms of n electrons consist of n + i particles and as we have seen, the elimination of the center of mass motion is easy: one singles out the nucleus, and the electrons are all treated on the same footing. This elimination is still clumsily carried out in nuclear physics.
88
So (3.4) can be translated possible.
As we have seen,
tromagnetic account
repulsion
0 < N - Z
into:
the most stable nuclei have as small
this is well verified
of protons
in negligible;
In the same approximation bars
(nuclei with the same number
Consider
in which A = Z + N
for light nuclei, where
when this repulsion
has to be a slowly increasing n, p
]Z - N]
function of
the elec-
is taken into
A = N + Z.
are considered
of nucleons)
as
identical,
iso-
should be identical.
Figure 3.2; it gives the energy spectrum of the known states
for
A = 15,
/2
>=
unobserved
.P 3
C 15
NI5
015
FI5
Z=6
7
8
9
N=9
8
7
6
is the spin
(j, 1/2 integer
FIGURE 3.2.
> 0)
and the parity
(+-)
of the state.
SPECTRUM OF STATES OF ISOBARS 15
States of C 15 have isospin m 3/2. Another conventional notation for nuclei is to use the chemical symbol of the corresponding atom (this gives implicitly the number Z) and write the number of nucleons A = Z + N in superscript.
89
and the known value of spin and parity of these states. spectra (at least for the low lying part) for nuclei) is striking. N 3 MeV
for
015
The similarity of the
1/21Z - N I = 1/2
(i.e. N 15
and
015
The essential difference is a shift upward in energy of
which has one more proton than
N 15.
The pairs of corresponding
states are called "doublets" of isospin 1/2 states. Let us explain in detail this point of view, which exploits the relations between the permutation groups
S(n)
and the unitary groups
U(k)
that we have
recalled in 1.4 and used in 2.9. Nuclear interaction does not distinguish between protons and neutrons. For a nucleus this implies a property of invariance under the permutations of its nucleons.
(E S(A))
We could also have translated this property as follows:
All nuclear physics observables
0
acting on
~C(I)
(space of the one
particle states for the nucleon) in Equation (3.3) are of the form (3.7), where ~(i)
= L2(x,t ) @ K
0 K
(3.2)
T
0 =
N
@ I
(3.7)
U(2) =
I
@ U(2)
(3.8)
i.e., they correspond to a trivial action on
K , the factor in this tensor product
which corresponds to Heisenberg's fifth degree of freedom "proton - neutron". The action of the group with every observable: observables".
The action of this
Nilbert space of the
U(2)
on
~(i), as defined by (3.8) , commutes
U(2) c {0} S , the commutant of the algebra of "one particle
A
U(2)
particle states.
can be extended to every
Therefore, in nuclear physics, when the
non-nuclear interactions are neglected, this group. S(A)
$C(A) and
$C(A) , (A ~ 0),
U(2)
is a subgroup of the invariance
has the same decomposition into spaces of factorial representation for
U(2)
and we use the same symbols (Young diagrams) for the corresponding
representations. Since Coulomb repulsion of the protons can be neglected only in light nuclei, it was not~expected that isospin conservation could be an interesting concept for heavier nuclei.
However, the progress of nuclear physics in the last five
years has shown that for nuclei with cept.
A
up to i00, isospin is indeed a useful con-
For a non-technical review of this question, see W. R. Coker and C. F. Moore,
"Isobaric Analog Resonances",
Physics Today, 3.3.
U(4)
22, no. 4, 53 (1969).
Invariance
In 1936 Wigner, in his paper "On the consequence of the Symmetry of the nuclear Hamiltonian on the Spectroscopy of Nuclei",
# Reproduced in Dyson's anthology: Benjamin, New York (1966).
Phys. Rev.,
51-106 (1937)t
Symmetry Groups in Nuclear and Particle Physics,
9o studied the approximation where not only isospin dependence of nuclear forces is neglected but also the spin dependence.
Then Equation
(3.7) and (3.8) can be re-
placed by 34(I) = [2(x) 8 K
In this approximation, U(4)
A
(3.2)
@ I
(3.9)
0 =
N
8 I
U(4) =
I
@ U(4)
(3.10)
nuclear theory is also invariant under the group
acting on the four dimensional
Hilbert space of
8 KT
space
K = K
@ K
and Equation
T
(3.3), for the
nucleon states can be replaced by ~(A) = $C(I) = p ~% ® K [iA] [iA] (L21%] [%]c )
where the
[%]
are representations
of
For the most stable states, two-nucleon possible,
force to be attractive
so
[%]c
U(4)
(acting on
as possible,
lowest lying states is
0, isospin 1.
%1 + %2 + %3 + %4 = A) as %1 = %2 = %3 = %4 = A/4. of
(Z < 17)
For nuclei with
A = 4n + 2, the
SU(2) × SU(2)
representations:
[%]c
representation
(spin i) and
He 6
decomposes
into the direct sum of
n = i) this gives correctly the spin of the and
Be 6
(spin = 0).
These last two
(spin 0 +) of
Li 6.
levels Whose spin are marked in Figure 3.3 belong to another equivalent U(4)
with an angular orbital momentum
degree of freedom) j = Z = 2
and
Z - s ~ j ~ Z + s Li 6
i.e., = j
with no correspon-
Be 6.
States belonging literature.
representa-
(i.e., angular momentum of the space
for the spin i, isospin 0 states i.e., those of He 6
The other
So the total angular momentum has the possible value
for the spin 0, isospin i states and
= 3, 2, i dents in
% = 2.
of
(~) = 6.
one of spin i, isospin 0, the other of spin
levels form an isospin triplet with the third level
tion of
U(4)
where Coulomb repulsion of
%1 - i = %2 - i = %3 = %4 = n; it has dimension
to the subgroup
Li 6
i.e., its Young diagram has
SU(2) × SU(2), yields a spin 0 and isospin 0
In Figure 3.3 (for which
lowest state of
as
As we have seen, the former result is observed for all such
protons is not too large.
two three-dimensional
(with
is as symmetrical
A/4 = integer this implies
the latter only for light nuclei
Its restriction
[%]
The restriction of this representation
Ko @ KT) to the subgroup
for the ground state. nuclei,
For
irrep has dim. i.
(used in 3.2) of the "residual"
implies now that
%1 a %2 a %3 z %4 a 0
nearly equal as possible. This
U(4). the property
is as antisymmetrical
its four lines of length
(3.11)
to a
U(4)
irrep are called supermultiplets
The study of Galilean invariance
in physics
of the theory of supermultiplets
very similar to that made in 2.9 for atomic physics.
is
94
2+
1+ 2+
0+
2+
,
2+~'"
0+
0+
3+
5 MeV
i+
He 6 Z=2
Li 6
N=4 jP
Z=3
N=
Be 6 3
Z=4
N=
2
is the spin (j integer a 0) and the parity p(!) FIGURE 3.3.
of the state.
SPECTRUM OF LEVELS OF THE NUCLEI WITH 6 NUCLEONS
7[
of the Galilean group
One has to consider the covering
G, ~ ÷ G
and
also the homomorphism
SU(2) x 1 c U(2)
÷
x
U(2) c U(4)
The invariance group of the theory is the direct product subgroup
G~
G × U(4)
with
G x U(4)
and
G
is the
i(g) = (~(g),~(g)).
The approximation of spin and isospin independence for nuclear forces leading to
U(4)
invariance is crude and could not be expected to be very useful
for nuclei with a number
A
of nucleons not very small.
approximation of isospin conservation, nuclei with
A
ground state: plet model",
U(4)
However, as for the better
invariance has been usefully applied to
up to i00 as shown by the statistical study of the energy of their P. Franzini and L. A. Radicati, "On the validity of the supermulti-
Phys. LGt%., 6,
in 3.3 and 4.)
322 (1963).
(Reproduced in Dyson's anthology, quoted
92
3.4.
Shell Model
We have seen that nuclei with seem more stable.
Z
or
Many properties of nuclei
N = 8, 20, 50, 82, and
(e.g. binding energy per nucleon, mag-
netic and quadripole moment) single out these numbers. shows that for
A survey of nuclear tables
A = odd nuclei, the parity and spin of the ground state vary in a
regular pattern which suggests very much the filling of shells physics).
Ground state, spin
number
or
Z
N = 126
N
j
and parity
±
(as in atomic
depend only on the value of the odd
so the order of the levels, with increasing energy, is the same
for protons and neutrons.
The order obtained can be deduced from the three-
dimensional harmonic oscillator spectrum
E
n
= n~
+ Eo
(where
~
is a constant)
with some modifications. Consider the set of nine operators Qj (i = 1,2,3)
Tij = PiPj + QiQj
where the
P'l
and
satisfy the canonical commutation relation [Pi,Qj] = ih~ij
Its use for the computation of the commutators form a representation up to
ih
[Tij,Ti,j,]
of the Lie algebra
U(3).
shows that the
Tij
The center
H = TrTij = TII + T22 + T33 is the harmonic oscillator Hamiltonian deduce that there is a tor similar to the spectrum of
H
U(3)
S0(4)
n
is
From
[H,Tij] = 0
we
invariance for the three-dimensional harmonic oscilla-
invariance of the hydrogen atom, 2.4, and we can find the
by a similar method:
The ground state E
(in convenient units).
n = 0
i/2(n + l)(n + 2).
= (-i) n, 0 ~ Z ~ n.
has energy
Eo.
The number of states of energy
Their orbital angular momentum
~
satisfies
This corresponds to the first column of Figure 3.4.
(-i) £ The second
column gives the spectrum of the Hamiltonian
~-lH' with
m, J , m tt
s = 1/2 > 0
and when
= ~H - , ~ 2
positive constants, w > ~ s
and
~ > 0, j = ~ + c 1/2, ~ = ±i
En, j ~ - E 1 = n~ - t ( ~ ( ~ ' 0 '7' 0
= 0 This Hamiltonian
_ a~''~
• ~
jt .
(3.12) Using Equation (2.55) for
one obtains the energy spectrum
1 + i) - ~ it(~ + 7) , s = sign(j - ~)
= nw Ht
(3.13 t)
is the one-nucleon Hamiltonian in the average potential pro-
duced by the whole nucleus. protons and neutrons.
(3.13)
As in 2.7 we can now "fill the successive shells" for
Such shell-model for nuclei was proposed in 1949 (see
M. Goeppert Mayer and J. H. D. Jensen, Wiley, New York (1955)).
Elementary Theory of Nuclear Shell Structure,
It is very successful in explaining the properties of the
95
n
Z
Enj ~ - Eo Z
6
5
4
or
~
11/2 +
6
i"~s
g/2 +
1 3
126 ~ ~.<__ I~ ,
1/25/2.13/2 + _ 3/2g/2 F/2-
5
I
4
// - 82
3/2+ 11/21/2 + F/2 +
.I
/
~'~
5/2 + 50
~ i
g/2 +
----~--- <"/ ~ f - -
3
~.
