This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
=, we get (24)
f (a I ) d7 In =
x
J x
QA *,3.
2.3 Sobolev Spaces of Forms Let X be an n-dimensional compact submanifold of u : X ---f R which are 7-ln L X-measurable and with
J Iulsd7-InLX
s>1
I[Yn+N. The class of functions
2.3 Sobolev Spaces of Forms
537
is denoted by LS (X) . LS (X) is clearly a Banach space, and for s = 2, L2(X) is an Hilbert space with the obvious scalar product. If (Vi, (pi), Vi : V -> X is
a system of charts for X, we define the class Wl,'(X), s > 1 as the class of u E LS(X) such that u o Wi E Wl,s(V), and we set II u ll wl,s (X) := 11 U 111s(X) + E f I D(u O Wi) IS dx.
(1)
Vi
It is readily seen that the definition of Wl,s(X) is independent of the system of charts, while the norm does depend. However different system of charts give rise to equivalent norms on W""s(X). An intrinsic norm on Wl,s(X) can be defined in terms of approximate tangential derivatives DX to X : one easily verifies that IIUIILS(X)+1
1
1/S
f IDXulsdHnLX)
is a norm equivalent to (1). Properties of the standard Sobolev spaces on open sets of Euclidean spaces are immediately transferred to analogous properties of W1"2(X), we in particular state (i) W"2(X) is a Hilbert space with the scalar product /
(u I V)Wi,2(X) := (u I v)L2(X) +
f(DXU I DXV)TX dfn LX.
LS(X) is compact (ii) The inclusion Wl,s(X) (iii) C°°(X) is dense in W""s(X).
A differential p-form w on X is said to be measurable if <w(y),v(y)> is 7-ln-measurable for any smooth vector field v, equivalently if for any local chart cp2 : Vi -> X, cp#w is L'-measurable, i.e., all coefficients of cp#w are measurable.
Definition 1. Let s > 1. The space of measurable p -forms on X such that
f IwlsdfnLXAPT,,X = <7r#w,7r#77>APRn+N
a scalar product on L22 (X) is given by (w,71)L;
f <w(y), 77(y) >APT,,X /Ln =
V w,7) E A°TyX,
/
<7r#W, rr#17 >npRn+N dH-.
538
5. Cartesian Currents in Riemannian Manifolds
Definition 2. The class of p -forms w in L4(X) such that cp#w have coefficients in W1'1 (V,1[2) for every coordinate chart (V,cp) are denoted by WI,"(X). (X) can be defined by choosing a coordinate system of charts A norm on (Vi, fpi) and setting II W II W;'s(X)
II W II LP(X)
+
#wI WP's(Vi).
Though the definition of WP' (X) is independent of the coordinate system, the norm does depend on it, but one easily verifies that different coordinate systems produce equivalent norms. Again properties of WP"°s(X) are inferred from properties of standard Sobolev spaces; we state
(i) W l ,' (X) is a Hilbert space.
L'(X) is compact. (ii) The immersion WP ,3 (X) (iii) Smooth forms are dense inWP(X). Finally we remark that similarly one can introduce the spaces Wp'3(X) and more generally WW ,r (X ), but we shall not dwell any further on this point.
2.4 Harmonic Forms Let X be a smooth compact n-dimensional submanifold of 1[in+k possibly with boundary. We assume that X is a CO° submanifold even if the claims on this subsection hold for Cl,l submanifolds with trivial changes. Let d be the differential operator on forms.
d : C°°(X,APTX) -+ C°°(X,AP+ITX). Definition 1. The codifferential operator
b : C' (X, APTX) ---> C' (X, AP-'TX) is the (formal) adjoint of d, i.e., is the operator defined for all a E CO° (X, APTX)
and p E C,- (X \ 9X, AP-'TX) by (Sa I )3)L2(X,AP-1TX) :=-(a I d,3)L2(X,AP-1TX).
Clearly b2 = 0, and S extends to all
WP,' forms.
Proposition 1. 8 : C°°(X,APTX) -+ C°°(X,AP-1TX) is a differential operator. If x1, ... , x'n are local coordinates and G = (gij) the metric tensor of X, then for w = (1)
aI =P wa
(6w). =
dxa we have Sw = E ,,I=P(Sw). dx-I where
E E E g7ao,(i,
Ial=p-1 i¢a 1QI=p
(a+2)wR)-
2.4 Harmonic Forms
539
Proof. Let 7) E C°° (X \ 8X, AP-'TX) be a (p - 1)-form with support in coordinate chart. We have
-(w,dr1)APTTx = - = gQ7wa(di1)7 IQI=I7I=P
( IQI=I'YI=P
1
\ a+i=ry
I$I=P Ial=P-1 i0«
Integration by parts on X then yields
-(4' dr7)L2 = - f (w, drl)f dx
-
a)Di(f9p(«+i)«)
0(2,
Igdx,
Ii3I=p i s
«1=p=1
Ial=pY_iOa o-(i, a)Dii.e. (1).
consequently g'Y«(Sw)ry =
Proposition 2. If X is oriented and * is the Hodge operator on X then S = (_1)n(p+1) * d * Proof. For a E COD (X, APTX) and /3 E C°° (X \ 8X, AP-1TX) we have
d/3A * a = d(/3A * a) = d(/3A * a)
-
(-1)P-1,3Ad
*a
(-1)p-1(-1)[n-(n-P+1)](n-p+l)/3n * *d * a
= d(/3A * a) - (-1)n(p+1)OA * *d * a. Therefore, by Stokes theorem and (24) in Sec. 5.2.2, we get
(Sa1/3) _ -(aId3)=-J d/3A*a (-1)n(p+l) f /3A * *d * a =
(-1)n.(p+l)(*d
* a 1)3).
Definition 2. The Laplace-Beltrami operator on Coo (X, APTX) is the operator
a=dS+Sd : COO(X,APTX) a form w is called harmonic if Zw = 0. The following claims follow at once
C°°(X,APTX);
540
5. Cartesian Currents in Riemannian Manifolds
Proposition 3. We have (i) A is formally self-adjoint, i.e., (La 1,3) _ (a Ai3) for a, 3 E Cc-(X \ 3X, APTX). (ii) If w E CC°(X \8X,A2TX) then Aw = 0 ifdw = bw = 0, in fact (w dw) _ J8wJ2
+ dwI2 (iii) A is well defined onW2'2(X,APTX). Definition 3. The Dirichlet integral associated to a p-form is D(w) := f(dwI2 + Ibw12) d7-ln
x 1 The Laplace-Beltrami operator on functions. First we compute b on 1-forms. In a chart U, if cp E C° ° (U) and w = > wi dxi we have
f f -
(6w I 0) =
(w, dcc) dy = -
f
-gUwi 28x
U
Dj(V,-gg2'wi)Wf dx
U
where (gij) is the metric tensor on X, gig Sw =
Dj
(gij)-1, and g := det (gij). Hence (-,fg-g2'wi)
and for any function f E C°°(X) A f = (bd + d8) f = bdf =
D; Voij -Lf
axi
Also in local coordinates, i.e., for f E C' (X) with compact support in a coordinate chart we have
D(f) = f Idf12dW2 = f X
8f of dx. /g2j 8xi axi
Notice that, by the maximum principle, any harmonic function on a compact and boundaryless manifold must be constant.
0 The Laplace-Beltrami operator on p-forms in IRn. Given w = E wadxa 1c1=p
we have dw
and, from (1)
=
E wa>z o-(2, a) dxa+i lal=p i a
2.4 Harmonic Forms
Sw = (-1)n(P+1) * d * w =
(2)
Q(i )w(j+i i
1
I01=p-1
541
J dxO.
i j3
We now compute the Laplace-Beltrami operator. Of course
d5w = E(dSw)a dxa, a
and, according to the above
(dbw)a =
/
o(i, ) (dw)a,i =
T Q(i, a)o,(j, Q)wa+j,ji
= E E o(i, a - i)o(j, a - i)wa-i+j,ij ; iE« jva-i now we split the sum in j as
>
joa-i
_ j=i
j$a
to get
o(i, a - i)o(i, a - i)wa,ii
(dbw)a = iEa
o(i, a - i)a(j, a - i)wa+j-i,ij.
+
(3)
iEa
j0a Also 8dw = E(Sdw)a dxa, and (Sdw)a =
o(j, a)(dw)a+j>j =
E a(j, a)o(i, )3)w0,ij j 0a (j+i=a+j
i 0a joaiEa+j
Similarly we split the sum in i as the sum of terms in which i E a and terms in which i = j to get (Sdw)« _
(4)
o(j, a)o(j, a)wa,jj j4za
o(j, a)o(i, a+ j - i)wa+j-i,ij
+ iEa
Summing (3) and (4) we then get (zlw)a = (dbw + Sdw)a = Lwa
+
[o (i, a - i)o(j, a - i) + a(j, a)o(i, a + j - i)]wa+j-i,ij iEa
ja
542
5. Cartesian Currents in Riemannian Manifolds
As fori<j (5)
u(i, a - i) = Q(i, a - i + j)
o-(j, a, - i) = -o-(j, a),
while for i > j (6)
o(i,a-i)+j)
o(j, a-i)=o(j,a),
we see that the term in square brackets in (5) vanish, therefore we conclude
L1w = E (Qwa) dxa
(7)
lal=p
i.e., a p-form in the flat manifold R n is harmonic if and only if its coefficients are harmonic functions. Obviously 1dwj2
+
lawl2
=
IDwal2 + L(w) lal=p
L(w) being a non-zero quadratic form of the derivatives of wa. Since the Euler equations of the two integrals Do(w, U) :=
fI&2 + 82) dx and
Dwa U
2
,
lal=p
as we have seen agree with Awfa = 0, we infer that (8)
Do(w, U) =
JU
IDwa12 dx
V WE W01,2 (U)
Moreover L(w) is a null Lagrangian; in particular fu L(w) dx depends only on
the values of w (wa) at the boundary 8U of U and vanishes if w E C' (U), compare e.g. Giaquinta and Hildebrandt [275]. Of course one can compute bw, L,w, I dw 12 + I dw 2 explicitly in terms of the metric tensor and of its derivative, but, since we do not need it in full generality, we postpone this topic to Sec. 5.2.7 and here we confine ourselves to the following remark. From (1) one sees that (dw).y has the form (9)
(8w).y(x) = api(G)wQ,i + bp(G, DG)wo
where aQi(G) and bp(G, DG) are smooth functions which pointwise depend re-
spectively on the metric tensor and on the metric tensor and its derivatives. Indeed api(G(x)) are Lipschitz functions and bp(G(x), DG(x)) are L°°-functions on X if the metric tensor G is Lipschitz. Moreover we notice that, in case G is the Euclidean metric,
2.5 Hodge and Hodge-Kodaira-Morrey Theorems
543
(8w)7 = E E o (i,,Q)DjwQ+i IQI=p-1 i0Q
consequently b Q (Id , 0) = 0. On account of the previous considerations it is then easy seen that in local coordinates the integrand of the Dirichlet integral has the form (10)
(IdwI2+ 5wI2)/dx = A«Qw«,iwQ,.7 +Bapiwa,iw,+C«QWaWQ
A" (G(x)), BaRi(x) = BaQi(G(x), DG(x)), CCQ(x) _ Cap(G(x),DG(x)) and
where
Aaa(Id)
BaQi(Id, 0) _ C«Q(Id, 0) = 0.
2.5 Representing Cohomology Classes by Harmonic Forms: Hodge and Hodge-Kodaira-Morrey Theorems We are now ready to prove the celebrated theorem of Hodge which states that every cohomology class of a compact smooth manifold without boundary contains exactly one harmonic form. More generally we shall prove a decomposition theorem for forms, Hodge-Kodaira-Morrey theorem, from which Hodge result follows. The proof, based on variational methods and more precisely on Dirichlet principle, shows the entire theory of linear elliptic partial differential equations at work.
In this subsection X is a compact, possibly non orientable, boundaryless Riemannian manifold of dimension n that is thought without loss in generality as isometrically embedded in some RN. We assume that X is a CO° manifold even if one can work on C""l manifolds with trivial changes. One of the key ingredients of the variational proof we are going to present is the following simple but important lemma
Lemma 1 (Gaffney). There is a constant C such that for each point y E X there exist a ball B(0, p) C ]R' and a chart (U,,, cp), cp : B(0, p) -+ Uy C X, cp(0) = y, such that
D(w) ?
2
f
I2
Diwa2 dx - C11 w L2(U,)
B(O,p) 2'a
whenever cp#w =
wa dxa has support in U,,.
Proof. Fix a point y E X and choose coordinates so that the metric tensor at y is the identity, gij(0) = bij. Then using the representation (10) in Sec. 5.2.4 for D(w) we infer D(w)
>
f A'Q(0)w«,iwQ,j dx - J[A(0) - f IBaQi(x)I w«,iIIwQldx - J
ICaQIIWal
dx IwQldx.
5. Cartesian Currents in Riemannian Manifolds
544
As
B(0, p)) = L(O,P)
if A,a(0)wa,tiwg,j dx =
IDwl2 dx, i,a
by (8) in Sec. 5.2.4 the result follows at once choosing p sufficiently small so that A'3 (0)1 be small. El A%3Q
Choosing a finite cover of X by charts for which Lemma 1 holds we readily infer that (i) D(w) +cll w lli2(x) is an equivalent norm on W1"2(X, APTX) for a suitable
constant c depending on the covering, and in fact there is a positive constant cl and a non- negative constant c2 such that (1)
2
2
D(w) >- cl 11 w IIw 1,2 - 0211 w IIL2
and, being ll w II L2(x) continuous with respect to the weak convergence in W1,2 by Rellich theorem, (ii)
D(w) is lower semicontinuous with respect to the weak convergence in W1"2 (X, APTX).
Definition 1. We say that a p-form w E W1,2 (X, JIPTX) is harmonic if dw = bw = 0. The linear space of harmonic p forms is denoted by HP.
Theorem 1. HP is finite dimensional. Proof. Otherwise we could find a sequence {wk} C HP which is orthonormal in L2. As ll wk IIL2 = 1 and D(wk) = 0 a subsequence will converge strongly in L2, which is impossible since wk is orthonormal in L2.
Theorem 2. There exists a positive constant c such that (2)
D(w) > cl w I12 2(x,APTx)
for any w E W1,2 (X, APTX) which is orthogonal to HP. Proof. Otherwise we find a sequence Wk E W1,2 (X, APTX) n HPl such that 1) (LOO < kll wk liL2.
If 77k := wk/I1 wk IIL2, we have 11 77k IIL2 = 1, D(77k) < 1/k. Passing to a subse71 weakly in W1'2, Ink - rl strongly in L2, hence 1171 IIL2 = 1, by quence 77k
semicontinuity 71 E HP' and D(77) = 0, i.e., 77 E HP fl HP' or in other words 71 = 0: a contradiction.
2.5 Hodge and Hodge-Kodaira-Morrey Theorems
545
Given a p-form f E L2 (X, APTX) we are now interested in finding a weak solution of the equation
aw = f, that is a p-form w E W1,2 (X, APTX) such that (3)
(
drl)L2 + (8w l'rl)L2 = (f 7)LZ
V 77E W1,2(X, APTX).
We have
Theorem 3. Equation (3) is solvable if and only if f belongs to the orthogonal in L2 (X, APTX) of HP, denoted by HP. Moreover there exists a unique solution wo in W1"2(X,APTX) n HP1 and the linear transformation from f to wo is a bounded linear transformation from L2 (X, APTX) into W I,2 (X, APTX). Proof. Clearly if w satisfies (3) then f is orthogonal to HP. Conversely assume
f E L2 (X, APTX) n HP1 and consider the integral
F(w) := D(w) - 2(f I U;) L2 defined on W1"2(X,APTX) n HP1. Trivially F is lower semicontinuous with W1,2, W1,2 respect to the weak convergence in n HP1 is weakly closed and, by Theorem 2,
F(W)>CIIwIIL2-21If
IL2IIWIIL2? 2IIWIIL2
- III IL2.
Therefore, taking also into account (1) we immediately infer that there exists a minimizer wo E W1"2 (X, APTX) n HP. From the minimality of wo we now infer (4)
0 = dtF(wo + t() _ (dwo I d() + (Swo 15()
- (f I
for all C E Wl°2(X, APTX) n HP' and hence for any E W1"2 (X, APTX) since any such C can be written uniquely in the form ( = h+(o where Co is orthogonal to HP and h E HP. From (4) we also get D(wo) < 11f IIL21IwoIIL2
which together with (1) and (2) proves the boundedness of the map f -+ wo. Finally, if w1 is another solution of (3) we have for (:= wo - wl I<12 + Ij(I2 = 0 i.e. wo - wl E HP, and if w1 also belongs to HP1, wo - wl = 0.
Equation (3) is the Euler equation of the integral D(w) - 2(f I w)L2
therefore on account of the representation (10) in Sec. 5.2.4, we see that in local
coordinates the coefficients of any solution of (3) are in fact solutions of an elliptic system. The regularity theory for elliptic systems then applies, and in particular we can infer
546
5. Cartesian Currents in Riemannian Manifolds
Theorem 4. Let X be of class C1'1 and let wo solves (3). We have (i) If f E L2 (X, APTX), then wo E W2'2(X, APTX).
If f E LP(X,APTX), then wo E W2'P(X,APTX), p > 1. If f E C°'N'(X, APTX), then wo E C2,4 (X, APTX).
Higher regularity follows, if we assume higher regularity of X and f. For instance, if X and f are of class C°° or analytic, then wo is C°° or analytic. (ii) In particular we see that w is weakly harmonic ifw is smooth and harmonic.
For each p set now
Imd = {da Im5 = {8a H = {a E
a E W1'2(X, AP-'TX) } a E Wi'2(X, AP+iTX)} Wi'2(X,AP-iTX) da = 8a = 0}. I
Then we have
Theorem 5 (Hodge-Kodaira-Morrey). The linear subspaces Im d, Im 6, H of L2 (X, APTX) are mutually orthogonal and closed. Moreover
L2 (X, APTX) = H ®Im 8 ®Imd, i.e., any w E L2(X, APTX) decomposes uniquely as
w=h+8a+d,Q
(5)
with a E Wi'2(X, AP+iTX), /3 E W12 (X, AP-'TX) and h E H, and (6)
IIhIIL2+ aIIw1.2+II/0IIw1,2
and in fact, if X is of class CO°, h E C°°, and for all k (7)
II a IIW1,k + IIQ IIW1,k
Proof. Let w E L2(X, APTX). Denote by h its projection on H which is finite dimensional. Then w - h E H -L and by Theorem 3 and Theorem 4 there is wo E W1"2 (X, APTX) such that
. wo=w-h. Writing a := dwo, /3 = 8w°, (5) and (6) follow at once. Let us prove now that the three subspaces Im 5, Imd and H are mutually orthogonal and closed, from which one easily infer that the decomposition (5) is unique. Let w E H, for a E Wi'2(X,AP-iTX) we have
(dalw)=-(a,bw)=0 and for 0 E W1'2 (X, AP+iTX)
2.5 Hodge and Hodge-Kodaira-Morrey Theorems
547
(5aIw)=-(aIdw)=0; while we also have
(da 15/3) = -(d2a /3) = 0 provided a belongs to W2'2. If a belongs to W1"2 we choose a sequence ak E W1,2
such that ak -> a strongly in W1,2 and we get (da 15p)
lim (dak I SO) = 0.
k
00
Finally let us prove that Imb is closed. Let wk := b77k, 17k E W1,2 (X, AP+'TX), and assume that wk -> w in L2. By (5) we have wk = hk + Sak + dQk,
the uniqueness of the decomposition yields Sak = Wk, moreover 11 ak I I W 1.2 <- C11 wk I
I L2 < cont.
Therefore we infer passing to a subsequence that
ak -k a weakly in W"2,
ak -> a strongly in L2,
and a E W"2. Consequently ba = w, which proves that Im S is closed. Similarly one proves that Im d is closed.
Before stating Hodge theorem, let us recall the definition of the De Rham cohomology groups . A smooth p-form a is called closed if da = 0, exact if there
exists 77 with d77 = a. Because of d o d = 0, exact forms are always closed. Two closed forms a,,3 are called cohomologous if a - /j is exact. This property determines an equivalence relation on the space of closed p-forms and the set of equivalence classes is a vector space over R, called the p-th de Rham cohomology group and denoted by HHR(X)
Theorem 6 (Hodge). In every cohomology class in HHR(X) there exists a unique harmonic p-form. Proof. Let wo be a closed differential form representing a cohomology class. All forms homologous to wo have the form
w=wo+d3; as wo E Im b1, we can write by the decomposition theorem wo = ho + d/30
where ho is harmonic, therefore every w cohomologous to wo has the form
w=ho+d(,3+,do).
548
5. Cartesian Currents in Riemannian Manifolds
It is worth noticing again that if w = h + d/3 is the decomposition of a closed form w, h being harmonic, we also have, assuming X of class C°° II 0Ilwk+l,r < Ck11 W Ilwk.r
J1 Q I1Ck+1,µ < CkI W 11C-,µ.
Later we shall be interested in closed forms in the product manifold f2 x X, where Q is an open set of an oriented Riemannian manifold. The product structure in ,(2 x X induces a canonical splitting of the exterior differential operator
din,(2xXas
d=d,+dy
with respect to local coordinates (x, y), x E ,(2 and y E X and the splitting of DP (f2 x X) as direct sum min(p,n)
Dp(,(2xX)_
®
Dp,k(,RxX)
k=max(O,p-n)
where Dp,k (,R x X) denotes the class of p-forms in the product ( x X with exactly k-differentials in y. A simple consequence of the decomposition theorem is the following result which will be used later
Proposition 1. Let w be a compactly supported p-form in 0 x X which has exactly f-differentials in y, w E Dp,e(.(2 x X), and is dy-closed, i.e., dyw = 0. Then w can written as s
w = 1 cps(x)no" (y) + dyrl (x, y) s=1
where i is a compactly supported (p - 1)-form in .(2 x X with at most £ - 1
[as], Qi harmonic, form a basis of
differentials in y, cps E VP-e, and the cohomology group HHR(X).
Proof. Choose an open covering {Ui} of U so that on each Ui there exists a
basis of simple tangent (n - P)-vector fields eay,i. Denote by i the dual form of ea,i and by {iii} a decomposition of unity associated to the covering {Ui}. On each Ui i jw can be seen as ,.Ijiw
77anwa
where w E CO° (Ui, V (X) ). Thinking of x as a parameter, Theorem 5 applied to yields s
(x) _
cpa's (x)os + dy?7a s=1
with cpa's E C:°(UU) and 77i E C°O(Uj,DP-1(X)). The result then follows summing on i and a.
2.6 Relative Cohomology: Hodge-Morrey Decomposition
549
2.6 Relative Cohomology: Hodge-Morrey Decomposition Let X be a smooth compact, possibly non orientable, Riemannian manifold of dimension n with smooth boundary. We think that X is isometrically embedded in an Euclidean space R' +N and we also assume that X is a C°° manifold with a C°° boundary aX even if our presentation extends to C1,1 manifolds with Lipschitz boundary with only trivial changes. For x E aX, let v(x) be the inward normal vector to aX at X, i.e., the unit vector v(x) E TTX orthogonal to TxX at x and pointing inside X. We infer the existence of special charts for X near each point x0 E 9X. For p > 0 set B+(0, p) := {x = ( x , , . .. , x.) I xI < p, x,, > o} F (O, P) :_ {x E R'-' (x', 0) E B(0, p), x' :_ (x1, ... I
I
Each point x0 E aX has then a chart cp : B+(0, po) --f X such that cp(0) = xo, : T(0, po) --> aX is a local chart for aX, and, for x' E F(0, po), Dgo(x', 0)e,,, = v(co(x', 0)) is the inward unit normal vector field to aX. We refer to these charts as to admissible charts. .
Tangential and normal traces of a form. Let us define the tangential and normal parts to an of a form. Denote by t the restriction map
r : C°° (X, APTX) --+ C°° (aX, APTX ) which associate to each w E C°° (X, APTX) its trace rw E C°O (aX, APTX), rw(x) = w(x), x E aX. As aX is a submanifold of X, the Riemannian structure on X yields a canonical decomposition of TxX in two orthogonal components TxX = Tx5X EB l v(x).
Such a decomposition then extends to p-vectors, and we can decompose every p -vector l; E APTXX in a unique way as
6 = i + 62nv,
61 E APTxax, S2 E AP-1TxaX.
Thus, give a form w E C°° (X, APTX) we can define the tangential and the normal parts of the trace of w, respectively tw and nw, as the functions in C°° (aX, APTX) defined by (1)
_
for
= 61 + 6 Av(x). Obviously we have rw = t w + n w and
(2)
IrWI2 =
ItwI2
+ InwI2
Also, if i : aX --> X is the immersion and it : X fl UEo -* aX where U60 is a neighborhood of aX is a smooth right inverse of i, then
550
5. Cartesian Currents in Riemannian Manifolds
tw(x) _ 7#ti#w(x).
(3)
In local coordinates (x', ... , X1, t) around xo E 8X we can write w E Dr (X) as w _ E1a1=P wao (x, t) dxa + E1a1=p-1 w,,,6 dxaAdt, then
t w(x, t) _ E wao (x, 0) dxa,
nw (x, t) =
wao (x, 0) dx' Adt. 1a1=P-1
IaI=P
We recall now that the restriction at the boundary f -> flan of smooth functions f defined on a Lipschitz domain .f2 extends to a continuous linear
operator from W',8(0) to LS (8Q) for s > 1. Based on that it is then easy to see using local coordinates that the trace operator r : C°° (X, APTX ) C°° (aX, APTX) extends to a continuous linear operator, the trace operator
r : Wl,'(X) -> L'(aX)
for anys>1. From (2) we then infer that the tangential and normal operators extend to linear continuous operators
t,n : WW'S(X) -* LP(X)
for all s > 1. In an oriented manifold, the associated Hodge operator * commutes tangential and normal operators. We have
Proposition 1. Let X be an oriented manifold, let * be the Hodge operator on X, and let S be the formal adjoint of d. Then we have (i) Forw E DP(X) *nw = t (*w),
*t w = n (*w),
t (dw) = d(t w),
n (bw) = 8(n w).
(ii) For wEDP (X),rlED4(X),q
which is also orientation preserving. In such a system, if w = wao dxaAdt as above, one easily computes
wao (x, t) dxa +
o(c , a)wao(x, 0) dxandt = n (*w)
*t W = Ia1=P
*nw =
Q(c , a)
O (x, 0) dxa = t (*w).
lal=P-1
Moreover, from (3) we infer
t (dw) = lr#i#dw = d(7r#i#w) = d(tw).
2.6 Relative Cohomology: Hodge-Morrey Decomposition
551
Equality nSw = Snw then follows as 6 = (-1)P("-P) * d*. Finally, again from (3) we infer t (WA * 77) = 7r*i# (wn * 77) = 7r#i#wAlr#i# * 77 = t wAt (*77) = t wA * n77.
Recall that if X is an oriented manifold, then one fixes an orientation on OX by choosing aX (x) as the unit (n - 1)-vector field orienting TXaX in such a way
that (x) = X(x) We then set for y E C°°(0X, A'-'TX) v(x)A
'Y
ax
f
for X E aX.
> dH"'
<'Y,
ax
Proposition 2. Suppose that X is oriented and let * be the associated Hodge operator. We have (i) If w E W,1,'11(X ), then fx dw = fax i#w (ii) If W E Wp,2(X) and 77 E W112 (X), then (
= fax t w.
I 77)Lp}1(X) + (w I S7I)LP(X) _ f twA * n?]. ax
(iii) If W E WP2'2 (X) and C E W 2 (X ), then (
d()L2P}1 (X) + (6w 16C)Lp_1(x)
= -(nw I OLP(y) + f {t(A *n &-,) + t Swn * n (}. ax
Proof. Since x"' LX-a.eT#W(x) =rW(x)IAPT ox and i#W(x) =tWIAPT,aX, (i) follows from a similar claim for smooth w, on account of the continuity of the trace and of the tangential part operators. To prove (ii) we observe that wA * 77 belongs to Wn' 11 (X), and d(wA * 77) = dwA * 77 + (-1)Pwnd(*77) = dwn * 77 + wA * 877.
Integrating we then get on account of (i) (dw 177)Lz + (w I S77)L2 =
d(wA * 77) = ax
f ax
J x
dwA * 77 + J wA * 677
x i# (WA * 77) = f i#wni# ex
* 77.
552
5. Cartesian Currents in Riemannian Manifolds
On the other hand i#w = t w I A,TaX, hence i#wAi# (*7)) = t wAt (*77) I APTX = twA * bn 77I ApTaX .
Finally (iv) follows at once integrating by parts, (zlw I () = (ddw I ()+(8dw I 11
We now observe that Proposition 2 extends to non orientable manifolds. In fact as any non orientable manifold X is nevertheless orientable in the small, the scalar function
(x) := ,
x E 8X
is well defined if either a E C°° (X, APTX) or 0 E Coo (X, AP+'TX) are supported in a coordinate patch U, and is independent on the local orientation on U. Therefore < a, /3 > is a well defined function in CO° (8X) even if X is non orientable.
Proposition 3. We have Wp,2 then t (dw) = d(tw), n (bw) = b(nw). (i) If W E then (ii) If w E WP(X) and 77 E
f
(dw I fl)LP+,(X)+(w I Si)L2(X) =
dHn-1
ax
(iii) If W E Wp'2(X) and r) E WP+2(X), then (dw I b?7)L2P+, (X)
=0
provided either t w = 0 or n7? = 0. (iv) If w c Wp'2(X) and 77 E Wp"2(X), then (dw I dr))LP+,(X) + (8w I drl)Lp_,(X)
_ -(nw I
77)Lp(x) +
J ax
{+ } d7-l'i-1.
Proof. (i) follows from (i) of Proposition 1 as the claim is local. (ii) and (iv) of Proposition 2 proves respectively (ii) and (iv) if we assume moreover that rl is supported in a local chart. The general cases of (ii) and (iv) then follows using a partition of unity. Finally if tw = 0 and 77j ri in W1"2, the r)j being smooth, we have by (ii)
(dwI drlj)_(dwI arli)+(wI
6277j)=
Jd?i'-1=0. ax
(iii) then follows letting j tend to infinity. Similarly one proceeds if nr) = 0 by approximating w.
2.6 Relative Cohomology: Hodge-Morrey Decomposition
553
The Hodge-Morrey decomposition theorem. A key step toward a decomposition theorem similarly to the boundaryless case is a Gaffney type lemma at boundary points. We have
Lemma 1 (Gaffney). For each xo E aX there are constants c, p > 0 and an admissible chart cp : B+(0, p) --p U C X, (0) = xo, such that
f
D(w, U) >
(4)
E JDiwa12 dx - cII w IIL2(U)
B+(O,p)
i,a
whenever cp#W = E wa dxa is compactly supported in B(0, p) and we have either
tw = 0 or nw = 0 on 6X. Proof. Proceeding as in the proof of Gaffney lemma in the interior, we can bound from below the Dirichlet integral D(w, U) by the flat Dirichlet's integral Do(cc#w, B+(0 p)) getting f D(w, U) > lDiwa12 dx - c1l w 11L2(U) p)) 2
J
B+(O,p) i'a
The coercivity inequality (4) then follows at once from the following lemma.
Lemma 2. We have
Do(w, B+(0 p)) = f
IDiwal2 dx,
i'a B+(o,p)
if w = E wa dx' is compactly supported in B (0, p) and moreover either t w = 0
ornw=0 onFp. Proof. From Proposition 2 (iv) we have (5)
Do (w, B+(0, p)) + f (dW I w) = B+(O,p)
r
J
{f wA * n dw + t bwn * n w}.
rv
On the other hand in Rn the Laplacian on forms is just the Laplacian on coefficients, i.e.,
LXW = E / wa dxa, a
consequently, integrating by parts and using Gauss-Green formula
f B+(o,p)
we infer
Digi dx =
f aB+(o,p)
E(-1)x_1gi
i
dxi
V g E C'(B+(O, p)),
5. Cartesian Currents in Riemannian Manifolds
554
(6)
(aw j w) dx
r B+(O,p)
IDw12
I
+
WaWa,n dxlA...
(-1)n-1
B+(O,p)
/ndxn-1.
1p
From (5) and (6) we then get (7)
Do(w,
B+(0, P)) -
IDw12 dx
J
B+ (O,p)
= f{twA *n dw+tbwn *nw+ (-1)n r, wawa n dx1A...
Adxn-1}.
a
rp
We now claim that the right hand side of (7) vanishes if either tw = 0 or nw = 0
on I''. In fact, if tw = 0, then, in I'p we have w = nw =
wa dxa, consequently a3n
t 8w
EE
o (i, a - i)Wa i
dxa-'=(-1)P-1 I
a-i3Jn a3n i
Hence t bwn * nW =
a3n
` 1` wQwa,nu(,a, ) dxa-nAdxQ.
(-1)P-1
3 n, we have dxc
(_i)P-1
twA * nw
wan dxa-n.
a3n
Adxa 54 0 if and only if a = 0, therefore
wawa na(a, a)o(a - n, a) dx1A... a3n
(-1)P-l+n-P
ndxn-1
wawa,n dx1A ... a3n
=
(-1)n-1
E WaWa,n dX1 A ...
ndxn-1
a3n which proves the claim in this case.
Ifnw=0, then in f wehavew=tw=
wadxa, a on
n6w =
E 1: a i, a)wa,i dxa+2 = (-1)P a+i3n aln
Wa,n dxa+n
aon
and
*ndw = (-1)P
Wa,nQ(U + n, a - n) aRfn
Therefore
d?-n.
ndxn-1
2.6 Relative Cohomology: Hodge-Morrey Decomposition
555
t wA * n dw = (-1)P E E wpwa,no-(a + n, a - n) dxandx«-n. )33(n
a
n, we have dxQAdxa-n = 0 unless a
hence
twn * ndw
_ (-1)P E wawa na(a,
a)(-1)n-p-10'(a,
a - n) dx1A... Adxn-1
aon
_
(-1)n-1
EWawa,n
dxlA...
Adxn-1
a5Jn
which concludes the proof.
Gaffney's lemmas in the interior and at the boundary now yield at once
Theorem 1. Let X be a compact Riemannian manifold with boundary. There exist constants cl > 0, c2 > 0 such that D(w) > cl II w IIwp.2(X) - c2II W II2P(X)
for every w E WP,2(X) satisfying one of the boundary conditions
nw = 0 on 8X,
tw = 0 on 8X.