2
N
0 2 4
0 /~ 2--'---'----~ -I
3
jP
0 2
_ 112-
5/2 3/2-
28
. F/2-
20 - I ~ - -
3/2+ 1/2+
~'~.
512+
1/2
i
_
i
_
s
I" 3/2
0
0
2 1/2 +
w' = ~
jP
II
= 0
J
= 0
is the spin (j half integer > O) and the parity
FIGURE 3.4.
(±)
of the state.
ENERGY SPECTRUM OF THE ONE NUCLEON HAMILTONIAN OF EQUATION (3.12)
94
low lying levels of nuclei, structure is that of
able antisymmetrical
of different
Z
(or
N) particle state is to a lesser degree a decompos-
tensor than it is for atoms
(see 2.8).
There is much more con-
i.e., the nuclear states are tensors which are linear combination
decomposable
Shell-model "same shell".
A more detailed book on nuclear shell
A. de Shalit and I. Talmi.
Note that a
figuration mixing,
and their decays.
tensors with the same quantum numbers.
computations
require a classification
of states inside the
(Phys. Rev.,
That was done between 1942 and 1949 by Racah
(1949)) who introduced many references
the seniority quantum number.
76, 1352
(See also work of Flowers and
in Dyson's anthology quoted in 3.3 and 4.3.)
Part of the problem is to label unambiguously
irreps
D. of SU(2) ] into a direct sum of irreps of the tensor power of a
appearing in the decomposition n given irrep 0 D. (where J is usually integral for atoms, half integral for J nuclei). The method is to find a chain of subgroups Ho = U ( 2 ~ + (where
SU(2),
(2J
by the matrices of
i) D H I = ... D ~
odd) or
SO(3),
(2J
D SU(2)
or
SO(3)
even) is the subgroup of
(3.14) U(2J + I)
The successive
restrictions
of the representation
of
U(2J + i)
n ® [] = ®% sh[h] (where
~h
is over all Young diagrams
the corresponding direct sums of example
irrep of
S(n))
of
n
only multiplicity
one.
U(5)
to the different
irrep
(see (l.15t))
squares and
SU (2) irreps with multiplicity
J = 2, n = 3, the
necessary
[I~
is the dimension of
Hi, (0 ~ 1 ~ k) must finally yield
one. and
sh
Table 3.1 illustrates ~
This is not the case for ~ .
restricted
to
a simple
SU(2)
One intermediate
DECOMPOSITION
OF
3 @ D2
(3 PARTICLES
IN D-SHELL)
[] representation
of
U(5)
dimension U(5)~ Restriction
(F~) 3 = ~
• 2 ~
+
=
one irrep
=
A @ B
S0(5)
Restriction ~I~
one irrep
to
S0(3)
= D6
D4 @ D3 @ D2 @ DO
A = D5
D4 @ D3 @ D2 @ DI
B = D2
=
H
53 = 35 + (2 × 40) + I0 to
~
= D3
DI
yields
group is
H I = S0(5).
TABLE 3.1.
Irrep of
formed
Dj) with the following property:
95 Neglecting the SU(2)
sh
multiplicity, every inequivalent irrep of
SO(3)/or
has a different genealogy of irreps of
ferent seniority quantum number.
H.. Racah says that they have difi For distinguishing the different seniorities Racah
had to introduce the exceptional Lie group n = 7)!
When
J
G2
among the
is half integer, one generally takes
H.l
(in the case
H I = Sp(2J + i).
nuclear physicists nowadays use more refined models of nuclei (e.g.,
Structure
I, ff, fff, by A. Bohr and B. Mottelson, Benjamin).
details is outside the scope of these lectures. paper by D. R. Inglis:
Of course,
Nuclear
To go into more
We refer again to a non-technical
Physics Today,
"Nuclear Models",
J = 3,
2 2 no. 6, 29 (1969) for a
recent survey.
3.5.
Although
SU(2)
The Hadrons
invariance, through isospin considerations, is more
familiar to physicists than
S(n)
invariance for the study of the property of
nuclear interaction not to distinguish between neutrons and protons, is it more fundamental?
If one had to deal only with nucleons, the answer is no; both mathe-
matical methods are physically equivalent.
However, there are many more particles
with strong interaction; they cannot be permuted with the nucleons but they can be attributed an isospin.
Let us give as example the ~-meson.
In 1935, Yukawa pre-
dicted the existence of mesons which are to the nuclear interaction what photons are to the electromagnetic interaction.
He predicted their electric charge
mass, their lifetime, their decay mode,
Soon the particles were discovered but it
was a case of mistaken identity with the ~-lepton! covered in 1947 and is called
~±.
±, their
The Yukawa particle was dis-
In 1937, physicists (e.g., Kemmer) showed that
3 states of charge were necessary for the meson, +, O, -.
Indeed, in order that
nuclear interaction preserve isospin, they have to be invariant under the corresponding
SU(2).
current.
This current transforms under
the representation
In Yukawa's theory the meson field is coupled with the nucleonic
(DI/2)2 = D 1 @ D 0.
SU(2)
as a tensor operator of the space of
Then the simplest
SU(2)
invariant Yukawa
coupling which can include electrically charged meson, is of the form
j(x) where
j(x)
and
~(x)
• +(x)dx
(3.15)
are vector operators for the isospin
interaction is the scalar product of these vectors.
SU(2)
and the
The T°-meson so predicted in
1937 was found in 1950. Already in 1947 two other strongly interacting particles had been found. The generic name "hadron" was given to particles with strong interaction.
The rate
of discovery of new hadrons has passed from 15 in the fifties to 250 in the sixties. We give their mass spectrum and their spin and parity when known, in Table 3.2;
96
TABLE 3.2.
JP
J=spin
SPECTROSCOPY OF HADRONS
p =parity
GJ P
G =isoperity
Baryons
Mesons
(Anti-Baryons ore not shown)
I I
Charge conjugation: 8*~--B, Y-~*-Y, Q.,-,.-Q
Charge cqnjugation
SeV
-3
11/2" 712-
7/24
.......7/2-
-
7/2;
5/24 512- ~:::i15/2+ ,~3f23/2~------112-
--3/24?
!
3/2+
"'1/2+ ::'-:--1/2~. 3/2-, ........ ':::::--5/2 1/2"512. . . . . . . . 3/2 . . . . . . . . . 1/2-
__:-----.3/2-
3/2" '- --------1/2 . . . . . .
1/2
......
.i.2"
__
4.
1/24 .... ~-2*
3/24
1/24.
41*?__ -1+
. . . . . . . . 1/2" n p ...... 112+
+I-
q~ x
....... O+ -1+0-
oJ Q q
-0-
2+ _
_
1 * ? - -
1-
-I+O-
O-
:rr,
0 0
112 -O"
-2
0 or 1 -0+
112 or 312 -0+ +*
Q
1/2 -0
0 or 1 -0+
0 I
+1 I
Y B
-1 0
0 0
-I
1
I
Q
E
Name E N various (Compiled from data of Rev. Mod. Phys., January, 1969.) --
A
112 O+ +1 0 K
97 different columns of this table correspond to different values of the quantum numbers preserved by the nuclear interaction also called strong interaction. There is a charge
b
which is for the strong interaction what the electric
charge is for the electromagnetic they are called baryons.
interaction.
The "charged" particles have
The "neutral" particles
b = 0
b = ±i;
are called mesons.
It
happens that all baryons have half integral spin and all mesons have integral spin. This will be explained in 4.2 (Equation 4.8). buting an isospin to every baryon.
There has been no difficulty in attri-
Particles inside the same isospin multiplet have
the same spin and parity, their masses are equal within 1% (exceptionally 3% for the q-mesons) and they have different electric charges.
Isospin conservation allows us
to predict some missing members of a multiplet which are then looked for and have always been found. 2t + i
Instead of using the value of their electric charge
q
states of the same isospin multiplet can be labeled by the value
isospin
SU(2)
different;
generator in the direction "3".
the difference
q - t3
of the
These two labels are equivalent but
depends on the multiplet;
are conserved by strong interaction,
the t3
since both
q
and
t3
their difference y = 2(q - t3)
(3.16)
is a new strong interaction quantum number which can be attributed to each isospin multiplet*;
it has integral values, and it is called the hypercharge.
one had enlarged the will see how
U(2)
SU(2)
invariance group to a
U(2)
groupt#,
sponding baryon what antiproton is to proton
each one is to the corre-
(see Dirae quotation in introduction).
Antibaryons are obtained from baryons by the involution called sign of the charges C
with values
in 5.1, we
was enlarged further.
In Table 3.2 we have left out the antibaryons,
detail on
To include it,
b, q, y
which changes the
and leaves invariant the mass and the spin (for more
and its relation with parity, see 4.6). t, y, b
C
To each isospin multiplet
correspond the C-conjugated multiplet
t, -y, -b.
tum number is then necessary for the self-conjugated multiplets I introduced it in 1953#t# and called it "isoparity".
A new quan-
(b = 0, y = 0).
It is denoted G in Table 3.2.
Taking into account all quantum numbers introduced for hadrons, iance group should be written baryonic charge, Z2(C) semi-direct product. the matrices of
U(1)
(U I × U2 ) D Z2(C )
and
U(2).
C
Irreps of this group when
#*
And not to
U I x SU(2)
C
and
D
means the
b = 0 = y, and by
Phys. Rev., 92, 833 (1953); see Phys., i0, 587 (1953).
because of the relation
See L. Michel reference LM I of 4.9.
N. Cim., i0, 319 (1953).
corresponds to the
is equivalent to the complex conjugation of
Relation (3.16) was guessed by M. Gell-Mann, also T. Nakano and K. Nishijima, Prog. Theo.
#*t L. Michel,
U1
is the two element group generated by The action of
*
(3.16).
where
the invar-
(-i) y = (-i) 2t
implied by
98
(3.16),
t
is integral,
isomorphic
are faithful
to the direct product Finally,
their decay modes interaction
S0(3)
let us emphasize
tion of the proton
representations
with a lifetime
are unstable.
of 10 -23 to 10 -22 sec.
because
4.4.)
stable for strong interaction
columns and
However,
they do not completely
to be stable,
Indeed if the mass difference of baryons
in Table
would be unstable
3.2
(invariance
of
~
under
between
(Ay = i)
their lifetime nor
fit this latter concept.