Set now
WPT(X) W12
(X)
HP (X) HPN (X)
{wEWP'2(X) I tw=0} :_ {w E Wp'2(X) I nw = 0} {WEWP"T(X) I dw = 0, dw=0} {W E WP, N (X)
I dw = 0, Sw = 0}.
Then, as in the boundaryless case, one proves
Theorem 2. HT and HN are finite dimensional vector spaces. Theorem 3. There exists a positive constant c such that D(w) > CII W II L2(X) for any w E WP'N(X) which is orthogonal in L2 to HN and for any w E WP'T(X) which is orthogonal in L2 to H. Given a form f E L2 (X), we can now seek a weak solution of the problem
J '6w = f in X (8)
nw=0 onBX
that is we seek a form w E WP,'N (X) such that
5. Cartesian Currents in R.iemannian Manifolds
556
(
(9)
dC)L2 + (SW I SC)L2 = (f I ()
or a weak solution of
V
C E WP,'N(X)
J aw= f inX
(10)
tw=0
on &X
equivalently w E WP 'T (X) such that (11)
(dw I d()L2 + (SW SC)L2 = (f C)LZ
V C E WP'T(X).
As in the boundaryless case, an easy consequence of Theorem 1 and Theorem 3 is the following
Theorem 4 (Hodge-Morrey). We have (i) Problem (8), equivalently equation (9), is solvable if and only if f is orthogonal in LP to HN . Moreover there is a unique solution wo E WP'N (X) n HN
of equation (9), and the linear transformation from f to w° is a bounded linear operator from LP (X) n HN into WP'N(X) n HN . (ii) Problem (10), equivalently equation (11), is solvable if and only if f is orthog-
onal in LP to H. Moreover there is a unique solution w° E WPT(X) n HT of equation (11), and the linear transformation from f to w° is a bounded linear operator from LP(X) n HT into WP'T(X) n HT . Problems (8) and (10) are easily seen to be elliptic boundary value problems, where one has zero Dirichlet data for suitable coefficients of w and zero Neumann
data for the others. Elliptic regularity theory then allows to complement the existence theorem in Theorem 4 with the following Regularity Theorem
Theorem 5. Let w in WP'N (X) or in W' 2 (X) be respectively a solution of (9) or of (11). Then we have (i) w E WP2'2(X) if f E LP(X). (ii) w E WP's(X) n CO, ` (X, APTX), a = 1 - P, if f E L; (X), s > n. (iii) w E C2,- (X, APTX) if f E C°Ia(X, APTX). (iv) Higher regularity of w follows from higher regularity of f, for instance w E C°° or w is analytic if f E C°° or is analytic. (v) In particular harmonic normal forms, i.e., w E HN, or harmonic tangential forms, w E HT, are C°° or analytic up to the boundary if X and aX are of class C°° or analytic. Define now
ImdN := {da ImSN := {Sa then we can state - ---------
a E WP'21,N(X)} ImdT aE
W1,2
N(X)} ImST
{da I a E WP'21,T(X)} {Sa I a E
---------------------------
- - - - ----_
2.6 Relative Cohomology: Hodge-Morrey Decomposition
557
Theorem 6 (Hodge-Morrey decomposition). We have (i) The spaces HN, Im dN and IM SN are mutually orthogonal and closed in
LP(X), and one has the decomposition
L2r(X) = HN ®ImdN ®ImSN, that is, every f E L2 (X) decomposes uniquely as
f =h+da+Sf
(12)
where h E HN, a E Wp 1,N(X), /3 E WP+1,N(X) and IIhIJW,,2+IIaIIw1.2+II13 IIWI,2 <_ cII f IILZ'
(ii)
The spaces HT, Im dT and Im ST are mutually orthogonal and closed in L2r (X), moreover
L22(X) = HT ®ImdT ®ImST, i.e., every f E LP2 (X) uniquely decomposes as
f =h+da+SQ
(13)
where h E HT, a E W l T(X), 3 E
and
IIhIIW1,2 +IIa1W1.2+IIQIIW1,2
Moreover in both cases h E C°° (X, APTX) and also a and Q are of class Cup to the boundary if f is of class C°° up to the boundary. Finally, SQ = 0 if
df =0. Proof. The proof follows the same path of the analogous proof of the decomposition theorem in the boundaryless case: one has only to take into account the formulas of integration by parts in Proposition 6 in Sec. 5.1.3. For the reader's convenience, we only prove (12). Let h be the projection (in L22) of f into HN and let wo E Wp'N (X) n HN be the unique solution of (14)
(dw I d()L2 + (SW I SS)L2 = (f I ()L2
V
E Wp'N(X )
which exists by Theorem 4. By the regularity theorem wo actually belongs to Wp'2, therefore (iv) of Proposition 3 yields (15)
(SW +J{
(dw I dC)L2
+
rr (QW I S)L2
+} dl-C-1
ax
for all C E WP'2(X). From (14), (15) we infer by taking C E C°°(X, APTX) that
5. Cartesian Currents in Riemannian Manifolds
558
Awo=f - h i.e.,
f = h + da + 60 if a := dwo, /3 := Swo E WP+1 (X) . As n a = n Swo = Sn wo = 0, we see that a E Wp'21,N(X). From (14), (15) we also infer where a E WW_21(X ),
J ax
< t C, n dwo > dx" = 0
V C E Wp,'N (X ),
and, being t( arbitrary on X, that n dwo = 0, hence no = it dwo = 0, i.e. 1,2
E WP+1,N(X)-
Remark 1. It is worth remarking that, since Hodge operator transforms pharmonic forms into (n - p)-harmonic forms and
*nw=t(*w)
*tw=n(*w),
actually on oriented manifolds Theorem 4 (ii) follows from Theorem 4 (i) and Theorem 6 (ii) follows from Theorem 6 (i). Remark 2. One might consider other boundary value problems for the LaplaceBeltrami operator and consequently infer other decomposition theorems. However we shall not pursue this topic any further. For k = 0, ... , n, define
Zk(X, aX) :_ {w E Co(X, AkTX) I i#w = 0, dw = 0} to be the vector space of relatively closed k-forms, i.e. the space of closed k-forms which vanish on aQ, and
Bk(X,OX) :_ {w = da I a E CO(X,.AkTX), i#a = 0, or w = 0 if k = 0}. the vector space of relatively exact k-forms. The relative de Rham cohomology groups are then defined by
HHR(X, aX) := Zk(X, aX)1Bk(X, aX). An immediate consequence of Theorem 6 is
Corollary 1. We have (i) Each cohomology class [w] E HHR(X) contains exactly one harmonic form h
with n h = 0. In particular HHR (X) is finite dimensional. (ii) Each relative cohomology class [w] E HHR(X, 9X) contains exactly one harmonic form h with t h = 0. In particular HHR (X, OX) is finite dimensional.
2.7 Weitzenbock Formula
559
2.7 Weitzenbock Formula In this section we derive a more intrinsic expression for the Laplace Beltrami operator on a Riemannian manifold X in terms of the covariant derivatives and of the curvature of X. Denote by r(TX) the smooth vector fields on X and if p : E -* X is a fiber bundle on X, denote by r(E) the space of sections over E. A connection V on E is an operator
denoted by Vxt; for X E V(TX) and
E r(E) which satisfies the following
properties
(i) V X t; is tensorial in X i.e.
Vx+Y = OX6+VYe,
OfXe = fvXe
for x, y E r(TX ), 6 E r(E), f E C°° (X ). (ii) Vx6 is linear in iii) Vxl; satisfies the following chains rule
fVxe
Vx(fe) =
(1)
for X E r(TX), f E C°°(X), 6 E r(E). Notice that (i) implies that Vxl; at a point x E M depends only on the tangent field X at point x and moreover that V can be also seen as an operator
V : r(E) -* r(E) ® r(A'TX). Moreover, (iii) implies that Vxe(x) depends only on t= near x. Given a connection V on a fiber bundle E one defines a connection V * on the dual bundle E* by the formula
w> _
(2)
for e E r(E), w E r(E*), X E r(TX) and also on all exterior algebra of E. In fact sections
E r(11pE) can be represented locally as
6=
S°(x)eai
(x)A...
Ae%(x)
aEI(p,n)
where a c Coo (X) and eat, ... , eaP are smooth sections in E. Consequently, by setting (3)
Vx f (x) :_,
(4)
Vx(t 1A... Abp)
P
i=1
f E C°°(X)
S1^ ... Si-1^OX6ti^Ci+1A ... Al p
560
5. Cartesian Currents in Riemannian Manifolds
E N(E) we define an operator
for
v: N(APE) ®T(TX) --> T(ApE)
which is easily seen to be a connection. Starting with V* on N(E*) a similar construction produces a connection on APE*
APE*
P
Vx(w'A... AW) :_ Ew1A... Awz-1AVxw'Awx+'A... nwP,
(5)
i=1
Proposition 1. Let 6 E N(APE), o E T(AgE), w E N(APE), 77 E N(APE). We have (6) (7)
OxeAo + eAV x o Vx(wAr7) =Vxwnr7+WAVxr7
(8)
<w, Vxe> += , X >.
Vx
Also, if p < q
Vx(e-J 77) _ CJ Vx?7+Vxl;J 17.
(9)
Proof. Let e1i... , e, be a local basis of sections near xo and fl, ... , fn be the dual basis in N(E*) so that < fi (x), ej (x) > = Jij
near x0.
If a E I (p, n), 0 E I (q, n), from (4) we infer Vx (e,,,Aep) = Vxe,,,Ae,3 + e,AVxeo hence (6) follows. Analogously (7) follows from (5). To prove (8) notice that
V*fa(ea) = f«(Vxeo) = 0 if a
,3
while for a = Q P
Oxfa(eo)+f, (Vxe«)
{Oxfck i(eai +fai(Vxe«i)} i=1
,X>=0. Let us prove (9). Using (6) and (8), for o E T(Aq_pE)
V_ 77)(o,) = -eJ 77(V xo) + < d<77, one >, X > = -77(OxoAe) +
X>
= -77(Vx(o-ne))+,X>+
_ C-J V*77(o) + Vxe-J 77(C)
2.7 Weitzenbock Formula
561
A basic result on Riemannian geometry is that there is only one connection V on TX, the Levi-Civita connection , which satisfies the following additional requirements (1) V is metric, i.e.
(VxYIZ)+(YIVxZ) _VX,Y,ZET(TX) (10)
(2) V is torsion free, i. e. for X, Y E T (T, X) (11)
T(x, y) =VxY-VyX - [X, Y] - 0
being [X, Y] the Lie bracket of X and Y. In this case, d
[X, Y] := X3
aXj
ei - Yj axi e2
where X := Xiei, Y := Yiei.
One can compute V in terms of the metric on T (TX ), and one has
(VxxlZ) = 2 {x(X Z) - Z(XJX) + x(XIZ) (12)
-(XI[X,Z])+(ZI[X,Y])+(XI[Z,X])}.
In local coordinates, denoting by Gij the metric tensor of X and by e1, ... , en a local basis for I'(TX), we have n
(13)
>rkj ek
Deyej
k=1
where {T kj} are the Christoffel symbols of the connection in the given basis; (12) is then equivalent to (14)
T = 2Gk1{Gil,j + Gj(,i -
Gij,1}.
The torsion free condition (11) implies
T =rk
di,j,k.
Proposition 2. Let X be a Riemannian manifold. Denote by V the Levi-Civita connection in X and also its extension to the algebra of p-vector fields and forms defined on an open set U c X and let dx...... dxn be the dual frame of p-forms. Then dX AVej W in U. (15) dW If moreover ej are orthonormal at xo E U then at xo (16)
dW = >ejJ Ve, w.
j
5. Cartesian Currents in Riemannian Manifolds
562
Proof. The idea is to prove the invariance of both sides in (15) and (16) with respect to changes of coordinates and then proves (15) and (16) in a suitable frame.
Let us prove (15). First we note the the expression
E dxj AVejw
j is independent of the choice of local frame In fact if fl,... , f., is another frame field, we have f j = A ek for some A E COO (U, ]R). For the corresponding dual frame f 1, ... , f' we then have f j = Bkdxk with B := A-1. Consequently
f3AOfjw =
j
8kdxkAVe=w = dxiAV,,w.
w= j,k,i
i,k
i
Since d is also independent on the frame field, it suffices to check (15) for a special frame field. We select around x0, normal coordinates for which we have Gij(xo) = 5ij, Gij,l (xo) = 0 V i, j,1. Hence for the coordinate frame fields ei,... , en, we have
Veiej(xo) = f (xo)ek(xo) = 0 from (13) (14). Thus, if w = codxa, a E I(p, n), f E C°O(X), we obtain at xo
EdxiAvej(cpdxa) _
dxa = d(c'dxa).
dxjA 9
j
To prove (16) we observe that the right hand side is independent of the chosen of
the normal frame at x0. In fact if fl, ... , fn is another orthonormal frame field, f j = A3 ek, A3 being an orthogonal matrix, A A3 = 5ik. Then we have
eiJ V.
Ve;w =
fi-J V f,w =
j
j,k,i
i
Again we chose normal coordinates at xo. The coordinate orthogonal at x0 and V ej = 0. Consequently
ejj Vet (pdxa) =
(17)
j
en) frame is
a e J dxa.
8xj
On the other hand from (1) in Sec. 5.2.4, taking into account that Gij(xo) _ Gti(xo) = V, Gij,a(xo) = 0 we have at x0 1:(5(Wdxa))7 dxl =
6(Wdxa)
(18)
_
a (i,'y) y+i=a
&p
(xi
aacpxe
dxa = iEa
o(i, a) dxa-i
2.7 Weitzenbock Formula
563
and the claim follows comparing (17) (18) since
ifi0a
0
ej i dxa = iS
o-(i, a)
dxa-2
if i E a. 71
In order to express the Laplace-Beltrami operator in terms of covariant derivatives, we introduce the second covariant derivative defined for w E F(APX) and X, Y E TX by
OXYw :_ VX(VyW) -Vvxyw and the curvature tensor associated with a connection (19)
(20)
R(X, Y) :_ VxVy -VyVX : APTX -4 APTX. is tensorial in X, Y, i.e. that OXY is linear in X
It is easy to check that and Y and for all xo E Y
oXY(W)(xo)
depends only on the values X (xo ), Y(xo) of the vector fields X and Y at xo. In fact, for f, g E C°°, X, Y smooth vector fields,
VfX(V9yw) = fVyw+fgVX(VYw) and
VVfx(9Y)W = V fX+f90xY = f Vyw + fgVvxyw so
v2fX,YW = f(x)g(x)VXYW. We need also the following algebraic lemma.
(21)
Lemma 1. Let e1,. .. , en E Al V, dx', ... , dx" the dual basis in A'V, w E APV. Then (22)
> {eiJ dxjAw +dxjA(eiJ w)} = 0
(23)
eiJ dxinw + dx1A(eiJ w) = w
ij
Proof. Assume first i
V i = 1, ... , n.
j if j E a, if j a, i 0 a 0 -i if j a, i E a. o, (j, a) o, (i, a + j - i) dxa+j
10
eiJ (dxjndxa) =
ifi a,
10 dx3A(eiJdxa)
ifiEa, jEa
0
v(i, a - i)o,(j, a - i)
dxa+j-z
if j
a, i E a.
5. Cartesian Currents in Riemannian Manifolds
564
a + j - i) = -o (i, a - i)o-(j, a - i), compare (5) in Sec. 5.2.4, (6) As v(j, in Sec. 5.2.4, we see that eiJ (dx3Adxa) + dx3A(eii dxa) = 0
`d a E I(p, n), consequently (22) is proved. To prove (23) it suffices to compute
ifiEa,
(0
eii dx"Adxa = {
dxa
if i
dxineiJ dxa =
a,
r dx' ifiEa
Sl
0
if i
,
c
,
thus
eiJ dx2Aw + dxiheiJ w = w.
Theorem 1 (Wietzenbock formula). Let X be a Riemannian manifold and let V be the Levi-Civita connection on forms. If el, ... , en is a local frame on X which is orthonormal at xo and dxl,... , dxn the dual frame on T*X, then
Qw = Deieiw + dxzA(ejl R(ei, ej)w)
(24)
Proof. By definition, Aw is intrinsic to the Riemannian manifold and depends on the Riemannian structure on X. Also we check that both terms in the right hand side of (24) are intrinsic. In fact, if fl, . . . , f, is another orthonormal frame field, then fj = A ek, A (xo) being orthogonal. Thus (21) implies Vfifiw = 1: AkAt Vekehw =
Dehehw. h
j,h,k
i
Also if fj is the dual basis of fj, fj = Bkdxk, B
R(fi, fj) _
A-1
AkA; R(ek, eh) h,k
fj i R(fi, fj) i
hkj,s
AkehJ R(ek, eh)
AsAi Aj esJ R(ek, eh) k,h
Therefore
ftin(fjJ R(fi, f3)) _
BsAkdxsn(ehJ R(ek, eh)) h,k,s,i
dxkn(ehJ R(ek, eh)). h,k
Consequently it suffices to prove (24) for a special frame. Choose normal coordinates around xo, and dual coordinate frames e1, ... , en, and dxl,... , dxn. At xo we have Veiej(xo) = 0, therefore Vei(dxj)(xo) = 0, Deieyw(xo) = VeiVeiw(xo) and R(ei,ej) =Ve;,Vej - VejVei
2.8 Poincare and Poincare-Lefschetz Dualities in Cohomology
565
Consequently, using Proposition 2 at xo (25)
dxin0ei(ejJ Vejw)
dSw =
dxin(ejJ DeyVejw)
i,j
i,j
and also using Lemma 1
Sdw = EejJ Vej(dx'AVeiw) i, j
E ejJ (dx'AVejVeiw)
- 1: dxin(ejJ Dej Veiw) i=j
i,j n
(26)
+ejJ (dxinvejVei.w)+dx2A(eiJ DeiVeiw)} i=1
E dx2A(ej-i Vej Ve, w) +
Dei Vei
i, j
Adding (25) and (26) we get the claim.
11
Remark 1. If p = 0, i.e. if w = f E C°°(X) then R(ei,ej)f = fR(ei,ej)1 = 0 because of the tensorial property of R, consequently for functions
If =Eoeieyf. 2.8 Poincare and Poincare-Lefschetz Dualities in Cohomology Compact manifolds without boundary. Let X be a compact manifold without boundary. Recalling Hodge's theorem Theorem 6 in Sec. 5.2.5 we first notice that Hodge operator * induces naturally an isomorphism between the homology classes HHR (X) and HHR' (X ), in fact since
*w=*,Aw we see that *w is harmonic if w is harmonic, hence the map * : w ---* *w defines an isomorphism
* : HaR(X) --> HHRP(X). However such an isomorphism depends not only on the orientation of X, but on the metric. An algebraic isomorphism independent from the metric between HHR (X) and the dual of H(X) is obtained as follows. Given two closed forms w E W1'2 (X, APTX) and 77 E W"2 (X, An-PTX) it is readily seen that
<w,77> :=
xJ
wnr7
does not change if we replace w and 77 by cohomologous forms, hence < , > defines a bilinear map over HHR (X) x H(X)
5. Cartesian Currents in Riemannian Manifolds
566
Poinc<, >: HaR(X) x HHRP(X) --> J R. We have
Theorem 1 (Poincare duality). The bilinear form Poinc<, > is non degenerate, that is for every non zero cohomology class [w] there is a non zero class [77] such that Poinc< [w], [77] > 54 0,
consequently HHR(X) is isomorphic to the dual (HaRP(X))* of HdRP(X).
Proof. For each non zero cohomology class that we may assume represented by an harmonic form w we see that *w represents a cohomology class in H.R P (X ) for which
J
wA*w=(w,w) #0.
The second part of the claim follows at once from Lemma 1 below as de Rham's cohomology classes are finite dimensional vector spaces.
Lemma 1. Let E, F be Hilbert spaces and let a(.,) : E x F -+ R be a bilinear map satisfying (i) Ia(u,v)I0 such that a(uo, v) > co I v I V v E F
(iii) a(u,v)=0VvEF implies u = 0 Then F ^ E*. Proof. Consider the linear map T : F -> E* defined by TT(u) := a(u, v). By (i) E- < k v IIF; also, by (ii)
T is continuous and T
IIVIIF <
IITV IIE
11uo
It follows that T has closed range, is invertible, and T-1 is continuous. It remains
to show that range T = E*. Assume not, then there is 0 E E*, 0 0 0 such that coo E E we then (OITv)E- = 0 V v E F. Representing 0 as 0(u) = (ulcpo), find 0 = (OIT,,)E` = Tv('po) = a('po,v)
(iii) then implies coo = 0, i.e.,
b v E F,
= 0: a contradiction.
Notice that in particular Theorem 1 yields
Proposition 1. Let w be a closed form in X. If fX wnrl = 0 for all rl with d77 = 0, then w is exact.
2.8 Poincare and Poincare-Lefschetz Dualities in Cohomology
567
Compact manifolds with boundary. Given a smooth compact Riemannian manifold X with boundary, we may define, as we have seen, three sets of cohomology groups depending on k = 0, ... , n: The de Rham cohomology groups k HaR(X) _ closed forms of degree k in X
exact forms of degree k in X '
no boundary condition is imposed; for w E Zk(X), we denote by [w]x the cohomology class of w,
The de Rham cohomology groups of the boundary aX k HaR(aX) _ closed forms of degree k in aX
exact forms of degree k in aX
for w E Zk(8X), we denote by [w]ax the cohomology class of w in HHR(aX),
The relative cohomology group of relative closed forms modulo relative exact forms i.e. HkdR (X
ax)
closed forms of degree k in X which are null on OX exact forms of degree k in X which are null on OX
that we have already defined in Sec. 5.2.6. The relative cohomology class of w E Zk (X, aX) in HHR (X, aX) is denoted by [w] rel.
The inclusion map i : OX -f X clearly induces a map i# : HHR(X) HHR(OX ). As forms in Zk (X, OX) and in Bk (X, DX) are just special closed and exact forms, it is easily seen that an inclusion map j* : HHR (X, OX) -# HR(X) is well defined by 7#([w]rel) = [w]x. Finally, one can define a coboundary operator b : HHR(aX) -, Hk+1(aX) by b([w]eX) _ [dw]rel
where ID is an extension of w in X. In fact, if w extends w to X, we have i#(di.-5) = d(i#w) = dw = 0
on aX,
i.e., dW E Zk+1(X, aX ); moreover, if w is another extension of w, then
and i#(w-w)=0, i.e., [dD]rel is independent of the choice of the extension w of w. Also, if w E Bk (aX ), w = da, a E Dk+l (aX ), and a E Dk+1(X) is an extension of a, we have dw = d(w - da)
and i#(w - da) = w - w = 0.
5. Cartesian Currents in Riemannian Manifolds
568
Recalling now that a sequence of vector spaces
+ Vi --f Vi+l - .. .
Vi-1
is said to be exact if for all i the kernel of fi is equal to the image of its predecessor
fi-1
ker fi = Im fi- 1,
we can state
Theorem 2. The long sequence
... ?#
HaR(X)
2#' HHR(aX) a ' HHR1 (X, aX)
'- Hk+1(X)
is exact.
Proof. For the reader's convenience we give the simple proof. (i) First we show that the composition of two successive maps gives zero. Let [w]re1 E HaR(X, aX). We have 2#(7#([w]re1))
= i#(1w)x) = [i#W]X = 0
as i#w = 0. Let [w]x E Hk(X). We have b(i#[w]x) = 6([i#w]ax) _ [di#w]re1 = [di#w]rel = [2#dw]re1 = 0.
Let [w]ax E Hk(aX). We have j#(8([w]ax)) = j#([dw]rel) = [dw]X = 0 as dw is exact in X (though not relatively exact). (ii) Suppose that i# ([w] x) = 0, we need to construct c:i so that j# ([W]re1) = [LO]X.
From the assumption [i#w]ax = 0, i.e., i#w is exact in OX, i#w = dcp in OX. We extend cp to cp in X, define W = w - cp, and verify ill (Lo
- dip) = 0,
j#([w]re1] = [W -
x = [w].
(iii) Suppose that [w]ax E Hk(DX) and 8([w]ax) = 0. Extend w tow in X, i#iv = w; since [dZ]re] = a([w]ax) = 0, dw is relatively exact, i.e., div = dcp with i#cp = 0. If we define = w - cp we have
dw=di3-dcp=0 i#[W]x = [i#w]ax =
[i#cp]ax = [w]ax.
(iv) Finally, let [w]rel E HHR(X,9X) and j#[W]rel = [w]x = 0, W being relatively
closed, du) = 0 and i#w = 0. Then w is exact in X, w = dep, and [i#coiax E Hk-1(aX), as di#cc = i#w = 0, 8(i#cP]ax) = [dcP]rei = [w]rel
2.8 Poincare and Poincare-Lefschetz Dualities in Cohomology
569
Returning to harmonic forms, we recall Remark 1 in Sec. 5.2.6. i.e., that Hodge * operator yields an isomorphism *:xN-'HTk
we easily infer that HaR(X) and HaR k(X, aX) are isomorphic. However the duality given by
* : HdR(X) - HHRk(X,aX) depends on the metric (and on the orientation) of X. On an oriented manifold, an algebraic isomorphism between HdkR(X) and the dual of HHRk(X,aX) is instead given by the map (w, 71) - JwAm
x It is in fact easily seen that fx (w + d17)A(C + do,) = f x wA if w E Zk (X), ii E Dk-1(X), e E Zn'-k (X, aX), and o, E Bn-k-1(X, aX). Therefore such a map induces a bilinear pairing Poinc< [w], [c] > : HaR(X) X HaR k (X, aX) -+ R
between the k-cohomology and the relative (n - k)-cohomology. We have
Theorem 3 (Poincare-Lefschetz duality). The pairing
Poinc<[],[]>:HHR(X) xHaRk(X,aX) _+R given by Poinc< [w], [l;] > :=
J x
wA
- HaR(X,oX)
'-
is non degenerate. Moreover the diagram i#
HHR(X)
'+
x Poinc b
HHRk(X,aX)
HHR'(OX)
x PoincT
'#
x Poinc i#
HHRk(X)
...
HHRk(aX)
b ...
is commutative. PoincT denotes the transposed duality, PoincT< 1771 r'!1' [w] > Poinc< [W], 1771 ]rel >
The commutativity of Al
xF1 A2
B1
xF2 B2
means F2(f(a,),b2) = F1(a,,g(b2)), Val E A1, Vb2 E B2.
570
5. Cartesian Currents in Riemannian Manifolds
Proof. If [w] is a non trivial cohomology class we can represent it by a non-zero harmonic form h with n h = 0. The form *h is harmonic and t (*h) _ *n h = 0,
consequently *h represents a relative cohomology class [*h] E H ' (X,BX), and we have
r
f
Poinc<[w],[*h]>=J hA*h=J Ih12#0. x
x
Similarly one proves that, if [S]rel ; 0, we can find [w] in HHR(X) such that Poinc< [w], 0. The commutativity of the diagram is trivial.
3 Currents and Real Homology of Compact Manifolds Let X be an oriented compact smooth submanifold of dimension n in Rn+N without boundary. We introduce in Sec. 5.3.2 for k := 0, 1.... the k-homology group Hk (X, R) of a X setting Zk(X) Bk(X)
{TENk(U) I sptTCX, 8T=0} {8S I S E Nk+l(U), spt S C X}
and
Hk(X,R) := Zk(X)/Bk(X) where U is a tubular neighbourhood of X, i.e., compare Sec. 5.3.1, as normal cycles in X modulo boundaries of normal currents on X. Two sets of pairings are naturally associated to homology and cohomology. One is the set of de Rham pairings between currents and forms which factors to homology and cohomology
<,>:Hk(X,lI)xHaR(X) defined by < [T], [w] > := T (w). The others are the pairings (w,71) E Dk(X) x
Dn-k(X) _
fwA7]
x which, by Stokes theorem, induce Poincare pairings in cohomology
Poinc<, >: HHR(X) X HHRk(X) -> R given by Poinc < [w], [ri] > := f wnr7. Classical theorems, respectively de Rham's
and Poincare's duality theorems, state in fact that de Rham and Poincare pairings are non degenerate; consequently there are isomorphisms, Poincare duality isomorphisms P : HHRk(X) -' Hk(X,
3 Currents and Real Homology of Compact Manifolds
571
between HHR k (X) and Hk (X, R). In Sec. 5.3.2 we show that the Poincare' duality
isomorphism is actually induced by the map again denoted by P, P : Z,-k(X) -a Zk (X) given by w -+ fX Aw. Moreover, by regularizing a k-cycle T E Zk (X) Dn+N-k(U), and integrating such a we construct an (n + N - k)-form w E form on the fibers of the projection map -7r, we construct a so-called Poincare
dual of T, PT E Dn-k(X). The cohomology class of PT turns out to depend only on the homology class of T, therefore T --# PT yields an explicit way to describe surjectivity of P. Non degeneracy of Poincare duality, which we already proved in Sec. 5.2.8 as consequence of Hodge decomposition theorem, then yields injectivity of the Poincare' duality isomorphism and also yields non-degeneracy of de Rham pairing, as < [T], [w] > = Poinc < [w], [PT] > =
fAPT,
x compare Sec. 5.3.2. We then deduce the weak closure of the real homology classes, hence in particular the existence of mass minimizing real cycles in each homology class.
A completely parallel theory holds in the case of compact manifolds with boundary. One has to introduce the relative homology groups, compare (1) in Sec. 5.3.3. The Poincare-Lefschetz duality isomorphism this time consists of three sets of maps
P# :HdRk(X,BX) -> Hk(X,lR) P' : HHR -k (X) -> Hk (X, 8X, IR)
P
:Hank-1(8X)
- Hk (BX, IR)
which actually give a morphism between the long exact sequences in cohomology and homology, compare (9) in Sec. 5.3.3. The duality forms-currents_ T(w) induces three sets of de Rham pairings at the homological level
<,>a:Hk(X,IR)xH,R(X) ---+
IR
< , >' : Hk (X, 8X, IR) x HdR(X, 8X) -> 1R < , > : Kk (BX, R) x HdR (8X) --+ 1R
which are related each other by the commutative diagram in de Rham's theorem, Theorem 5 in Sec. 5.3.3. As consequence we deduce the closure of absolute and relative homology classes and consequently the existence of mass minimizing cycles in each homology class. Still another set of pairings is relevant in homology: these are the so-called intersection indices of cycles of complementary dimension.
In Sec. 5.3.4 we shall define Kronecker's index for two'arbitrary currents S E Nk(X) and T E N,,-k(X) for which spt 8S f1 spt T = spt 8T f1 spt S = 0 and see that it induces a pairing Hk(X,1R) X Hn_k(X,8X,1R) -> R.
572
5. Cartesian Currents in Riemannian Manifolds
It turns out that
ix(S,T) = J Ps APT,
x therefore non degeneracy of Poincare duality is equivalent to non-degeneracy of the intersection index.
3.1 Currents on Manifolds Let X be a smooth manifold of dimension m. As already mentioned, by Whitney embedding theorem we may and do assume that X is a smooth (CO°) subman-
ifold of some l1P', X C R'. Moreover we can assume that X has a tubular neighbourhood U D X, U C Rn, and there is a C°° retraction map 7r : U -> X, rr(U)=X,ir(y)=yVyEX.
Definition 1. A k-dimensional current T E Dk(U) is said to be a current in X if T(w) = 0 for any null form w E Dk(U) to X. The class of k-dimensional currents in X is denoted by Dk(X). From the characterization of null forms to X, compare the end of Sec. 5.2.1, we easily infer
Proposition 1. Let T E Vk(U). Then T E Dk(X) if and only if for some E C°°(U) with ( = 1 on X we have T = 7r# (T we have
T=7r#(TLc)
Actually, if T E Dk(X),
b'(EC°°(U), (=1 on X.
Proof. In fact for any w E Dk(U) the form w - (7r#w E Dk(U) is a null form
to X, hence T(w) - T((ir#w) = 0. Conversely if q is a null form to X, then 7r#w = 0, hence T(77) = a# (T L () (y) = T ((7r#77)=0.
Remark 1. Proposition 1 extends to the case of Ck manifolds, k > 1, provided T is a vector valued distribution of order -k. In this case one in fact can assume that the retraction map is Ck hence 7r#(TLC) is well defined. In particular Proposition 1 holds true for flat chains on C' manifolds. Trivially sptT C X if T E Dk (X) . In fact, if w is supported outside X and
w is a null form to X, we have T(w) = 0. But in general it is not true that spt T C X implies T E Dk(X). Consider for instance the 1-dimensional current in R2 integration of 1-forms along the real axis el against the normal vector e2 L(w) :=
J
<w, e2 > dxl.
Clearly spt L C l[8xi, while for example on the null form W(x1) dx2 we have T (,p(xl)dx2) =
f(x1)dx1.
Notice that M(T) < oo, while M(8T) = +oo, compare 10 in Sec. 2.2.4. However we have
3.1 Currents on Manifolds
573
Theorem 1 (Flatness theorem). Let X be a closed submanifold of Pn. If T is a flat chain in U, T E Fk(U), and sptT C X, then T E Dk(X). Proof. Theorem 1 is a consequence of Proposition 7 in Sec. 5.1.3. In fact the X and the identity map id : U -+ U trivially agree on retraction map 7r : U
X, hence on spt T. Proposition 7 in Sec. 5.1.3 then yields T = 'r# (T L () V ( E C°°(U), C = 1 on X, and the conclusion follows from Proposition 1. By Theorem 1 normal currents on a submanifold X on Rn may be identified to currents in the ambient space IRn with support in X. We shall now see that a constancy theorem holds also for currents on submanifolds. Let X be an m-dimensional C°° submanifold of RTh with locally finite area. Suppose that X is oriented by a smooth rn-vector field X. Then the current
Q X := T(X,1,..) is a locally m-rectifiable current in R'.
Theorem 2 (Constancy theorem). Let .fl be an open set in ][MTh such that .flflX is connected and letT E D,(X), m = dim X, be such that sptaTfl,fl = 0. Then there exists a constant c E IR such that T= c Q X
in D'(Q f1 U)
U being a tubular neighbourhood of X. Moreover c is an integer if T is i.m. rectifiable.
X be the retraction map, and let ( E C°° (U), = 1 Proof. As usual let -7r : U ,(l2 C D on X. We cover .fl fl X by a system of coordinates Vi : V -> Di f1 X
so that cj#U = X; the maps cpi have inverses, and more precisely there are smooth maps z/ii : U fl it-' (.Qi fl X) -> V which are onto and such that W o Oi = ir,
Oi o wi = id,.
Therefore the currents Wi# (T L () belong to D,,,, (V) and, as spt aT fl .fl = 0, we
have fti# (T L () = 0 on Vi. It follows from the constancy theorem in the flat case, Theorem 1 in Sec. 4.3.1 that
'bi#(TL() =G;QV on V, hence
T = it#(T LO) =Wi#4Ji#(T L() = ciEX
on ,fli fl U,
since cpi#U = X. As the .(li's overlap on X, we infer that the constant ci are all equal to a c independent of the chart T = c QX ]J
on Dm(.fli fl u),
and the conclusion follows by means of a decomposition of unity associated to the covering {.Qi}.
5. Cartesian Currents in Riemannian Manifolds
574
Remark 2. Theorem 1 holds also for flat chains on Cl manifolds, compare Remark 1. More generally we have
Theorem 3 (Constancy theorem). Let E be an r-dimensional oriented, connected and embedded submanifold of U of class C', where U is an open set in ]Rn, and let T E F;°` (U). If (spt T) \Z is closed relative to U and spt 8T C U\E, then there exists a real number p such that
spt(T-pjZ]) CU\E. Moreover p is an integer if T is locally integer rectifiable.
If f : X --+ Y is a Lipschitz map between C' manifolds, every T E Nk(X) has a Lipschitz image f#T E Nk(Y). Also if fo, f, : X - Y are two maps and h : [0,1] x X - Y is an homotopy map between fo and fl, h(0, x) = fo(x), h(1, x) = f, (x), then the homotopy formula holds (1)
fl#T- fo#T=Bh#(Q0,11 xT)+h#(Q0,1J x8T)
for T E Nk (X) . Also a corresponding homotopy formula for forms holds (2)
fl#w - fo#w = d[h#(w)](o,i) + [h#(dw)] (o,l),
[i7] (o,l) being the integration along the fibers of the components of rl which are not
tangential to X. Also the standard estimates for homotopies hold true, compare Sec. 2.2.3.
3.2 Manifolds Without Boundary: Poincare and de Rham Dualities By Poincare duality isomorphism one usually refers to the duality between the k-homology group and the (n - k)-de Rham cohomology group of an oriented compact smooth manifold without boundary. In this subsection we shall describe such an isomorphism in terms of currents and forms.
Poincare duality isomorphism on open sets. Let U be an open set in R". For 0 < k < n consider the vector spaces of currents
Zk(U, R) = {T E Dk(U) 18T = 0, M(T) < +oo} Bk(U,R) = {8S I S E Dk+l(U), M(S) <+oc} and set Hk(U, R) = Zk(U, R)lBk(U, R) Analogously we define de Rham's cohomology groups in U by HdR(U, R) := Zk(U, R)/Bk(U,
3.2 Poincare and de Rham Dualities
575
where
Zk(U,R) _ {w E L1(U,AkRn) n£k(U) I dw = 0} Bk(U,1R) _ {dr) 177 E L' (U, AkRn) n £k-1(U)}
Then we have
Theorem 1.
HaRk(U,R) and Hk(U,R) are isomorphic.
To prove this we consider the map P : Dk(U) I M(T) < +oo} given by
L1(U,An-kRn)
p(W) :_ (-1)k(n-k)Rn LW = As 0(1Rn L W) _
(-1)k-1Wn
J
n £n-k(U) , {T E
Au.
L dw the map P factorizes to a map still denoted by
P P : H a R k (U, R) --+ H k (U,1 [ 8) .
We shall prove that P is an isomorphism, indeed the isomorphism of Theorem 1. Of course this amount to prove
Proposition 1. We have (2)
Let T be a cycle of finite mass, aT = 0. Then there is a (n - k) form PT E An-kR2) n £n-k(U) such that for some S E Dk+l(U), M(S) < +oo L' (U, we have
T= 1 APT+aS.
(1)
UU
(ii) Let w E L'(U, An-kRn) n £n-k (U) be such that 1R L W = aS for some S E Dk+1(U), M(S) < +oo. Then w = dr7, 77 E L1 (U, An-k-lRn) n £n-k-1(U).
Proof. We prove both claims by regularizing T. (i) is in fact already proved in (-1)k(n-k)f. Proposition 7 in Sec. 5.1.2, by taking PT :_ To prove (ii), we choose a convolution kernel p and a regularized distance
d. For0<E
Wi L wE = (Rn L W)E = (aS)E = as'.
On the other hand by Proposition 6 in Sec. 5.1.2 (3)
SE = IEBn L 17e
i 6, E L'(U,
An-k-'Rn) n £n-k-1(U)
From (2) and (3) we conclude that wE = (_l)n-kdr)E. As w6 and w are also cohomologous to a form L1(U,
An-k-lRn) n £n-k-1(U) the claim follows.
576
5. Cartesian Currents in Riemannian Manifolds
Remark 1. We observe that the same reasoning of Proposition 1 implies the following:
(i) if the current T in (i) of Proposition 1 has compact support in U, then the decomposition (1) holds with spt PT and spt S arbitrarily close to spt T; (ii) if the form w and the current S in (ii) of Proposition 1 have compact support in U, then one finds a potential for w with support arbitrarily to spt w U spt S.
(iii) In particular we infer that de Rham's (n - k)-cohomology group of smooth forms with compact support is isomorphic to the k-homology group of currents with compact supports {T E Dk(U) I OT = 0, sptT CC U}/{ 8S I S E Dk+1(U)}.
From now on we shall refer to the map P : Hd k (U, R) - Hk (U, ]R) as to the Poincare duality isomorphism and to (n - k)-forms PT satisfying (1) as to Poincare's dual forms of T. Poincare's dual forms PT, T E Zk(U) are constructed mollifying T, compare Sec. 5.1.2, in particular there are not unique. However they preserve some of the properties of T, e.g., PT E L' (U, do-kin) if
M(T) < oo, sptPT CC U if sptT CC U. Also (ii) of Proposition 1 states that the cohomology class of any Poincare dual form PT E Zn-k (U) of a given cycle T E Zk(U) is uniquely defined by the homology class of T. Let X be a compact oriented and boundaryless submanifold in U of dimension k. A form w E Zn-k(U) with compact support in U such that Q X (7l) =
f7lAw U
for all closed forms ?] E Zk (U) is classically referred as to a Poincare's dual form of X. In principle therefore our notion of Poincare dual form is slightly stronger,
as we require equality (1). Actually this is not the case, in fact one can show that every classical Poincare form is actually a Poincare form in our sense. This amounts to show that for every (n - k)-form w for which (4)
R' Lw(a) = fwAa = 0
V a E Dk(U), da = 0
U
(in particular w is closed) we can find a current S E Dk+1 (U) such that 1R L w = 8S. This is equivalent to say that the linear functional S' : Bk+1(U) --+ Ifs (5)
S'(,(3) :=
J
was
if 3 =
doe
extends as a linear continuous functional on Dk+1(U). We have in fact
Proposition 2. Let U be a smooth bounded open set in Rn and let w be a form LP(U,An-kJn) n £n-k(U), p > 1, such that in
3.2 Poincare and de Rham Dualities
fwAa=0
577
b a E Dk(U) with da = 0,
U
in particular dw = 0. Then there exists S E Dk+1 (U) such that 11 n L W = 3S. Proof. First assume p > 1, and let w = >2 wa (x)dxa. We solve the problem
dQa = wa O6
.Qa E Wo'p(U).
By the elliptic regularity we find a unique .fla which actually belongs to Coo (U) n W2,P(U). Consequently the form (1 := >2 Qa dxa solves
d8ffl + Sdfl = w,
compare (7) in Sec. 5.2.4, 8 being the adjoint of d with respect to zero values on W. Since dw = 0, we infer SdD E ker d n ker 5 and by the assumption, taking into account Proposition 2 in Sec. 5.2.4, Sdfl = w - dSS? E ker J', hence 8d.fl = 0. Therefore we find w = dij, rl E 6,_k-1(U) n W1,P(U, An-k-lRn), and consequently ][8n LW = IRn L drl = (-1)n-ka(Rn L 77).
If p = 1, we can still solve (6), compare Stampacchia [609], and find a weak solution which belongs to C°°(U) n W1,9 (U) for all q < n/ (n - 1). The proof is then concluded as previously.
Integration along the fibers and the lift of currents. Let X be a C2 compact, oriented n-dimensional manifold without boundary in IEPn+N and denote by X : X -* 11nRn+N and by N : X -> ANIRn+N respectively the unit tangent n-vector field to X and a unit normal vector field to X so that
X/vV has the same orientation of Rn+N. Recall now that X is a tubular neighbourhood retract, and more precisely we can find an Eo > 0 such that each point in
U = {x E Rn+N
I
dist (x, X) < Eo }
has a unique point of least distance ir(x) in X, and that -7r : U -+ X is a retraction. Moreover, we have a system of charts cpa : fla -> Va C X of X, and on each chart a system of N independent vectors via), ... , VNa) in such that via)n...nvN) =N on K,; consequently the maps
Ua x B(0, Eo) C Ua x RN -+ 7r- 1(V,) given by N
(7)
Y a (x, t) _ 0a (x) + E tiv(a) (x) i=1
578
5. Cartesian Currents in Riemannian Manifolds
are diffeomorphisms. Of course, taking into account the special form of 0a and denoting by El, ... , EN the standard basis of RN, we see that ANDca (x, t) (Eln ... AEN) = v1A ... AvN (co(x)) = N(W(x))
for x E .(la, t E B(0, Eo). Also note that 1 ° ga (x), A(x)t),
1(x, b) =
where A(x) E MN x N and det A > 0, or in other words 7r : U --> X is an oriented fiber bundle over X. To any (k+N)-form w E Dk+N(U) we now associate a k-form 7r*w in Dk(X) by "integrating w along the fibers of the fibration it : U --> X". We define 7rw by (8)
<7r*w, v > :=
r
J 7r-'(X)
<w(z), vAN(x) > dH'(z) V v E AkTXX.
Clearly 7r* defines a map from Dk+N (U) into Dk (X). In the special case in which
X = .(l is an open set of R'", U = .(2 x RN introducing coordinates el, . .(2 and El, ... , EN in R", we have
X = e1A... hen
N = E1A... AEN,
and any form w E Dk+N l/ f7 x RN) has the form
E
w«Q(x, y) dxandyQ.
aI+j8I=k+N
Then 7r (x, y) = x and (9)
7c*(w)(x) := >
(fWa(X,S)dS)dXa.
Notice that the formula for a product of currents gives (10)
for T E Dk(fl), W E Dk+N(Q X RN)
T x RN(w) = T(7r*(w))
Proposition 3. We have
11
(i) 7<* d = d7r* . (ii) 7r*(7r#77Aw) = 77A7r*w, d w E Dk+N(U), V 77 E DP(X), p + k < n. (iii) For 77 E D,-k (X) and w E Dk+N (U)
f x
7)^7r*w =
f 7#77nw, U
i.e., 7r# (IRn+N L w) = (-1) N(n-k) Q X L -7r*w.
. .
, e,,, in
3.2 Poincare and de Rham Dualities
579
If Px is a Poincare dual form of X in U then 7r.Px = 1.
(iv)
Proof. The first two claims are local, consequently it suffices to prove them for forms 77 and w respectively with compact support in a chart V and in 7r-1(V), we omit here the index a, compare (7). Also we first prove the claims in the case that V =: S2 is an open set of Rn so that -7r-1(.f2) = S2 x RN. In this case we can write 7r(x, y) = x and
(fWa(X)d) dxa.
7r.w := F, Ial=k
Now each component of w takes the form = a(x, y) dxandyO. Therefore
d = ax, (x, y)dxindxandya + v(i, 0) (-1)ka, (x, y)dxandy,3+a hence
0
(i,
7r.dt; =
Q)(_1)k
( l
\
faye (x, y) dy) dxa = 0
(fax(x)d) dxiAdxa
1/31
2.=1
and we conclude n
7r.dw = E T, (
f
axi (x, y) dy) dxindxa = d7r.w.
i=1 dal=k
Writing 7r#77 = E ,7.y (x) dxy the second claim follows easily as 17I=n
7r* (7r# ('y) Aw) (x)
E U 77-y (x)aao (x, y) dy I dx'yAdxa
171=h
Ial=k 77(x)A7r.w(x).
The general case follows from the flat case because (11)
7r.(O#w) = V#(-7r*w),
and cp being the diffeomorphisms in (7) (index omitted). In fact co#[d7r.w] = dco#7r*w = (by (11)) d7r.O#w
= (by (i) in the flat case) 7r.do#w
= 7r*0#(dw) = (again by (11)) cp#(7r.d(.1)
580
5. Cartesian Currents in R.iemannian Manifolds
which proves that in any coordinate chart d7r* - 7rd = 0. Similarly one proves (9).
Let us prove (11). We have ca#w(x, t) = Ak+NDC,(x, t)w(q5(x, t)), therefore for v E AkRn, < 0#W, vAe1A ... ^EN > = <W (O(x, t) ), Ak+NDO(x, t) (vAe1 A ... /BEN) >.
But on the account of the particular form of 0 AkDc(x, t) (v) = AkDcp(x)(v)
ANDO(x, t)Eln... AEN = I ANDO(x, t) I N(((x, t))
... Err) are orthogonal, hence
AkDO(v), ANDO(ElA
< O#W, vAe1 A ... AEN > = < w (q5(x, t) ), AkD p(x) (v) NV (cp(x, t)) >,
consequently <7r* (0#w), v >
r
_ / <w(o(x,t)),AkDV(x)v/
J
<w(z), AkDW(x)vAN(V(x)) > d7-lN (z)
7-1(rv(x))
_ <7r*w(cp(x)),AkDco(x)v> =. Proof of (iii). We use the general coarea formula
f gIVNfI dH =f dnn f g(z)dxN(z) x
x
7r-1(x)
where f : U --+ X and VN f (x) = projection of V f onto the level line of through x. In this case we have for r) E D,-k(X)
f
71A7r*(W) =
f
f (#Aw) = x
x
x
dH f d7-lr'<7r#7]Aw, ex/vV> 7r-1(x)
= f <7r#7)nw, e1A... /den+N>IVN7rI dz = f 7r#7?Aw U
U
since ON7rI = 1. This proves the claim together with
QXj L7r*(w) = (_I)("-k)k
f
r)n7r*(w)
x
_
(-1)(n-k)k f#()Aw
_
(-1)(n-k)k7r#(IRn+N L W).
=
U
(-1)(n-k)[k+N-k]
¶
Ull(WA7r#(7/))
f
3.2 Poincare and de Rham Dualities
581
Finally (iv) follows at once as for 77 E Dn(X) by (iii) and (1) we have
f
fi7ir*(Px) = f 7r#(r1)APx = x U X
771
hence 7r* (Px) = 1.
Fig. 5.3. The lift of a current.
Remark 2. Let cp :.Q -+ L be a diffeomorphism of two open sets of 1R
and let
0: (1xRN -,A x R N be of the type h : 1? x RN __ RN smooth and ah (x, t) > 0.
(12) 0(x, t) = (cp(x), h(x, t)),
Then it is easily seen that cp#7r* simply by changing variables. In particular we obtain that if 7r : E --* X is an oriented fiber bundle, then the integration along the fibers is well defined and gives a map 7r*
: Sk+N(E) -' Ek(X)
since the transition functions of E are maps of the type (12).
One then defines a lift operator 7r* : Dk(X) -> Dk+N(U) on currents by duality *(S)(W) := S(7r*w).
In case X = 1l x {0} C IRn+N and 7r : Rn+N __,, R x {0} is the orthogonal projection on the first factor, 7r(x, y) = x, then
7r*S=SX RN for SEDk(lR X{0}).
582
5. Cartesian Currents in Riemannian Manifolds
Proposition 4. We have (i)1
a7r*s = it*as in U,
(ii) (iii) If M(S) < +oo, then M(7r*S) = vol(BN(0,1))Eo M(S). (iv) If spt S C X, and S is i.m. rectifiable, then 7r*S is i.m. rectifiable in U. Proof. (i) and (ii) follows from (i) and (iii) of Proposition 3.
(iii) For ( E AkTTX,
f <w(x, y), S(x) > dfN(y)
1<7r*w(x), (>I =
7r-1(x)
Taking the sup, we get
7r*w(x)
w(x) vol(BN(0,1))EO hence the conclu
sion.
(iv) We claim that if S = 7-(M, 0, S), M C X then 7r*S = rr(.7r-1(.M), 0 o
7r, F-1(M) ). In fact if w E Dk+N (U) we have
7r*(S)(w) = 7'(M,0, )(7r*w) = =
f<*w(x),(x)>9(x)dflc(x) = M
ff
<w(x, y), S(x)NV(x)>0(x) d7{N(y) d-Hk(x) _
M 7r-1(x)
f
<w(x, y), S(x)NV(x)>0(x) d?jN+k(x, y)
it 1(M)
and the claim follows as SAN(x) is a unit vector orienting 7r-I(M) at (x, y) E 7r-1(M).
Poincare duality isomorphism on compact manifolds without boundary. Let us discuss now Poincare duality between (n - k)-cohomology and khomology groups of oriented, compact and boundaryless submanifolds of dimension n in Rn+N and more precisely prove the following
Theorem 2. Let X be a smooth compact oriented and boundaryless smooth submanifold of dimension n in Rn+N Then the map P : w E D,-k(X) -> P(w) :=
f
Au) E Dk(X)
x induces an isomorphism between the (n - k) de Rham cohomology group of X and the k-homology group of currents Hk (X, R) := Zk (X)/Bk (X) where
Zk(X) = {T E Nk(X) jaT = 0}, More precisely
Bk(X) = {aS j S E Nk+1(X)}.
3.2 Poincare and de Rham Dualities
583
(i) for any cycle T E Zk (X) one can find a closed (n - k) -form PT E Zn-k (X )
such that
T=
(13)
r APT + aS J
X
for some S E Nk+1(X), (ii) if for some w E Zn-k(X) we have fx Aw = as, S E Nk+l(X), then w is
exact in X.
We refer to P as to the Poincare duality isomorphism in X and to any form PT satisfying (13) as a Poincare dual form for T. Proof. Let T E Zk (X). Then, if WT is a Poincare dual of T in U, we have
T(a)=faAwT+aE, EENk+l(U), sptECCU. U
Consequently, taking into account (iii) of Proposition 3, we get
T(a) = 7r#(T)(a) = J '7r#(a)AwT+a7r#E(a) U
r aA-7r*WT + 197r#E(a), J X which proves (i). (ii) From QX ] LW =
(-1)k(n-k) fx
. AW = as, S E Nk+l(U), spt S C X we get
inU by Proposition 4. We then infer from Proposition 1 that 7r#w = d/3, /3 E en-k-1(U) n L1(U,An-k-1Rn+N) which in turns implies that w is exact on X.
Remark S. It is worthwhile noticing that yet another proof of the non degeneracy of the Poincare duality in cohomology on a compact manifold X can be recovered from Proposition 2 using the integration along the fibers. In fact, assume that
fwAa = 0
Va E Zn-k(X).
x Then for /3 E Zn+N-k(U) we have f 7r#c.wA/3 = U
f wAir*/3 = 0 Jx
Then -7r#w is exact in U by Proposition 2 and consequently w is exact in X.
584
5. Cartesian Currents in Riemannian Manifolds
Remark 4. Classically a Poincare dual of T is a closed (n - k)-form in Zn-k(X) such that on Zk(X). T = J . Aa
x
If PT is a Poincare dual of T we also have
T=
J x
APT + 8S,
S E Nk+l (X).
therefore
J?1A(a - PT) = 0
Vrl E Zk(X).
x Consequently a - PT = dQ, and
T=J Aa+BS x
J x
Ad/3=J .Aa+8S1i x
which proves that the classical notion of Poincare dual and the one we have used in this section agree.
de Rham theorem. As we have seen in Sec. 5.2.8, Hodge decomposition theorem yields an isomorphism between HHR(X) and Horn (HaR k (X), IR),
HdR(X) -> Hom (HdR k(X ), l).
On the other hand Poincare duality isomorphism gives also an isomorphism between HaR k (X) and the k-homology group Hk (X, R) of X
P : HHRk(X) _ Hk(X, Therefore we can state
Theorem 3 (de Rham). HHR(X) and Hom (Hk(X, IR), IR) are isomorphic. Actually we shall now show that the induced non degenerate pairing between homology and cohomology HaR (X) x Hk (X, IR) -> R is the integration of closed forms over cycles. In fact the bilinear map
: Zk(X) X Zk(X) -+ IR,
:= T(w)
clearly depends only on the homology class of T and the cohomology class of w, therefore it induces a bilinear map, de Rham's duality <[T], [w] > : Hk(X,IR) x HaR(X) -> IR
and we have
3.2 Poincare and de Rham Dualities
< [T], [w] > = T(w) =
(14)
J X
585
WAPT = Poinc< [w], [PT] >.
where PT is a Poincare dual form of T. In particular (14) clearly shows that de Rham's duality is non degenerate i. e. (i) If w E Z"(X) and T(w) = 0 for all T E Zk(X), then w is exact. (ii) If T E Zk(X) and T (w) = 0 for all w E Zk (X ), then T = 8S
for S E
Nk+l(U), sptS C X. De Rham's theorem can be equivalently stated in terms of periods. Let 'Yi...... s E Zk(X) be cycles such that their homology classes ['y],... , [ys] form a basis of Hk (X, IR). If w E Zk (X ), the numbers 'Yi (w)
are said to be the periods of w with respect to the basis [yi]. Theorem 3 can be equivalently stated as
Theorem 4 (de Rham). Let {yi}, i = 1, ... , s be a basis of k-cycles of X. Then (i) If w is closed and all its periods of w are zero, then w is exact. (ii) Given cu,... , as E R. Then there exists a closed form w E Zk (X) with given
periods al, ... , as, i. e., yi (w) = ai, i = 1, ... , s.
In particular homology classes are identified by the periods of a basis of harmonic forms. This yields in particular the closure of real homology classes.
Proposition 5. Bk (X, IR) is weakly closed. The following corollary is then an immediate consequence of Proposition 5 and of the lower semicontinuity of the mass
Corollary 1. In each class -y E Hk (X, JR) there is a mass minimizing current, that is T E Zk (X, III), [T] = -y such that
M(T) = inf{M(S)
[S] = y}.
I
In particular
M(y) := inf{M(T)
I
[T] = y}
defines a norm on Hk(X,IR), called the sphere norm, consequently
M(y)>0 for all yEHk(X,R), y and the coset map Zk (X, IR) -+ Hk (X, R) is continuous.
0
586
5. Cartesian Currents in Riemannian Manifolds
Normal currents and singular real homology. Another path to get the closure of homology classes which also yields a representation of homology classes by Lipschitz chains can be followed by using the deformation theorem of Sec. 5.1.1. Let X be a compact smooth submanifold without boundary of dimension n
in IRn+N. As it is well known X is a neighborhood C' retract, i.e. there is an open set U X and a C' map 7r: U --> X which retracts U onto X and actually one can think of U as a tubular neibourhood, {x E IRn+N I dist (x, X) < Eo }
(15)
for some positive Co and of 7r as homotopic to the identity in U. Fix now E < EQ/Vn and denote by L ,,x the collection of all standard cubes then Q C U because of (15), of size E which meet X. Of course if Q E and one can modify 7r by composing 7r with a projection onto £n,x to another retraction map again denoted by 7r, 7r : L ,,x --> X which is homotopic to the identity in G,,,,X. Given an integer k, 0 < k < n, we denote by Lk,x the collection of all k-faces of cubes in L,,,x, and by Lk,x the k-skeleton of Ln,x
Lk,x := U{F I F E
If F E Lk,x we orient F in the natural way, i.e., if F is parallel to the plane associated to eal A ... Ae, k, Ic = k, then f := ea, A ... Aeak . Finally we denote by Lx the CW-complex of cubes in Ln,,x. Simplicial homology. Given an integer k, 0 < k < n, the simplicial homology of Lx is defined as follows. A chain on Lk,x is simply a polyhedral chain
P := E Oij Fi j,
Oi E
R
F2 E Lk,X .
Then we set Zk,simpl(Lx, IR)
{P I P is a polyhedral chain in Lk,x, aP = 0}
Bk,simpl (LX, IR)
{P = 8Q I Q is a polyhedral chain on Lk+i,X }
and the k-simplicial homology of Lx is defined by Hk,simpl(LX,IR) := Zk,simpl(LX)/Bk,simpl(Lx).
Clearly Hk,simpl(LX, IR) is a vector space over R of finite dimension, since the collection Lk,x is finite.
Homology groups of Lipschitz chains. Similarly to the simplicial homology or to the real homology, by considering Lipschitz chains in X, i.e., Lipschitz images of polyhedral chains, we can define the homology groups Hk,Lip(X,IR). We have
Theorem 5. The spaces Hk(X,IR), Hk,Lip(X,IR) and Hk,simpl(Lx,IR) are isomorphic. In particular Hk (X, IR) has finite dimension.
3.2 Poincare and de Rham Dualities
587
Proof. Let 7r : U ---> X be the retraction map. If P is a cycle in the k-skeleton Lx then 7r# (P) is a Lipschitz cycle on X, and, as any Lipschitz cycle is also a normal cycle, we have the following maps
Hk,simpl(LX, R) -i Hk,Lip(X, R) -s Hk(X, R) Let T E Zk (X, R). By the deformation theorem we can write T = P + OR where P is a k-cycle on Lx and R E Nk+l(U). We then have T = 7r#T = it#P + 87r#R,
i.e. T and i* o 7r*P are homologous, consequently i* o 7r. is onto. Suppose now that L is a Lipschitz k-cycle in X which is a boundary regarded as a current in U, L = OR, R E Nk+l (U), spt R C X. Deforming R into Lx we get
R=Q+8R1+S, with Q E Lk+1,x and S is a Lipschitz chain in U. Therefore L = OR = 87r# (Q + S) is a Lipschitz boundary which clearly proves injectivity of i*. Finally let us prove that ir, is injective. Let L E Lk,X such that 7r#L is a boundary in X. Then being 7r homotopic to the identity in L,,,x, we have
L = lr#L + 8S,
S E Nk+1(U), spt S C 4,.,x
hence Lisa boundary in L ,,,X, L = OR, R E Nk+l (U), spt R C C,,,x. Deforming R into Lk+l,X we get
R=Q+OR1+S, with Q E Lk+l,x and S = 0, being OR = L E Lk,x. Therefore L = OR = ar# (Q) is a simplicial boundary, which clearly proves injectivity of 7r*.
Remark 5. If A is a local Lipschitz retract, one can do the same constructions and define simplicial homology of the skeleton L,,,x, and Lipschitz homology. However it is easy to find examples in which no retraction homotopic to the identity exists. The previous proof applies however partially and we get Theorem 6. The spaces Hk(A, R), Hk,Lip(A, ]R) are isomorphic. Moreover the retraction map 7r yields a surjective linear map ir* : Hk,simpl (LA, III) -> Hk (A, JR) Again Hk (A, IR) has finite dimension.
As we have seen, compare Proposition 5, one of the advantages in dealing with homology groups defined in terms of normal currents is that the cosets are closed with respect to the weak convergence of currents, and this allows of course the use of variational methods. Another proof of Proposition 5 can be done using the following rough isoperimetric inequality, which works also for compact neighbourhood Lipschitz retracts.
588
5. Cartesian Currents in Riemannian Manifolds
Proposition 6. Let A be a compact neighborhood Lipschitz retract. Every current T E Zk(A,1R) which is the boundary of a normal (k + 1) -current in U,
T = aS1 with S1 E Nk+l(U), is in fact a boundary in A, i.e., there exists S E Nk+1 (U) with spt S C A and T = as, and moreover
M(S) < r(k,n)M(T). Proof. Since simplicial boundaries on LA form a finite dimensional space, we
can easily construct a linear map
T : Bk,simpl(LA, R) -> Nk(U)
which is continuous with respect to the mass norm. For that, choose a basis Pi, ... , Pnr E LA and for each element Pi select a simplicial (k + 1)-chain T(Pi) in Lk+1,A so that af(PP) = Pi, then extend T by linearity. For any P in Bk,simpl (LA, R) we of course have
aP(P) = P,
M(F(P)) < ri (k, n)M(P).
Applying the deformation theorem to T = 8S1 we find R E Nk+l (U), spt R CC U, such that
T = P + aR = a(F(P) + R) and
M(F(P) + R) < M(F(P)) + M(R) < riM(P) + c(k, n)EM(T) < c(k, n)(E + ri)M(T). It is now readily seen that S :='r#(r(P) + R) has the desired properties. Remark 6. Notice that in fact we have proved the existence of a current S in Nk+l (U), spt S CC U such that
T = aS,
M(S) < r(k,n)M(T)
independently from the fact that A is a Lipschitz retract. We have then used the last property to retract S to A and find S :_ 7r#S with
T = as,
M(S) < r(k, n)M(T),
sptS C A.
We can now state
Theorem 7. Bk(A, R) is weakly closed.
Proof. Let Ti E Bk(A, R), Ti T. Of course T, = 8Ri, spt Ri C A. Proposition 6 then yields Si E Nk+1(U), spt Si C A such that Ti = asi and M(Si) < r(k, n)M(Ti) < 7- (k, n) supM(Ti) < oo. i
Passing to a subsequence we get Si -s S, spt S C A, T = as.
3.3 Poincare-Lefschetz and de Rham Dualities
589
The following corollary is an immediate consequence of Theorem 7 and of the lower semicontinuity of the mass
Corollary 2. In each class y E Hk(A,IR) there is a mass minimizing current, that is T E Zk (A, ]R), [T] = y such that
M(T) = inf{M(S)
[S] = y}.
I
In particular
M(y) := inf{M(T)
I
[T] = y}
defines a norm on Hk (A, ]R), called the sphere norm, consequently
M(y)>0
for all yEHk(A,1R), y#0
and the coset map Zk (A, I[8) --+ Hk (A, I[R) is continuous.
3.3 Manifolds with Boundary: Poincare-Lefschetz and de Rham Dualities The Poincare-Lefschetz duality isomorphism. Let X be a compact, oriented smooth n-dimensional submanifold of Rn+N with boundary 8X. We define relative cycles in X as
Zk(X, 0X) := IT E Fk,X(r+N)
I
spt 8T C 8X or aT = 0 if k = 0},
relative boundaries as
Bk(X,3X) :_ {R+8S I R E Fk,ax(1Rn+N) S E Fk+l,X(Rn+N)}, and the real relative homology groups as (1)
Hk (X, 0X, II8) := Zk (X, OX) lBk (X, 0X),
k = 0, ... , n.
Since X and 8X are compact and Lipschitz neighbourhood retracts in Rn+N, Federer's flatness theorem yields Fk,ax(Rn+N) = Fk,X(Rn+N) n IT I sptT C 8X};
also observe that Fk X (R'+N) = Fk,X (U),
Fk,ax (Rn+N) = Fk,aX (V )
U and V being neighbourhoods respectively of X and aX which retract onto X and aX. A few words are in order to explain why we are working with flat chains while when dealing with the homology of boundaryless manifold we worked with normal currents. First we should remark that in the boundaryless case, though cosets defined in terms of flat chains are larger than cosets constructed
590
5. Cartesian Currents in Riemannian Manifolds
with normal currents, the homology groups defined in terms of flat chains or of normal currents are the same. This is not difficult to be seen on account of the density of normal currents among flat chains and the closure properties of homology classes. However, while i.m. rectifiable cycles are obviously normal currents, relative cycles, which in general do have boundaries, are not in general
normal as in principle we have no control on the masses of their boundaries: this is the reason why we are forced to work with the more general flat relative cycles. In the next section we shall illustrate relative homology in an even more general setting. Paralleling the similar situation in cohomology, we therefore may define three homology groups depending on k = 0, 1,... on X. The homology group Hk(X,]R)
flat cycles in X flat boundaries in X' if T E Zk(X) is a flat cycle, we denote by [T] x its homology class in Hk(X,R). The homology group Hk (OX, R)
flat cycles supported on aX flat boundaries in ax
Hk(aX,R)
for T E Zk(aX), we denote by [T] ax its homology class in Hk(aX,
lR
The relative homology group Hk (X, 9X, ]R)
Hk (X, aX, R) :=
flat relative cycles flat relative boundaries'
for T E Zk (X, DX) we denote by [T] 1el its relative class in Hk (X, aX, R). The inclusion map i : aX -> X induces a map
i# : Hk(DX,R)
Hk(X,1R).
As cycles are special relative cycles and boundaries are special relative boundaries an inclusion map .7# : Hk (X, RIB) - Hk (X, ,9X, IR)
is well defined by
#([T]x) = [T]rei.
Finally, the boundary operator a : Fk,x(R"1') - Fk-l,x(Rn+1`') induces naturally a map a : Hk (X, aX, R) -4 Hk_ l (aX, ]R)
defined by
a([T]rei) = [DT]ax.
Then it is not difficult to prove
3.3 Poincare-Lefschetz and de Rham Dualities
591
Theorem 1. The long sequence in homology
#Hk(X,R),#Hk(X,aX,R)
... a Hk-,(aX,R)->Hk-,(X,R)-+ # j#
is exact.
It turns out that the long sequence in homology and the long sequence in cohomology are in duality, such a duality being given by Poincare-Lefschetz duality. In the sequel of this section we shall discuss this topic which will lead us to a de Rham theorem for manifolds with boundary. For k = 0, 1, . . . , n we first consider the pairing
<, >: given by
<w, r7> :=
(2)
En-k(X) ,
Ek-l(aX) X
r
J ax
wnr7.
As we saw in Sec. 5.3.2 (2) induces a map P:
Zn-k(aX)
, Zk-1(aX)
given by
ax which factorizes to the Poincare duality isomorphism
P : HHRk(aX) -> Hk-1(aX,
IR
compare Theorem 2 in Sec. 5.3.2. Of course there are other pairings on X
<, >: Ek(X) X En-k(X) --> ][8 again given by
<w,7)> =
(3)
xJ
wAr7.
But, as by Stokes theorem (4)
JX
do-n77 + (-1)k f o-Adrl =
x
f
J ax
oni7
a E Ek-1(X), rl E En-k(X)
we see that there is no way of inducing a map between the absolute cohomology and homology of X unless aX = 0. However, if 77 E Zn-k (X, aX) the k-dimensional current
592
5. Cartesian Currents in Riemannian Manifolds PO77(w)
(5)
(-1)k(n-k) ((X
fwA77 =
IL
w E Ek (X )
L 7? (W),
x is a cycle in X, Par? E Zk(X) by (4). Also the homology class [PO77]x of P077 depends only on [77]rel. In fact, if 77 = da with i#a = 0, then by (4) Pbr7(w)
= J wAda = (-1)k+1
f dwna = (-1)k+1apta(w).
In conclusion we can say that the map Pt : Z,-k (X, aX) -+ Zk (X) induces a map again denoted by Pd
Pt : Hdn k(X,aX) , Hk(X,ltd).