>m k
y = 0, b = i (E)
The Other Particles
columns
the highest of these lowest states
The
E
is stable only because
were satisfied,
and that
(See
does not seem so fundamental!
the lowest states of two neighboring
were
m A - mN < m
3.6.
Neither
of the column
is due to the fact that both P, see 4.7a)
in Table 3.2 with the excep-
are the lowest states of the
or unstable
for strong interaction.
< m ; if for instance The stability
which is
They are often called resonances
in Table 3.2 and the first excited states
b = 0 (n).
D Z2(C)
Most of them are even unstable by the strong
instead of particles The particles
S0(3)
x Z2(C ).
that all particles
(and the antiproton) are indicated.
of
the
A
m E - mA
would also be stable.
~
and
~
have spin 0, parity -
q-isoparity
= +
while
~-isoparity
= -.
and the Other Interactions
There are only nine known "elementary"
particles which are not hadrons,
i.e., have no strong interaction. The photon,
y, with mass zero, spin i (see 4.4). + + The 8 leptons ~ , ~-; e , e (electrons) and their associated
neutrinos
~ +~ _~ +~ ; they all have spin 1/2. ~ e e ±
+ ÷e-+~
Only the
± +~
~
zero mass
are unstable.
+
e because
m
= 207 m . e All particles
electric
charge
Amp~re's
hypothesist
electromagnetic
(e.g.
have electromagnetic
interaction
even when they have no
q = 0; baryons have magnetic moments) that the whole electromagnetic
current
j~(x)
is well verified;
but it seems that
interaction
is through the
the interaction
Hamiltonian
is (3.17)
Hem = e ~ A~(x)j~(x)d3~ where
A~(x)
is the photon
field
which
h = c = i, the universal
2.4).
Electromagnetic
(electromagnetic
constant
interaction
is about i00 times weaker
This is also the order of magnitude
t Called nowadays
"minimal
e
potential vector). In units for 2 is given by e = ~ = 1/137.039 (see
coupling"
of mass difference
in the jargon.
than strong interaction.
in an isospin multiplet.
99 There is another universal interaction,
shared by all particles
the photon), characterized by a universal constant
(except
G, introduced by Fermi#, whose
value is G = 1.01 x 10 -5 x m 2 /~ P
(3.18)
Since it is much weaker than the electromagnetic interaction, action is simply called the "weak" interaction.
this inter-
Fermi postulated as early Ns 1934
that this interaction has some analogy with the electromagnetic interaction. instance there are four electrically charged
(!)
For
"weak" hadronic currents, respec-
tively vectors and axial vectors for the Lorentz group, v (+~(x),, a (±)(x) -interacts with the leptons through a leptonic current
which
~+)(x),~_ and the interaction
Hamiltonian being H w
= G /~
~ j" ~ ( E ) ( x ) h ~=±i
(~)(x)d3~
(3.19)
with h (c)(x) = v (s)(x) - a (E)(x), (s = ±i) Equation (3.19) has indeed some similarity with (3.17).
(3.20)
The fact that
h
is a
linear combination of a vector and an axial vector will explain the parity violation of the weak interaction
(see 4.7.b).
R. P. Feynman and M. Gell-Mann (Phys. Rev., 109, 193 (1958)) have found a very deep relation among the three interactions. U(2)
on
From the unitary representation of
~, the Hilbert space of hadrons, one obtains the representation
U(2)-Lie algebra on
~.
The operators corresponding to the observables
F y
of the and
t3
are the self-adjoint operators Y = F(y) Since
q = t 3 + 1/2 y
and
T 3 = F(t 3)
(3.21)
(Equation 3.16) for all hadronic states, this relation has
also to be true for the self adjoint operators representing these observables, i F(q) = Q = ~ j°(x)d3~ = T 3 + ~ y Note that
~ jP(x) = 0
~ Q
is time independent
~=~ [H,Q] = 0.
so (3.22)
However, Q
here
is the total electric charge of the hadronic part of the world, it is not conserved since weak interaction can t r ~ s f e r
it to the leptonic part of the world.
It is
conserved only in the approximation which neglects weak interaction. The beauty of the discovery by Feynmann and Gell-Mann is that, when electromagnetic and weak interactions are neglected, hadronic currents
vt(~)(x)p
the vector part of the weak
(Equation 3.20) and the electric current
j (x)
t E. Fermi, "Versuch einer Theorie der B-strahlen", Z. Physik, 88, 61 (1934).
of the
JOO
hadrons are images of the same tensor operator for the strong interaction for, respectively, of the complexified
the vectors
t±
U(2) and
group of invariance of q
of the vector space
U(2)-Lie algebras t± A y = 0 = y A t 3, t± A t 3 = ±t±
(3.23)
T+ = F(t+) = f v'0(_+)(x)d3~
(3.24)
This implies that
The isospin group, which was introduced in 3.2 in such a formal and abstract f a s h i o ~ becomes a physical reality since it is generated by the space integral of the weakly interacting hadronic currents! full
U (2)
3~v~(x)
group.
as well as
becomes:
i)
The addition of the electric charge generates the
When electromagnetic and weak interactions are not neglected, 3 j~(x)
do not vanish and the representation of
time-dependent for the physicists
any time with time-dependent
P, Q); 2)
Coleman and other physicists have shown). in Equation
(3.24)?
(just as
[P,Q] = i h l
Have you noticed the
U(2)
vt
on
is true at
undefined for the mathematicians
I have shortened a long story.
esis really needs the enlargement of the
U(2)
(as
instead of
v
The Feynman-Gell-Mann hypoth-
group to
SU(3)
as we will explain
in 5.1 and 5.3. To stay inside
U(2), one has to decompose
h (e)(x) = h~(c)(x) where
hi(e)
has hypercharge
Cabibbo angle #. the
a (s)(x)
IAyi = 1
y = 0
and
h
of Equation (3.19)
cos e + hlt(e)(x) hSe(s)
has
sin e
y = s
(3.25)
and
The same decomposition appears separately for the part of
h
(Equation (3.20)).
The angle
weak transitions are slower than those with
0
s
and
is a vector-operator for the h~.
SU(2)
isospin group.
is the and
has a value 15 ° so the
iAyl = 0
by a factor
They have also a different "selection rule" for isospin. v
e
v (E)(x)
tg20.
As we just said,
This is also true for
as
Hence weak transitions
IA i = 0
satisfy
I Ti = 0
while weak transitions
iAYI = 1
satisfy
[ATI : 1/2
i.e., het , v It , a el
are
or
1
SU(2)-spin0r operators.
We have also to mention two other charges conserved by all known interactions
(as the baryonic and electric charges).
which seem separately conserved: and zero for e zero for
They are the two leptonic charges
that of the e-type with value
~, ~ ; that of the ~-type:
e, ~ . e
# N. Cabibbo, Phys. Rev. Lett., iO, 531 (1963).
with value
~ = ±i
~ = ±i for
for ~ , ~
e s, and
4.
RELATIVISTIC INVAPXANCE.
4.1.
THE DISCRETE SYMMETRIES C. P. T.
The Poincar& Group and its Automorphisms;
Zeeman Theorem
Physicists call Poincar& group the inhomogeneous Lorentz group#. denote its connected component by connected Lorentz group
L0
PO"
by the translation group
It is a i0 parameter real Lie group. product
The group law of
We call flection
T.
T.
T [] L
P
P0
~0
is the semi-direct
T(r,t) = (-r,t), D
We denote by
P(r,t) = (-r,t)
the group of dilations
P, P+, F +, F
Z2(P ) × Z2(T) , Z2(P) , Z2(P)
and
T
the time re-
{a > 0,~(r,t) = (ar,~t)}.
the group generated by respectively
the groups generated by
and
m = "the
is given explicitly in Equation (4.10).
the space reflection
Z2(P), Z2(T), Z2(P) × Z2(T)
of the
It has a trivial center.
Its universal covering
T[]SL(2,C), whose center is a two element group generated by
rotation by 2~".
call
It is the semi-direct product
We will
D, Z2(P) × Z2(T)
and
We
P, T, P
P0
and respectively
D.
We call
P
and
the full
Poincar& group. It can be provent# that all automorphisms of these groups are oontinuous and, if
Aut G
is the automorphism group of
G, (4.1)
Aut PO = Aut P+ = Aut P = Aut F + = Aut F = F Given any group Out G that
G, we denote by
the quotient F
In.Aut G
the group of inner automorphisms and by
Out G = Aut G/In.Aut G.
Note that here
PO = In.Aut PO
and
is the semi-direct product F = PO [] (Z2 × Z 2 × D) = P0 [] Out P0
Binary Relation on Space Time
Given
x ~ y
(4.2)
E
two distinct points of
E, we define the notations:
x T y = (y
is inside the light cone of
s L y = (y
is on the light cone of
x)
x)
Lorentz transformations were introduced by Vogt in 1882 and applied by Lorentz to electromagnetism. H. Poincar& (C. R. Acad. Sci., Paris, 140, 1504 (1905)) required that they form a group with the rotation group and, from it deduced physical consequences. In Rend. Circ. Mat. Palermo, 21, 129 (1906) he included the translations and studied ~hysical implications of invariance under the group we call here Poincar& group. ~# L. Michel, "Relations entre sym&tries internes et invariance relativiste", lectures published in Application of Mathematics to Problems in Theoretical Physics, Carg~se 1965, Lur~at editor, Gordon and Breach (1967) referred to as LM III. We will also refer to my lectures in Istanbul (1962) and Brandeis (1965) as LM I and LM II. They are both published by Gordon and Breach (Book of the lectures, for each school).
'102
x S y = (y
is outside the light cone of
x < y = (y
is inside the future light cone of
x < • y = (y
is on the future light cone of
Given a permutation binary relations
x R y
x)
if
f
of the points of
x) x).
E, it is said to preserve the
x R y ~ f(x) R f(y).
E. C. Zeeman (J. Math. Phys., ~,
490 (1964)) proved the following theorem:
Theorem i.
The necessary and sufficient condition that E, preserve the relation
x < y
or the relation
f
and
f-l, permutations of
x < • y, is
f E F +.
Zeeman also established the corollary (proof published in LM II, p. 297):
Corollary i.
The necessary and sufficient condition that serves the three relations
4.2.
x T y, x L y, x S y
is
f, permutation of f E F.