(6)
Finally, in a similar way, one proves that for 77 E Z,-k(X) the current Pb :_ (-1)k(n-k) QX] L7)
(7)
is a relative cycle, P577 E Zk(X, aX), and moreover the map Pb : Zn-k(X) --> Zk (X, aX) induces a map Pb : HHR(X) , Hk (X, aX, ltd).
(8)
We have
Theorem 2 (Poincare-Lefschetz isomorphisms). The maps P, Pt, Pb are isomorphisms, and the diagram S
H,Rk(X,8X)
HdRk(X)
1 Pt
I Pb
(9)
Hk (X, R) i#
-i
Hk (X, aX, IR)
£
Hdn k (aX)
-> .. .
IP
a
Hk-1(aX, R)
#
.. .
is commutative.
We shall refer to the maps Pt, Pb, P as to the Poincare Lefschetz duality isomorphisms. It is quite simple to prove the commutativity of the diagram, though quite long. We already known, compare Sec. 5.3.2 that P is an isomorphism, Poincare duality isomorphism. We shall therefore prove only that Pt and Pb are isomorphisms. This will be done in the next propositions below. Let X be a compact, oriented, smooth n-dimensional manifold with smooth boundary 8X. To be precise we think of X as a C2 manifold with a C2 boundary.
Observing that aX has codimension 1 in X, and X is oriented, the tubular neighbourhood theorem on aX yields
Theorem 3 (Collar theorem). There is an eo > 0 such that the normal geodesic flow to TxaX at x E 0X, g(t, x) : [0, co] x aX --> X is a difeomorphism onto a neighbourhood of aX in x with g(0, x) = x for all x E aX.
3.3 Poincare-Lefschetz and de Rham Dualities
593
Fig. 5.4. The collar map.
A consequence of the collar theorem is then
Theorem 4. We have HHR(X, aX) = H pt(X \ aX) where
HPt(X\8X)
{wEC-(X\aX,AkT(X\aX)) I dw=0} {w=da I aEC0(X\9X,AIT(XQ,aX))}'
In fact
Proposition 1. We have (i) Each relative cohomology class contains a form 6 E Zk(X,aX) which is compactly supported in x \ ax. (ii) Each homology class of X contains a cycle T E Zk(X) of finite mass such
that spt T n ax = 0.
Let Eo and g be as in the collar theorem, Theorem 3. Define r(t) [-Eo, 0] -> [-eo, 0] to be a non-decreasing function such that r(-eo) = -6o, r(-Eo/2) = r(0) = 0, r'(-eo) = 1. Of course (1 - s)t + sr(t) yields a homotopy Proof.
with the identity which is fixed at -Eo and 0. Consider then the mapping O(x)
x g(r(t), y)
if x E X_Eo
if g(t, y) = x E X \ X_E0
which maps X \ X_E0 onto aX and satisfies 0 o i = i on aX. Then h(s, x) := (1 - s)x + so(x) yields a homotopy of ¢ with the identity which keeps the boundary fixed. Consequently, if ri E Zk(X, aX), the form w = 0#rl vanish on X \ X_E0/2i and by the homotopy formula w - 77 = d[h#il](o,l)
5. Cartesian Currents in Riemannian Manifolds
594
This concludes the proof of the first claim as i# [h#r)] (o,l) = 0 being h(s, x) = x
d x E aX and V s E [0, 1]. Analogously, if T E Zk(X), we infer from (i) of Proposition 4 in Sec. 5.1.3 that T is homologous to a relative cycle T' E Zk(X) of finite mass and consequently r#T' E Zk(X) has finite mass. Moreover r#T' is homologous to T' and spt r# TI f1 aX = 0. Actually without loss of generality we can and do assume that, in case aX # 0, X is actually a compact piece of a larger manifold Y and the geodesic flow on X normal to aX extends to both sides of 8X in Y. Denoting such a flow by g(t, X), the collar theorem Theorem 3 then yields, possibly for a smaller EO, a diffeomorphism g(t, x) : [-EO, Eo] x aX - * Y
onto a neighbourhood of aX in Y such that g(0, x) = x for x E aX and g(t, x) E X if t > 0. The equation g(t, x) = y means that x is the point of least geodesic distance in aX from y, Itj is the geodesic distance of y from 9X, and t > 0 if
yEX. For 0
EU we now set
(10)
X_E
{x E X I x 0 g([0, e] x 8X}
X+e
g([-E, Eo] x ax) U X
and X0 = X. Of course any XE, E E [-EO, EO], is a compact oriented manifold with boundary. Given any smooth increasing function r(t) : [-60, Eo] -p [-60, 60]
with r(-Eo) = -so, r(eo) = eo, r'(-eo) = 1, r(-EO/2) = 0, and r(0) = Eo/2, it is convenient to introduce the map e : X+EO -r X+ep
defined by e(x)
fx
if x E X_E0
g(r(t), y)
if g(t, y) = x E X+E0 \ X_60.
It is easily seen that e yields an isomorphism which is the identity on X_EO, maps X \ X_E0/2 onto X+,012 \ X, &X_.012 onto aX and aX onto aX+eo/2 We can also assume that e is homotopic to the identity with a homotopy which keeps the boundary values fixed. The inverse map of e denoted by r r : X+E0 --4 X+EO
which maps the collar X+,012 \ X onto X \ X_,012 and 09X+EO/2 onto OX, is then also homotopic to the identity with a homotopy which leaves aXE0 fixed. Finally, we shall also use the distance function from X. Setting Ue := {x E R"`+N we have
I
dist (x, X) < E}
3.3 Poincare-Lefschetz and de Rham Dualities
595
Proposition 2. There exists eo > 0 such that any point x E U 0 has a unique point 7r(x) of least distance on X. Moreover the map 7r : U6p -- X is Lipschitz in UE and smooth in U6 \ X. Finally, for x E X \ aX, 7r-1(x) is just a flat ball of center x and radius 60 lying in the normal space to T, X . For every 6 in (-60, 60) we consider the enlarged manifold X6 and U6
{x E Rn+rr
dist (x, X6p/2) < 6}.
I
Choosing el so that eo = 461 and taking 61 smaller if necessary a least distance projection map it : UU1 --> X261
is well defined: it maps each x E UU1 to the least distance point in X2,1.
Proposition 3. P is injective, i.e., if w E Zn-k (X, aX) and (11)
S E Fk+l,x (U6y ), then w is relatively exact, that is there exists a E
Dn-k-1(X
)
with i#a = 0 such that w = da. Proof. On account of (i) of Proposition 1, we can and do assume that w = 0 on X \ X_,1. Moreover on account of Proposition 4 in Sec. 5.1.3 we also assume that S has finite mass. Consequently, as in the boundaryless case, we infer Q7r-1(X)J L7r#w = 7r*(f X Lw) = a7r*S
in U,1,
spt,3 C 7r-1(X_61/2), compare Proposition 1 in Sec. 5.3.2 and Remark 1 in Sec. 5.3.2. In particular w is exact on X and zero near aX. hence 7r#w is exact, 7r#w =
Proposition 4. PO is surjective, i.e., if T E Zk(X) there exist a relative closed form w E Zn-k(X, aX) and a current S E Fk+1,x(U61) such that
Proof. First notice that one can suppose that M(T) < oo and spt T C X_61
on account of Proposition 1. We then regularize T in U61 with a parameter 6 << 61/2. This way we have
T=f n6+aS' UE1
where l; E Dn+N-k (U,1 ), spt C 7-1(X_61/2), S' E Fk+l,opt 7r-1(X_61/2), M(S') < oo. Projecting onto X we get (12)
T = 7r#T = 7r#(
f
Ue1
-
a7r#S'
spt S' C
596
5. Cartesian Currents in R.iemannian Manifolds
with it#S' E Fk+1,x(UE,); as for 77 E Sk(X) we have fx 7r#17AC = fx 7 A7r , the conclusion follows since (12) yields also dC = 0, hence ir.dC = 0, and
e = 0 near
1(aX), hence 7r.C = 0 near aX.
Proposition 5. Pb is injective, i.e., if w E Z,-k(X) and
QX]ILw=R+aS
(13)
with R E Fk,ax (UE, ), S E Fk+l,x (UE, ), then w is exact, that is there is a E Bn-k-1(X) such that w = da.
Proof. On account of Proposition 4 in Sec. 5.1.3 we can and do assume that M(R) + M(S) < oo. Applying the extension map e we get Q Xe, I L r#w = R'+ aS'
(14)
where R' = e#R E Fk,aX,012 (U), S' = e#(S) E Fk,XE012 (U), M(R') + M(S') < oo. Since rIx is homotopic to the identity and dw = 0, the homotopy formula for forms yields that w - r#w is exact, hence it suffices to prove that r#w!x is exact too. For that notice that (14) yields
1 XE,12 Lr#w = aS'
in U, fl 7r-1XEO/2.
We then proceed as in Proposition 3 and conclude that 7r#r#w is exact in Ue, n 7r-1Xeo/2 This implies that r#w is exact on X.
Proposition 6. Pb is surjective, i.e., if T E Zk(X, aX) then there are w E Z,-k (X), R E Fk,ax (U,,), S E Fk,x (UU,) such that Proof. We may and do assume that M(T) < oo, compare Proposition 4 in parameter, we infer that Sec. 5.1.3. Regularizing T with a
T= I A +aS
(15)
U.,
Dn+N-k(Uel) S E Nk+1(Ue, ), spt S CUE, fl 7r-1(Xep/2). Therefore, with E projecting on X,.
T = 7r#T =
J
r# () AC + a7r# S
U.,
=
f
xEo/2
. n7r. + a7r#s =
f
. A7r.C + R' + aS'
x
where R' := fxEO/2\x Air.C, S' := -7r#S. As X is a deformation retract of X0012i
denoting by p :X,.12- X the retraction, we infer
3.3 Poincare-Lefschetz and de Rham Dualities
T = p#T =
f
597
. A1r*C +p#R' + ap#S'
X
with R' E Nk+l,ax(UE1), p#S' E Nk+l,x(UE1). The claim now follows observing
that being aT = 0 in X \ 8X, (15) implies that d6 = 0 on 7r-1(X \ aX) which in turns implies dint = 0 on X \ aX and by continuity dir 1; = 0 on X. Remark 1. We emphasize the fact that surjectivity of P4 amounts to the existence for a given absolute cycle T E Zk (X) of a relative Poincare dual form PTE Zn-k(X,aX) such that Pp[PT]rei = [T], i.e.
J T(w) = ox wfPT Vw E Zk(X ) 1
T-
i#
It is reasonable to find for T E Zk (X) a Poincare dual with null value at aX since one can deform with an homotopy T into T' E Zk(X) in such a way that spt T' n aX = 0. Also the surjectivity of Pb means the existence for a given relative cycle T E Zk (X, aX) of a Poincare dual form PT E Zn-k (X) such that PL[PT] _ [T]rel, i.e.
r
T(w)=J wAPT+R+aS x
`dwEZk(X)
where R E Fk,ax (UE1 ), spt R C aX, S E Fk+l (UE1)
Remark 2. The proof of Theorem 2 can follow a different path. First one proves
that P and P are isomorphisms, as previously, then that the diagram (9) is commutative, and finally one proves the bijectivity of Pb using the so-called five-lemma
Proposition 7 (Five Lemma). Suppose we have the following commutative diagram (16)
Al
a+
1h Bl
A2
a2
IA B2
A3
a3
B3
a4
I f4
j f3 _+
A4
G
B4
A5 Lf5
_+
B5
consisting of vector spaces Ai, Bi, i = 1, ... , 5 and of linear maps ai, bi, i = 1, ... , 4 and fi, i = Im bi = Ker bi+i then
1, ... , 5. If the rows are exact, i.e. Im ai = Ker ai+i,
(i) if f2, f4 are surjective and f5 is injective, then f3 is surjective, (ii) if f2, f4 are injective and fl is surjective, then f3 is injective (iii) if fl, f2, f4, fs are isomorphisms, then f3 is an isomorphism.
598
5. Cartesian Currents in Riemannian Manifolds
de Rham duality on manifolds with boundary. As consequence of the non-degeneracy of Poincare-Lefschetz pairings, compare Theorem 3 in Sec. 5.2.8 and taking into account Poincare-Lefschetz isomorphisms, we get
Theorem 5 (de Rham). The pairings <[T]x, [cp]x>9 : Hk(X,R) x HakR(X) --.lI8 < [T]rel, [co]rer >b : Hk (X, aX, R) x Hag.(X, 9X) --> R
induced by the pairings: Ek(X) X Ek(X) ---> R, := T(cp) are non degenerate. In particular
HHR(X,aX) =Hk(X,aX,R)*.
Hk(X) =Hk(X,R)*, Moreover, if we also set
< [T]ax, [W] ax > := T (cp) for [T]ax E Hk (OX, R) and [cp] ax E HaR (5X ),
the following diagram is commutative
...
Fa--
HaR(aX)
HHR(X)
x<, > ...
't-
x<, >a
- Hk (aX, R) * Hk (X, R)
...
HHR(X,aX)
x<, >b '*+
Hk (X, (9X, R)
--a+...
Let Y 1 , .. ,.ys be cycles in Zk(X) such that [Yi]x, [Y2]x, - -, [rys]x form a basis of Hk (X, R). Given a form w E Zk (X), the periods of w with respect to -
yl...... s are the numbers yi(W),
i=1,...,s.
Also if 81i ... , 8, be relative cycles in Zk(X, aX) such that [8i]rel,
- - - ,
[8r]rel form
a basis for Hk (X, aX, R), and w E Zk (X, OX), the relative period of w with respect to 61,. .. , Jr are the numbers 8i (w),
i = 1, ... , r.
Then it is easily seen that equivalently de Rham theorem can be stated as follows
Theorem 6. Let [yi], i = 1, ... , s be cycles in X such that [yj]x forms a basis in homology. Then (i) If w E Zk(X) and all the periods of w with respect to yi are zero, then w is exact. (ii) Given numbers ai E R, i = 1, ... , s. Then there is a form w E Zk(X) with
given periods ai, i.e. 7i(w) = ai, i = I,-, s.
3.4 Intersection of Currents and Kronecker Index
599
Also, if {8i}, i = 1, . . . , s are relative cycles in Zk(X, OX) such that [8i]rei, i = 1, ... , s form a basis for Hk (X, aX, IR), then (iii) If w E Zk(X, aX) and all the periods of w are zero, 8i(w) = 0,
i = 1'. .. , s,
then w is relatively exact, i.e. w = da, a E Ek(X) and i#w = 0. (iv) Given numbers ai E R, i = 1, ... , s. Then there is a form w E Zk (X, aX ) with periods ai, i. e. 8i(w) = aj, i = 1, ... , s. Also, the dual statements holds true, in particular
Theorem 7. We have
(i) T EBk(X) ifT(w)=0VWEZk(X) (ii) T E Bk(X, aX) if T(w) = 0 V W E Zk(X, aX) Of course for T E Zk (X) we have T (w) = 0 b w E Zk (X) iff T (wi) = 0 d wi E
Zk(X), i = 1, ... , s, [wi]x being a basis of HHR(X), and for T E Zk(X, OX) T(w) = 0 V w E Zk(X, aX) if T(wi) = 0 V wi E Zk(X, aX), i = 1, ... , r,
[wi]re1
being a basis of HHR(X, OX). A consequence of Theorem 7 is then
Theorem 8. Bk (X) and Bk (X, aX) are closed with respect to the weak convergence of currents respectively in Zk(X), Zk(X, ax). Consequently we can state Theorem 9. Homology classes can be represented by mass minimizing currents. That is, given y E Hk (X, R) and 8 E Hk (X, aX, IEF) there are currents T E Zk (X) and R E Zk (X, ax) such that
M(T) = inf{M(S), S E Zk(X), [S]x = y} M(R) = inf{M(S), S E Zk(X, aX), [S]rel = y}.
3.4 Intersection of Currents and Kronecker Index The aim of this subsection is to discuss some extensions to currents of the classical notion of intersection of oriented surfaces in generic positions and to introduce the intersection index of homology classes.
Intersection of currents in R'. Let U be an open set of Rn and let S E Dk(U) and T E Dh(U), where h + k > n. Denote by f : Rn x I[8n --> Rn the map f (x, y) := x - y, and let 9: U --> U x U be the diagonal map
g(x) := (x, x). Suppose that the slice, compare Sec. 2.2.5, < S x T, f , 0 > exists and defines a current in Nh+k_n(U X U). Then its support lies in {(x, y) I x = y} = Img,
5. Cartesian Currents in Riemannian Manifolds
600
consequently, by Federer flatness theorem, see Sec. 5.3.1, there exists a unique current, denoted s n T and called the intersection of S and T , in Nh+k-n(U) such that (-1)(n-k)h<S x T, f, 0>. g# (s nT) _ (1) Actually it suffices that <S x T, f, 0> be a flat chain, compare Sec. 5.1.3 below. In the next proposition we collect a few simple properties of the intersection.
Proposition 1. We have (i) If s n T exists, then spt (S n T) C spt s n spt T. S=(-1)(n-k)(n-h)S (ii) S n T exists iff T n S exists, and T n n T. (iii) Assume that S n T, and either S n aT or aS n T exist. Then
a(S n T) = S n aT + (-1)h-k 3S n T.
(2)
Proof. (i) follows as
spt <S x T, f, 0> C spt S x spt T) n {x = y} = g(spt S n spt T). (ii) Using the map r(x, y) = (y, x), which reverse the role of the factors, we infer
for = -f, r#(T x S) = (-1)hkS x T, consequently (-1)hk<S x T, f, 0> == r# = (-1)nr# = (-1)n .
It follows s n T = (-1)'T n s where r = -hk + n + (n - h)k + (n - k)h = (n - h)(n - k) (mod 2). (iii) From 8(S x T) = 8S x T + (-1)kS x 8T, S E Dk(U), we infer that
8<S x T, f, 0> _ (-1)n = = (-1)n<3S x T, f, 0> + (-1)n-k<S x 8T, f, 0> provided < S x T, f, 0 > and either < 8S x T, f, 0 > or < S x 8T, f, 0 > exist. Therefore 8(S n T) = (-1)a8S n T + (-1)T S n aT where
Q=n+(n-h)h+(n-k-1)h=n-h (mod 2) -r=n-k+(n-h)h+(n-k)(h-1) =0 (mod 2). Intersection theory of currents is in general quite complicated as the intersection of currents may fail to exist. In fact while the existence of the slice < S x T, f , z > is granted for a.e. z by the Lebesgue differentiation theorem, it may fail to exist at a single point, e.g. z = 0. The following examples show some of-the difficulties. -------------
3.4 Intersection of Currents and Kronecker Index
601
0 It is clear by the definition that the intersection defines a bilinear map on Nk(U) x Nh(U) whenever it exists. It is also worth noticing that in general the density of intersection current is the product of the densities of the factors. For example let W E LT'(Rn), 0 E Lq(Pn) be two functions. Consider the ndimensional currents S :_ Rn L W and T = lR L' &. Then the current S x T E Den (RI x JRn) is given by
S x T(w) = f (P (x)'b(y)w(x,y),
w E D2n(Rn x llgn).
R'' xR-
If f (x, y) = x - y, Op (t) = XB, (t)dt' n ... dtn, for
w = E waQ(x, y) dxandyQ I&I+IQI=n
we have 1
wnpn
dx ¢ W(x)o(y)
S x T(f#(Qp)nw)
a(Y, Y)w7ti(x,y) dy h'I
Bp(x)
li
Since
O('Y,7)w77(x,y) - b(x)g#(w)(x)
O(y)
I+
in Lq(PJ) we infer that, assuming 1/p + 1/q < 1, <S x T, f, 0> exists,
<S x T,f,0>(w) = lim
- S x T(f#(S?p)nw) =
f
hence
s n T(w) = f O(x) (x)w(x)
w E Dn(Rn).
Simple examples of course show that s n T may fail to exist.
0 Let Q = Q (0, 1) x (0, 1) 1 E 12 (R2) be the integral current integration over the square in JR2 and let 7r(x, y) = x be the projection on the first factor. Then it is easy to compute 8x1(0,1)1=
18:,: x Q
0
(0,1)
if0<x<1 if x = 0,1 otherwise
which shows that at specific values the slice need not be rectifiable. Notice that
such a phenomena would not appear if we project into a plane which is not perpendicular to any of the edges of Q.
0
5. Cartesian Currents in Riemannian Manifolds
602
However, by slightly moving one of the two currents, we can and do ensure the existence of the intersection. In fact denoting by Ta, (x) = x + a, x, a E 1I ' we have
Proposition 2. Let S E Nk(U) be a normal current compactly supported in U, and let T E Nh(U), h + k > n. Then for a.e. a E ll close to zero (i) S n Ta,#T exists in Nh+k-n (U), (ii) a(S n Ta,#T) = S n Ta#oT + (-l) k (as) n Ta#T, (iii) if moreover S E Ik (U), T E Th (U) then s n Ta#T E Zh+k-n (U) . Proof.
From the slicing theory <S x T, f,z> exists in Nh+.k_n(U) for a.e.
z E Rn close to zero. On the other hand,
<S x Ta#T, f, 0> = < S X Ta#, f o (id x Ta), 0> = id x Ta#< S x T, f, -a> whenever <S x T, f, -a> exists. This proves (i). (ii) then follows from (iii) of Proposition 1 and (iii) follows similarly to (i) on account of <S x T, f, z> E Zh+k-n(U) for a.e. z E R. Also the intersection is well defined, essentially if we are able to prove continuity of < S x T, f , z > in z. We have in fact
Proposition 3. Let S E Nk(U) and T E £n-h(U). Then S n Rn L T exists and S n (R' L T) = S L T
in U.
Proof. Let w E Dk+h-n(U X U). Consider the k-form O(x, y) = r(y)Aw(x, y) E Dk(U X U). As f #(dtlA... Adtn) = (-1)I o (ry, y) dx"Adyry we have
T(y)Af#(dtlA...Adtn)Aw(x,y) _
'Y)dxryAO(x,y)ndyry, 171<_n
where b=(n-j)+j(n-h)+(n-j)(h+k-n), j Writing 0 = E 0«p dx«AdyQ, we get as p --+ 0
f
B(x,p)
0
if )3
o'('Y,'7)0«y(x, x)
if j3=ry
ry
by continuity. Consequently the averages
I T(y)Af#(dtlA... Adtn)Aw(x, y) B(z,p)
converges, as p goes to zero, to (3)
(x x) dx«Adx1" _ (-1)(h+k)ng#(0)(x) 171
3.4 Intersection of Currents and Kronecker Index since 8 + I ry I
603
I a I = 8 + j (k - j) _ (h + k) n mod 2. Since now S X Rn LT(f#(QP)nw)/wnpn
=
(-1)(n-h)kS w Pn x 1[8n(TAf#(dt'A...ndtn)nw)
I
_ (-1)(n-h)kS(
T(y)Af#(dt'A...Adtn)Aw)
B(x,p)
from (3) we infer that < S X I[2n L T, f, 0 >(w) exists and agrees with (
1)(n-h)kS(g#0) = (_1)(n-k)hS(Tng#w). In other words
g#(sn1
LT) =g#(SLT).
As immediate corollary we notice
Corollary 1. Assume o- E
£n-k
£n-h
(U), T E
(R'
(U), h + k > n. Then
R'
in U.
Of course one is also interested in intersecting currents when both factors are not integration against form, the case of intersection of rectifiable currents being relevant. We shall not go into these topics and confine ourselves to mention the following result taken from Federer [226] which applies for instance to the intersections of smooth surfaces in general position. We have
Theorem 1. Let S E Tk (Rn), T E Zh ([$n) be two integral currents, h + k > n. Assume that spt S and spt T are contained respectively in two smooth submanifolds X and Y of dimension respectively k and h so that
S = r(M,0,X),
T =T(N,cp,y)
where M C X, N C Y. Assume furthermore that (i) 0 is bounded on M, go is bounded on Al. (ii) For each x E X n y, TTX n Ti,)) has dimension h + k - n. (iii) Either k = 0 or spt aS n spt T has a neighborhood E C
Rn such that
E n sptaS is covered by a finite family r of (k - 1) -dimensional smooth submanifolds G of X with E n clos G C G
TXGnTxyhas dimensionh+k-n-1dxEGny. (iv)
Either h = 0 or spt S n spt aT has a neighborhood F C 1R? such that F n spt aT is covered by a finite family A of h - 1-dimensional smooth submanifolds D of y with F n clos D C D and
TXDnTXX has dimension h+k-n-1V X EDnX.
604
5. Cartesian Currents in Riemannian Manifolds
Then the (h + k - n) -vector ((x) given by
9#,.(((x)) = X(x)ny(x) L (df'A... Adfn)(x,x) is non zero and simple, whenever x E spt T n spt S \ (spt aT U spt aS). Moreover Hh+k-n (spt T n spt S \ (spt aT U spt aS)) = 0 and
SnT =T(.tit We omit the proof of Theorem 1 and we refer to Federer [226]. We just make a few remarks on the role of transversality of X and Y. If X (x), y(x) are the two
unit tangent vectors to X and y at x, then X(x)ny(x) generates T,(X) ®T.(y) in the product R' X I. As Im g = ker f , we have TTX n Txy has dimension h + k - n iff
f : Tx X ® T. y has a kernel of dimension h + k - n iff
f,TTX®TTY has maximal rank n.
From Proposition 3 in Sec. 5.2.2, the last claim is equivalent to say that
X(x)AX(X) L (df'A... ndfn) 0 0. Intersection in real homology. In contrast to the many difficulties in defining the intersection of currents, homologically the intersection of cycles is always well defined. Let X be a compact, oriented, n-dimensional submanifold of IIBn+N possibly with boundary. We first consider the intersection of relative and absolute homology classes. Let [S] rel E Hk (X, aX, ]IP) and [T] E Hh (X, Ii), h + k > n, and let P. be a Poincare dual form of T, PT E Zn-h(X, aX). It is easy to see that
the absolute homology of the (h + k - n)-cycle
S('nPT) depends only on the homology classes [S]rel of S and the absolute homology class [T] of T. Then we define the intersection map
nX : Hk(X,aX,R) x Hh(X,]R) -> Hh+k-n(X,I) by (4)
[S]rei nX [Ti :=
Similarly one defines the intersection map
(-1)(n-k)(n-h)[S(
APT)]
3.4 Intersection of Currents and Kronecker Index
605
nx : Hk(X,R) x Hh(X, R) --i Hh+k-n(X,IR) (-1)(n-k)(n-h) for [S] E Hk(X,R), [T] E Hh(X,R), by [S] nx [T] := h+k > n. Finally denoting by PT E Zn-h(X) a Poincare dual form of a relative cycle T E Zh(X, aX), the formula
(-1)(n-k)(n-h)
[Sire, nX [T]rel :_
[S(' APT)]rel
defines a third set of intersection maps nX : Hk (X, aX, ]R)
X
Hh(X, aX, R) -4 Hh+k-n (X, aX, R).
From the definitions it is easily seen that intersection of cycles in homology is Poincare dual of the wedge product in cohomology. For instance for the map nX one has the following commutative diagram Hdn k(X)
X
Hdn h(X,aX) A
Pb
I Pd
I Hk (X, aX, IR)
x
Hh (X, lid)
-n
Hen-h-k(X, l aX)
t
Pd
Hh+k-n (X, l)
where [w]A[77] := [wArf].
Kronecker index in 1EBn. Finally, let us introduce the Kronecker index S E Nk (Rn) and T E Nn-k (Rn) be two normal currents of complementary dimensions. Assume that either k = 0 or spt S n spt aT = 0 and either h = 0 or spt T n spt aS = 0, so that spt (a(S x T)) does not meet the diagonal of J2n, i.e., 0
f (spt a(S x T)).
Then the current (-1)n-kf#(S x T) belongs to Nn,(Rn) and 0 does not belong
to spta((-1)n-kf#(S x T)). The constancy theorem then yields a constant k(S, T) E ][8 such that in a neighborhood of 0
(-1)n-k.f#(S x T) = k(S,T)QRnj The number k(S, T) is called the Kronecker index of S and T, and it is actually an integer if S and T are i.m. rectifiable. (5)
For any cp E C°(JR) and p sufficiently small, applying (5) to cp(t)xBP dt' n ...
ndtn =: p(t)Qp(t), we get
(_1)n-k IBpI
S x T L f #(.Q p )(cp o f) = k(S T)
I B(°,p)
-(t) dt.
606
5. Cartesian Currents in Riemannian Manifolds
Fig. 5.5. SnT =5a-bb, k(S,T) =0. Taking cp = 1, we then conclude in the case that S n T exists that
(S nT)(1) = g# (S nT)(1) =
(-1)(n-k)2+(n-k)<S
x T, f, 0>(1) = k(S,T).
A special interesting case is that of two oriented submanifolds S and T in generic position. By Theorem 1, in this case S n T exists and S n T = >2 njbp. where the ni are defined by S(PZ)nT(Pi) = nzejA ... nen. Therefore we conclude
k(S,T) _ 1: n2, i.e., Kronecker index of S and T is the sum with sign of the intersections. Another interesting case is that in which Dn-k(Rn) S = (-1)k(n-k)Rn La, uE T E Dk(Rn). T = (-1)k(n-k)Rn LT In this case, as we have seen, S n T exists and, by Proposition 4 in Sec. 5.3.2, k(S, T) = S n T(1) = I[gn L (Tnv) (1) = frAa. lin
Kronecker index on manifolds. Of course one can extend the notion of Kronecker index to currents on oriented manifolds. More precisely let S E Nk (X), T E Nn-k(X) where X is a n-dimensional, compact oriented submanifold of lRn+N Assume
k = 0 or spt a8 n spt T = O and k =nor spt aT n spt S = 0. Recall that the lift 7r* (T) on a tubular neighborhood U of X of bounded crosssection has finite mass, and that
3.4 Intersection of Currents and Kronecker Index
607
spt a7r*T n spt S = 0
since spt a7r*T lives on it-1(spt aT) or outside U. Then we can set
Definition 1. The Kronecker index of S and T in X is defined by
kx(S,T) := k(S,lr*T), i.e. as the number such that
(_1)n-kf#(S X lr*T) = kx(S,T)QRn+1v
(6)
near the origin.
From (6) it is easy seen that (i) (ii) (iii)
kx(S,T) is integer-valued if S and T are i.m. rectifiable kx (aR, T) = kx (S, aR) = 0. kx (S, T) = 0 if spt s n spt T = 0.
In particular, if S E Zk (X, aX) is a relative normal cycle and T E Zn-k (X) is an absolute normal cycle which do not meet aX , k(S, T) is well defined and depends only on the relative homology class of S and on the homology class of T, thus it actually defines a map
ix(S,T) : Hk(X,aX,R) x Hn_k(X,]R) -i ][P which is the intersection index of S and T T. As we have essentially seen, if s n lr*T exists, then
s n 7r*T(1) = kx(S,T) = ix(S,T).
In particular if S = the boundary of X, and T = T, then
(-1)k(n-k)
(7)
Q X I LPs, i.e. PP is Poincare dual of S null to (-1)k(n-k)
Q X ] L PT, i.e. PT is a Poincare dual of
1 PAP
x Therefore we can state that the intersection index is Poincare-Lefschetz dual of Poincare Lefschetz duality on forms. Of course one has also an intersection index on the compact oriented manifold without boundary aX iax : Hk (aX, R) x H,,,-k (aX, R) -> R defined by iax (S, T) := Kax (S, T). This yields another way to state non degeneracy of Poincare duality of forms.
608
5. Cartesian Currents in Riemannian Manifolds
Theorem 2. Let X be a compact oriented manifold. The intersection index
ix : Hk(X,8X,R) x Hn_k(X,R) ->
II8
is non degenerate. Moreover, denoting by iX (T, S) := ix (S, T), the following diagram Z#
Hk (X, aX, R) -
Hk(X,R)
Hn_k (X, 8X, IE8)
<.7#
Hn-k (X, Ilk)
s .. .
x iax
X ix
X iX a
Hk-1(3X, R)
4---
Hn-k (8X, lR)
z#
a
is commutative.
3.5 Relative Homology and Cohomology Groups We conclude this section reporting partially and briefly on Federer and Fleming approach to relative homology groups in the setting of local Lipschitz neighbourhood retracts. Let A and B be local Lipschitz neighbourhood retracts in R. Suppose B C A and B relatively closed in A. Denote by 7r : U -> A and p : V B the retractions of neighbourhoods U of A and V of B, and assume, without loss in generality,
that V C U and moreover that (1 - t)x + tp(x) E V whenever x E V and t E [0, 1]. For k = 0, ... , n, we define relative cycles by
Zk (A, B, R) := IT E Fk,cpt (U) I spt T C A, spt 8T C B, or k = 0}, relative boundaries by
Bk(A,B,R) ._ {R + 8S I R E Fk,cpt (V), spt R C B, SE Fk+l,cpt (U), spt S C A}, and the real relative homology groups by Hk (A, B, R) = Zk (A, B, R) /Bk (A, B, R).
We point out that all currents involved have compact supports and that the retractions 7r and p induce continuous maps 7r# : Fk,cpt (U)
Fk,cpt (A),
p# : Fk,cpt (V) - Fk,cpt (B)
The relative homology groups defined above satisfy the Eilenberg-Steenrod axioms for a homology theory in the local Lipschitz category. We recall that the local Lipschitz category consists of couples (A, B) of local Lipschitz neighbourhood retracts, with A D B, B relatively closed in A, and , morphisms f between couples
3.5 Relative Homology and Cohomology Groups
609
(A', B') are Lipschitz maps from A to A' satisfying f (B) C B'. It is not difficult to prove using Federer's flatness theorem that any such morphism induces a map f : (A, B)
f# : Hk (A, B, R) -+ Hk (A', B', R)
compare Federer [226, 4.4.1]. Also if C C B C A are local Lipschitz retracts in W1, the boundary operator acts as
a: Zk(A,B) -* Zk-1(B, C), 8: Bk(A,B) --> Bk-1(B,C), thus induces a map again denoted by 8
8 : Hk(A, B) -* Hk_1(B, C). Here we understood the dependence on R, Hk (A, B) := Hk (A, B, R).
Theorem 1. We have (i) If f : (A, B) -> (A, B) is the identity map, then f* : Hk (A, B) --> Hk (A, B)
is the identity map. (ii) For f : (A, B) -> (A', B') and g : (A', B') --+ (A", B") we have
(g o f)* = g« o f*.