Relativistic Invariance and Internal Symmetries#
A physical theory is relativistic if its automorphism group P0"
We are also interested in other s y m e t r y
internal symmetry groups. tions
P0
groups, subgroups of
G
contains
G, and called
Note that if we consider "passive" invariance,
the dila-
D c G. If
of
E, pre-
P0
is a subgroup of
in
G = {g E G, p C P0' gPg -i G = {g E G, p E P0, gpg E PO}. That
Aut ~ = F
G, one can consider C = CG(P0) , the centralizer -i = p}' N = NG(P0) , the normalizer of P0 in
is the semi-direct product
(2) and that
P
has no center
imply that
N = Po = (NJ
o)
(4.3)
,
and there is a canonical homomorphism f (4.4)
N/P 0 * Out P0 = Z2(P) × Z2(T) × D And for instance also see that
Z2(P ) c Im f
D N Im f will give information on the mass spectrum.
theory of mass zero particle has if there is a particle of mass ties and any
means that parity is preserved in the theory.
m > 0
D # 0
in its automorphism group.
If
We
Indeed, a ~ c Im f
and
then there are particles with the same proper-
for the mass value.
# We also refer the reader to the paper with the same title: 137B, 405 (1965).
L. Michel, Phys. Rev.,
~03
O'Raifeartaigh (Phys. Rev. Lett., 14, 519 (1965)) has proven the following theorem when
G
is a connected Lie group:
Theorem 2.
If the restriction of an irrep (= unitary irreducible representation) of G
to
PO
has an isolated point in the mass spectrum, it is the whole mass spectrun~ There have been too many papers written by physicists proving "theorems"
much weaker than the following trivial lemma (LM III, p. 450).
Lamina.
Let that
P0
be a subgroup of
Vg E G, gpg-i ~ PO" then
the homomorphism
f, G i
gP0 ~
G
xgP0
of
G.
If there exists
(G/PO)
on its homogenous space PO
p E PO' p ~ T c PO' such
is an invariant subgroup of
permutations of
is an invariant subgroup of implies
P0
containing
G.
giving the action
G/P 0.
Then
p; it is
p E Ker f
PO
and
Indeed consider Vx, g E G, so
PO N K e r f
PO c Ker f; that
Vq E P0' Yg E G, qg = gPo" In my opinion, the preceding considerations are physically very poor,
indeed
P0
acts on space time so if
G ~ PO
is an automorphism group of the theory,
Zeeman's theorem implies that in order to preserve causality, G time only through a quotient subgroup either of This led us to consider
G
as an extension of
F
A
P
if we forget dilations.
P.
We are interested in quantum mechanics. the *-algebra
or
can act on space
So we must use the existence of
of observables.
We refer the reader to the remarkable paper of Haag and Kastler, "An Algebraic Approach to Quantum Field Theory", J. Hath. Phys., ~, 848 (1964); there are physical arguments for
A
to be a C*-algebra. t
Let
A
(obtained by a Gelfand-Segal construction) by operators on states, A t
its commutant, Att
The spectral resolution of spectrum is d i s c r e t e , ~ =
Z
the enveloping W*-algebra, Z = A ~ A Act yields superselection rules t#.
@h~h
are those belonging to one of the Assume that
P0
be its representation ~, the Hilbert space of
a n d the only vectors of ~ ~h"
The
is a subgroup of
its elements can be realized by operators of
~%
its center.
For instance, if the which represent states
are called superselection sectors.
Aut A, which is implementable (i.e., L 0 0).
This proposition was made by I. E. Segal, more than ten years earlier. *~ Concept introduced by G. C. Wick, A. S. Wightman, E. P. Wigner, Phys. Ray., 88, i01 (1952). See the preprint~of Doplicher, Haag and Roberts for the most recent study of this question.
10z~
Let U(p) E L O 0 .
U(p)
be a realization of the automorphism
tion
p.
by an operator
It has to be unitary in order to be an automorphism:
(UAU-I)* = (UA*U-I). operators of
p 6 PO
If
A', U(p)V
V
is any element of the group
U(A')
VA E A;
of the unitary
is just as good for representing the Poincar4 transforma-
So the set:
E = {U(p)V, p E PO' V 6 U(A')}
(4.5)
,
forms a group of unitary operators which is a "central extension" of
PO
by
U(A')
i.e.,
E/U(A') = PO' quotient group
(4.6)
and
W
6 U(As), FU 6 E, V ~+ UVU -I
is an inner automorphism of
U(A') •
(4.6')
One can prove (see Moore's lectures), that any Polish topological group satisfying (4.6) and (4.6') is either the direct product
U(A') × PO
E
or are of the
form (see also LM II):
Eo~ = (U(A') x P'o)/Z2(c~,~o )
(4.7)
where the two element group is generated by the element
(~,m)
group of square roots
U(Z)
of
~
e)
of the unit, in the group
U(A'). t Which is the extension
E
chosen by nature?t%
with
a 6 2U(Z), the
which is the center The answer is the
extension defined by (4.7) with iz(B+EiLi) ~=e
*
2 ,a
(4.8)
= I
Equation (4.7) implies some topology as explained in Moore's lectures. In "Sur les extensions eentrales du groupe de Lorentz inhomog~ne connexe", Nucl. Phys., 57, 356 (1964), I have studied the same problem for abstract groups: any abelian group A is the direct sum A = D • K where D is the maximal divisible subgroup and K is a reduced subgroup (no infinitely divisible elements ~ i). One has the relations:
H2(Po,A) = H2(SL(2,~),A), H2(PN,A) = H2(Lo,A) = 2K + H2(Lo,D),
H2(SL(2,~),D) Aut ~ = 0
and of course
H2(Lo,D)AUtVC = 2 D.
group of automorphisms of the complex field act on Aut LolL 0 = Aut ~).
So it acts on
H2(SL(2,~),A)
SL(2,C)
Indeed and on
Aut ~ L0
the (exactly
(through a trivial action on
A) since the group of inner automorphisms of
L_ acts trivially. Following the U usual convention, also used in Chapter i, H2(SL(2,~),A) Aut ~ is the subgroup of fixed elements. 8o if H2(SL(2,~),A) # O, the automorphisms of ~ do not pass the non-trivial extensions and the corresponding extensions are very pathological. I found this a sufficient argument for considering in physics only the extensions of Equation (4.7).
%t This was the question that Lur~at and myself asked and answered in N. C~m., 21, 57 (1965) and Comptes Rendus of the Conference of Aix-en-Provence, p. 183, C.E.A. Saclay editor, (1962).
~o5
where
B
is the baryonic charge operator and
L.
the (different) leptonic charges
i
(see 3.6).
Indeed this choice of extension implies the observed relation between
spin and charges: (-i) 2j = (-i) b+Ei~i where
j
,
(4.9)
is the angular momentum of any state and
b, %. are its baryonic and i Note that Equation (4.9) shows that the integer or half
different leptonic charges.
integer nature of spin form a superselection rule.
4.3.
Irrep of
g0
All irreps (= unitary linear irreducible representations) known. to T TO
of
PO
are
In 1937, Wignert showed, by extending Frobenius' methods for finite groups
g0' that irrep of
gO
are characterized by an orbit of
T0
on
and an irrep of the corresponding little group (= stabilizer). invariant symmetric bilinear form on
products) yields an &somorphism of
T0
To be explicit, we denote by elements of
T
and
_a in a basis of group of
2 x 2
T.
SL(2,$) = TO .
_a, b, ...
T
and
and
T = R4
scalar
TS.
A, B, ...
(aO ÷,a) = (aO,a i ,a2,a 3)
Let
the dual of
(= Minkowski pseudo-Euclidean
space between
Consider the isomorphism between
respectively the the coordinates of
and the additive
hermitian matrices
(Z 0 +
_a +--+a = As we saw, the group action on
T
TI
The non-degenerate
gO
a
a
i + ia 2
-
aO
is the semi-direct product
ia 2
a3 / T~SL(2,C)
with the
SL(2,C)
T A E SL(2,C)
,
a +--+ ~ ~-+ A A~aA* +--+ Aa
The Minkowski pseudo-Euclidean scalar product is (_a,b) = aOb 0 - alb I - a2b 2 - a3b 3 and the Minkowski "length" of a We denote by
(~,A)
2
a
is
= (a,a) = determinant
the elements of
gO
with
a ~-~ (~,i)
the canonical injection
t E. P. Wigner, Ann. of Math., 40, 149 (1939) reproduced in F. J. Dysou, Sy~etr~j Groups in Nuclear and Particle Physics, Benjamin, New York (1966). Wigner,s paper was the first one giving a complete family of irreps of a non-aompact non-semisimple Lie group.
~06
T ÷ ~0
and
A %+ (0,A)
an injectiont of
%
= SL(2,C) %÷ P0"
The
~0
group law
is (~,A)(b,B) = (a + Ab, AB) We will use the same notation for elements of
T
(4.10)
and
TS.
It is useful to introduce the notion of stratum. a set
When a group
M, all the points with conjugate stabilizers form a stratum:
G
acts on
in other words,
a stratum is the union of all orbits of the same type (i.e., isomorphic as Ghomogeneous spaces).
The action of
%,
decomposes
T
or
T"
in four strata.
See
Figure 4.1.
It
The point 0 is stratum IV. x
FIGURE 4.1.
STRATA ON
T
BY THE ACTION OF
50
# This injection is unique up to a conjugation in P0" Indeed Wigner, in his paper of 1939, showed that for the abstract groups (with the action of T 0 on T just defined), HI(T0,T) = 0.
~07
Table 4.1 lists these strata, the corresponding little groups (defined up to a conjugation in
~,) and the corresponding irreps of %/
TABLE 4.1.
Stratum
I ~
2
= m
STRATUM OF
Tt
~. U
DUE TO THE ACTION OF
Little Group 2
P II p-- = 0 # 0
~0; IRREPS OF
Irrep of
P0
g0
0 m > 0 (i.e., p > 0), 2j integer ~ 0 a 0 Ib m < 0 (i.e., p < 0), 2j integer ~ 0
SU(2)
I
E(2)
II a sign of p0, 2h integer
(2-dimensional Euclidean group)
IIb sign of p0; H positive number, ~ = 1 0 II sign of p ; Z positive number, m = -i
S~(2~P0 = S~(2,1)
III a m 2 < 0, ascendant and descendant discrete series, ±j
C
III 2
< 0
N = double covering
III b m 2 < 0, principal series io,p E R lllc m2 < 0, supplementary series, 0 ~ o ~
IV p = 0
SL(2,~) = g0
IV
Irreps of SL(2,~) (see Stein lectures) are irreps of ~0 with T trivially represented.
is the non-trivial element of the center of (-i) 2j Those of
and in II by
(-I) 2h.
i
~0; in
I, ~
is represented by
Wigner constructed the irreps of type I and II.
needed for type III were given by Bargmann, Ann. of Math.,4_~8 568
S~(2,R)
(1947) and those of
SL(2,~)
(type IV for
g0 ) were first given by I. M. Gelfand,
M. A. Naimark, Acad. Sci. USSR J. Phys., iO, 93 (1946) and Isv. Akad. Nauk SSSR Ser.
Mat., ii, 91 (1947). Wigner's method is a particular case of Mackey's theory of induced representations.
Since the (measurable) axis
t10t
and
0x
are a set of representatives
of the orbits, a theorem by Mackeyt insures that this method yields all irreps of P0"
As we will see, the only irreps of
those of mass
m ~ 0
g0
which correspond to known particles are
(I a and II a in Table 4.1).
Wigner, in his paper, has given
the following realization: An d3~/p 0 = d~.
To
invariant measure on the orbit
Consider the functions
f
~:
defined on
p 2 = m 2 ~ 0, pO > O, is ~
with value in a
2j + i
# G. W. Mackey has described his theory in a book for physicists: Induced Representations of Groups and Quantum Mechanics, Benjamin, New York (1968). The needed theorem is Theorem B, p. 43.
108
K., such that ]
dimensional Hilbert
llfl12 =
J'f2
= E f (~)f (~)
where
~ P
< =o
,
is the hermitian scalar product in
(U(~,A)f) (~) = E6ei--a'PQ~B(~,A)fB(A-I~) where the unitary
(2j + i) x (2j + I)
matrices
Q(p,A)Q(A-I~,B) When
j = 0
module.
or when
Q
,
Q
(4.12)
(4.13) are complex numbers of unit p, Equation (4.13) show
[ . P
Particle States and Irrep of
What is a particle?
Then
satisfy
A, B E [p, the little group of
form an irrep of
4.4.
Q
Kj.
= Q(~,AB)
m = O, dim K 0 = i, and the
In all cases, when
that the
(4.11)
PO
This word is used very much by physicists.
This word
is attributed to the electron, the photon, and the 270 hadrons of Table 3.2 in 3.5, and also to nuclei (an "e-particle" for instance means a Helium nucleus) and even to atoms or ions.
The meaning of this word is in full evolution; it was clear up to
ten years ago.
Let us try to define this word.
"A particle is a physical system which can be isolated and cannot be decomposed into subsystems without destroying it." This concept is very clear for stable particles: electrons and positrons + ± e-, protons and antiprotons p , photons ¥, neutrinos and antineutrinos ~, ~ and also stable nuclei (deuteron, m-particle, 6C12 ) and the fundamental states of atoms or molecules formed with these nuclei. Invariants which can be attributed to these particles are the
P0
invari-
ants; mass and spin, and the Hilbert space of the states of a particle is the carrier of the irrep of
~0
of mass
example, proton or antiproton (0,-1/2), antineutrinos
m, spin
j
(or for
m = 0, of helicity
(mp,i/2), electron or positron
%).%
For
(me,i/2) , neutrinos
(0,1/2).
To good approximation this concept of particle can be extended to unstable particles whose lifetime is long enough to study them isolated
(T > 10 -21 sec).
Strictly speaking, because of Heisenberg's uncertainty relations between energy and time, the Hilbert space of states carries the
t The
~O
invariant
%
rep~(m,j~m
where
P
is a curve
is called helicity by elementary particle physicists but was
called circular polarization by its discoverer, Fresnel, in the 1820's. It just happens that the photon is the only known particle whose space of states is the carrier of a reducible representation (m = O, % = i) • (m = O, % = -i) of PO"
109
with a mean spread of decays,
Am.
For weak decays,
gm/m < 10 -14
and for electromagnetic
Am/m < 10 -5 . It is not clear that this concept of particle
the strongly unstable deed their lifetime the p-meson)
resonances •
(most of the "hadrons"
can be usefully extended to of Table 3.2 in 3.5).
might be as short as 10 -23 sec (and
Am/m
In-
reach 0.15 for
so they do not exist isolated since the range of strong interaction
10 -13 cm (of the same order as
3 x i0 I0 x 10 -23 = ~c).
is
However, we shall here call
them particles. We have used also the word particle as a collective name for different particles with the same spin and not very different masses and similar properties, so they can be considered
approximately
as identical.
This was the case of "the
nucleon" with the isospin degree of freedom corresponding + 0 n; the T-meson with the three states ~ , ~ , ~ . There is another degeneracy is expected)
for most particles.
the two values of the corresponding
and "antiparticle". conjugated, (examples:
p%, M~V
relativistic
of
%'
Such pairs are called charge-conjugate
Particles with all charges zero are called self-charge-
Elements of the enveloping
P0
and
degree of freedom are labeled "particle"
algebra
E(P 0)
observables which we will call the kinematical
The
(m,j)
indeed there is no degeneracy under charge conjugation 0 0 0 0 0 0 T,~ ,q ,p ,~ ,~ ,x , etc.).
The
p
It has been found (or it
that they come in pairs with the same representation
but all charges are opposite within the pair. pairs,
to the two states
Lie algebra is (with
C
for them
of the Lie algebra of
observables
P0
are
of the particle.
M v = -Mvv)
[iPU,ip ~ ] = O, [iPh,iM ~v] = iP~g h~ - iPWg I~
(4.14)
[iMVV,iM p°] = iM~Pg v° + iMV°g ~p _ iM~°g vp _ iMVPg v°
(4.14 t)
are the self-adjoint angular momentum.
operators
Pauli
Nat. Acad. Sci., (1967)) considered in
on
~
(unpublished)
representing
energy momentum and
and Bargmann and Wigner
(Proc.
E(Po):
1 1 M~VP p = (*M Wh = - ~ s~h~p P~Mvp = - 2 sharp " P)I
(4.15)
which satisfies [WX,P ~] = 0, [Wh,W ~] = i ¢ % ~ P P
W v 0
,
(4.16)
[WI,M ~v] = iW~g Xv _ iWVg %~
(4.161 )
P • W = PhWh = 0
(4.17)
Note that
110
The center of
E(P 0)
is generated by
P$ = PhWx
and
2W
= WXWh.
For irrep of
~0
these operators are the following multiples of the unit: m > 0,
p2 = m 2 i, 2W
m = 0
lla
m = 0
lib
or llc
4.5.
= -m2j(j + i)
p2 = 0,
W$ = 0
W = %P, % helicity
p2 = 0,
W 2 = -E < 0
Particle Polarization
In most experiments, the energy and momenta of the particles are measured (i.e.:
monokinetic beam, target at rest, bubble chamber track curved in a magnetic
field).
The additional observables to be measured in order to have a complete
knowledge of the particle state are called "the polarization". with
P, it is the polarization operator.
mute with each other.
However, the
W
Since
W
commutes
components do not com-
Equations (4.14), (4.14'), (4.15), (4.16), (4.16') show that
(4.18)
P0' PI' P2' P3' W3' W2 generate a maximal abelian subalgebra of observables of a particle
(m,j)
E(~0).
It is easy to interpret t the
when one remarks that
~(m,j), the one-particle
space of states, is a direct integral
3C(m,j) =
iO
Kj(~)
d3~ 0
'
(4.19)
p of
2j + 1
dimensional Hilbert spaces
K.. Given 3 tetrad of vectors in the Minkowski space (i.e., in
~, introduce an orthonormal
T')
_n(0) = £m-l,~(i)(i = 1,2,3,~,B = 0,i,2,3)
(4.20)
Minkowski scalar product n(~),n (B) = g
,
(4.20')
right hand orientation e%~Pn(~)n(B)n(V)n A
~
~
(~) =
p
_saBx ~
Introduce then the self-adjoint operators on S i = _ i n (i) . W(£) m
(4.20'') Kj(~) (4.21)
--
% See for instance L. Michel, N. Cim. Suppl., 14, 99 (1959) for more details and the treatment of the case m = 0.
111
where
W(£)
is the integrand of
f
=
Of course,
Equation
~(£)
d~ 0
(4.21')
p
(4.17) implies S O = _ ! n (0) . W(£) = 0 m
The S (i)
satisfies
the commutation
relations
[s(i),s (j)] = ieijkS(k) so they generate a = (m,0)
SU(2)
(4.21'')
--
Lie algebra,
,
(4.22)
that of the little group of
(particle at rest), S (i) = M0i(~).
~.
For
This shows the relation between angu-
lar momentum and polarization. For a particle of energy momentum R(~)
is a
rrR(~) = i, hermitian
which is transformed by
~, the polarization
density matrix
R*(~) = R(~) ~ 0, (2j + i) × (2j + i)
L 6 SU(2)
(defined by Equation
matrix
(4.22)) as
Lp = p, R(p) ~-> Q(p,L)R(p)Q*(p,L)
,
(4.23)
Let
R(/!) = (2j + i) -I i +
2j Z R(~)(e)
,
(4.24)
~=i be the decomposition The
R(%)(£)
and the
R(£)(£)
R(£)
into a sum of irreducible
1
= - ~ s
covariant
,R(Z)(£) = (-i)£ s mR
is a completely
s
SU(2)-tensor
of the polarization
can be given a completely
R(1)(~)
Where
of
are called the multipoles
matrix.
operators.
Note that
R(£)
form.
al~2"''eZ
W a ~ ~2. .W ~
(4.25)
syrmnetrical %th order tensor satisfying
al...~£ (partial trace) = SaBy... a This is obtained#
= 0, ~ s B ya. . " = ~ • s = 0
from the equivalent
form of relation
(4.26) (4.21)
W(~) = mZ.S(i)n (i)
(4.27)
i
From now on, we shall consider only the dipole polarization only one existing
for a spin 1/2 particle).
Its evolution
(which is the
is given in a macroscopic
# For more details see C. Henry and E. De Rafael, Ann. Inst. H. Poincar~, 2A, 87 (1965).
112
(N slowly variable in space time) electromagnetic field = EijkBk
and
*F
the polar tensor of = ~ M • s, with
- -
and
m
F
F i = Ei,F ij
(with
F) by the equation*
M = F + P ( ~
F + g'*F)P
,
(4.28)
- -
P = I - u g u = P
= ~/m, the quadri-velocity, which satisfies the
where
¢
Lorentz equation: N~ - -
The
F • u m
(4.281 )
- -
means the proper time derivative; M
and
F
are skew symmetric tensors so
(4.28) and (4.29) correspond to infinitesimal Lorentz transformations which of course preserve the Minkowski products: 2 = i, u • s = 0, 0 ~ 6 = (_ 2)1/2 ~ 1 where
~
(4.29)
is called the degree of (dipole = vector) polarization.
moment of the particle is dipole moment. g = 2
,
~ = (g/2)e/m(h = c = i)
The magnetic
and
~
= g
e/m
(For neutral particles, write directly
~
and
~.)
is a remarkable value of
g
is its electric Note that
which simplifies Equation (4.28).