(iii) IfCCBCA, C'CB'CA', f:(A,B)-+(A',B'), fB:(B,C)->(B',C') and k > 0, then
fiB*a = of*. (iv)
If i : (B, C) -* (A, C) and j : (A, C) --+ (A, B) denote the inclusion maps, then the sequence of homomorphism
a, Hk (B, C) (v)
Hk (A, C) '-', Hk (A, B)
Hk-1(B, C) A .. .
is exact. I f h : ([0,1] x A, [0,1] x B) --> (A', B') and h* : (A, B)
(A', B'),
h* (x) = h(t, x), 0 < t < 1, x E A,
then ho* = h1*. (vi)
If i (A, B) --. (A', B') is an inclusion map such that B = A f1 B' and clos (A'\B')f1(clos (A'\A) = 0, or equivalently B C A C A', B' = BU(A'\A) :
and clos (A \ B) n clos (A' \ A) = 0, then i* is an isomorphism. 'vii) If x E R', then Ho({x}) ^ Z and Hk({x}) _ {0} for k > 0.
Suppose now that A and consequently B are compact. Then one can construct a CW-complex LA surrounding A and contained in U which has a sub complex LB surrounding B and contained in V by selecting cubes not far from A of a standard subdivision of IR in cubes with a suitable small mesh size. As in the case B = 0 one proves
610
5. Cartesian Currents in Riemannian Manifolds
Theorem 2. Let A be a compact. Then Hk (A, B, IR) is isomorphic to the relative homology of Lipschitz chains of (A, B). Moreover Hk,,impi (LA, LB, R) can be mapped linearly onto Hk (A, B, R). In particular the vector space Hk (A, B, R) has finite dimension. In this more general context, following the same path of Sec. 5.3.2 one derives an isoperimetric inequality
Theorem 3. Let A D B be compact neighbourhood Lipschitz retracts. There exists a constant r > 0 such that, if T E Zk (A, B) is a relative homology cycle of finite mass, M(T) < oo, with T = X+aY with X E Fk,,pt(V), Y E Fk+1,cpt(U), then T = Bk(A, B), i.e.,
T=R+aS,
with R E Fk,,pt(V), spt R C B, S E Fk,C t(U), spt S C A, and
M(R) + M(S) < TM(T); in particular S E Nk+l (U)
As already remarked it is useful to work with homology group defined as equivalence classes of flat chains as no control on the mass of the boundaries is required. We remark however that as far as homology groups are concerned this is quite irrelevant: replacing flat chains by normal currents leads in fact to isomorphic relative homology groups. This follows from the following proposition, compare Sec. 5.1.3,
Proposition 1. If T' E Zk (A, B), then there is T E Zk (A, B) n Nk (U) such
that T-T'EBk(A,B). As in the boundaryless case, a consequence of the isoperimetric inequality is the weak closure of relative homology classes. In fact we have
Theorem 4. Let {Tz} C Zk(A, B) be a sequence which is equibounded in mass and weakly converging to T. Then T L (A \ B) E Fk,,pt(U) and T L (A \ B) belongs to Zk (A, B). Moreover if the Ti E Bk (A, B) are relative boundaries, then T L (A \ B) is a relative boundary. Proof. First notice that by semicontinuity T has finite mass, hence T L (A \ B) is well defined. In order to prove that T L (A \ B) is a flat chain we set g(x) _ dist (x, B), x E U, and for r > 0
H(r) := {x E U I g(x) > r}.
Being spt aT C B = U \ H(0), from the slicing theory we have for a.e. r,
0
= a(T L H(r)) - aT L H(r) = a(T L H(r)) and
3.5 Relative Homology and Cohomology Groups
1/i
1/rj
f
611
M(a(T L H(r))) dr =
0
J0
M() dr < cM(T).
Consequently we can find a sequence of numbers {rj }, M(a(T L H(rj)) < oc, i.e., we have
rj - 0, such that
T L H(rj) E Nk(U). Also
M(T L H(rj) - T L (A \ B)) = lim M(T L (H(rj) - H(0))) 00 jlim 3-00 hence F(T LH(rj)-T L (A\B)) - 0, which proves that T L (A\B) E Fk,,pt(U) As trivially a(T L (A \ B)) C B, the first part of the claim is proved too. If the Ti's are relative boundaries, the isoperimetric inequality yields
Ti = Ri + 8Si
where Ri E Fk,cpt(V), sptRi C B, Si E Nk+1(U), sptSi C A, and
M(Rj) + M(Si) + M(aSi) < cM(Ti). Passing to the limit on a suitable subsequence, we then infer T = R + aS with R E Dk M, spt R C B, S E Nk+l (U), spt S C A; therefore
T L (A \ B) = R - T L B + aS. Being S normal in U with spt S C A, S is a flat chain, hence aS E Fk (U); by the first part of our claim R-T L B E Fk,cpt(U). Finally, since spt (R - T L B) C B, we conclude that T L (A \ B) E Bk(A, B).
We notice explicitly that we cannot expect in general that T belongs to Fk,Cpt(U). As corollary of Theorem 4 we can now state
Theorem 5. We have (i)
In each relative homology class y c Hk(A, B, ]R) there is a homological minimizer, that is a T E Zk(A, B, IR), [T] rel = y such that
M(T) = inf{M(R) I R E Zk (A, B, IR), [R]r.1 = y} (ii) In particular the function
M(y) := inf{M(R) I R E Zk(A, B, ]R), [R]rel = 71 defines a norm on Hk (A, B,1R) .
612
5. Cartesian Currents in Riemannian Manifolds
Proof. First we observe that in every relative homology class there is a current with finite mass, hence A := inf{M(R) I R E Zk(A, B, R), [R]rel = 'Y} < oo.
The claim in (i) then follows at once from the semicontinuity of the mass and from Theorem 5. The claim in (ii) is trivial.
Actually the previous results extend to the quite more general situation in which one does not assume A compact, but with a number of changes and technical difficulties. For instance homology groups are vector spaces but in general of infinite dimension, in fact they are inductive limits of Banach spaces endowed with a flat norm. Also, in dealing with such a general situation one cannot avoid some compact control on the location of the supports of involved currents. In some sense the homology vector space can be exhausted by a sequence of finite dimensional vector spaces. In fact one has that the vector subspace of Hk (A, B, R) consisting of those homology classes which meet Fk,K(R), K fixed compact, has finite dimension. To keep our extension simple we shall not discuss this general approach and we refer the interested readers to Whitney [674], Federer [226] [228].
Relative de Rham cohomology. Let A, B local Lipschitz neighborhood retracts, B C A, B closed relatively to A, and let it : U -> A the retraction. One then defines the de Rham cohomology groups HHR(A, B)
H4R(A)
Zk(U, B) Hk (A, 0)
where
Zk(U,B):={wEZk(U)1w=0 near B} and for w, 77 E Zk(U) w - 77 if there exist an open set V, A C V C U and a form
a E Zk-1(V,B) such that w - 77 = da in V. One has of course the immersion map j# : HHR(A, B) -> HHR(A), the restriction map i# : HaR(A) -* HHR(B) which restrict forms in HHR(A) to their values in B, and a coboundary operator 8 : HHR(B) - HaR 1(A, B) defined as follows. Choose an open set U' such that B CC U' CC U n V and extend w1u, to w E Dk(U). Of course dw = 0 near B, hence dw E Zk+1 (A, B); one defines b([w]B) :_ [dW]rel
noticing that the relative homology class of w depends only on the cohomology of w.
It is easy to prove that in case A is a compact oriented submanifold X and B = 8X, that the above definitions agree with the ones in Sec. 5.2.6, compare Proposition 1 in Sec. 5.3.3. Moreover a rewriting of Theorem 2 in Sec. 5.2.8 yields
3.5 Relative Homology and Cohomology Groups
613
Theorem 6. The long sequence
? HaR(A) i ' Hax(B) a' HaR1(A,B) - HaR'(A) is exact.
de Rham duality in the Lipschitz category. In the attempt to generalize de Rham duality to the Lipschitz category, Whitney and then Federer introduced the notion of flat cochains which play the role of non smooth differential forms. In order to give a simple presentation we assume A compact, and refer to Federer [228] for general results and proofs. Set Fk(A)
Fk,A(U),
we would have to write instead Fk(A) := U{Fk,K(U) I K C A, K compact} for general local Lipschitz neighbourhood retract A.
Definition 1. We define (i) flat cochains as linear continuous functionals on Fk (A)
Fk(A) :_ {a : Fk(A) -+ R I a linear and continuous} (ii) relative cochains, cocycles and coboundaries as
Fk(A, B)
Zk(A,B)
{a E Fk(A) {a E Fk(A)
Bk(A,B){aEFk(A)
a = 0 on Fk(B)} a = 0 on Bk(A,B)} a = 0 onZk(A,B)}
(iii) cohomology groups as
Hk(A, B) = Zk(A, B)/Bk(A, B). We recall, see Sec. 5.1.3, that Fk (A) endowed with the flat norm is a Banach space. One sees immediately that Zk (A, B) is closed in Fk (A), and, using the isoperimetric inequality, that also Bk (A, B) is closed in Fk (A). It follows that inf{Fk (S) I S E 'y},
'y E Hk (A, B, ]R)
defines a norm on Hk (A, B, I[8) with respect to which the coset map p : Zk (A, B, IEP)
-' Hk (A, B, R)
is continuous, being Hk (A, B, R) finite dimensional. It is worth to notice that instead relative boundaries with equibounded masses are in general not compact
5. Cartesian Currents in Riemannian Manifolds
614
with respect to the flat norm. For instance take A :_ R2 and B :_ R2 fl {x I x2 = 0}. Then the sequence of flat relative boundaries
T i :=iQ(Pl,P2)Jj+iQ(P2,P3)1 +iQ(P3,P4)I
i = 1,2,...,
where
P1 := (0, 0), P2 := (0, i-2), P3 := (i-1, i-2), and P4 :_ (i-1, 0),
converge to the current boner which is not flat. Clearly every relative cocycle a E Zk(A, B) induces by restriction a linear map a : Hk (A, B, R) -* R. Consequently there is a natural morphism Hk (A, B) -* Horn (Hk (A, B, Ilk), R)
which is obviously injective. On the other hand if 2 E Hom (Hk (A, B, R), R), 8 o p defines a linear continuous map from Zk (A, B) into R which extends by Hahn-Banach theorem to a linear continuous map still denoted 2
t: Fk(A) -* R. Therefore we conclude
Proposition 2. Hk (A, B) f-- Horn (Hk (A, B, L), R) Notice that every smooth form w defines by iv(T) := T(w)
a flat cochain iv E Fk(A). The main question becomes now that of realizing Hk(A, B). A key point in this respect is the following
Theorem 7. We have
Bk(A, B) _ {d1 10 E F-1(A, B) On the other hand, a regularization procedure, compare Federer [228], then allows to prove
Proposition 3. We have Hk(A, B) = HdkR(A, B)
Consequently Theorem 7 yields the following extension of the de Rham theorem
Theorem 8. We have
4 Integral Homology (i)
615
For every v E Hom (Hk (A, B, IR), R) there exists a closed differential form w E Ek (U) which vanish in a neighbourhood of B such that v([T])
(ii)
There exists neighbourhood U' C U and V' C V such that, if w E Ek(U), dw = 0, w = 0 in a neighbourhood of B, and d T E Zk (A, B)
T (w) = 0
then w is exact in U', i.e. w = dcp,
WE Ek-1(U') and spt cp C U'\ V'.
In particular the duality between forms and currents yields to a non degenerate bilinear form
<,>:Hk(A,B,lR)xHHR(A,B) -> R. Finally the diagram
...
*-
HHR(B)
{-
HHR(A)
x<, >
x<, > Hk (B
F-
Hk (A, IR)
Hk(A, B)
4 .. .
x<, >
4
Hk (A B R)
is commutative.
4 Integral Homology In this section we discuss the integral homology of a compact, oriented manifold in terms of currents.
4.1 Integral Homology Groups In this subsection we discuss the integral homology of a compact, oriented manifold in terms of currents, in particular we shall prove that homology classes are weakly closed. Actually we shall see that the cosets of homology classes are in fact connected components of integral cycles.
Integral absolute homology groups. Let A be a compact neighborhood retract of IEBn, let -7r
: U --- A be the retraction. We can and do assume that U= {x E ][8'Z
I
dist (x, A) < eA}
for some CA > 0. By decomposing IR' into cubes of size eA/2\, and denoting by G the finite family of those cubes which cover A, we can see that
5. Cartesian Currents in Riemannian Manifolds
616
U{Q I QE.C} cc U. From now on we denote by LA the CW cubical complex over cubes in L and by Lk,A the k-skeleton of LA, k = 0, ... , n. Similarly to Sec. 5.3.2 an integral chain P on Lk,A is simply a polyhedral chain with integral coefficients
/9 EZ, FiELk,A. Setting Zk,simpl(LA, Z) Bk,simpl (LA, Z)
{P P integral polyhedral chain on Lk,A, aP = 0} {P = aQ I Q integral polyhedral chain on Lk,A}
we define the simplicial integral k-homology of LA by Hk,simpl(LA, Z) := Zk,simpl (LA, Z)/Bk,simpl (LA, Z)
which is of course a finitely generated Z-module. We then set for k = 0, ...
,n
Zk(A,Z) :_ {T E Zk(U), 8T = 0, sptT c Al Bk (A, 7L) :_ {aS, S E Zk+1(U), spt S C Al
Then the integral k-homology group of A is defined by
Hk(A,Z) := Zk(A,Z)/Bk(A,Z). Proceeding as for Theorem 6 in Sec. 5.3.2, taking into account that the deformation theorem produces i.m. rectifiable decomposition of i.m. rectifiable currents, we get
Theorem 1. The module Hk(A,Z) is isomorphic to the k-homology of Lipschitz chains in A with integral coefficients. Moreover the retraction map 7 induces an epimorphism 7r,, : Hk,simpl (LA, Z) --y Hk (A, Z). In particular Hk (A, Z) is finitely
generated.
Of course, if A is a C' manifold then Hk(A, Z) and Hk,simpl(LA, Z) are isomorphic, compare Theorem 5 in Sec. 5.3.2. Again a relevant advantage in representing the singular homology in terms
of homology groups of currents is that the cosets in Zk (A, Z) turn out to be closed with respect to the weak convergence of currents. Actually in the case of integral homology we are dealing now, we shall prove that in fact the cosets are connected components of Zk (A, Z). The key step in doing that is an isoperimetric inequality which is more stringent than in the real case. 1 The following example shows that one cannot expect isoperimetric inequality for general currents on manifolds, a size condition is necessary.
4.1 Integral Homology Groups
617
Let S2 = {x E R31 IxI = 1} be the two sphere and let E > 0 be small enough. Let
M =Stn {x E JR3 I x3 >-I+ E} and C = S2 n {x E R3 I x3 = -1 + 2e}. Clearly C is a 1-cycle in M and bounds the surface S := S2 n {x E ][83 I x > -1 + 2e} in M, C = aS, but M(S) 47r,
while M(C) ^-, O(/). If one consider
M=S2nxER3Ix<1-E { } then the cycle C does not bound any surface lying in M.
Theorem 2. For k = 0, ... , n there is a constant e = E(k, n, EA) such that if T E Zk(A,Z) with M(T) < e then T E Bk(A,Z); more precisely T = aR, R E Zk+l(U), sptR C A and M(R) < cM(T) 1+1/k where c = c(k, n).
Proof. Using the deformation theorem one can deform T over Lk,A and write T = P + 8R' where R' E Zk+1(U), spt R' cc U, P = > /i Q Fi , Fi E Lk,A, Qi E 7 G and M(P) < cM(T), M(R') < co-AM(T). As /3 i > 1, M(P) a Therefore choosing e < oA/c, we infer that if M(T) < e then P = 0. Consequently, setting A.
R = 7r# (R'), we infer
T=7r#T=aR, REZk+l(U), sptRcA and
M(R) < c1M(R') <
cl(cM(T))l+11k
The following theorem yields the key decomposition argument, which is relevant also in the context of flat chains, compare Theorem 2 in Sec. 5.1.3.
Proposition 1. Under the assumption of Theorem 2 in Sec. 5.1.3, if moreover all Ti and T are i.m. rectifiable, we can find integral currents Ri E Zk+1(A) and Si E Zk (A, Z) such that
Ti-T=Si+MRi.
M(Ri)+M(Si) -> 0,
Proof. We proceed as in the proof Theorem 2 in Sec. 5.1.3 observing that all currents Pi, Q. , Fi, Gi which are involved in the proof of Theorem 2 in Sec. 5.1.3 this time are i.m. rectifiable and the convergences k
EFj j=1
are in the mass norm.
00
j=1
k
00
E Gj
EGi
j=1
j=1
618
5. Cartesian Currents in Riemannian Manifolds
From Theorem 2 and Proposition 1 we readily infer
Theorem 3. Let Ti, T E Zk (A, Z) be integral k-cycles, Ti -+ T. Then Ti - T = 8R1i Ri E Zk+1(A) for large i. Moreover M(Ri) -- 0. Proof. In fact using Proposition 1 we can write
Ti-T=Si+8R2 Si E Zk(A), R' E Zk+1(A) and M(R'i) +M(Si) -+ 0. The isoperimetric inequality of Theorem 2 implies that Si = OZi, Ei E Zk+l(A), M(Ei) < cM(SS)l+1/k for large i. Thus
Ti-T=a(R'+E'i)
and the claim follows. The following corollary collects some important consequences of Theorem 3.
Corollary 1. We have (i) Bk (A, 7Z) is weakly closed. (ii) Zk (A, Z) \ Bk (A, Z) is weakly closed. (iii) In each integral homology class y there is a mass minimizing i.m. rectifiable cycle, that is, there is T E Zk (A, Z), [T] = y, such that
M(T) := inf{M(S) (iv)
I
[S] = y}.
Given c > 0 there are at most a finite number of integral homology classes y such that inf{M(S) [S] = y} < c. I
(v) For a non trivial integral homology class
inf{M(S)
I
[S] = y} > 0.
Proof. (i) and (ii) follows trivially from Theorem 3. (iii) trivially follows from (i). To prove (iv) assume on the contrary that one has a sequence {Ti} of cycles of equibounded masses. Passing to a subsequence Ti - T and by Theorem 3, all
the Ti but a finite number belong to the same homology class, a contradiction. (v) is just the isoperimetric inequality, Theorem 2.
Relative integral homology. A parallel theory for integral relative homology can be developed. Let A, B be two local Lipschitz neighbourhood retracts in Rn, B C A, B relatively closed in A. F o r k = 0, ... , n, we define the integral relative cycles by
Zk (A, B, Z) :_ IT E
. k (IR')
I spt T C A, spt aT C B, or k = 0},
the integral relative boundaries by
Bk (A, B, Z) := IT + aS I T E .Fk (Rn), spt T C B, S E Fk+l (Rm), spt S C Al,
4.1 Integral Homology Groups
619
and the integral relative homology by
Hk(A,B,Z) = Zk(A,B,Z)/Bk(A,B,Z) One shows that the Z-modules Hk (A, B, Z) satisfy the Eilenberg-Steenrod axioms for an homology theory with integral coefficients in the Lipschitz category, compare Federer [226, 4.4.1], and, being A compact that we have as consequence of the deformation theorem Theorem 4. Hk (A, B, Z) is isomorphic to the k-homology of integral Lipschitz chains in (A, B). Moreover one has an epimorphism of the k-simplicial homology Hk (LA, LB, Z) of (LA, LB) onto Hk (A, B, Z). In particular Hk (A, B, 7G) is finitely generated. From the deformation theorem we also infer the following isoperimetric inequality, compare Federer [226, 4.4.2]
Theorem 5. Let T E Rk(Rn) fl Zk(A, B) fl Bk(U,V). There exists an integral current S such that spt S C A, spt (T - 8S) C B, and M(S)k/(k+l) + M(T - 8S) < cM(T), c being an absolute constant.
Proceeding as in the real case and taking into account Federer-Fleming closure theorem one infers Theorem 6. Suppose that Ti E Rk (U) fl Zk (A, B, Z) are relative cycles converging weakly to T. Then T belongs to Rk (A \ B) and T L (A \ B) E Zk (A, B, Z). If moreover the Ti's are relative boundaries, then T is also a relative boundary, T E Bk(A,B,7G) and consequently, compare Federer [226, 4.4.2]
Theorem 7. We have (i) In each relative homology class -y c Hk(A, B, Z) there is a homological mass minimizer, i. e. a T E Zk (A, B, Z) with [T]rel,z _ 'y and
M(T) = inf{M(R) I R E Zk (A, B, Z), [R]rel = 7} (ii) There are finitely many integral relative homology classes which meet the set
{T E Zk(A,B,Z), M(T) < c}. Real and Integral homology. Let us discuss briefly relationships between integral and real homology in terms of cycles. Let A, B be compact neighbourhood retracts of R'. According to Theorem 1 the integral homology groups are Abelian groups of finite type. The general presentation theorem for such groups says that Hk (A, B, Z) splits as a direct sum of groups with a single generator; more precisely, one can choose exactly q elements
5. Cartesian Currents in Riemannian Manifolds
620
yl...... q E Hk (A, B, Z), where q is the rank of Hk (A, B, Z), independent over Z such that q
Hk(A,B,Z) _
(1)
Z'Y, ®Tk,
where Tk is the torsion subgroup of Hk(A, B, Z), i.e.,
Tk:={7EHk(A,B,Z) 1 2pEZ, py=0}. In other words each y E Hk (A, B, Z) decomposes uniquely as
y = 1: ni'yi + w,
ni E Z, w E Tk.
i
It is also a very well known fact of homological algebra that the integral homology groups fix the homology with any group coefficient. In fact the universal
coefficient theorem, compare e.g. Spanier [607, p. 222 and p.243] express the homology with coefficients in G as
Hk(A,B;G) =Hk(A,B,Z) ®G®Tor(Hk_l(A,B,7G),G),
(2)
where the Tor groups are the "torsion" subgroups. In case G = IR (2) reduces to Hk (A, B, R) = Hk (A, B, Z) ®R,
(3)
which in terms of the generators of Hk (A, B, Z) means
Theorem 8. Let q = rank Hk (A, B, Z) and let S1,.. . , Sq be q integral relative cycles in Zk (A, B, Z) such that [Si]rel,z,
,
[Sq]rel,z
are independent over Z and q
Hk (A, B, Z)
Z[Si]rel,z + Tk (A, B).
Then [Sl]re1,II2,
,
[Sq]rel,R
yield a basis for Hk (A, B, R). In particular dim Hk (A, B, R) = rank Hk (A, B, Z)
A simple consequence is
Theorem 9. Let dim Hk (A, B, lid) = q. Then we can always choose q integral cycles Ti in Zk (A, B, Z) such that
4.1 Integral Homology Groups (i)
621
[Tl]a,...,[Tq]z are independent over Z and generate the free part of the
homology group Hk (A, B, Z). (ii) [T1]lk,... , [Tq]R forma basis for Hk (A, B, ]R).
For the sake of completeness, we give a direct proof of Theorem 8 in terms of currents without any mention of the algebraic theory in the case of the absolute homology. Modulo additional technicalities, the general case of relative homology can be treated in the same way, compare Federer [228, 5.7]. From now on we assume B = 0. It is enough to prove Theorem 8 for the simplicial homology Hk,simpl(LA,Z). For that we need the following lemma
Lemma 1. Let N > 1 and E C QN be a subset of QN, and let r E IRNbe such that (r I a) = 0 d a E E. Then for any e > 0 one can find q E cQN such that
Iq - rI < Eand(gI o,)=0Vo'EE. Proof. We denote by S1, ... , Sk a basis for the Q span of r1, k
. . .
, rN and we write
1 Q, i = I,-, N. sj E IIB, aij E
ri = E aij sj, j=1
Then EN1 riai = 0 is equivalent to EN aij o i = 0 j = 1, ... , k. Choosing now T = (Ti, ... , Tk) in such a way that N
k
EIax7I ITj -Sjj <E i=1 j=1
Ej=1 aijTj has the required properties.
we see that q E QN given by qj w`
0
Notice that ee can choose qj = ri if ri happen to be rational. Next proposition contains one of the key points in the proof of Theorem 8.
Proposition 2. Let T E Zk (LA, Z) fl Bk (LA, ]R), i. e. T is an integral polyhedral cycle, which is a boundary in the real sense, T = 8R, R E Pk+l (LA, R). Then, for each E > 0, one can find a polyhedral chain R' with rational coefficients such
that T = 8R' and M(R - R') < E.
Proof. Denote by {Fi}, i = 1, . . . , I, and {Gj }, j = 1, ... , J respectively the k-faces and the (k + 1) faces of LA with some orientation. We must have I
BQGj j =
E Z. i=1
Moreover, if we write
T=>n4QFjI,
nje7G,
R=I:rjQGjl, rjER
622
5. Cartesian Currents in Riemannian Manifolds
T = aR means ni = Ejj=1 rj Bji, d i = 1, ... , I. Consequently by Lemma 1, given e > 0, one finds sj E (0 such that
ni = E sj8ji. j-1 Therefore, if R' := Ej sjIGj , we have 8T = R' and
N(R - R') < E N(Q Gj Irj - sj I < ce-
j
Remark 1. We observe that T is a polyhedral chain with rational coefficients if for some p E Z, pT is an integral polyhedral chain. Accordingly we could restate Proposition 2.
Denote by ik : Hk (LA, Z) --+ Hk (LA, R) the group homomorphism which associate to each integer homology class [T]z the corresponding real homology class j*([T]a) = [T]R.
Corollary 2. The kernel of j*, ker j*, is the torsion subgroup of Hk(LA,Z), i.e., for T E Zk(LA,Z) [T]R = 0
if
[T]z is a torsion class.
Proof. If [T]a belong to the torsion subgroup of Hk (LA, Z) then for some p p[T]z = 0. Consequently
0,
[T]1 = pp[T]R = p[pT]n = pj*([pT]z) = 0.
Conversely assume that [T]R = 0, that is T E Zk(A,Z) and T = aR, R E Pk+1(LA, R) By Proposition 2 we find R' E Pk+1(LA, Z) and an integer p # 0 such that
pT = aR' therefore [pT]z = p[T]z = 0.
Proof of Theorem 8. We split the proof into three steps. Step I. [T1]R, ... , [TT]R are independent over R.
Assume 9=1 ri [Ti]R = 0. Since R is a vector space over Q, denoting by $1, ... , sk a basis for the Q-span of r1,... , rq, we write ri = Ekj=1 ai7 E Q, sj E R. Consequently q
aij[Ti]IIP=0 i=1
b j=1,...,k.
4.1 Integral Homology Groups
623
Denote by p the 1.c.m. of the denominators of the aid. The current T k
Ek
qaijTi is an integral cycle and k
aii [Ti]R = 0.
[T ]R = q i=1
By Corollary 1 [T] z belongs to the torsion subgroup of Hk (LA, Z) hence
0 = [T]z = j:(gaij)[TT]z. Consequently aid = 0, V i, j and therefore ri = 0. Step 2. Real homology class of rational polyhedral cycles are linear combinations o f [T1]R, ... , [Tq]R with rational coefficients.
In fact we can write any polyhedral cycle T with rational coefficients as
T= 1 S
p where p E Z and S E Zk (LA, Z) . Since [S]z =
ni [Ti]z + w, ni E Z, w E Tk we
get q
[SIR =
57, ni[Tilpp. i=1
Consequently q
[S]R =
n2
[Ti]R-
i=1 p Step 3. Real homology classes of rational polyhedral cycles are dense in Hk (A, R). In fact for T E Zk (A, l) we find, using Proposition 2, a sequence of rational polyhedral cycles Tj such that M(Tj - T) --> 0. As
II[T]R-[Tj ]jII: M(T-Tj) --+ 0 the proof is concluded.
Combining Theorem 8 and de Rham's theorems, compare Sec. 5.3.2 and Sec. 5.3.3, we also get
Theorem 10. Let X be a n dimensional oriented compact manifold. Assume one has q cycles TI, . . . , Tq E Zk (X, Z) such that the corresponding homology classes are independent over Z and generate the free part of Hk (X, Z). Then one can find q closed k-forms w1, ... , wq E Zk (X) such that
Ti(wj) = -5 .
Analogously, given q' relative cycles Ti, ... , T9, E Zk (X, 8X, Z) such that the corresponding relative homology classes are independent over Z and generate the freeqq,part of Hk (X, 8X, Z), then one can find q' relatively closed k - forms 771, ... , E Zk(X, 8X) such that Ti,
51
5. Cartesian Currents in Riemannian Manifolds
624
Proof. By Theorem 8, [Ti]3t generate the real homology Hk(X,IR) of X. By de Rham's duality HaR(X, Ill) is dual to Hk (X, R) therefore choosing dual cohomology class [w1],..., [Wq]
of [Tl]R, ... , [TT]a yields the right forms. The second part of the claim is identical.
In terms of the forms w1, ... , wq, 771, ... , q ' of Theorem 10 we can state
Corollary 3. We have (i) Let T E Zk(X,IB). Then T is (real) homologous to an integer cycle if
T(wi)EZdj=1,...,q.
(ii) Let T E Zk (X, Z). Then T (wi) = 0 V j = 1, ... , q if and only if [T]z belongs
to the torsion subgroup of Hk (X, Z). (iii) Let T E Zk (X, 3X, Ill). Then T is (real) homologous to an integer cycle if
T(wi)EZbj=l,...,q.
(iv)
Let T E Zk (X, aX, Z). Then T (rp) = 0 V j = 1, . belongs to the torsion subgroup of Hk (X, aX, Z).
.
.
, q if and only if [T]Z,rel
4.2 Intersection in Integral Homology Let X be a n-dimensional oriented compact submanifold of I[Bn+N The intersection maps in real homology that we have defined in Sec. 5.3.4 extends to integral homology. The idea is to define [S] n [T] by taking the homology class of the intersection of S with almost any translation Ta#T of T. We first do that in the case of boundaryless manifolds.
Intersection of cycles on boundaryless manifolds. Assume that aX = 0 and for convenience set for T E Dk (X) and x, a E Rn+Iv T. := Ta,#7r*T.
Proposition 1. Let S E Zk(X, R) and T E Zh(X, IR), h + k > n be two normal cycles. Then we have (i) For a. e. a close to zero S n Ta exists in Zh+k-n (X) and [S n Ta]IlZ = [S]l[t nx [T]R-
(ii) If moreover S and T are integral, then for a. e. a close to zero S n Ta is an integral cycle, SnTa E Zh+k_n(X,Z) and its integral homology class depends only on the integral homology classes [S] E Hk (X, Z) and [T] E Hh(X,ZG). Proof.
(i) Let PT E Z,,,-h(X) be a Poincare dual form of T so that T =
fx APT + 3E, E E Nh+l(X), or, setting w :=
T=QXLw+aE.
(-1)h(n-h)PT
4.2 Intersection in Integral Homology
625
Taking into account (iii) of Proposition 3 in Sec. 5.3.2 7r*(QX I LwT) =I U I L7r#wT
and therefore
7r*T= QUj L7r#wT+a7r*E By slightly translating we then infer Ta, = Q U L 7 a7r#wT +
aEa
in U.
near X.
By Proposition 2 in Sec. 5.3.4 we then infer that S n Ea, S n BEa,, SnTa, exists as normal currents on X for a.e. a close to zero. Consequently being aS n Ea, = 0 we infer
a(SnEa)=SnBE,
and (1)
SnTa = S n I U J1 L 7- a7r#wT + a(S n Ea)
near X.
for a.e. a close to zero. Since on the other hand 7- a7r#wT is cohomologous to
7r#WT with compact support in X \ BX, we obtain from (1) that s n Ta is homologous to
SnQU]IL7r#WT=SL7r#wT=SLWT. (i) now follows if we compute
SLWT =
(_1)('n-k)(n-h) S(-APT)
(-1)(n-k)(n-l,)
(ii) We already known, compare (iii) of Proposition 2 in Sec. 5.3.4, that S n Ta
is an integral current for a.e. a close to zero. We now prove that the integral homology class of s n Ta is independent of a for a.e. a close to zero. In fact, being To, and T homotopic,
Ta = To + BE
near X,
E E -Th+N+1(U). Therefore for JbI close to zero, Ta+b = Tb + BE'
near X,
E' E Zh+N+1(U) Proposition 2 in Sec. 5.3.4 then yields that for a.e. b the currents SnTa+b, S n Tb, S n Eb exist as integral currents in X, and B(S n Eb) _ S n BEb. Consequently
SnTa+b-SnTb=a(SnE') inX for each a and a.e. b close to zero. In particular the integral homology of s n Ta is independent of a for a.e. a.
On the other hand, if T = aR, R E Zh+1(X) or S = aQ, QE Zk+1(X), Proposition 2 in Sec. 5.3.4 again yields that for a.e. a close to zero S n (BRa) = B(S n Ra),
being aTa = BS = 0.
(8Q) n Ta = (-1)k+1 8(Q n Ta)
5. Cartesian Currents in Riemannian Manifolds
626
The claim (ii) of Proposition 1 clearly allows us to define the intersection map in integral homology
nx : Hk(X,7L) X Hh(X,Z) -4 Hh+k-n(X,Z) setting
[S]z nx [T]z := [S n Ta]z
for a.e. a,
while (i) of Proposition 1 show that such a map is compatible with the real version of the intersection map defined in Sec. 5.3.4.
Intersection of cycles on manifolds with boundary. More tricky is the
definition of the intersection of cycles for manifolds with boundary. Let us start with the intersection of a relative with an absolute homology class. Let X be a compact oriented manifold of dimension n in ]Rn+N possibly with boundary. Let X1, X2 be two smooth domains in X and let Al C X1 and A2 C OX2 be compact. Since we deal with the relative homology, we suppose even if this is not important for the sequel that Al and A2 are compact Lipschitz neighborhood retracts. Proposition 2. Let S E Zk(XI,A1), T E Zh(X2,A2), h+k > n be two relative normal cycles, and assume that (2)
XlnA2=X2nA1=0,
sptTnX1n8X2=0.
Then for a.e. a close to zero (i) S n Ta exists as an absolute normal cycle, S n To, E Zh+k-n (X1).
(ii) T has a Poincare dual PT E Zn-h(X2) such that PT = 0 on 8X2 n X1 and S n Ta = (-1)(n-k)(n-k) S(. APT) + 8E,
on X1,
Z E Nh+k-n+1(X1) (iii) If S and T are i.m. rectifiable, then S n Ta E Zh+k_n(X1, Z) and its integral homology class depends only on the homology classes [S] z,rel E Hk (Xl, A1i Z) and [T]z,rei E Hh (X2, A2, Z)
Proof. (i) follows from Proposition 2 in Sec. 5.3.4 taking into account that for
a close to zero spt s n spt 8Ta is still empty being X1 n A2 = 0.