This, as
first noted by Thomas in 1926, is characteristic of the Poincar~ group and does not happen for Galilean invariance
(as we commented in 2.6 and 2.9).
Consider a reaction between particles energy momenta are exactly known. polarization states is
where all
Then the transition probability between pure
% = II 2
S(PA,PA,PC,PD...) More precisely, S
A + B ÷ C + D + ...
E Horn 0CA g 4 '
~C ® ~
where the isometry
g "" ")
(4.30)
is the restriction of a unitary operator, defined on
Hilbert space of physics and called S-matrix in the physics literature.
X
the
In the
general case of given polarization density matrices X(pA,PB;Pc,PD,p...) where
RA,B(PA,PB)
@ KjB(PB)
= TrRc,D...(pc,PD. ..)SRA,B(pA,pB)S*
(reap., RC,D(Pc,PD.. .)
one of the particles, Equations
*
are hermitian operators** on
(reap., Kjc(PC) g KjD(PD) @ ...)
(etc.) when no polarization is observed.
which reduce to
(4.30') KjA(PA)
~(2JA + I)-I(2JB + I)-I
If one observes the polarization of only
(4.24), (4.25) and (4.30')show that
V. Bargmann, L. Michel, V. Telegdi,
,
%(pA,PB;Pc,~...)
Phys. Rev. Lett., 2, 435 (1959).
*T Practically, for all experiments, there is no correlation between the states of initial particles so RA,B(PA,PB) = RA(PA ) g ~ ( p B ).
113
depends linearly on the different polarization tensors of this particle (Sa,Sa~,Sasy,...) .+
4.6.
Invariance Under
p x Z2(C); PCT Theorem
If a physical theory is invariant under a group, say form the theory by an automorphism If
~
~ E Aut P0' (replace everywhere
is an inner automorphism, by definition of
theory is equivalent.
If
~
an active invariance of physical theories
P
and
T
invariance,
P, T
by
e(~).
the transformed theory
It is obvious that dilations are not
(except when only zero masses occur).
(and their product
and also under
g E P0
the transformed
If it is, then one can enlarge the invariance
group, in order to include this automorphism.
under
P0
is not an inner automorphism,
might not be physically equivalent.
What can be said about
P0' one can trans-
PT)?
We will assume invariance
C, the charge conjugation,
and in the next section,
see if these invariances are respected in nature. It seems a reasonable assumption that mutant of (4.5)).
A, the representation on We do know the action of
an antiunitary operator operator and well as
K
~ T
P
does not act on
on
A t , because
(see 1.2), i.e., by
is a complex conjugation
T
where
V(T)
leaves
U(A)
Note that if and
U(Z)
U = (U-I) *
is unitary, so is
KUK.
U(C), the charge conjugation operator on
tion
P0
C
L00
K (as
i.e.,
,
(4.31)
We assume that
T
globally invariant, but acts as an anti-linear automorphism.
Finally, we can introduce acts trivally on
physical properties of
is a unitary
(whose choice cannot be canonical).
KXAK = ~KAK, KABK = KAK KBK, K(A + B)K = KAK + KBK K 2 = ~.
(see Equation
has to be represented by
U(T) = V(T)K
U(T)) induces an anti-linear automorphism on the algebra
since
A ° , the com-
of the algebra of observables
C
and anti-commute with all charges.
tell
tially to internal symmetry.
us
how it must act on
At
~.
By defini-
More generally,
which corresponds essen-
Let D = Z2(P) × Z2(T) × Z2(C)
,
(4.32)
Pc = P x Z2(C) = P0 o D
(4.33)
In LM I, I gave the proof kindly tailor made by J. P. Serre for us physicists, (Theorem i, p. 183). H2(Pc,U(Z))
= H2(D,U(Z)) • 2U(Z) D
(see a similar theorem in Moore's lecture), where
2U(Z) D
,
(4.34)
is the group of the
t If the polarization of more than a final particle is observed one has also to introduce polarization correlations.
square roots of the unit of U(Z) invariant under every element of D. We check (iT (B+ ~iLi) ) (-i~ (B+ EiL i) ) that e = e is such an element so relation (4.9) is preserved. What is the extension in (4.34) chosen by nature? will see in the next section because
P, C, PC
phisms of the physical laws of nature.
Probably none, as we
(and probably
T) are not automor-
However, we can consider for
D
in (4.34),
a subgroup of that of (4.32). Let us first consider parity. deduced from those of
P0"
(0,0,±)
For
Irreps of
m > O, and
P+, (m,j,±)
or
tion to
P0
is irreducible.
(0,I~ I)
because their restriction to
P+ = P0 [] Z2(P)
m = 0, ~ = 0, there are two irreps of
with opposite parity (eigenvalue of For mass zero, ~ ~ 0 P0
are easily
irrep of
U(P)) whose restricP+
are denoted by
reduces to
(o,lxl)Po (o,1~1) • =
(O,-Ik I )
(4.35)
Note that, as projective representations of P0' (m,s,+) and (m,s,-) are equiva2 lent. More generally, since g ÷ g is a surjective homomorphism of U(Z),
H2(Z2(P),U(Z)) =
0.
So to speak of the parity of a state is not a canonical state-
ment; only relative parity can be defined for states in the same superselection sector.
By convention, the parity of the vacuum is taken +i, as well as that of the
proton, the neutron, the electron, the
A 0.
Wigner in his Istanbul lectures in 1962 (same reference as LM I) has studied the projective irreps of
P
(and even
P ).
This study can be easily
C
transferred to the study of the extension of
P
by
U(Z)
(and then by
U(As), from
general results of group extension by a non-abel±an kernel, as explained in LM I). This is not the case for
P
because
U(C), as unitary operator, does not act on
C
the phase of the projective representation, but trivally on
Z.
are canonical (since E 2U(Z).
C
as charge conjugation acts non-
We just give here the following results:
U(Z)
is divisible and
U(Z) ~
g ÷ g
U(T) 2, U(CPT) 2, U(PT) 2 2
is surjective) and are
For non-zero mass states, a choice different from U(T) 2 = U(PT) 2 = U(CPT) 2 = (-i) 2j
will require that irrep of
Pc
restricted to
P~Z2(C)
,
are not irreducible.
(4.36) This
would correspond to a new degree of freedom for particles which is not observed in nature.# In usual quantum field theories, relations
(4.36) are always satisfied.
This is related also to the two following theorems:
* See Wigner discussion in his Istanbul notes and for a recent review see H. Goldberg, N. C~m.j 60, 509 (1969).
115
Theorem 3. The good connection between spin and statisticst is a consequence of the Wightman axioms:it
covariance under
P0
of finite component quantum fields,
existence of vacuum, positivity of energy and "locality".
Theorem 4.
These axioms also imply invariance under CPT.ttt
4.7.
4.7.a.
How to Observe Violation
Action of
P~ T~ C
on Observables
Let us summarize in Table 4.2 the action of the automorphisms of
P0' on invariants of this group.
cles and antiparticles.
C
P, T, PT
acts trivally on them, but exchanges parti-
The self conjugated particles are eigenstates of
instance, consider quantum electrodynamics; C
C.
For
is an automorphism of this theory.
The electromagnetic interaction Hamiltonian is: Hem ='~ j~(x)A (x)d3~ By definition of
(4.37)
C,
utc)j~(x)U(C) -I = -j~(x) i.e., the electromagnetic current changes sign.
So
(4.38)
, H
em
is invariant under
C
if
also U(C)A (x)U(C) -I = -A (x)
(4.39)
i.e., integral (resp. half integral) spin fields describe particles which satisfy Bose = Completely symmetrical (resp. Fermi = antisymmetrical) statistics. This was proven by Pauli; his last publication in the subject is "Exclusion principle, Lorentz group and Reflection of space time and charge", p. 30 i__n_n Niels Bohr and the Development of Physics, Pauli editor, Pergamon, New York (1955). There he also proves the CPT theorem, first proven by LHders and Schwinger.
tt See R. F. Streater and A. S. Wightman, PCT, Spin and Statistics and All That, Benjamin, New York (1964); R. Jost, General Theory of Quantized Fields, American Mathematical Society, Providence (1965). ttt From weaker axioms (Haag-Araki theory of local observables), H. Epstein, J. Math. Phys., 8, 750 (1967) , has proven the CPT invariance of the S matrix. For infinite component fields, neither the connection between spin and statistics, nor the CPT invariance are implied by P0 invariance. For a counter example, see e.g., I. Todorov, 8th Nobel Symposium, Wiley (1968).
116
By definition, U(C)0> = 0>
where
0>
is the vacuum.
U(C)A (x)0> = -A (x)0> i.e., a photon has charged conjugation -i. formation of the electromagnetic F For
So ,
(4.40)
We have added in Table 4.2 the trans-
field
(x) = (~AA(X)) ~ = ~ A ( x )
T, time reversal, the space part
- ~ A (x) j
(4.41)
of the e.m. current
changes sign (as a velocity) while the time component
jP(x)
(whose space integral is the
electric charge) does not. variant), of
F 0i = E i
Hence the time reversal property of A , of H .° p em (electric field) and F IJ = Sijk Bk (magnetic field).
Consider Equation (4.28). transforms under pole) Equation with both
P
t
as
d/dt mu.
The quadrivector
and
mF • u = (-B "P,÷ - P 0 ~ _ p ÷ × ~ )
Hence, except for the term in
(4.28) is invariant under
P, T, PT.
(in-
The term in
gt
(electric di-
gt
is incompatible
T.
TABLE 4.2.
COVARIANCE UNDER P,T,C, OF THE INVARIANTS OF AND THE ELECTROMAGNETIC FIELD
Physical Observable
P0
P
T
PT
C
CPT
+
+
+
+
+
~'~
+
-
+
(pi,ej ,~,e~), (~,pj,~,~_~)
-
+
+
-
_
+
+
-
+
-
pj,~
~
(~,pj ,~,~_~),
helicity
(~,~_j ,~,~_~)
+
%
+
(electric field)
-
+
-
-
+
(magnetic field)
+
-
-
-
+
(a,b,c,d)
means determinant of the four components of four vectors.
4.7.b.
Parity Violation
The consequence of invariance under Consider
two s t a t e s
S 1,
I
S1
of a physical
through an "active" plane symmetry
~, and
P
is called parity conservation.
system corresponding t
$2, S 2
to each other
two other states of the same
117
system also symmetric to each other through be
the
respective
probabilities
E.
Let
of transitions
If an experiment I
EI12 = t 1 2 ,
2t12 = t12,
it
1 , a = ~(XI2 + 112)
=
I
(4.42)
I
yields
112 ~ 112, it proves parity violation.
means t h a t
t12
=
a + b,
Since
I
t12 = a - b, where
1 , is a scalar, b = ~(XI2 - I12 )
is a pseudoscalar.
So in a two particle decay of a polarized particle
the angular distribution
I
£+
£i + ~ 2 '
(4.43) (or more
P conservation
of decay products depends only on the even polarization
s B , s Bya , . . . . % In 1957, the following experiment was performed.
Co 60
nuclei at rest
(~ = m,O)
were polarized in a magnetic field B; this gives them a dipole polar±+ zation only: ~ = (0,1B). So P is a symmetry of Co 60 state P(m,O) = (m,~), P(0,XB)
= (0,1B).
electrons
Those nuclei decay spontaneously
of energy momentum
K(q0,$)
• ~ = -IB • q = -IBq cos 8. Similarly,
(B-
radioactivity)
with an angular dependence
emitting
proportional
to
This decay proved parity violation.
in the spontaneous ±
decay of zero spin
+ +p-+~¥
into a spin 1/2 p-lepton and a massless
~±
~
mesons
(
= m ):
,
(- for neutrinos,
+ for ant±neutrinos).
The p-lepton has a polarization s (which can depend only on the observed quanti2 P tatives ~ = ~ + ~ , pv = 0; remember sp pp = 0; see LM II).
S
=
$
m
--~
m 2 mp -
m f --m2
~
~
'
(4.44)
+
depends on the sign of the (electric charge of)
where also
P
violation
(by observation
of a pseudoscalar
~-.
p~ - s p
This proves
compatible with CP invariance. 2 satisfies s = -i
Note that in T-decay,
emitted in a pure helicity state suggest that
% See Equation s, s~, s~
~_ +
has helicity
requires
% = $i. +
for both
p
Then
P0
and
are
s (given --~ invariance
that the accompanying
All observation ~
and
the p-polarization
(complete polarization).
(through angular momentum conservation)
C
in the decay).
By the same type of argumentation we verify that those experiments
by (4.44t))
~$
is
on neutrinos helicity
~ . e
(4.25) and, at the end of 4.5, the property ....
for
X
I
112 = t r ~ R 2
112
generally in a decay where only two energy momenta are observed)
multipoles
I
and
1 ÷ 2.
Parity conservation ~ 112
I
112 = trRiR 2
to be linear in
118
This shows that the set of neutrino states in not invariant under C, and it implies that all reactions with neutrinos violate neutrino-less
(in fact, non-leptonic)
and
C.
or
But
decays of hyperons also violate parity.
Example: A 0 ÷ p+ + ~-, the angular distribution depends on (since
P
P
&A " ~
= -~A "
~A " ~ A = 0).
4.7.c.
Time Reversal Invariance
It would be better to call it "velocity reversal" since Let
SI
and
S2
(~/pO =
~)
T S1,
the
T
be two states of a physical system and
T S2
÷
- -v.
0
÷ ÷ (p0 a,+ p , p a~÷ - p ,
corresponding states obtained by a T active transformation + % ~+ %, etc.). s 0 ~+ s o ' ÷s ~+ -s, Then T T ~ XI2 = X21
(4.45)
Note the reversal of time ordering for the two transitions.
A precise experiment
comparing the cross section of the two inverse reactions
and
y + d+ ++ p+ + n is in progress.
+
+
,
(4.46)
(The rates are equal for pure states; since polarization is not
observed one has to divide the rate by the dimension of the polarization space for the particles
÷ (2 1/2 + 1) 2 = $, ÷ 2 × 3 = 6.)
Consider an elastic process -
+ p
+
÷ ~
-
+
p+
K. 3
(same initial and final particles)
such as
t
and compare the final polarization
s of the proton with the --p initial polarization of the proton target in another experiment. We must have ÷
÷
÷t
X(p~,pp
÷t
÷l
+t
÷ p~,pp,Sp)
+$
÷1
= X(-pw,-pp,-S
There is an approximate condition of
T
+
÷
p
+
-p
invariance,
,-pp)
(4.47)
in perturbation theory which is
based on the following expansion of the "S-matrix". S = I +iH where H).
H
+ 0 (H 2 )
has to be a self-adjoint operator
,
(write
(4.48)
SS* = S*S = i, in first order in
In this form, we have for orthogonal states (i.e., RiR j = 0) TrR.SR.S* N TrR.HR.H = TrR.HR.H 3 x 3 x x 3
,
(4.49)
i.e., in this approximation %.. = %.. x3 31 Then in this approximation,
(4.45) reads
T ~12 = %12"
(4.50) Even in this approximation
there is no positive evidence of violation of time reversal in physics, with perhaps the exception of K0-decay
(next section).
li9 Note that in Equation (4.28), using Table 4.2, the term in dipole) is not compatible with time reversal invariance
(or with
P
g'
(electric
invariance).
So the existence of an electric dipole for an elementary particle would prove violation of both I
P and 10-12.
T.
e gneutron
Experimentally
is known to be
~i0 -9
and
gelectron Note that
PT
invariance has a simple formulation. I
For example
I
h(Pl + P2 ÷ Pl + PC2) : h(Pl + P2 -~ Pl + P2 )
'
for spinless particles or for pure states (then change ~ ÷ -~, Sal..._ak ÷ (-l)ks l...~k)).
This is known in physical literature as the "principle of
detailed balancing". 4.8.
CP
CP
Violation
violation was first observed by I. H. Christenson, C. W. Cronin,
V. L. Fitch and R. Turley, Phys. Rev. Lett., 13, 138 (1964) in K0-decay.
Many
experiments have confirmed it. The state of a Hilbert space
K0
or
~ ( 1 ) = L2(R3,t ) @ K2 where
defined on the two element set while
P
PC = CP
is of the form so
(= anti-K 0, Y = -I) can be described by the
(PC) 2 = i
K2
i s the v e c t o r s p a c e of f u n c t i o n s
(Y = i, Y = -1).
P ® I
so
PC = P @ C.
Then
C
is of the form
We assume that
of
K's).
So we can write
JC(I) :JC~ 1) ~JC(I)_ ~
~+
(i) _
=
,
(4.51)
~ (i) ~±
Now it is easy to deduce the action of
(4.511 ) CP
on states of two
are two identical self-conjugated particles hence any state of C.
p2 = i, C 2 = i,
(as we have seen in 4.6, for spin 0, another assumption will
increase the degree of freedom
with value +i for
I @ C
The tensor product of the representation
270
0.
These
is eigenstate
(m,0)
of
P+, by
itself yields
®(m,0) =
~ ~=0
2
I2m
(m,~)dm
,
(5.52)
with symmetry I-I-] for even
~, ~
for odd
~
(4.52 I)
0nly the ~I] = symmetric states are allowed by Bose statistics. essentially similar to that yielding Equation (2.40), P Equation (4.51) by multiplication by 2~ 0
are eigenstates of
C, P, CP
(-i)~
And (by an argument
acts in the space of
in each direct summand.
with eigenvalue +i.
So states of
120
For states of
~+ + ~ , one has to consider these two particles as identi-
cal in order to apply Bose statistics, but in the two different possible states of + and -). So states of ~ ~ of total spin ~, are eigenstates of C, P,
charge (+ CP
with eigenvalue
(-i) ~, (-i) ~, i.
was predicted that states of states of
~C(I)
--~!)
of the two spaces
3~ 0
--~!i) and
CP
was believed to be preserved, it
in Equation (4.51) would decay into
would decay into
eigenvalue -i (as e.g. all
When
3~
states which are eigenstates of
states).
~(i)
2~
while
CP
with
This was exactly observed and the states
were also called "short" and "long" because the
"I"
2~-decay is faster. In 1964, the above quoted experiment proved that the long lived meson also decays into
2~
(with a rate
N 106 slower than the short lived).
We do know that the universe around our galaxy is not
CP
invariant, but
the influence of this asymmetry (which could depend on the relative velocity of the K-meson with respect to the galaxy, or the earth) seems to be ruled out by more precise experiments. Must we conclude that there is a small violation ((10-6) 1/2 = 10 -3 amplitude) of
CP
in the transition
K ÷ 2~?
in
Another possibility could be that
CP
is conserved in this transition but that the two observed meson with exponential decay:
short-lived
tively a large
KS
cS
and long-lived
and a small
cL
~
are non-orthogonal states with respec-
component in
K S ÷ 2~0 bS =
and
b~
(Ks ÷ 2~+ + ~-)
~il).
Then the branching ratio
~ ÷ 2~ 2 ÷ + ~ +
,
should be equal, since they would be the branching ratio of all the states in The value of
bS
is
N 1/2 (as predicted by the selection rule
The first measured values of another experiment.
b~
~i)
A~ = 1/2, see 3.6).
were around i0 to 12, but a value zero appears in
The present experimental evidence is still an incompatible set
but "optimists" say it is compatible with So it is possible that
CP
b L N b S ~ 1/2.
violation is due to a still undetected inter-
action, to which no particle transition or spontaneous decay can be attributed, and which has to be superweak. CP ~£ = ~
or
violation has also been observed in
~
÷ ~± + kS + ~
(where
e) decay; there is a relative difference of 3.10 -3 in the two
conjugated rates.
But
CP
C
or
violation has not yet been observed anywhere else.
Of course physicists have proposed many theories (about thirty not yet ruled out by the meager experimental data) to explain
CP
violation.
There is no
possibility to give more details here. To conclude, let us just remark that there is no evidence against violation and there is one fact which suggests that invariance than
CP:
a small upper limit of the
CPT
K 0 - -~ K
CPT
is a "much better" mass difference is well
CP
121
known.
It is
m
< 10-14ink .
- m
K0
Such a perfect equality cannot be due to
K--0
chance and suggests an invariance in nature which contains seen that with the
C, CT, CP CPT
are ruled out, so
CPT
C.
However, we have
is the likely candidate in agreement
theorem 4.6.
Remark on Galilean Invariance
We dealt in Section 4 with relativistic P~incar4 invariance only.
Although
we sometimes spoke in Chapters 2 and 3 of Galilean invariance, such invariance was not thoroughly used in atomic and nuclear physics. characterized the irreps of
E. In~nH and E. P. Wigner
3, the covering of the Galilee group, in 1952
(N. Cim.,
9, 705). Their results did not fit with physics.
V. Bargmann
i (1954)) shewed that for central extensions of the For each irrep of parameter
m
G
(Ann. of Math., 59,
Lie algebra
8, H2(g, R) = R.
G, this yields a family of projective irreps depending on one
which corresponds to the mass of the particle.
See also 0'Raifeartaigh's
lectures where it is shown that projective
irreps of an invariance group also appear in classical mechanics.
5.
THE INTERNAL SYMMETRIES OF HADRONS
5.1.
SU(3)
5.l.a.