(ii) Since spt T n X1 n 8X2 = 0 by (2) then clearly one constructs a Poincare (-1)h(n-h)PT, and for dual form PT with spt PT = 0 on 8X2 n X1. Setting w = 8 := dist (X1, A2),
Xd := {x I dist (x, X1) < 6121,
w extends as a smooth form on X8, w = 0 on X1 \ X2,
T=QX111Lw+R+aZ onX where R E Nh(X2), spt R n Xs = 0, Z E Nh+1(X2). In particular
4.2 Intersection in Integral Homology
627
T=QX11Lw+6E onX5. Arguing now as in the proof of (ii) of Proposition 1 we then conclude that for a.e. a close to zero (3)
SnTa,=SnI7r-1(X1)]Lwa+SnEa,=SLWa+SnEa
(4)
S n aE = a(S n Ea),
where we have denoted by wa the form wa On the other hand wa - w = d,13, 0 E D'-'-1(X2), spt 0 C X2 \ X1. We then have as L Q = 0 and S L (wa - w) = (-1)ka(S L (3) on X1
(5)
Putting together (3) (4) (5) we conclude
onX2 as required. (iii) If now S and T are i.m. rectifiable, then SnTa is an i.m. rectifiable current for a close to zero by Proposition 2 in Sec. 5.3.4. Being Ta and To clearly homotopic we have
Ta:=To+R+aE where E E Nh+N+1(7C-1(X2)), R E Nh+N(7r-1(X2)), sptRn7r-1(X5) = 0, X5 as in step (ii). Consequently for b close to zero Ta+b = Tb+Tb#R+Tb#E near X2, -rb#E E Nh+N+1(7r-1(X2)), Tb#R E Nh+N(7r-1(X2)) and spt75#R n Xl = 0. On the other hand S n Tb# E, S n aTb# E , as n Tb# E, S n Ta+b, S n Tb exist for a.e. b as integral currents,
SnTa+b=SnTb+SnaTb#E onX1
(6)
and
SnaTb#E = a(S n Tb#E) + as n rb#E = a(S n Tb#E)
(7)
being spt rb#E n spt as = 0 for sufficiently small a and b. Putting together (6) and (7) we then conclude that for a.e. a and b S nTa and S n Tb are homologous. To conclude the proof it suffices to note that that for a.e. a close to zero clearly (8)
aR n T a = (-1)k+la(R n Ta),
S n 5Ra = a(S n Ra).
The claim (iii) of Proposition 2 shows that by setting [S]re, n [T] := [S n Ta]
for a.e. a close to zero, [S]rd E Hk(X, aX, Z), and [T] E Hk(X, Z) defines actually the intersection map
628
5. Cartesian Currents in Riemannian Manifolds
nx : Hk(X,8X,Z) x Hh(X,Z) - Hh+k-n(X,Z) between the relative and absolute homology classes. Also (ii) of Proposition 2 shows that such a definition is compatible with the intersection map in real homology, compare Sec. 5.3.4.
In order to deal with the intersection of two relative cycles, we need the following variant of Proposition 2. As before we suppose that X1, X2 are two smooth domains in a compact oriented manifold X of dimension n that we think to be a submanifold in Rn+N. Let Al C Xl and A2 C 8X2 be compact Lipschitz neighborhood retracts. Proposition 3. Let S E Zk(X1, Al), T E Zh (X2, A2), h+k > n, be two relative normal cycles, and assume that (9)
X1nA2=0,
sptTnXlnaX2=0.
Then for a.e. a close to zero (i) S n To, exists as a normal current and S n Ta, E Zh+k-n (XI, Al)
(ii) T has a Poincare dual PT E Zn-h(X2) such that PT = 0 on aX2 n Xl and S n Ta, =
(-1)(n-k)(n-k) S(, APT)
+ R + 8E,
in Xl,
R E Nh+k_n(X1), sptR C Al, E E Nh+k-n+1(X1). (iii) If S and T are i.m. rectifiable, then S n Ta0 E Zh+k_n (X1, Al, 7G) and its
integral homology class in Hh+k_n(Xl, Al, Z) depends only on the homology classes [S]z,rei E Hk(Xl,A1iZ) and [T]z,Lel E Hh(X2iA2,Z).
Proof. The proof is quite similar to the proof of Proposition 2. Because of the different hypotheses, we may substitute (4), (5) and (7) respectively by (10)
SnEa=a(SnEa)+(-1)kaSnEa-aE'+R'
R' E Nh+k_n(X1), spt R' C spt S C Al, E' E Nh+k_n+1(Xl), by (11)
S L (Wa - W) = aS L + (-1)ka(S L,8) = R" + aE"
R" E Nh+k-,,,(X1), sptR" C sptS C Al, E" E Nh+k-n+l (XI), and by (12)
SnaTb#E=a(SnTb#E)+aSnTb#E=R'"+aE"'
R"' E Nh+k-n(X1), sptR"' C sptS C Al, E"' E Nh+k-n+1(X1). In order to define the intersection of two relative cycles of X, we recall the notation of Sec. 5.1.2, particularly the extension map
e:X-*Xao,
eo>0.
which is homotopic to the identity on X. Let [S] E Hk (X, 8X, Z), [T] E Hh (X, 8X, Z) be two relative cycles. Then e#T E Zh (XE0 , aXE0,, Z), and x n
4.2 Intersection in Integral Homology
629
aXEO = 0. We then can apply (iii) of Proposition 3 to S and T with Xl := X, Al := aX, X2 := XEO, A2 := OXEO and infer that the homology class of s n (e#T)a,
in Hh+k_,,, (X, aX, 7L) depends only on the relative homology class of S and T for a.e. a close to zero. This then defines the intersection map between relative cycles, nx : Hk (X, aX, 7L) X Hh, (X, aX, 7L) --b Hh+k-n (X, aX, 7L) by
[S]re, nx [T]r.1
[S n (e#T)a]rel
Moreover (ii) of Proposition 3 shows that this definition is compatible with the intersection of relative cycles in real homology that we have defined in Sec. 5.3.4.
Intersection index in integral homology. Let
be two sys-
tems of integral cycles which are independent and generate respectively the free part of Hk (X, OX, 7L) and of Hn_k (X, 7L). Then the intersection index Iij := ix (Si, Tj) takes values in 7L, compare Sec. 5.3.4, and non degeneracy is equivalent to the non degeneracy of the real bilinear form
riSi,Esj Tj) -> ix (S, T) = E ri sj lij. i,j
Thus, being non degeneracy of Poincare duality and of the intersection index equivalent, we have
Proposition 4. Non degeneracy of Poincare duality is equivalent to
dimHk(X,aX,1R) =dimHn_k(X,R) =: q,
(13)
detIii
0.
We can also state
Proposition 5. Non degeneracy of Poincare-Lefschetz duality is equivalent to one of the following equivalent statements: (i) rank Hk(X,OX,Z) =rankHn_k(X,Z) =: q, detIi.7 o 0. (ii) given S E Zk (X, aX, 7L) ix (S, T) = 0 V T E Zn_k (X, 7L) if [S] z,rei is a torsion class. (iii) given T E Zn-k (X, 7L) ix (S, T) = 0 `d S E Zk (X, aX, 7L) iff [T]z is a torsion class.
In fact in case of (ii) non degeneracy of the intersection index on integral cycles amounts to the non degeneracy of the map
Zp X V -> 7L,
(n, m) ---
nimjIi1,
630
5. Cartesian Currents in Riemannian Manifolds
that is again to p = q and det IZj 54 0. Therefore we have [S]R,rel = 0, consequently [S]z,rel is a torsion class. The same argument applies for the third claim (iii).
We emphasize the fact that a good basis for Hk (A, B, R) can be obtained only starting from free generators of Hk (A, B, Z). In fact considering a generic basis of H k (A, B, IR), [T1]x, ... , [Tq]R of integral cycles we cannot be sure in principle that [Ti]z,..., [Tq]z generate the free part of Hk(A, B, Z).
An algebraic view of integral homology. Although integral currents are useful in finding representatives of integral homology classes by means of variational integrals, their use in proving the integral version of the Poincard duality isomorphism between homology and integral cohomology is not completely clear. In describing the integral homology of a given manifold X, one usually starts with a triangulation of a manifold and then define the integral chains of k - cells of that subdivision, a boundary operator, hence the homology groups
Hk(X,Z),
Hk(X,aX,7G).
Then, by duality one defines cochains as morphisms on chains of k - cells, a notion of coboundary, and the integral cohomology Hk (X, Z),
Hk (X, aX, 7G)
of the integral cochains. One then proves
Theorem 1. H" (X, Z) = Homz(Hk (X, Z), Z) ®Ext (Hk_ 1(X, Z), Z) Hk (X, aX, Z) = Homz (Hk (X, OX, Z), Z) ® Ext (Hk_ 1(X, OX, Z), Z) In other words Homz (Hk (X, 7G), 7G) and Homz (Hk (X, aX, 7G), 7G) are the free
parts respectively of Hk(X, Z) and Hk(X, aX, Z). If moreover Hk_1(X, Z) (resp. Hk_1(X, aX, Z)) is free, then Hk (X, Z) (resp. Hk (X, OX, Z)) is free, too. Next point consists in defining a pair of morphisms L : Hk(X, (9X, Z) x L : Hk (X, aX, 7G) X
Hn_h
(X, OX, Z) -+ Hh+k_n (X, Z)
Hn-h
(X, Z) -' Hh+k-n (X, aX, Z)
known as the cap product which mimic the role of the interior product between forms and currents in real homology. Applying such maps with the fundamental cycle Q X E Hn (X, OX) as first factor, one constructs the two PoincareLefschetz morphisms PO :
Hn-h(X, aX, Z)
Hh(X, Z) Pb : Hn-h(X, Z) -i Hh(X, aX, Z)
The celebrated theorems of Poincar6 and Poincard-Lefschetz then state
5 Maps Between Manifolds
631
Theorem 2. PO and Pb are isomorphisms of Z-modules. Finally, combining Theorem 1 and Theorem 2 we get Homz(Hk(X, Z), Z) = Hk,free(X, Z) = Ilnfrek(X ax Z) Homz (Hk (X, aX, Z), Z)
= Hk,free (X, aX, Z) = Hnrek (X, Z)
In other words defining the intersection index on X
ix : H,,-k(X, 3X, Z) x Hk(x, Z) -> Z by (14)
ix(S,T) :_ [S]re1 L (P)-1[T],
we conclude that
Theorem 3. ix is unimodular. It turns out that the intersection index that we have defined in terms of currents agrees with the one defined in (14). From the current theory we already known that the intersection matrix {III} for two given systems of generators {S} of Hk ee (X, OX, Z) and } of Hnrek (X, , Z) have integral non zero determinant, 0 # det I2j E Z. The algebraic integral homology theory adds to this information that det I2j = ±1. In particular for k = 0, ... , n one can choose of the free part of Hk (X, aX, Z) and a basis {T3c -k) } of the free a basis {T('L-k)
part of Hn_k(X,Z) such that ix(,(kT(n-k)) =
a?.9.
5 Maps Between Manifolds In this final section we discuss Sobolev and Cartesian maps and currents between Riemannian manifolds focusing on some topological properties of those maps. In Sec. 5.5.1 we deal with Sobolev maps, reviewing some of Bethuel's results
on density of smooth maps and of White's results on the d-homotopy type of Sobolev maps.
After introducing Cartesian maps and currents between Riemannian manifolds and stating a few simple extensions of results we have already proved in the previous chapters for Cartesian currents in Euclidean spaces, we illustrate in Sec. 5.5.3 and Sec. 5.5.4 the homological content of Cartesian maps and currents. In particular we shall prove that, similarly to the case of smooth maps, every Cartesian current induces sets of maps in cohomology and homology.
632
5. Cartesian Currents in Riemannian Manifolds
5.1 Sobolev Classes of Maps Between Riemannian Manifolds Let 0 be a bounded domain in R'. If aQ is regular, say Lipschitz-continuous, and ,f2 » d2, then every map u E W',P(Q, RN) can be extended to a map u : !2 -4 RN in such a way that ii E W1,P(d2, RN), ii = u in d2, and even u E Wo'P (dl, RN), with spt u CC S?. A sequence of mollifiers of u is then easily seen to converge in the norm of W1,P to u. Therefore we deduce
Theorem 1. Suppose that a12 is Lipschitz-continuous. Then every map u E W1,P(dl,RN) is the W',P-limit of a sequence ui E C' (f2, RN) Actually, the following Meyers-Serrin H = W theorem holds independently of the smoothness of ad2 Theorem 2. The class COO(d2,RN) n W1,P(.f2,RN) is dense in W1,P(Q,RN)
This is proved by taking weighted mollifiers according to the distance of points x c d2 from the boundary a1, compare also the proof of (ii) of Theorem 1 in Sec. 4.1.1. Introducing the space
H1,P(Q,RN) := strong closure of Cl(d2,RN) nW1,P(Q ISBN) (1)
in W1,P(d2,RN)
we can state Theorem 2 as (2)
W1,P(Q,RN) = H1,P(0,RN)
Moreover, recalling that convex sets of a Banach space are sequentially weakly closed if and only if they are strongly closed, one easily deduces that H1,P(d2, RN), also agrees with the sequential weak closure of C, (Q' RN) n W1,P(,(l, RN), i.e., with the notation of Sec. 3.4, (sw-cl- and strong-cl standing respectively for sequential weak and strong closure in W1,P), we have (3)
H1'P(dl, RN) = sw-clw,,p (C1(d2, RN) n W1,P(.fl,
= strong-clwi,p (C1(Q, RN) n W',P(f2 RN)) This is what makes the space H1"P(d2,RN), equivalently W1°P(1?,RN), very useful in finding minima of energy functionals. Let now d2 be a bounded domain of an n-dimensional Riemannian manifold X and let us replace the flat space RN by a compact and boundaryless submanifold y of RN of dimension m. We set (4)
W"P(Q, y) := {u E W1"P(d2, RN) I u(x) E y for a.e. x E Q}.
We shall see that in general
5.1 Sobolev Classes of Maps Between R.iemannian Manifolds (5)
W1,P(Q, y)
633
strong-clwi,, (C1(,f2, y) n W1°P(rI,RN)), D
the obstruction to equality being the topology of Y. Of course it is to be expected that
Hs ong(Q, Y) := strong-clw,,, (C' (f2, Y) n W1,P(u RN)) has better properties compared to W1,P(Q, Y). On the other hand this class in principle lacks nice compactness properties with respect to the weak convergence and therefore it is not suitable for finding minima of energy functionals. For this reason it is convenient to introduce also the class (6)
Hweak(Q,Y) := sw-clwl,p (c' ((2,Y) n W1,P(Q,RN)),
i.e., the sequential weak closure of Cl (.f2, y) n W1,P (f2, RN) in W l'P (Q, IRN). In general one finds then , Y D Y)) D H1,P Wl'PQ, ( weak ( )
H1,P
strong
Q Y)
In the sequel of this subsection we shall report on some results about the classes W1,P(fl Y), HWeak(,f2, Y) and H t ong(.f2, y) in principle omitting proofs.
The class W1'P(X7, ym) and density of smooth maps. Let X and y be two oriented Riemannian manifolds of dimension respectively n and m. As usual we shall assume that Y is compact and without boundary, while X may or may not have boundary. Moreover, without loss of generality, we think of Y as of a submanifold of RN, and, if we want to emphasize the dimensions, we write X1
and ym instead of X and Y. For any real number p > 1, but actually we shall only consider the case p > 1, we set (7)
W"p(X, y) :_ {u E W1,P(fl, ]RN) I u(x) E Y for a.e. X E X}.
The nonlinear space W1'P(X,Y) depends in general on the embedding of Y into N. This is not essential in case X U 8X and Y are compact, because different embeddings give rise to homeomorphic spaces W1'P(X, Y). However, since the definition (7) works also for non compact X and Y an intrinsic definition should be more suitable. But we shall not pursue this point. Our first question is whether C°° (X, y)nW1"P(f2, JRN) is dense in W1"P(X, Y). Suppose for a moment that X is a domain of Rn and, for the sake of simplicity,
that X = R'. Given u E W1"P(X, y), the easiest and natural approach to our question is to consider the mollifiers uE
J W,, (x - y) u(y) dy , SoE - E-n P( )
where cp is a standard C°°(IR") kernel. Then the maps uE converge in W 1,P (X, RN) to u. But in general the range of uE is not Y. Assuming however
634
5. Cartesian Currents in Riemannian Manifolds
that u6(x), x E X, lies in a normal neighbourhood VE of Y in IRN, which shrinks to y as E tends to zero, we might consider the maps
.7ouE:X-->y where n is the nearest point projection map from VE into Y. In this case 17 o uE is readily seen to converge in W1,P to u. Unfortunately, in general we cannot expect that uE(x), x E Si, lies in a small normal neighbourhood of y; for instance, any mollification with a symmetric
kernel of the map xllxl from B3 into S2 has value zero at zero. The previous argument however does work, with minor changes, if p is larger than the dimension n of X because, by Sobolev-Morrey embedding theorem, we know that every u E W 1,P (X, y) is actually Holder-continuous and the mollifiers uE of u converge uniformly to u. In fact, essentially the same argument works also for p = n, and we have
Theorem 3 (Schoen-Uhlenbeck). Let X' be a Riemannian manifold with smooth boundary and let y' be a compact Riemannian manifold without boundary. Then C'(X, y) is dense in Wl°n(X, y).
Proof. By standard extension theorems we may assume that X is compactly contained in a Riemannian manifold k and that any u E W""n(X, y) is the restriction of a map u E W""n(X, y) to Y. We can also assume that X is isometrically embedded in some IRk, k > n. Let U be a normal neighbourhood of X in Rc and let V be a normal neighbourhood of y in RN. First we observe that, for E < dist (3X ), the functions
f
:=
Iduln dvolx
,
B'''(x,e)
where Bn (x, E) denotes the geodesic ball in X around x with radius e, is a continuous function of x E ,f2. Also, GE (x) clearly decreases when s decreases, and limE.0 GE(x) = 0 for all x E X. Therefore it follows, by the so-called Dini's theorem, that GE converges uniformly to zero in X. Now we extend u to a map u E W"°n(U,RN) by setting
u(x) = u(Px) where P : U --> X denotes the nearest point projection onto X. Since the metric on U is uniformly equivalent to a product metric on X x R71-n, we deduce for
x E P-1(X) (8)
Duln dxl
... dxk
< CEk-n
f Bn (Px,e)
Jduln dvolx = cek-n GE(x).
5.1 Sobolev Classes of Maps Between Riemannian Manifolds
635
Note also that u(x) E y for a.e. x E P-1(X). Let now V be a standard mollifier on Bk(0, 1), cpE(x) := e-kc (E), and
f(x_y)ii(y)dy
uE(x)
for x E X. By Poincare inequality and (8) we deduce
e-k f
(9)
Iu(y) - u6(x)In dy
Bk (x,E)
< cen-k
f
I D4LI In dxl
... dxk < cGE(x).
Bk (X,E)
Since ii: P-1(X) - y, this inequality implies that for all x E ,f1 we have (10)
dist(uE(x),Y) < cG.In(x)
Finally, let H : V -> y be the nearest point projection map, and observe that (10) and the fact that GE converge uniformly to zero. Then we can define a smooth map vE : X -* y by setting vE(x) := H o uE(x). It is readily seen that limo II vE
- u Ilw" (X,37) = 0
Remark 1. Taking into account the proof of Meyers-Serrin's H = W theorem,
it is not difficult to adapt the proof Theorem 3 to prove that coo (X, y) n W1,' (X, y) is dense in W1,' (X, y), independently of the regularity of an. Actually one then shows that each u E W""n(X, y) can be obtained as W1'n-limit of a sequence of functions in C°° (X, y) n W1,1 (X, y) which all have the same trace of u on aX. Next example shows that Theorem 3 does not hold for p < n, in particular we have 1 The map u(x) := x/IxI from B3 into S2, which belongs to W1,2 (B3, S2) is not a W1"2-limit of a sequence ui E C'(B3, S2). To see this suppose that such a sequence {ui } exists. Then for almost every
r E (1, 1) we would have W1,2(aB3(0, r), S2)-convergence of ui to the map xllxI. In particular the sequence of maps from S2 into S2 vi(x) := ui(rx) would converge to the identity map of S2 and each vi would have degree zero, since each ui provides an homotopy of vi to a constant map; in particular for the Jacobian determinant we have (11)
fJ2(dvi) = 0 S2
Vi
636
5. Cartesian Currents in Riemannian Manifolds
On the other hand, by taking a subsequence, we could assume dvi converge
pointwise a.e. to the identity. Thus J2(dvi) -+ 1 a.e. on S2. Since IJ2(dvi)1
<_
1Idvil2
and I dvil2 converge in L' to a limit, Lebesgue's dominated convergence theorem would imply that J2 (dvi) = 47r lim ioo J s2
a contradiction to (11).
The previous example rises the question of finding in which circumstances smooth maps between two manifolds are dense in the Sobolev classes W1'P(X, y) for 1 < p < n. The answer to this question is provided by the following two theorems.
Theorem 4 (Bethuel). Let 1 < p < n. Then smooth maps between X" and y'n are dense in Wl'P(Xn, ym) if and only if the homotopy group of order [p] of Y'n is trivial, i.e., 7r[p](ym) = 0. Here [p] denotes the largest integer less or equal to p.
For the proof of Theorem 4 we refer the reader to Bethuel [87]. We shall present the proof of Theorem 4 and of related results in the special case in which y'n is the two-dimensional standard sphere S2 of 1l in Vol. II Sec. 4.2.2. A variant of Theorem 4 in the case XI has a non zero boundary, in particular when Xn is replaced by a subdomain of Xn with smooth boundary is given by
Theorem 5 (Bethuel). Let 8Xn
0, 1 < p < n, and
ir[P]
(yn) = 0. Sup-
pose that u E W1,P(X' , ym), is such that its trace belongs to Wl'P(8Xn, y'n) f1 Co (BXn, y), respectively C°° (BXn, y'n), and that there exists v E C°(X', ym), respectively v E CO°(BX', y'n), such that u = v on BXn. Then u can be approximated in W"'P(Xn,Ym) by maps in W1,P(Xn, y'n) fl C°(Xn, y71), respectively C°°(Xn, ym), which coincide with u on BXn. As an immediate corollary of the previous result we can state in the case y'n = S2 in which 7r2 (S2) = Z
Proposition 1. We have (i) Suppose n = 2. Then Coo (XI, S2) is dense in Wl'P(X2, S2) for all p > 1. (ii) Suppose n = 3. Then Cm (X3, S2) is dense in Wl'P(X3, S2) for all p < 2
and C°°(X3, S2) is not dense in Wl'P(X3, S2) for all p with 2 < p < 3. When 7r[P](ym) 0, smooth maps are not dense in W1,P(Xn, y'n) for 1 < can be approximated p < n. However, in this case, each map u E W 1'P (X n, by maps which are regular except on low dimensional sets.
y,)
5.1 Sobolev Classes of Maps Between R.iemannian Manifolds
637
Definition 1. We say that u E W1,P(Xn, y-) is in RP, (Xn,ym), respectively in R' (Xn, Y) if and only if u is of class C°°, respectively continuous, except on a singular set r
E(u) = U Fi'i i=1
where, for i = 1, 2, ... , r, Ei is a subset of an (n - [p] - 1)-manifold and the boundary of Eiis smooth; if p > n - 1, Ei is a point. In the case that Xn is some domain of R n' we also require that each Ei is a subset of an affine subspace of Rn of dimension n - [p] - 1, and aEi is a subset of an affine subspace of dimension n - [p] - 2.
Theorem 6 (Bethuel). We have (i) let Xn be compact and 8Xn = 0. For every p with 1 < p < n, R°°(XI, y,,,,),
respectively RP (Xn, ym) is dense in W1,P(Xn, ym). (ii) Suppose 8Xn 0 and that the assumptions of Theorem 5 hold. Then each u E W1,P(Xn, y'n) can be approximated in W1'P(Xn, ))m) by maps in RR(Xn,Y'n), respectively in Rp (Xn,yt), which coincide with u on 8Xn.
In particular every map in W1'2(B3,S2), which is regular on aB3, can be approximated in W1'2 by maps from B3 into S2 which are regular except at a finite number of points in B3. Notice that in principle the number of singular points may tend to infinity. The class H''P(Xn, ym). A suited class of Sobolev functions for finding minima of energy functionals is clearly given by the sequential weak closure of smooth functions in Wl'P(X1, y"n) that we denote by (12)
1'P (X
y)
Hweak,n C1(X,Y)nW1'P(X,Y)). := sw-ClWi(
In general we have
(13)
W1,
P(X, y)
1,P
Hweak(X,
y) D HP (X Y) strong( ,
As we have seen, if p is larger than or equal to the dimension n of X or if 1 < p < n and ir[,] (y) = 0, the inclusions in (13) are actually equality.
0, we shall prove in Vol. II Ch. 4 that for integer In the case 7r[p] (Y) p and y = SP every element in W 1,P (X, Y) can be weakly approximated by a sequence of smooth maps in W1'P(X, Y). This in particular shows that for
X=B3,Y=S2,p=2 W1,P(X,Y) = Hw ak(X, Y) 1 Hs ong(X, Y).
(14)
It is an open question to decide whether (14) holds for all integer p, all X' and 3)
with 7r[P] (Y)
0.
For non integral p's we instead have
5. Cartesian Currents in Riemannian Manifolds
638
Theorem 7 (Bethuel). Let 8X = 0, 7r[p] (y) 0 and p be not an integer. Then smooth maps are not sequentially dense for the weak topology in W1,P(X, Y). Moreover every map in W',P(X, Y) which is a weak limit of smooth maps, i.e., which is in HI'P weak (X, y), is in the strong closure of smooth maps in W"'P(X, Y)
Theorem 8 (Bethuel). Let aX
1r[p] (y) 54 0 and p be not an integer. Suppose that u E W1'P(X,Y) is the weak limit of a sequence of smooth maps uk with traces all equals to u, ukjaX = u, and uklaX E W1'P(aX,Y) n C°, and all homotopic to a constant map. Then u can be approximated strongly in HW ak(X,Y) by a sequence of smooth maps which coincide with u on oX. 0,
In particular, for non integral p's and 7r[p] (Y) (15)
W 1,P(X y) D
0, we deduce
P (X, y) = Htlong(X, y). H" weak
Sobolev maps and homotopy. A natural question to ask is about the homotopy properties of Sobolev maps. It turns out that the homotopy class of a generic map in W1'P(X, y) is not well defined. However, as B. White has shown in White [667] White [668], Sobolev maps between manifolds X and y, that we shall assume for the sake of simplicity to be compact and without boundary, have a well defined k-homotopy type for suitable k.
Definition 2. Two continuous maps u, v : X --a Y are said to be k-homotopic if and only if their restrictions to the k-skeleton of a triangulation of X are homotopic.
If E is a polyhedral complex of dimension k and cp : E' --+ X is continuous, by an argument similar to the one in the deformation theorem one sees that, whenever X is a k-skeleton of X, cp is homotopic to a map cp such that cp(E) C X. Consequently the k-homotopy of continuous maps is well defined, and u, v induce the same homomorphism 7ri(X)
-. ii (y)
i = 1,..., k
whenever they are k-homotopically equivalent. The two classes H t' ong(X, y) and HW ak(X, y) play a different role also with respect to homotopy considerations.
Let d be the greatest integer strictly less than p. By Sobolev embedding theorem the restriction of a map in W1'P to a dimensional skeleton of (a triangulation of) X which is in generic position give rise to a continuous map which has a d-homotopy type. This however depends a priori on the triangulation and on the position of the k-skeleton. B. White shows that for each K > 0 there is an e > 0 such that if u1, u2 E C1(X, y) n W1"P(X, Y), II u1 - u2 II LI < e and 11 Dui II LI < K, then ul and u2 have the same homotopy type, consequently, compare White [668],
5.1 Sobolev Classes of Maps Between Riemannian Manifolds
639
Theorem 9 (White). Let d be the greatest integer strictly less than p. Then each u E Hs' ong(X, y) has a well defined d-homotopy type, and d-homotopy types are preserved by bounded weak convergence in H1"P(X, Y). He also shows, compare White [669]
Theorem 10 (White). The infimum of f I duI P dvolx Xn
among all smooth maps homotopic to a given smooth map cp depends only on the [p]-homotopy type of W. In particular the infimum is zero if and only if cp has the [p]-homotopy type of a constant map.
For the weak classes the following instead holds
Theorem 11 (White). Let d be the greatest integer less or equal to p - 1, d = [p - 1]. Then each u E HWeak (X, y) has a well-defined d-homotopy type, and d-homotopy types are preserved by bounded weak convergence.
Notice that both theorems say that the (p -1)-homotopy type is well defined and preserved by bounded weak convergence if p is integer. While for non integer p's the [p] and [p - 1]-homotopy classes respectively in HWeak and H" ong are well defined and preserved by the bounded weak convergence. Next examples show that the previous two theorems are optimal; in particular, if p is an integer, the p-homotopy type is not preserved by bounded weak convergence.
2 Let 1 < p < n and let ui : aB3 - S2 -+ 8B3 _S2 be a sequence of conformal diffeomorphisms that converges almost everywhere to a constant map, see 03 in Sec. 4.2.5. Then f s2 Jdu i I P is uniformly bounded and the ui's, each of which
is 2-homotopic to the identity map, converge to a constant map, which is not homotopic to the identity. 3 Consider the map u : B2 -+ aB2 - Sl, u(x) = x/1xI E HWeak(B2,DB2) for p < 2. By Theorem 7 u is not in H9'ong(B2, aB2). Moreover there are curves in B2, for instance two circles one enclosing the origin and one not enclosing the origin, which are homotopic in B2 but whose images under u are not homotopic in Sl.
Remark 2. The previous results imply also that each u in HWeak (X, Y) or in Hs 'ong (X, Y) induces homeomorphisms of the k-homotopy groups for 0 < k < d, d being the number defined respectively in Theorem 11 and Theorem 9. In
particular if p is an integer u induces homeomorphisms between the (real or integral) k-homology groups of X and y for all k with 0 < k < p - 1.
640
5. Cartesian Currents in Riemannian Manifolds
There seems to be deep relationships among the approximability results of Bethuel and the k-homotopy type of maps, in fact as remarked in Baldo and Orlandi [61], from their works it follows
Theorem 12. Suppose that p > 1 is not an integer, and u E W1"P(X, y). Then the following statements are equivalent
(i) u is the limit in W1,P-norm of smooth maps between X and y, (ii) u has a well defined [p] -homotopy type (iii) u induces well defined homeomorphisms u* : -7ri (X) -* 7ri (y), 0 < i < [p].
5.2 Cartesian Currents Between Manifolds Extending the notions of Cartesian currents and maps to the case in which Rn and RN are replaced by Riemannian manifolds X and Y, as well as generalizing local properties is quite simple. However it is worth to state explicitly a few facts and fix notation in this more general setting.
Approximate differentiability and the area formula. In Sec. 3.1.4 and in Sec. 3.1.5 we have already discussed these notions, clearly they extend to manifolds using local charts. Let X be a n-dimensional submanifold of Rn+k and let u : X -> R be an Nn measurable map. we say that u is approximately differentiable at x E X is for some chart cp : U -> X, cp(0) = x, u o cp is approximately differentiable at 0. The approximate differential DXu(x) is the linear map DX.u(x) : TAX -+ R characterized by
apD(u o cp)(0) = DXu(x)Dcp(0).
One easily checks that the notion of approximate differentiability and of approximate gradient is independent of the chosen local charts. We then also denote by AX (u) the set of points of X where u is approximately differentiable. Federer's theorem, compare Sec. 3.1.4, extends to manifolds.
Theorem 1. Let X be a n-dimensional submanifold of facts are equivalent
Rn+k. The following two
(i) u is a.e. approximately differentiable on X, 7-ln(X \ A' (u)) = 0. (ii) There exist a decomposition of X into ?-ln-measurable pieces 00
X=UX'UX0 i=1
such that ?in (X0) = 0 and for i = 1, 2, ..., uI X, is Lipschitz.
5.2 Cartesian Currents Between Manifolds
641
Moreover
AD (u) C 00 U Xi i_1
and Dx(ulx)(x) = Dxu(x) for a.e. x c Xi. Let X be a n-dimensional submanifold in ][8,+x and let u : X --> RN, N > n, be an a.e. approximately differentiable map on X. The approximate Jacobian is defined at every x E AD (u) by Ju (x) =
det (Dxu)*(Dxu)
AnDxu(x)
II.
If A is a measurable subset of X one defines also the Banach indicatrix of A Nx (u, A, y) = #{x E A n AD (u)
I
u(x) = y}
for every y E RN. The area formula, Theorem 1 in Sec. 3.1.5, then extends to manifolds.
Theorem 2. We have (i) the map y --* Nx (u, A, y) is xn-measurable in ]ESN and the equality
Nx (u, A, y)
Ju (x) d Hn
f
A
(y)
RN
holds.
(ii) The set {y E RN
I
Nx (u, A, y) # 0} = u(A n AD (u)) is countably n-
rectifiable,
(iii) Ju E L' (A;,Hn) iff Nx (u, A, ) is 9-1'2-summable in RN.
Moreover for any non negative ?- n-measurable function f : RN -* R, or any f : RN -* R xn-measurable such that either (a) f ()Nx (u, A,.) is 7-tn-summable or
(b) f
is ?-ln L A-summable
we have the area formula
f f(u(x))Ju (x)d?-ln= f f(y)Nx(u,A,y)d?-ln. A
RN
Proof. First assume that A is contained in a coordinate chart cp : V -* U A C U and let B := co-1(A). Then u o V is ?-LI a.e. approximately differentiable on V and xn a.e. y E V. apD(u o cc) (y) = Dxu(So(y))Dco(y) Therefore J,,,o,(y) = Ju (co(y)) J,(y). The classical area formula and Theorem 1 in Sec. 3.1.5 then yield
5. Cartesian Currents in Riemannian Manifolds
642
fJcdnn A
= f Jv (,p)J,(y) dxn B
B
r
J
J...,d?{n
r NX (u, A, z) d70 (z).
NX (u o cp, B, z) d'h' RN
RN
as required. Then we prove by approximation the second part of the claim for functions which are supported in a coordinate chart. For general A's it suffices to use a partition of unity associated with a system of coordinate charts.
Graphs of approximately differentiable maps. Let X be a n-dimensional oriented submanifold of Rn+k and denote by X it orienting unit n-vector field.
Let D be an open set in X, let u : (2 -> RN, N > 1 be an 7ln L d2 a.e. approximately differentiable map, and let AD (u) be the set of points of 12 where u is approximately differentiable.
Definition 1. The graph gu,fl of u is defined as the subset of ,f2 x I[ RN given by
{(x, u(x))), x E 0 n AX (u)}.
The area formula, Theorem 3 in Sec. 5.1.3, then says that 9u,f2 is countably rectifiable set and x' (9u,.?) =
fJdfl2LS2. ;au
Moreover it is also clear that Theorem 5 in Sec. 3.1.5 extends to this more general situation, hence we conclude
Proposition 1. The following holds (i) 7-ln (c ,{2) < +oo and 9,,,, l is n-rectifiable iff An (id >a DX u(x))X (x) E Ll (S2).
(ii) If 7-tn(9u,n) < +00 then the unit n-vector in T 12 x ISBN
A,(id
orients the n-tangent plane to 9u,f2 at (x,u(x)). As in the case .2 C IIBn we introduce the class
Al(0, RN) = {u E LI(,f2,RN) I u is a.e. approx. diff., 7{"(c'u,, 2 < oo}.
Of course for the maps in Al the conclusions of the previous Proposition 1 hold. Also to each u E Al (.(2, IRN) we can associate the rectifiable n-current Gu E Dn (.(2 x RN) given by Gu = 'r(CJu,d2, 1, u,12)
Clearly
5.2 Cartesian Currents Between Manifolds
M(Gu)
(1)
/
643
11 A, (id D< DXu(x)) 11 dxn,
and, using the area formula, we can express Gu as integration over the base domain. In fact if w E D, (f? X RN) we compute Gu(w)
<w, 02L,f2 > dRn L 9u,f2
_
f<W,An(idDXU(X))>d1ThLQ
=
(2)
<W, au, .fl >J dHn L .f2
J J
d7-ln L.(2
(id x u)#w
92
which expresses, in the case u is Lipschitz, that Gu = (id t< u) # Q ,f2
1.
The weak continuity theorems in Sec. 3.3.2 on minors extend trivially to maps defined on manifolds. For example, we have
Theorem 3. Let {uk} be a sequence of maps in A1(ul,RN) such that A, (id x DXu(x))X(x) E L'(,f7). Assume that weakly in L'
Uk -s u
and that
A, (id X DXUk(x))X(x) -
weakly in L'.
If sup M(aGuk L (.f2 x Rn)) < +oc
(3)
k
then
u E Al(.0, RN)
and An(id D< DXu(x))X(x) _ C(x).
This theorem applies in particular to maps in the class cartl(.f2, IRN) := {u E Al(fl, RN)
I
9G u L .fl x RN = O}.
Actually a stronger result, compare Proposition 2 in Sec. 3.3.2, holds true. Denote by .Fp the class of (n - 1)-forms which are linear combinations of forms of the type w(x)Ag(yz)
forO-min(n,N),
dya-2
wEDk(Q), k=n-1i I,andgEC°°(R,R).
Then one can substitute the boundary control (3) by (4)
sup sup{aGu(w)
I
w E .7p, jwj < 1} < +oo
k
and the claim of Theorem 3 holds true.
5. Cartesian Currents in Riemannian Manifolds
644
Cartesian currents in X x R'. Let X be an n- dimensional oriented submanifold of Rn+k and let .f2 be an open set in X of finite volume. It is a trivial matter to single out the class of Cartesian currents in X X IISN and prove for this class closure and structure theorems similar to Theorem 3 in Sec. 4.2.2 and Theorem 1 in Sec. 4.2.3. But in order to fix notation and for completeness let us state a few simple facts. Let us begin with a few remarks on forms in the product ,f2 x Y where more generally we take as Y any oriented, compact and boundaryless Riemannian manifold. The product structure of ( x Y yields a way to decompose for x E (, y E Y, the vector space of tangent n-vectors to T(:,,,y) (.f2 x Y) which are linear combination of n-vectors of the type v1A... vn_kAwln... nwk, with vi, E An_kTT.f2 and wi E AkTyY. Set Vk := An-kTTQ ®AkTyY.
These vector spaces are independent and yield a decomposition of AnT(y,y) ((2 x Y), min(n,m)
A,,T(x,y) (,fl x Y) =ED An-kTTQ ® AkTyY k=0
as we can verify using local basis in T,, f2 and T.Y. If v1, ... , vn, w1, ... , wn are basis respectively for TTQ and TyY, one can uniquely express E AnT(x,y) (.f2 x Y) as
n
£,,pvanwQ = > e(k)
e=
k=0
where e(k) :=
&,,,VO \WR E Vk. IuI+IQI=n 1P1=k
The decomposition then yields by duality a decomposition of covectors and forms and consequently of currents by setting for w E AnT(y,y)(.f2 x Y) min(n,m)
W=
W(k),
C <W(k), S > := <W, C(k) > V e E AnT(.,,y) ('i2 x Y),
k=0
forwEDn(Qx31)
w=
<w(k) (x, y), e> := <w(x, y), e(k) >
and for T E Dn (,f2 x Y), T =
V f e AfT(,,y) (Q x Y)
'
j:ron,m) T(k) where
T(k) (w) := T(w(k)),
W E Dn(f? X IIBN).
5.2 Cartesian Currents Between Manifolds
645
Let us denote by X the orienting n-vector field of X, by f? the dual n:.(l x RN RN the two orthogonal
volume form, and by -7r : S2 x RN
projections on the factors. Let T E 1),(Q x IRN) . We define
IIT111:=sup{T(f(x,t)Iyl r#Qx(x,y))
I
f EC,(Qx1RN), Ifl <_ 1}
and the class of Cartesian currents in .(l x RN as
cart(, x IRN)
T E D"(,(l x IRN), T is i.m. n-rectifiable, M(T) < +oo, II T II < +oo, aT L f2 x 1RN = 0, 1
it#T=
TL7r#Qx >0}.
We have, compare Theorem 1 in Sec. 4.2.2
Theorem 4. Let Tk E cart(.f2 x RN). Assume that Tk -k T in D"(f2 x RN) and supk(M(Tk) + Tk 111 < +oc. Then T E cart(. x IRN). Let now T = -r (M, 0, T) E cart (!2 x RN). Define
M+ = {(x, y) EMI> 0} and for j = 1, 2, ... , N, the measures
:=T Lyi7r#nx then we have Theorem 1 in Sec. 4.2.3
Theorem 5. Let T E cart (f2 x IRN). The measures ui (T) are absolutely continuous with respect to the . l' L ( measure, ui (T) = uT (x) dl-l" L .f2. Setting UT (x) := (uT (x), ... , UN (x)) T
for a. e. x E fl
we have
(i) UT is a. e. approximately differentiable and T L MT = GUT, in particular T(o) = GUT(o)
.
(ii) UT E BV(.f2,IRN). Thus defining also
ST:=TL(M\M+) we have a canonical decomposition of T as T = GUT + ST,
M(T) = M(GUT) + M(ST)
with UT E BV (,(l, IRN) and ST(o) = 0. Consequently we can state
Proposition 2. T E cart(S1 x R N) if and only if one can write T = GU + S where
5. Cartesian Currents in Riemannian Manifolds
646
(i) uEBV(Q,RN)nAl(f2,RN) (ii) M(S) < +oo, S(o) = 0 'iii) 8(G. + S) = 0. Moreover the decomposition is unique and
Gu=TLM+
S=TL(M\M+). Cartesian maps and currents between manifolds. It is now natural to extend the definition of Cartesian maps and currents when the target RN is replaced by a m-dimensional oriented submanifold y of IRN: it suffices to require
that the corresponding graphs be currents in Q x Y. Since our currents are rectifiable, we can impose such a condition just by restricting the support of the graph current, on account of Federer's flatness theorem, Theorem 1 in Sec. 5.3.1. Therefore we set
Definition 2. We set {u E carts(,R,RN) {u E cart'(,R,l[8N)
cart'(f2,Y)
I
I
Gu E Dn,(.fl x y)}
u(x) E Y H''-a.e.}
and
cart(fl x y)
:= cart(.f2 X RN) n Dn(,f2 x y) sptT C fl x y}. {T E cart (0 x IRN) I
Let 12 be an open subset of an oriented, n-dimensional submanifold X in Rn+k and let y be a closed submanifold of RN of dimension m. In dealing with the Dirichlet integral for mappings between 0 and y, u :.2 --p y, the class cart 2,2,1,...,l (n x y) := briefly
cart2,2,1,...,1(Q
x R 'V) n { T I spt T C ,f2 x Y}
cart 2,1(o X Y)
will be relevant
Proposition 3. We have
cart 2'1(1x3))={T=GuT+ ST Ecart(Qx3))IUT E W1,2(Q
Y),
ST(o) = ST(1) = 0} = {T = Gu +S I u E W1,2 (f2' y), M(S) < +oo, S(o) = S(1) = 0 and a(Gu + S) = 0 on 80 x y}
Proof. Since the cartp(.f2,RN)-norm of T is finite, being p = (2, 2, 1), we deduce from Sec. 4.2.3 that M (T) is absolutely continuous with respect to Lebesgue's measure and
5.2 Cartesian Currents Between Manifolds
uT(x) E S2 a.e.,
T00(0(x, y)) =
J
647
0(x, uT (x)) dx
TZ' (O(x, y)) _ (-1)z-1 f 0(x, UT (x)) DiuT dx n hence u E W"1(1I, S2), since 8T L .(2 x S2 = 0. As <_
I1TIIcart2.1
IIu1IW1,2
we infer UT E W1"2(.(2, S2) . The isoperimetric inequality
IJ2(DuT)I = IM(2)(DuT)I < 1IDUT12 yields UT E A' (0,183) .
The class cart2'1((2 x y) has the following closure property
Theorem 6. We have We have Suppose that {Tk} is a sequence of currents in cart2'1(.(2 x y), Tk = Guk + Sk, for which
supM(Tk) = sup(M(Guk) + M(Sk)) < +00 k
k
sup II uk IIW=.2 < +00k
T as currents and Uk -k u in W1"2 then T E cart2'1(Q x y) and If Tk T = Gu, + S, with S(o) = S(1) = 0. Proof. Passing to subsequences
Guk - P,
Sk -4 R in Dn (.fl x Y)
and obviously R(o) = R(1) = 0. On the other hand, since DUk -s Du in L',
Guk(o) - Gu(o),
Guk(1) - Gu(1)
so that
Gus - P = Gu + Q,
and
Q(o) = Q(1) = 0.
Hence T = Gu + S, S = Q + R and S(o) = S(1) = 0 as required. Limiting graphs with such vertical pieces S may occur as weak limits of maps with bounded Dirichlet energies. We have already discussed several examples in Sec. 4.2.5. However in the special case of dim .(2 = dim y = 2, such a concentration may occur only if vol(y) < +oo. In fact we have Theorem 7. Let T E cart2>1(f2 x y) where dim .(2 = dim y = 2 and R2 (y) _ +oo. Then T = Gu, U E W1"2(.(2, y).
648
5. Cartesian Currents in Riemannian Manifolds
Proof. We already know that T = G,, + S, u E W1,2 (.f2, y) and S(o) = S(1) = 0. Since dim .2 = 2, aG, = 0 hence aS is zero. Being S a 2-current, it suffices to prove only that S(2) = 0. Let a be a 1-form in 0 x 3) and f E Cc' (f2). From
d(f a) =dfna+ fAda, we infer
S L f (da) = S(d(f a)) - S(df Aa) = aS(f a) - S(1) (df Aa) = 0 i.e. V f E C°° (.fl), 3(SL f) = 0. Consequently by the constancy theorem
#(SL f) = Since M(S) < +oc and M(QYJ) = 7.12(3)) _ +oo we infer
#(SL f) = 0
V f E C°° (.Q ), hence
S(f(x)w(y)) = 0 By density
S(w)=0
Vf E C°°(Q), Vw E D2(y)
VwED2(.f2xy), w=w(2).
i.e. 8(2) = 0.
5.3 Homology Induced Maps: Manifolds Without Boundary In this section we briefly discuss different ways of tracking the homological content of a given Cartesian current between oriented, compact and boundaryless Riemannian manifolds. This is actually a classical topic in homology theory, and
in fact in this section we shall describe a number of constructions in homology that one can perform starting from a given homology class in the product manifold. Let X, Y be two compact, oriented and boundaryless Riemannian manifolds
of dimension n and m respectively. As usual we denote by it : X x Y --> X and 5 : X x Y -> Y the orthogonal projections of the product manifold X x Y onto its factors.
If u : X -+ Y is a smooth map between X and Y, a way of tracking its homology is to consider the homology maps associated to u, i.e. the maps
u : Hk (X, Z) - Hk (Y, Z) defined for k = 0, ... , n by
u*([R]) := [u#(R)] for any k-cycle with finite mass R E Zk (X, Z); representing homology classes by k-cycles with finite mass is particularly convenient. If T E cart(X x Y), then T
5.3 Homology Induced Maps: Manifolds Without Boundary
649
is an n-cycle of finite mass, T E Zn(X x Y, Z). In the sequel we shall see how we can associate to any class y = [T]a E Hn (X x Y, ]R) a homology map y* : Hk (X, If8) ---> Hk (Y, R)
k = 0, ... , n
and toy = [T]z E Hn (X x Y, Z) a homology map
y* : Hk(X,Z) -
%
Hk(Y,Z)
k = 0,...,n
which in particular extend the homology maps from the classical case of smooth maps to Cartesian currents. Such a map turns out to be stable with respect to the weak convergence of Cartesian currents just by definition. As we have proved in Sec. 5.3, on any compact manifold without boundary one has the two non degenerate pairings of de Rham and Poincare <[T], [w] > Poincx < [w], [n] >
:= T(w),
T E Zk(X), w E Zk(X),
,w E Zn-k(X ),
:=
17 E Zk (X),
X
the two being related to the Poincare isomorphism
Px : Hk(X,R) -' HHRk(X) by
< [T], [w] > = Poincx < w, Px [T] >.
(1)
Let now -y E Hp(X x Y, l ), equivalently let -y = [T]R, T E Zp(X x Y). Then T yields for k = max(p - n, 0),... , min(p, m) bilinear maps
T : Zp-k (X) X Zk (Y) -> R given by T(w, ri) = T (7r#wA:R#r7). It is easy to check that the value of T(w, 77) depends only on the cohomology classes of w and 77 and on the homology class y = [T]R of T. This way each y E Hp(X x Y, ]R) fixes the bilinear maps (2)
'Y([w]
-' R,
k E (max(p - n, 0), min(p, m)) [77]) := T(7r#wA'r#17) if w E Zp-k(X), 77 E Zk(Y)
HaRk(X) x HaR(Y)
which are the pairings in cohomology induced by y. Starting from the pairings : HHRk(X) x HHR(Y) --+ R one easily obtain equivalent maps by combining with Poincare and de Rham dualities. (i) One defines the cohomology map y* : HHR(Y) -' HaRn-p(X) by (3)
Poincx<[w],'Y*[77]> :='Y([w], [77]) = T(7r#wA77)
for 77 E Zk (Y), W E Zp-k (X ).
(ii) The so called D-map yD : HHR(Y) -4 Hp_k(X,R) by
650
5. Cartesian Currents in Riemannian Manifolds <'YD[71],W> :='Y([w}, [rl])
(4)
for w E Zp-k(X), 77 E Zk(Y). Of course for rl E Zk(Y), 7D[77] is the real homology class of the current w --> T(rr#wnF#r7), w E Dp-k(X) and P(-YD)[rl] ='Y*[rl]-
(iii) Also one defines the homology map induced by -y, y* : Hk+n_p(X,R) Hk (Y, JR) by duality, i.e. < y* [S], [r)] > _ < [S], y* [71] >
(5)
for S E Zk+n_p(X ), ri E Zk (Y).
Proposition 1. Let T E Zp(X x Y) and y = [T] E Hp(X X Y, R). Then, for S E Zk+n_p (X ), y* [S] is the real homology class of the k-current (-1)(n-p)(p-k)T(?#rlnrr#Ps),
77 -,
Ps being a Pozncare dual form of S. Proof. If r) E Zk (Y) and e represents the cohomology class of y* [77], [l ] = y* [77],
then we compute
fAPs =
<[S],'Y*[77]> = S(l;) =
x (6)
_
(-1)(k+n-p)(p-k)
f PsAl; =
(-1)(p-k)(k+n-p)Poin([Ps],'Y*[7l])
x
=
(-1)(k+n-p)(p-k)T(7r#Psnrr#r7)
= (-1)
#Ps)
We now show that y* and y* are extensions of the classical homology and cohomology maps induced by a smooth map
Proposition 2. Let u : X -* Y be a smooth map. Then [Gu]* and [G,,]* are respectively the ordinary homomorphisms in (real) homology and cohomology [Gu]* = u*,
[Gu]* = u*.
Proof. Let 77 E Zk(Y), R E Zk(X) and let PR be a Poincare dual form of R.
Then
f u#rltPR = R(u#7l) = u#(R)(77). X
Thus < [Gu] * ([R] ), [77] > = < [u#R], [77] > = < u* [R], [r7] >. Analogously for n E
Zk(Y), w E Zn-k(X) Gu(rr#wn?#r)) = fx wnu#r) and therefore
5.3 Homology Induced Maps: Manifolds Without Boundary Poincx < [w], [Gu] * [7l] > =
J
651
wnu#rl = Poincy < [w], ,> [u#7)]
X
Poincy < [w], u* [77] >.
It is worthwhile noticing that if one starts with an integral cycle T E Z'(X X Y, Z) and -y := [T]a, the above construction yields actually an homology map ry* : Hk+n-p(X, Z) -' Hk (Y Z), -y :_ [T]1 provided Hk (Y, Z) is free. In fact from Proposition 1 in Sec. 5.4.2 the (real) intersection class [T]ie flxxy [S X Y]R
S E Zk+n-p(X,Z), contains an integral cycle obtained by intersecting T with a suitable translation of S x Y. Consequently 7-[S]R := *([T]R flxxy [S X Y]IIe) contains an integral cycle, too. Being Hk (Y, Z) free, y [S]a defines a unique integral homology class -y. [S] E 'Hk(X, Z). This way we have defined a map ry*
: Hk+n-p(X,Z) --> Hk+n,-p(X,]R) ---> Hk(Y,Z)
In the general case in which the target manifold Y have some torsion, one can construct nevertheless an homology map between the integral homologies of X and Y associated to an integral cycle T E Zp (X X Y, Z). In fact, given -y = [T]z E Hp(X x Y, Z), one defines y* : Hk+n_p(X,Z) Hk(Y,Z) by
7*[T]z :_*([T]z nxxy [S X Y]z) where fl is the intersection operator in integral homology for X x Y, compare Sec. 5.4. It is easy to show that if u : X -> Y is a smooth map, then [G,L]*
:
Hk (X, Z) ---> Hk (Y, Z) agrees with the standard homology map u*. In fact, since
G, is smooth the intersection current Gu nxxy Ta#(S x Y) is well defined for all a close to zero and G. nxxy Ta#(S x Y) = (id
ry*[T]z := 7r*[(id x u)#(S)]z = [u#S]z = u*([S]z)
Also such a homology map extends the one in real homology we have defined before, on account of (ii) of Proposition 1 in Sec. 5.4.2 and of the relations between intersection in real and integral homology.
Kiinneth formula. As shown by their definitions (2) (3) (4) and (5) the maps ry*, yD and ry* can be recovered each other and all depend on ry. Actually ry can be recovered from y, hence the previous five ways of describing the (real) homological content of a p-cycle in X x Y are equivalent. This is a consequence %y,
652
5. Cartesian Currents in Riemannian Manifolds
of the Kiinneth formula. This formula yields an explicit relation between the homology of the product X x Y and the homology of the factors, modulo torsion. We first state and prove Kunneth formula for the simplicial integral homology where it reduces to a simple combinatorial argument. So let us assume that X and Y are two finite CW cubical complexes in some RN and RNA and assume that their integral homology groups H,(X,7G),
H*(Y,Z)
are torsion free, i.e. free Abelian groups. Then one can choose polyhedral rec E Zk (Y, 7G), in such a way that the corretifiable cycles R?h) E Zh (X, Z), sponding homology classes yield respectively a basis for Hh (X, Z) and Hk (Y, Z). Kunneth formula states that the cycles x
E Zp(X x Y,Z),
h+ k= p
yield a system of independent generators for Hp (X x Y, Z). More precisely
Theorem 1 (Kunneth formula). Let T E Z1,(X x Y). Then T decomposes uniquely as
T=
a ]kR2h) X S(k) + 8E 2,j, h+k=p
where a k E R and E E Bp+l (X x Y, ]R). If moreover the integral homology groups of X and Y are free and T is rectifiable, T E Zp(X X Y, Z), then the coefficients a J E Z and E is rectifiable. In the proof we will use the following simple lemma of commutative algebra whose proof is straightforward, see Bott and Tu [102, 14.13.1, 14.17]
Lemma 1. Let A, B be free Abelian groups, where A C B and the factor group B/A is also free. Then B -- A ® B/A. In particular each basis of A (an independent system of generators of A) can be completed by elements from B/A in such a way that the resulting set forms a basis for B.
Proof of Kunneth theorem. We first prove the claim for integral cycles, T E Zp(X X Y,Z). Denote by Bk (X, Z), Zk (X, Z), Pk (X, Z) respectively the sets of k-simplicial boundaries with i.m., k-cycles with i.m. and k-chains with i.m. and by Bk(Y,Z), Zk (Y, Z), Pk (Y, 7G) the corresponding ones for Y. First we construct suitable
bases for all Pk(X,Z) and Pk(Y,Z). We start with k = n. Note that B,, (X, Z) = 0, choose a basis {San) } for Zn, (X, Z). Complete now {San) } with ft-( ) } C P,,, (X, Z) so that {San) , CA) } be
a basis for Pn (X, Z). To do this we use Lemma 1. Since Zn (X, Z) -- Hn (X, Z) and Pn (X, Z) is free we need to show that the factor space Pn (X, Z) /Zn (X, Z) is
5.3 Homology Induced Maps: Manifolds Without Boundary
653
free. But this is trivial since from C E P, (X, Z) \ Z7, (X, Z) and k C E Zn(X, Z) for some k E Z it follows 0 = 8(kC) = kaC,
i.e. aC = 0.
C( form a basis for P,,, (X, Z) {8CQ generates B,_1 (X, Z). we shall now show that they are independent. Assume that for some {bp} C 7L we Since
have bp8CQ) = 0,
{bp} non-trivial.
Then > bpCp) is a cycle in Zn (X, Z) and, using the basis {S,) } for Zn (X, Z), we have bpCQ)
= L. a0, an )
for some integers ac,. But this contradicts the fact that {S«n), CQ form a basis for Zn(X,7Z).
We now proceed by downward induction. We select for k = n, n - 1, .. . cycles SQ) E Zk (X, Z) and chains Cyk) E Pk (X, Z) such that {aCCk+1) } is a basis for Bk(X, Z),
(7)
{aC,k+1) SQ) }
(8)
is a basis for Zk (X,71,),
{aCak+1) S0 ), Cyk) } is a basis for Pk (X, Z),
(9)
and we prove that (i)
{eCCk) } are independent. If not
caC(k) = 0,
c, E Z,
{C} non trivial
,k+1) + then E c-,dyk) E Zk (X, 7G) and by (8) we have cyCryk) _ bpSak) for some integers a,,,, bp: a contradiction with (9). (ii) Since Zk_1(X, Z)/Bk_1(X, Z) ^- Hk_1(X, Z) and Bk_1(X, Z) free, we can choose {SQ -1)} C Zk_1(X,Z) by means of Lemma 1, to obtain a basis aaaC'(
{aq(,k),
(iii)
for Zk_1(X, Z)
We complete the basis for Pk_1(X, Z) by adding suitable polyhedral chains {CC.k-1)}, which is possible again by Lemma 1 since Pk_1(X, 7G)/Zk_1(X, 7G)
_ _ _ S0 ), Cryy } for Bh(Y, Z), Zh(Y, Z) Similarly we construct basis and Ph (Y, Z) respectively. Then each cycle T E Z, (X x Y, Z) can be expressed as a linear combination (with integer coefficients) of products of basis elements in Pk (X, Z) and Ph (Y, Z), k + h = p. We now observe that the terms involving products of the type is trivially free.
654
5. Cartesian Currents in Riemannian Manifolds aCak+l)
X
8C(&h+1)
aCak+l) X §(h), S(k) X aCah+l)
are boundaries in Bp (X x Y, Z), and terms of the type aCak) x C(O,h) are homologous to terms like C(k) x aC(O,h) e.g.
a(cak) XCQ))=acak) xCQh)+(-1)kCak) xacQ) Therefore, modulo a boundary, we get the decomposition of T,
T-aR =
aahSak)XSR)+ +
baQSak) x C(h) +
caQCak) x
dapCak)
e,aC(k) x 3CQ+1)
x cp) +
where coefficients are integers and R E Pp+1(X x Y, Z). Taking the boundary, we obtain
O=aT =
1:(-1)kbkhSak) x aCQh) +
+ +
daQaCak) X
C)( ,h)
eaQacak) x
capaCak) X SQ) +
+ (-1)kda%Cak)
X aCQh) +
8C-(O,h+1)
Since all the products on the right hand side are independent, we infer kh kh kh kh baQ - CaQ = daQ = eaQ = 0 hence
T - aR =
aapSak) x SQ ),
aaQ E Z.
This concludes the proof of the Kiinneth formula for integral cycles. If now T E Zp (X x Y, IR) is a real cycle we follow the same path and we get
T-aR=I:a,pS(k) X g(h) for some aaQ E R and R E Pp+1(X x Y, R). Finally in order to show that the decomposition is unique, assume that S-131(h)
aa(jSak) x
= aR
for some R E Pp+l (X x Y, Z) or R E Pp+l (X x Y, IR). Representing R in the
same basis as before and taking the boundary we obtains that aR does not contain terms of the type Sak) X S(113hS. Using the independence of the elements of the basis (9) we conclude that akh 0,0 = 0. '
Since that simplicial homology of a cubical complex inside of a tubular neighborhood of a compact deformation retract X is equivalent to the current homology of X, we infer at once from Theorem 1
5.3 Homology Induced Maps: Manifolds Without Boundary
655
Theorem 2. Let X, Y be two compact deformation retracts in RN, and assume that the homology groups Hk (X, 7G), Hh (Y, Z), k, h = 0.1.... have no torsion, i.e. are free Abelian groups. Then there exist rectifiable cycles SZ k) E Zk (X, Z), (Y, Z), k = 0,... , n, h = 0,.. . , m such that the homology classes form bases in Hk (X, Z) and Hh (Y, Z) respectively, and such that each current T E Zp (X x Y, Z) can be uniquely decomposed as
T = > a hSik) x
aR
with integers a. Si( Now we shall assume both X, Y without torsion and fix generators {k) as in Theorem 2. Let {Wak) }, {wQ ) } be dual generators in HaR(Y) and
HHR.(Y), i.e.
S(k) (Wak)) _ ji.,
sjh) (Wph)) = 5
.
By duality we obtain the following decomposition of closed forms on X x Y.
Proposition 3. For each closed p-form 0 E ZP (X xY) there exist real numbers bkh and a form E DP-1(X x Y) such that Ceo
=
bkh7r#W(Ckk)no#(D
(h)
+dpi.
on forms just defined it is easy to check Using the dual basis that the coefficients in Kunneth formula are given by (10)
ak=
y([Wih)] [rl(k)])
if y := [T]. In particular the coefficients in Kunneth formula depends on ry and we conclude
Proposition 4. Let T E Z p (X x Y, l ), then the real homology classy of T, the pairings in cohomology ry, the real homology maps y* and the cohomology maps -y*, are all equivalent ways of describing the (real) homological content of T. Proof. In fact as we have seen ry, 'yD, y*, y* can be recovered one from the other.
Now since the coefficients in Kunneth formula for T in (10) depend only on y, then the homology class y of T depends only on ry. It is also interesting to compute the homology map y* in terms of Kenneth's coefficients. We have
Proposition 5. Let T E Zp(X x Y, R), y = [T]R and let R E Zk+,l_p(X,1R) be a cycle in X. Then we have (11)
y*[R]
ap-k
being the intersection index of R(p-k) and R. In particular if T and R are both integral cycles, then -y* [R] has a representative which is integral.
656
5. Cartesian Currents in Riemannian Manifolds
Proof. Let us assume for a moment that R is one of the basis cycles R(k+m-p) in Zk+f.-p (X, Z), and let us denote by PR a Poincare dual form in ZP-k (X) of R. Then we have for 77 E Zk (Y)
T(:r#rjnlr#PR) = (_1)k(p-k)T(rr#PRA #77)
=
(-1)k(P-k) 1: ap-k,kR(P-k) (PR)'S(k)
(77)
i, j
Since, compare (7) in Sec. 5.3.4, R(P-k) (PR) = J PEAPR( -k)
=ix(R,R(i -k)),
X
(11) is proved for the cycles of the given basis and by linearity for all cycles. If now T and R are rectifiable, then aP k E Z, and ix (R, Rip-k)) E Z, consequently -y* [R] contains a rectifiable cycle.
0 Degree. Let X and Y be two closed, connected, oriented manifolds of the same dimension n and let T E Z,, (X X Y,Z). One can computes the homology map induced by T. Being X connected, H,, (X, Z) has one generator, the fundamental class of X, that we can represent by [Q X II] . Then, taking into account that a Poincare dual of Q X II in X is the constant 1, one computes for i E Zn (Y) [T]*(IXf)(77)
F#(T)(71)
and by constancy theorem, [T]*(QX 1) = degTQY]J. The number degT is called the mapping degree of T and evidently deg T E Z if T E Z,,, (X x Y, 7G). Notice that deg T = 0 if vol(Y) = oo. We conclude this section with a few remarks on the cup product in the context of the de Rham cohomology.
Let X and Y be two compact manifolds and let n := dim X. As we have already seen in Sec. 5.3.4, the wedge product of closed forms actually induce a product on the de Rham cohomology ring, the cup product, [w]A[rj] := [wnrj] E
HdRk(X)
for w E HHR(X), 77 E HHR(X). If U : X ---p Y is a C1 map, then clearly u#wnu#rj = u#(wnrj), hence u*[w]Au*[n] = u*([w]A[77])
for w E Zk(X), rj E Zh(X), h, k > 0. Since in this case [Gu]* = u*, we then conclude that
5.3 Homology Induced Maps: Manifolds Without Boundary
657
Proposition 6. If -y E H, (X x Y, Z) is representable as a graph of a C1 map v : X --> Y,
'y then
'Y*[wJA-y*[rl] ='Y*([W]n[77]),
VW E Zk(X), b'rl E Zh(X).
Definition 1. Define
cart(X x Y) :_ {T E cart(X x Y)
1
2v E C'(X,Y) such that [T] E Hn(X X Y,Z)}
Proposition 7. cart(X x Y) is closed for the mass bounded weak convergence of currents.
Proof. In fact, let Tj E cart(X x Y), sups M(TV) < oc, and let Tj T. From Theorem 4 in Sec. 5.5.2, taking also into account that Y is compact, we infer that T E cart(X x Y). On the other hand, Theorem 1 in Sec. 5.1.3 shows that Tj and T belong to the same homology class for large j hence [T] = [Ti] = [G,,; ]
for large j.
In particular defining
Cart(X x Y) as the smallest set containing the graphs of C' (X, Y) maps which is closed under the mass bounded weak convergence of currents, we conclude from Proposition 7
that
Corollary I. If T E Cart(X x Y) then there is a C1 map v : X -+ Y such that [T] in H,(X x Y,Z). The same reasoning yields also that
Proposition 8. Let u E Cart' (X, Y). Then there exists v E C' (X, Y) such that [G,,] = [Cr] in H, (X x Y, Z). 2 Not every homology class represented by a Cartesian current can be represented by a smooth graph. Let T2 denotes the standard torus in R3 with its usual orientation. Consider the product manifold S2 x T2 and denote as usual by 7< : S2 x T2 -+ T2 the orthogonal projection onto T2. We have
Proposition 9. If T E cart(S2 x T2) then #T = 0. In particular if u E Cart1(S2 , T2) then deg u = 0.
658
5. Cartesian Currents in Riemannian Manifolds
Proof. Recall that any two form w in T2 can be decomposed as w = 771A712,
771,2 E Dl(T2). Since the homology class of T can be represented by a smooth function by Corollary 1, and HdR(S2) = 0, we infer that T* [w] _ T*[771]AT*[772] = 0. Consequently
r#T(w) = T (F#w) = Poinc< 1, T* [w] > = 0. If now u E Cart1 (X, Y) then for 77 E D2 (Y), < (deg u)T2, [77] > = < [Gu] * S2, [r!] > = Gv, (1 '#r!) = 0.
The current T := Gc +8q x T2, c E T2, q E S2 belongs to cart(S2 x T2), but #T = T2. In particular T ¢ cart(S2 x T2).
5.4 Homology Induced Maps: Manifolds with Boundary Let X, Y be two compact, smooth oriented submanifolds respectively of dimension n, m possibly with boundary and let B C Y be a closed Lipschitz neighbor-
hood retract (relevant cases are B = 0, B = aY, B = Y). As usual we think of X and Y as submanifolds of orthogonal Euclidean spaces and we denote by 7r : X x Y -+ X, : X x Y - Y the orthogonal projections on the factors. Let T E Fp(X xY) with M(T) < oc and spt aT C aX x B, and let -y:= [T]rei be its homology class, y E Hp(X x Y, aX x B). It is easy to see that the values of
T (7r#wnF#77)
when either w E Zp-k (X), 77 E Zk (Y, B) or w E Zp-k (X, aX ), 77 E Zk (Y) depend only on the cohomology class of w and 77 and of the relative homology class -y of
T. Therefore one can define three sets of bilinear forms ryb
: HHRk(X) x HdR(Y, B) -> III
ya : HHRk(X,8X) X HHR(Y)
ay : Hdpik 1(5X) x HHR(B)
III
R
respectively by [77]rel) := T(7r#wA? 77), (1)
7p([w]re1, [1])
:=T(7r#wnF#77),
a'Y([w], [77]) :=
w E Zp-k(X), 77 E Zk(Y, B), W E Zp-k(X,ax), 1 E Zk(Y), wE
Zp-k-l(ax), r] E Zk(B).
The last one is well defined since 0T E Fp_1(aX x B) by Federer's support theorem. The maps rya, yb, ay are compatible with the long exact sequences in cohomology of the pairs (X, aX) and (Y, B); in fact we have
5.4 Homology Induced Maps: Manifolds with Boundary
659
Proposition 1. The diagram
-4 Hank-1(aX)
HaRk(X,aX)
x ay
(2)
-.
HHRk(X)
X9o j#
b
HaR(B)
HHR(Y)
HHR(Y, B)
'-b
is commutative.