Symmetry
The Octets
Table 3.2 of "elementary particles" in 3.5 is reminiscent of similar tables of atomic and nuclear spectra. So, before a dozen of baryons and as many mesons were known, physicists were searching for a larger symmetry than that of which we have described in 3.5.
U2
(isospin and hypercharge)
There is no point and no time to tell here about the
ill-fated choices except to mention that of Sakata, with a U(3) mental representation was spanned by Sakata, Prog.
group whose funda-
p, n, A, the first three known baryons.
(S.
Theor. Phys., 16, 686 (1956).)
Just as Heisenberg proposed to consider neutron and proton as two states of the same spin
1/2
particle, the nucleon, by neglecting their very small mass dif-
ference (or more precisely attributing it to an electromagnetic self-mass effect), the eight known spin ½ + baryons p, n, A 0 , Z-, Z 0, Z+, Z-, ~0 could be considered as eight states of the "same" particle although the mass difference is of the order of 15 percent instead of 0.15 percent.
"122
By 1961, seven pseudoscalar mesons
( 0 - ) were known, with the same group-
ing in isospin and hypercharge y = i, t = 1/2, K+KO; y = -i, t = 1/2, K-K0; y = O, +0 t = i, ~ ~ ~ but the spread in mass was much larger. M. Gell-Mann and Y. Ne'emann independently proposed to use SU(3) as a 1+ classifying group; the eight ~ baryons and, predicting a y = 0, t = 0 pseudoscalar meson which was discovered a few month's later and called mesons form two octets = eight dimension space
E8
0,
the eight
0-
of the adjoint representation of
SU(3) ([I] in Young diagram notation). For instance the Hilbert space of states of U i+ ~] i+ one baryon is the tensor product L(m, ~ ) @ K ( h _ ; ) where L(m, ~ ) is the space of the irrep SU(3)
l+
(m, ~ )
of
P, the Poincar~ group and
K(
)
the octet space
is an exact symmetry when the baryon mass differences are neglected.
E 8.
We can
say that strong interactions will be decomposed into two parts:
a strong
SU-3
in-
variant part and a semi-strong part invariant under the subgroup
U2(T,Y)
only.
This fits the reduction*
F
of su(3)Iu(2)° dim
8
=
(01) (1½) (yt) 2
But, would it be possible to consider the
+
i
+
SU(3)
a perturbation of the very strong interaction?
3
+
51)
2
breaking semi-strong interaction as Surely, if you are an optimist.
After all 15 percent (effect in baryon mass) is small compared to i. Let us now study the mass splitting within the SU(3)
5.l.b.
multiplet.
The Mass Operator
The simple hypothesis for the mass operator
M
is that it can be decom-
posed into M = M 0 + M~(y) where
M0
algebra of
is a "scalar" tensor operator and
,
(5.2)
Mr(y)
is the image of
SU(3)) by an octet = E8-tensor operator.
irreducible representation of
SU(3).
Because
Let
SU(3)
ly has two zero roots (which are zero weights for
E
y
(of the Lie
be the space of an
is of rank two, or equivalent-
~
)##
dim Hom(E @ E8,E)SU(3) ~ 2
For
u E U(2), the black column means
(det U) -I
(5.3)
while
r-1 U means
(det u).
t+ If %1 m %2 m 0 are the number of squares in the first and second line of the Young diagram of an irrep of SU(3), one also uses the notation (%1 - %2,%2 ) for the irrep of SU(3). The contragredient of (p,q) is (q,p), so (p,p) is self-contragredient,
as in
has for contragredient
~ ~
= (i,i), while denoted
J J l]
= (3,0) of dimension I0,
i-~ by the physicists.
~23
More precisely, is
1
it is
2, except for the trivial irrep, for which it is zero, and it
for the irreps whose Young diagram has only
equal lines
X1 = %2' (i.e.,
(0,%2)):
1
line (i.e.,
for example,
~
(%1,0)), or two
and its contragredient
-
which are also denoted This is also true for ~ ] = (2,0), ~
= (0,2)
i0
= (i,0)
denoted
and
6
SU(3)
parameters
and
the
= (0,i)
denoted
3
and
U(2)
and weak interactions)
SU(3)
is an exact symmetry
the particle masses
Explicitly,
is the representation
~, the Hilbert space of hadronic
for any
p , p E K ~i ~2 operator T
and for any octet-tensor
K
depend on three
Ml(Y)).
octet-tensor
From 1.5, operators
(up to the factor states;
i) of
it satisfies
the space of an irrep of
SU(3)
in
= p + SD(x))p > ~i ~2 <~i '(~F(x) ~2 In the physics (see Equation subspaces
of
literature
e/~
is called the
F/D
~
carriers of inequivalent
M0, MI, M 2
and
are
SU(3)
SU(3)
scalar operators.
T(T + i), the generators
By definition
If the octet part
MI(y)
(5.2)) of the mass operator has no matrix elements between two
commute and their common eigenspaces Y
(5.4)
ratio.
irreps,# this implies that
M = M 0 + MIF(Y) + M2D(Y) where
of
operatorg
(i.e., neglect of elec-
in a multiplet
E, as linearly independent
x ~--+ F(x)
Lie algebra on
F A F = iF.
7; and
octet-tensor
(one, the expectation value of M0, and two at most for
D = F V F, where
i0.
(5.3) is to say that on the Hilbert space
we know that we can take for each F
~
because they are of dimension
there are at most two linearly independent
Thus, in the approximation where tromagnetic
i0
and
Another way to interpret an irrep of
and
F(y)
are
U(2)
,
(5.5)
The operators multiplets,
F(y)
to
D(y)
so they are functions
of the center of the enveloping
is proportional
and
Y, the hypercharge
of
algebra of
operator,
U (2). Y and by compu-
tation one finds D(y) = ~2 _ ~1 y2 - ~1 K where
K
is the (quadratic)
in the definition
Casimir operator of
of the scalar operators,
state of hypercharge
y, isospin
on a
,
(5.6)
SU(3). SU(3)
So with a convenient multiplet
the mass of a
t
s i 12 m = m 0 + m I + m2(t(t + i) - ~ y
Applied
to the octet of Baryons
change
N, A, E, Z
(5.7)
this yields a relation between their
four masses i (Gell-Mann,
i (mN + m~) = ~ (3m A + m E )
,
(5.8)
Okubo mass relation) which is well verified within few MeV (for mass
> 10 3 MeV!). t There are exceptions to this rule: see 5.l.d, the vector mesons. main idea, we simplify here too much.
To convey the
124
For mesons (zero baryonic charge) because of the charge conjugation between I
particles and antiparticles, M 1
must be zero.
The Gell-Mann Okubo mass relation for
pseudo-scalar-mesons m K = ~i ( m is verified only within 50 MeV, about
+ 3m )
i/i0
(5.9)
of the
K
and
~
mass.
Optimistic
physicists have found good reasons why this relation should be better verified by 2 (instead of m).
m
5.1.c.
' E
When
SU(3)
was proposed as symmetry group in 1961 only the first N and 3+ 3+ A(j p = ~ ,t = 3/2), Z*(j p = ~ ,t = 1)I" were_~known. Gell-Mann
excited states,
putting them in a excited state of seen in the
The First Baryon Decuplet
I0
meters (one for
i0 E)
representation, predicted a
M0
~(jP = ~3+ ,t = 0,y = -2.
and finally a particle
(i.e. , ~ - ~
E*, (jP = ~',t = i/2,y = -i, As we have
) irrep, the mass must depend linearly on two para-
and only one for
Okubo r.elation predicts for the mass
M1(y)), so in this decuplet the Gell-Mann and m
of the states of hypercharge
y,
Y m Y
=
mE,
(mE,
-
A few months later (in 1962) the predicted
-
mA)y
E*
(5. i0)
was found with a mass of 1530 MeV (to
be compared to the predicted value (1385 + (1385 - 1236) = 1534 MeV!). established that its spin is the
3/2
It was later
and it has the same relative parity as
E.
But
g-, which should be stable against strong and electromagnetic decay, since it
would be the lowest hadronic state with and not found ... immediately. tion why the
~
b = i, y = -2,
was frantically looked for
Many physicists had given up hope and given explana-
did not exist, before it was found in 1964 after two and a half
years of feverish impatience. dicted 1677 MeV).
The
~
mass is 1672 MeV, (to be compared to the pre-
Its spin has not yet been measured since less than a score of
particles have been observed up to now. predicted, when would the
~
If it had not been looked for where it was
have been observed by chance?
5.l.d.
Other
SU(3)
Multiplets
The known experimental data at a given date give a deformed view of the SU(3)-multiplets. known yet, although
In the baryon case for instance, no excited states of the ~ are 5+ ~ and other decuplets probably exist. Some octets have been
tentatively identified, although too few excited
Z
states are yet known and their
quantum numbers are not measured.
# Is often called also Y*,
We denote by
jP
the spin
j
and parity
p.
125 The mesons seem to prefer to occur in nonets. t = 0, 0-
meson is known in addition to the octet of
K*, K*, is very well known. and the
@
i>
is a
The
~
and
@
singlet and
8>
I-: p, ~, ~,
are orthogonal states of "mixed con-
is the octet vector
is also well established and an octet of ( ~ )
of baryons
SU(3)/Z 3
is likely.
A nonet of
2+
There is some possibility
It is to be noted that only irreps of the ad-
do appear.
5.l.e.
SU(3)
1+
q = y = t ~ 0.
(not indicated in the Table 2.3, for the experi-
mental data are still preliminary). joint group
A nonet of
q = y = t = 0, m = i> cos ~ + 8> sin ~, ~ = 8> cos a - i> sin a, where
SU(3)
of a 27-plet
0-.
q = 0, y = 0,
The mass formula could not apply to the known "octet"
was predicted.
figuration"
Indeed a
Cross-Sections
and Decays of Resonsances
invariance implies ratios of resonances decay rates
(measured by
the natural width and the different branching ratios) into lighter hadrons.
This
yields remarkably good predictions and explains strange facts such as the small branching ratio for the decay of
~
into
2w.
For two octet-particle reactions
A + B ÷ C + D, one can deduce that the
scattering amplitude belongs to the representation
(5.11) 8
@
8
27
8
1
symmetric which yields seven arbitrary parameters.
8
i0
i0
antisymmetric There are less in
8 ~ 8 + 8 @ i0.
to correct for the mass difference is not obvious and the predictions
The way
are not
spectacular. An anthology of original papers in
SU(3)
has been published by Gell-Mann
Eightfold Way, Benjamin, New York (1964). M. Gourdin, Unitary Symmetry, North-Holland,
and Ne'emann, The this subject by
5.2.
There is also a