Proof. Recall that the commutativity of Al
A2
xF1
xF2
B1 - B2 means F2(f (al), b2) = Fi(a,, g(b2)) `d al E A1, d b2 E B2. (i) Let w E Zp-k(X,aX), 71 E Zk(Y,B). Then j#[w]rel = [w]x, j#[?7]rei = [?7]Y, consequently yd([w]x, [7)]re1) = T(7r#WA7r#77)
_ b([w]rel, [?7]Y).
(ii) Let now w E Zp-k-1(5X) and 77 E Zc(Y). Then by definition 8[w] where w is an extension of w to X, i.e. i#w = w. Therefore 76([bw]rel, [17]Y) := T(7r#di,)n:r#77)
=
=
[dl.,)]rei
T(d(7r#wnir#77))
a7([w]ax,i#[,1]Y)
(iii) The third case in which w c Zp-k-I (aX) and 77 E Zk(B) is similar to (ii). Representing the bilinear forms ry0, ryb, 5y by means of the Poincare-Lefschetz
duality in cohomology on X, compare Theorem 3 in Sec. 5.2.8, one obtains the three sets of induced maps in cohomology y0
'yb
HdR(Y) -+ HaRp-n(X ) : HHR(Y, B) -' HaRp-' (X, 5X )
(ay)* : HHR(B)
HHR
l+p-n(5X)
defined by
:=T(7r#wn'r#7)), (3) Poincx < [w],'yb [77]rel >
T (1r#wn7r#77),
w E Zp-k(X,aX), 7) E Zk(Y) W E Z,-I(X), 77 E Zk (Y, B) W E Zp-k-l(aX),77 E Zk(B)
Poincax<[w], (5'Y)*[?7]> := aT(7r#wA7#7l),
One can show easily that the three sets of maps yd, yb, (ay)* yield actually a morphism of the two long exact sequence in cohomology of the pairs (X, ax) and (Y, B), i.e.
660
5. Cartesian Currents in R.iemannian Manifolds
Proposition 2. The diagram Hk 1+p-n (aX )
4-
T (a7)*
(4)
6
T7 b
4-
HdR(B)
4-
HdRp-n(X)
4--
HdR(Y)
is commutative. Proof. In order to prove the claim let us prove the following simple fact. Let Ai, Bi, Ci, i = 1, 2 be finite vector spaces and for i = 1, 2, let Fi : Bi x Ci --+ R be
a bilinear map, Gi : Ai x Bi -4 IR, a duality and fi : Ci --+ Ai be a representation of Fi by means of the duality Gi, i.e.
Gi(fi(ai), bi) = Fi(bi, ai)
V ai c Ai V bi e Bi.
Given now three linear maps g : Al -, A2, h : B1 -j B2 and k : C1 - C2, form the the following diagram Al
-- A2
<9
xGl
fi
xG2 hf
B1
x F2
x F1
C1
B2
+
k k
C2
We claim that if the small squares are commutative, i.e., if G1(g(a2), b1) = G2 (a2, h(bl))
bbl E B1, b a2 E A2
Fl (b1, k(c2)) = F2(h(bi), c2)
b b1 E B1, b c2 E C2
then the big square Al
Tfi
t-9
A2
TA
C1 '- C2 is commutative too, as in fact for c2 E C2 and b1 E B1, G1(f1 o k(c2), bl)
= Fi(bl, k(c2)) = F2(h(bl), c2) = G2(f2(c2), h(bl)) = G1(g o f2(c2), b1)
Putting together the diagram of the Poincare-Lefschetz duality Theorem 3 in Sec. 5.2.8 and (4) we obtain the diagram
5.4 Homology Induced Maps: Manifolds with Boundary 6
HdkR+n-p (aX)
r
Hk+n-p
-->
x Poinc
Hank-1(9X)
Hk+n-7'(X aX) I , dR
(X)
x Poinc 6
x Poinc
HHRk(X, aX)
HHRk(X)
x a-y 6
661
X yo i#
HdR(B)
--4
K
HaR(Y)
HHR(Y, B)
6
'
Since all the small squares are commutative the claim follows from the above observation.
It is also easy to show that in case y = [Gu]rei, u being a smooth function u : X -> Y with u(aX) C B, the maps [Gu]O, [Gu]b and (0[Gu])* agree with the standard induced maps in cohomology by the pullback of forms. Let us prove the claim for [Gu] b, the others being similar. Fix 77 E Zk (Y, B) and let E Zk (X, aX ) be a form representing the cohomology class of [Gu] b [r7] rel. Then
fwA
= Poincx< [w], [Gu]b [77]rel > = Gu(7r#wAF#17) =
x
#77
x
for w c Z,-k(X). Therefore 6 and u#r7 are in the same relative cohomology class, i.e. [Gu]b[7l]
= [S]n1 = [u#77]rel
Starting again with the bilinear maps in cohomology
yb, ay, one then
constructs the induced maps in homology
Hk+n-p(X, R) - Hk(Y, R) b Hk+n-p(X, aX, R) --# Hk(Y, B, R) (ay)* : Hk(OX,R) --4 Hk(B,R) yq
setting respectively
SE Zk+n-p(X) (5)
yb([S]) := T( AP.), S E Zk+n-p(X, aX) (&y) *[S] := aT( APs), s E Zk+n-p(aX)
and Ps., PP, Ps denotes respectively a Poincare dual of S in the three cases, S E Zk+n-p(X), S E Zk+,,-p(X, aX) and S E Zk+n-p(8X). One can prove, but we leave to the reader, that the couples of maps yd, yq, yb, yb and (ay)*, (ay)* are dual maps via the de Rham duality,Y = < [S], y# In] (6)
>Y, S E
Zk+n-p(X),
l. _ <[S],yb[71]>X SE
<(ay)*[s], [77]>B = <[S], (ay)*[77]>ax, S E
77
E zk(Y)
Zk+n-p(X,8X)
77 E Zk(Y,B) 77 E Zk(B).
Zk+n-p(aX),
662
5. Cartesian Currents in Riemannian Manifolds
It is also easy to show that in case 'y = [Gu]rel, u being a smooth function u : X --> Y, u(ax) C B, the maps [Gu]p, [Gu], and (a[Gu])* agree with the standard induced maps in cohomology by the push-forward of currents. Finally the maps in homology define a morphism between the long exact sequences in homology for the couples (X, ,9X) and (Y, B). In fact one proves
that Proposition 3. The diagram Hk (X, aX,R)
Hk(X,R) (7)
Hk(Y,lf2)
Hk-1(aX, R) (a Y)+
yb
.Lyp
2#,
a
A#.
Hk(Y,B,R)
Hk-1(B,R)
-a->
is commutative.
Finally starting with an integral cycle T E Zp (X x Y) with spt T C aX x B one sees that the induced real homology maps To, TT, (aT), extend to maps on the integral homology. Such extensions can be found as in the boundaryless case in terms of intersection of cycles in X x Y. In order to do that, recall the notations of Sec. 5.1.2 and particularly the extension diffeomorphisms e:X-+XEO,
e:Y-+YEO,
60>0
which are homotopic respectively to the identity on X and Y, e(aX) = aXEO, e(OY) = aYEO, and their inverses
r:XEO -->X
r:YEO-+Y.
Let S E Zk+n,_p (X, 7G) be a rectifiable cycle. Evidently r# S is in the same homology class of S in Hk+n_p(X,Z) and sptr#S n OX = 0. Given now an integral relative cycle T E Zp(X,B,Z) one can apply (iii) of Proposition 2 in Sec. 5.4.2 to the currents T E Zp(X x Y,Z),
r#S x e#QY E
X YE,,,X X OYEO,Z)
and find that for a.e. a close to zero, T n (r#S x e# Q Y I),, is an integral current with no boundary in Zk (X x Y, 0, Z) whose integral homology class depends only on the integral homology classes of S and T. In this way one defines [T]p : Hk+,,-p(X, Z) - Hk(Y, Z)
setting (8)
[T]p[S]z
[F#(T n (r#S X e#QUDa)]z.
Notice that (ii) of Proposition 2 in Sec. 5.4.2 shows that such a definition is compatible with the definition of T. on real homology. Similarly one checks that for [S] E Hk+n,_p_ 1(aX, 7G),
6 Notes
(a[T]).[S]z
663
[7r# (T n (S x e#QY
for a.e. a close to zero yields a boundary homology map
(a [T])* : Hk+n-p(aX,Z) --+ Hk(B,Z) which is compatible with the real one. Finally, if S E Zk+,,_p (X, aX, Z) is a relative normal cycle, then one can apply Proposition 3 in Sec. 5.4.2 to the currents T E Zp (X x Y, aX x B, Z), e#S x e#QY E Zk+n+m-p(XEO X YO,a(XEO X YEO),Z)
and obtain for a.e. a close to zero a relative cycle T n (e#S x F# Q Y ]I) E Zk (X x Y, aX x B, Z)
whose homology class depends only on the homology classes of S and T. In this way one has a homology induced map [T] y : Hk+n_p (X, aX, Z) --> Hk (Y, B, Z) setting [T]b[S]r.1 := [7#(Tn (e#S x for a.e. a. Using (ii) of Proposition 3 in Sec. 5.4.2 one checks that such a definition is compatible with the corresponding map in real homology. Finally it is easy to
check that the diagram
Hk(X,Z) -+ Hk(X,3X,Z) a' Hk-l(aX,Z) i#+... 1 [T]a
(9)
Hk(Y,Z)-#
1 [T],
Hk(Y,B,Z)
-a
1 (3[T]),,
Hk-i(B,Z)
...
is commutative.
6 Notes 1 The proof of the deformation theorem in Sec. 5.1.1 follows quite closely Simon [592], compare also Federer and Fleming [230], Federer [226]; for improved versions with applications to Q-valued functions see Almgren [21]. An application of the same ideas to the problem of non parametric area functional can be found in Mucci [498]. Sec. 5.1.2 and Sec. 5.1.3 follow closely Federer [228] and [226].
2 Hodge decomposition theorems are a classical subject and go back to Hodge [382] [381], De Rham [189] [188], Weyl [661], Kodaira [414], Friedrichs [247], Morrey [490] [491]. We followed the variational approach of Morrey [490]. For further information on this topic the reader may consult e.g. Schwarz [582] and its bibliography.
Weizenbock formula in Sec. 5.2.7 gives a link between bounds on curvature and non existence of harmonic forms, this way producing global vanishing theorems in cohomology, see e. g. Bochner [95] [96] and Goldberg [322].
5. Cartesian Currents in Riemannian Manifolds
664
Hodge theorem has been probably the first example of the relations between analysis and topology, another one is describing homology in terms of currents. It is out of question to quote even a few more contributions in the development of the interplay between linear and nonlinear analysis, topology and geometry. This includes among others invariant forms, K-theory, Atiyah-Singer index, Bott periodicity theorem, differential topology, Yang-Mills equations, topology of lower dimensional manifolds, Floer cohomology etc.. Probably these relations are one of the most fruitful research field in the mathematics of this century. The reader may start e.g. with Gilkey [303], Freed and Uhlenbeck [246] and their bibliographies. 3 The approach to homology evolved into a purely algebraic theory based on a set of axioms, the Eilenberg-Steenrood axioms, which characterizes the homology groups for triangulable pairs of spaces. A key point in this approach, even from a computational point of view, is the Meier-Vietoris sequence. Poincare and Poincare-Lefschetz theorems on the unimodularity of the intersection index can be proved in this way, compare e.g. Bott and Tu [102] Dold [195] Eilenberg and Steenrood [206] Greenberg [325] Griffits and Harris [326] Spanier [607] Vick [645].
We instead preferred to concentrate on a more analytic point of view to the real homology and cohomology, stressing the duality forms-currents at work, in the spirit of De Rham [189], Whitney [672], Federer [226] and [228]. Thus we define the homology as classes of currents and the Poincare and Poincare-Lefschetz duality isomorphism for the real homology are proved by mollifying currents and using Hodge theory. An obvious consequence is then de Rham theorem. Another analytic-combinatoric proof of de Rham's theorem, at least for compact manifolds without boundary, can be found in Whitney [672], see also Singer and Thorpe [600].
de Rham theorem implies in particular the weak closure of the homology classes. An alternative proof which also gives a representation of homology clases by Lipschitz chains is given in Sec. 5.3.2 following Federer and Fleming [230], Federer [226]. For the integral homology we confined ourselves to represent integral homology
classes in terms of currents, and in fact by least mass integral currents, and discuss following Federer [228] the relationships between real and integral cycles. This way we proved that the intersection index is an integer, however the unimodularity of the intersection index apparently seems to require the algebraic theory. 4 There are several results concerning least area representatives of homology classes, see also Vol. II Ch. 6. General partial regularity results has been proved by Almgren [18]. For a brief discussion of properties of optimal representatives of integral homology classes we refer the reader to Almgren [23].
There has been quite some work in finding explicit least area representatives of homology classes. In this context the calibration theory turns to be useful. For a short survey the reader may consult e.g. Morgan [484]. 5 As general references for Sobolev spaces on 1[8" or on Riemannian manifolds the reader may consult e.g. Adams [4], Maz'ja [463] and Hebey [364]. Theorem 3 in Sec. 5.5.1 was proved in Schoen and Uhlenbeck [577]. For the proofs of Theorem 4 in Sec. 5.5.1 to Theorem 8 in Sec. 5.5.1 we refer the reader to Bethuel and Zheng [94], Bethuel [87] Bethuel [86], Helein [369]. When CC°(X, y) is not dense in W1'P(X, y), i.e. when dimX > p and 7r[ ](X) # 0, the strong closure of Cf°(X,y) in W1'P(X, y) has been studied in Bethuel [86] Bethuel, Coron, Demengel, and Helein [91] Demengel [191] Giaquinta, Modica, and Soucek [295], Isobe [393], Zhou [685], and in particular in Isobe [395].
As already mentioned for the proofs of Theorem 9 in Sec. 5.5.1, Theorem 11 in Sec. 5.5.1 and Theorem 12 in Sec. 5.5.1 we refer respectively to White [667] and White [668], and for the proof of Theorem 12 in Sec. 5.5.1 to Baldo and Orlandi [61].
6 Notes
665
6 The results of Sec. 5.5.2 are simple extensions of similar results proved in previous chapters. The discussion of the homology maps associated to Cartesian currents in the ;enerality of Sec. 5.5.3 and Sec. 5.5.4 should be considered known to experts though we have no specific references. The proof of Kenneth formula is taken from Griffits and Harris [326].
7 Spaces of currents:
{k-dimensional currents in 0}
Dk ((2) Ek(f2)
{T E Dk(Q) I sptT C .f2} {real polyhedral chains} Pk(.f2) {T E Dk(Q) I M(T) Goo} Mk(S2) {T E Dk(.f2) I M(T) +M(8T) < oo} Nk(.f2) {real flat chains} Fk((2) Pk (0) = Pk (.f2, Z) := {integral polyhedral chains} Rk(() {rectifiable currents with integer multiplicity } {T E Rk(92) 18T E Rk-1(,f2)} Zk(,fl) -rk(0) {integral flat chains} Pk
n Pk
Rk C .Pk n C Fk C Nk C
C
Zk
C
n
C Dk
Bibliography
[1]
[2]
[3]
[4] [5] [6]
ACERBI, E. and DAL MASO, G.: New lower semicontinuity results for polyconvex
integrals. Calc. Var., 2, 329-371, 1994. ACERBI, E. and Fusco, N.: Semicontinuity problems in the calculus of variations. Arch. Rat. Mech. Anal., 86, 125-145, 1986. ACERBI, E. and Fusco, N.: An approximate lemma for W" functions. In Matereal instability in continuum mechanics, edited by Ball, J. M. Oxford University Press, Oxford, 1988. ADAMS, R.: Sobolev spaces. Academic Press, New York, 1975. ALBERTI, G.: Proprietd fini delle funzioni a variazione limitata. Tesi di laurea, 1988, University di Pisa. ALBERTI, G.: A Lusin type theorem for gradients. J. Punct. Anal., 100, 110-118, 1991.
[7]
ALBERTI, G.: Rank one property for derivatives of functions with bounded variation. Proc. Roy. Soc. Edinburg, 123, 239-274, 1993.
[8]
ALBERTI, G., AMBROSIO, L., and CANNARSA, P.: On the singularities of convex
[9]
functions. manuscripta math., 76, 421-435, 1992. ALBERTI, G. and MANTEGAZZA, C.: A note on the theory of SBV functions. preprint, 1996.
[10] [11]
[12] [13] [14] [15] [16]
[17]
[18]
[19]
[20]
ALEXANDROFF, P. and HOPF, H.: Topologie. Springer, Berlin, 1935. ALIBERT, J. J. and DACOROGNA, B.: An example of a quasiconvex function not polyconvex in dimension two. Arch. Rat. Mech. Anal., 117, 155-166, 1992. ALLARD, W. K.: First variation of a varifold. Ann. of Math., 95, 417-491, 1972.
ALLARD, W. K.: First variation of a varifold, boundary behaviour. Ann. of Math., 95, 417-491, 1975. ALMGREN, F. J.: The theory of varifolds. Mimeographed notes, Princeton, 1965. ALMGREN, F. J.: Plateau's problem. An invitation to varifold geometry. Benjamin, Inc., New York, 1966. ALMGREN, F. J.: Some interior regularity theorems for minimal surfaces and an extension of Bernstein's theorem. Ann. of Math., 84, 277-292, 1966. ALMGREN, F. J.: Existence and regularity almost everywhere of solutions to elliptic variational problems among surfaces of varying topological type and singuarity structure. Ann. of Math., 87, 321-391, 1968. ALMGREN, F. J.: Q valued functions minimizing Dirichlet's integral and the regularity of area minimizing rectifiable currents up to codimension two. Buli.AMS, 8, 327-328, 1983. ALMGREN, F. J.: Q valued functions minimizing Dirichlet's integral and the regularity of area minimizing rectifiable currents up to codimension two, 1984. Notes, Princeton, NJ. ALMGREN, F. J.: Optimal isoperimetric inequalities. Bull. Amer. Math. Soc., 13, 123-126, 1985.
668
[21]
[22]
BIBLIOGRAPHY
ALMGREN, F. J.: Deformations and multiple-valued functions. In Geometric measure theory and the calculus of variations, Proc_ Sympos. Pure Math., 44,, pp. 29-130. Amer. Math. Soc., Providence, R.I., 1986. ALMGREN, F. J.: Optimal isoperimetric inequalities. Indiana Univ. Math. J., 35, 451-547, 1986.
[23]
ALMGREN, F. J.: Questions and answers about area minimizing surfaces and geometric measure theory. In Differential geometry: partial differential equations
on manifolds, Proc. Sympos. Pure Math., 54, Part 1, pp. 29-53. Amer. Math. Soc., Providence, RI, 1993. [24]
[25]
[26]
[27] [28]
ALMGREN, F. J., BROWDER, W., and LIES, E. H.: Co-area, liquid crystals, and minimal surfaces. In Partial differential equations, Lecture Notes in Math., 1306,, pp. 1-22. Springer, Berlin, 1988. ALMGREN, F. J. and LIEB, E. H.: Singularities of energy-minimizing maps from the ball to the sphere. Bull. Amer. Math. Soc., 17, 304-306, 1987. ALMGREN, F. J. and LIEB, E. H.: Singularities of energy minimizing maps from the ball to the sphere: examples, counterexamples, and bounds. Ann. of Math., 128, 483-530, 1988. ALMGREN, F. J. and SIMON, L.: Existence of embedded solutions of Plateau's problem. Ann. Sc. Norm. Sup. Pisa, 6, 447-495, 1979. ALMGREN, F. J. and THURSTON, W. P.: Examples of unknotted curves which bound only surfaces of high genus within their convex hulls. Ann. Math., 105, 527-538, 1977.
[29]
ALT, W.: Verzweigungspunkte von H-Flachen. Part I. Math. Z., 127, 333-362,
[30]
ALT, W.: Verzweigungspunkte von H-Flachen. Part II. Math. Ann., 201, 33-55,
[31]
AMBROSIO, L.: New lower semicontinuity results for integral functionals. Rend. Accad. Naz. XL, 11, 1-42, 1987. AMBROSIO, L.: A compactness theorem for a special class of functions of bounded variation. Boll. UMI (3-B), 7, 857-881, 1989. AMBROSIO, L.: Variational problems in SBV. Acta Applicandae Math., 17, 1-40,
1972. 1973.
[32] [33]
1989. [34]
AMBROSIO, L.: Corso introduttivo alla teoria delle funzioni By, le correnti ed i varifolds. Unpublished, see also: Corso introduttivo alla Teoria geometraca della misura e alle superfici minime. Scuola Normale Superiore, Pisa, 1997.
[35] [36] [37] [38] [39]
AMBROSIO, L.: On the lower semicontinuity of quasi-convex integrals in SBV. Nonlinear Anal., 23, 405-425, 1994. AMBROSIO, L.: A new proof of the SBV compactness theorem. Calc. Var., 3, 127-137, 1995. AMBROSIO, L. and DAL MASO, G.: A general chain rule for distributional derivatives. Proceedings Am. Math. Soc., 108, 691-702, 1990. AMBROSIO, L. and DAL MASO, G.: On the representation in BV (.f2, Rtm) of quasi-convex integrals. J. Funct. Anal., 109, 76-97, 1992. AMBROSIO, L., MORTOLA, S., and TORTORELLI, V. M.: Functionals with linear
growth defined on vector valued BV functions. J. Math. Pures Appl., 70, 269[40]
[41]
323, 1991. AMRANI, A., CASTAING, C., and VALADIER, M.: Methodes de troncature ap-
pliquees a des problemes de convergence faible ou forte dans Ll. Arch. Rat. Mech. Anal., 117, 167-191, 1992. ANTMAN, S.: Geometrical and analytical questions in nonlinear elasticity. In Seminar on Nonlinear Partial Differential Equations, edited by S.S.Chern. Springer, Berlin, 1984.
BIBLIOGRAPHY [42] [43] [44] [45] [46]
669
ANTMAN, S. S.: Nonlinear problems of elasticity. Springer, New York, 1995. ANZELLOTTI, G.: Dirichlet problem and removable singularities for functionals with linear growth. Boll. UMI(C), 18, 141-159, 1981. ANZELLOTTI, G.: Parametric and non-parametric minima. manuscripta math., 48, 103-115, 1984. ANZELLOTTI, G. and GIAQUINTA, M.: Funzioni BV e tracce. Rend. Sem. Mat. Univ. Padova, 60, 1-21, 1978. ANZELLOTTI, G. and GIAQUINTA, M.: Existence of the displacements field for
an elasto-plastic body subject to Hencky's law and von Mises yield condition. manuscripta math., 32, 101-136, 1980. [47]
ANZELLOTTI, G. and GIAQUINTA, M.: On the existence of the fields of stresses
and displacements for an elasto-perfectly plastic body in static equilibrium. J. Math. Pures et Appl., 61, 219-244, 1982. [48]
ANZELLOTTI, G. and GIAQUINTA, M.: Convex functionals and partial regularity.
[49]
ANZELLOTTI, G., GIAQUINTA, M., MASSARI, U., MODICA, G., and PEPE, L.:
Arch. Rat. Mech. Anal., 102, 243-272, 1988. Note sul problems di Plateau. Editrice Tecnico Scientifica, Pisa, 1974. [50] [51]
[52]
[53] [54] [55] [56] [57] [58] [59] [60]
[61] [62]
[63] [64] [65] [66]
ANZELLOTTI, G. and SERAPIONI, R.: Ck-rectifiable sets. J. Reine Angew. Math.,
453, 1-20, 1994. ANZELLOTTI, G., SERAPIONI, R., and TAMANINI, I.: Curvatures, Functionals, Currents. Indiana Univ. Math. J., 39, 617-669, 1990. ARONSZAJN, N.: A unique continuation theorem for solutions of elliptic partial differential equations or inequalities of second order. J. Math. Pures et Appl., 36, 235-249, 1957. AUBIN, T.: Equations differentielles non lineaires et probleme de Yamabe concernent la courbure scalaire. J. Math. Pures Appl., 55, 269-296, 1976. AUBIN, T.: Nonlinear analysis on manifolds. Monge-Ampere equations. Grundleheren math. Wiss. 252. Springer, Berlin, 1982. AvILEZ, P. and GIGA, Y.: Variational integrals on mappings of bounded variation and their lower semicontinuity. preprint, ???? BAKEL'MAN, I. Y.: Mean curature and quasilinear elliptic equations. Sibirskii Mat. 2, 9, 752-771, 1968. (English transl. Siberian Math. J. 9 (1968) 752-771). BAKEL'MAN, I. Y.: Geometric problems in quasilinear elliptic equations. Uspehi Mat. Nauk, 25, 49-112, 1970. BALDER, E. J.: A general approach to lower semicontinuity and lower closure theorems in optimal control theory. SIAM J. Cont. Optim., 22, 570-598, 1984. BALDER, E. J.: On weak convergence implying strong convergence in L' spaces. Bull. Austral. Math. Soc., 33, 363-368, 1986. BALDES, A.: Stability and uniqueness properties of the equator map from a ball into an ellipsoid. Math. Z., 185, 505-516, 1984. BALDO, S. and ORLANDI, G.: Homotopy types for tamely approximable maps between manifolds. Calc. Var., 4, 369-384, 1996. BALL, J. M.: Constitutive inequalities and existence theorems in nonlinear elastostatics. In Nonlinear Analysis and Mechanics: Heraot-Watt Symposium, Vol I. Pitman Research Notes in Math. 17, London, 1977. BALL, J. M.: Convexity conditions and existence theorems in nonlinear elasticity. Arch. Rat. Mech. Anal., 63, 337-403, 1977. BALL, J. M.: Global invertibility of Sobolev functions and the interpenetration of matter. Proc. Roy. Soc. Edinburgh, 88A, 315-328, 1981. BALL, J. M.: Discontinuous equilibrium solutions and cavitation in nonlinear elasticity. Philos. Trans. Roy. Soc., A 306, 557-611, 1982. BALL, J. M.: Differentiability properties of symmetric and isotropic functions. Duke Math. J., 50, 699-727, 1984.
670
[67]
BIBLIOGRAPHY
BALL, J. M.: On the paper "Basic calculus of variations". Pacific J. Math., 116, 7-10, 1985.
[68]
[69]
[70]
BALL, J. M.: Does rank one convexity imply quasiconvexity? In Metastability and Incompletely Posed Problems, edited by S. Antman, J.L.Ericksen, D. Kinerleherer and I. Muller, pp. 17-32. Springer, New York, 1987. BALL, J. M.: A version of the fundamental theorem for Young measures. In PDEs and continuum models of phase transitions, Lecture Notes in Phys., 344,, pp. 207-215. Springer, Berlin-New York, 1989. BALL, J. M., CURRIE, J. C., and OLIVER, P. J.: Null Lagrangians, weak conti-
nuity and variational problems of arbitrary order. J. Funct. Anal., pp. 135-174, 1981. [71]
[72]
[73] [74]
[75]
[76]
BALL, J. M. and MURAT, F.: WI'p-quasiconvexity and variational problems for multiple integrals. J. Funct. Anal., 58, 225-253, 1988. Erratum, J. Funct. Anal. 66 1986, 439. BALL, J. M. and MURAT, F.: Remarks on Chacon's biting lemma. Proc. Amer. Math. Soc., 107, 655-663, 1989. BANACH, S.: ThMorie des operations lineaires. Monografje Matematyczne, Warsawa, 1932. Republished by Chelsea, New York 1978. BAUMAN, P., OWEN, N., and PHILLIPS, D.: Maximal smoothness of solutions to certain Euler-Lagrange equations from nonlinear elasticity. Proc. R. Soc. Edinb., 119, 241-263, 1991. BAUMAN, P., OWEN, N., and PHILLIPS, D.: Maximum principles and a priori estimates for a class of problems from nonlinear elasticity. Commun. Partial Differential Equations, 17, 1185-1212, 1992. BENCI, V. and CORON, J. M.: The Dirichlet problem for harmonic maps from
the disk into the Euclidean n-sphere. Ann. I.H.P., Analyse Non Lineaire, 2, 1985, 1985. [77] [78] [79] [80] [81]
[82]
BENNETT, C. and SHARPLEY, R.: Weak-type inequalities for Hp and BMO. Proc. Symposia Pure Math., 35, 201-229, 1979. BERKOWITZ, L. D.: Lower semicontinuity of integral functionals. Trans. Am. Math. Soc., 192, 51-57, 1974. BERNSTEIN, S.: Sur les surfaces definie au moyen de leur courbure moyenne et totale. Ann. Ecole Norm. Sup., 27, 233-256, 1910. BERNSTEIN, S.: Sur les equations du calcul des variations. Ann. Sci. Ecole Norm. Sup., 29, 431-485, 1912. BERNSTEIN, S.: Sur un theoreme de geometrie et ses applications aux equations aux derivees partielles du type elliptique. Comm. Soc. Math. Kharkov, 15, 38-45, 1915-1917. (German transl.:Uber ein geometrisches Theorem and seine Anwendung auf die partiellen Differentialgleichungen von elliptischen Typus, Math. Z. 26 (1927) 551-558). BERS, L.: Isolated singularities of minimal surfaces. Ann. of Math., 53, 364-386, 1951.
[83] [84] [85]
BESICOVITCH, A.: On the fundamental geometric properties of linearly measurable plane sets of points III. Math. Ann., 116, 349-357, 1939. BESICOVITCH, A.: On the definition and value of the area of a surface. Quart. J. Math., 16, 86-102, 1945. BESICOVITCH, A.: Parametric surfaces. Bull. Amer. Math. Soc., 56, 288-296, 1950.
[86]
BETHUEL, F.: A characterization of maps in Hl'2(B3, S2) which can be approximated by smooth maps. Ann. Inst. H. Poincare, Anal. Non Line'aire, 7, 269-286,
[87]
BETHUEL, F.: The approximation problem for Sobolev maps between two manifolds. Acta Math., 167, 153-206, 1992.
1990.
BIBLIOGRAPHY [88]
671
BETHUEL, F.: On the singular set of stationary harmonic maps. manuscripta math., 78, 417-443, 1993.
[89]
BETHUEL, F. and BREZIS, H.: Regularity of minimizers of relaxed problems for
[90]
harmonic maps. J. Funct. Anal., 101, 145-161, 1991. BETHUEL, F., BREZIS, H., and CORON, J. M.: Relaxed energies for harmonic maps. In Variational methods, edited by Berestycki, H., Coron, J., and Ekeland, J. Birkhauser, Basel, 1990.
[91]
[92] [93]
[94] [95]
BETHUEL, F., CORON, J. M., DEMENGEL, F., and HELEIN, F.: A cohomological criterion for density of smooth maps in Sobolev spaces between two manifolds. In Nematics, Mathematical and Physical Aspects, edited by Coron, J. M., Ghidaglia,
J. M., and Helein, F., NATO ASI Series C, 332, pp. 15-23. Kluwer Academic Publishers, Dordrecht, 1991. BETHUEL, F. and DEMENGEL, F.: Extensions for Sobolev mappings between manifolds. Calc. Var., 3, 475-491, 1995. BETHUEL, F., MODICA, G., STEFFEN, K., and WHITE, B.: Lectures notes on geometric measure theory and geometrical variational problems. Dipartimento di Matematica, University di Trento, 1995. BETHUEL, F. and ZHENG, X. M.: Density of smooth functions between two manifolds in Sobolev spaces. J. Funct. Anal., 80, 60-75, 1988. BOCHNER, S.: Curvature and Betti numbers I. Ann. of Math., 49, 379-390, 1948.
BOCHNER, S.: Curvature and Betti numbers II. Ann. of Math., 50, 77-93, 1949. BOMBIERI, E.: Regularity theory for almost minimal currents. Arch. Rat. Mech. Anal., 78, 99-130, 1982. BOMBIERI, E., DE GIORGI, E., and GIUSTI, E.: Minimal cones and the Bernstein [98] problem. Inv. Math., 7, 243-268, 1969. BOMBIERI, E., DE GIORGI, E., and MIRANDA, M.: Una maggiorazione a priori [99] relatives alle ipersuperfici minimali non parametriche. Arch. Rat. Mech. Anal., 32, 255-267, 1965. [100] BOMBIERI (ED.), E.: Seminar on minimal submanifolds. Ann. of Math. Studies n. 103. Princeton Univ. Press, Princeton, 1983. [101] BONNESEN, T. and FENCHEL, W.: Theorie der konvexen Korper. Ergebnisse der Math. u. ihrer Grenzgebiete 3. Springer, Berlin, 1934. [102] BOTT, R. and TU, L. W.: Differential forms in Algebraic Topology. SpringerVerlag, New York, 1982. [103] BREZIS, H.: Large harmonic maps in two dimensions. In Nonlinear variational problems, Res. Notes in Math., 127, pp. 33-46. Pitman, Boston, Mass.-London, [96] [97]
1985. [104]
BREZIS, H.: Sk-valued maps with singularities. In Topics in calculus of Variations, edited by Giaquinta, M., Lecture notes 1365. Springer-Verlag, Berlin, 1989.
[105]
BREZIS, H.: Convergence in D' and in Ll under strict convexity. In Boundary values problems for PDE and applications. Dedicated to E. Magenes on the occasion of his 70th birthday, edited by Lions, J. L., pp. 43-52. Masson, Paris, 1993.
[106] [107] [108] [109]
BREZIS, H. and CORON, J. M.: Large solutions for harmonic maps in two dimensions. Commun. Math. Phys., 92, 203-215, 1983. BREZIS, H. and CORON, J. M.: Multiple solutions of H-systems and Rellich's conjecture. Comm. Pure Appl. Math., 37, 149-187, 1984. BREZIS, H. and CORON, J. M.: Convergence of solutions of H-sistems or how to blow bubbles. Arch. Rat. Mech. Anal., 89, 21-56, 1985. BREZIS, H., CORON, J. M., and LIES, E. H.: Harmonic maps with defects. Comm. Math. Phys., 107, 649-705, 1986.
672
BIBLIOGRAPHY
[110] BREZIS, H. and LIES, E.: A relation between pointwise convergence of functions
and convergence of functionals. Proc. Amer. Math. Soc., 88, 486-490, 1983. [111] BREZIS, H. and NIRENBERG, L.: Positive solutions of nonlinear elliptic equations
involving critical Sobolev exponents. Comm. Pure Appl. Math., 36, 437-477, 1983. [112] BREZIS, H. and NIRENBERG, L.: Degree theory and BMO I. Compact manifolds
without boundaries. Selecta Math., 1, 197-263, 1995. [113] BREZIS, H. and NIRENBERG, L.: Degree theory and BMO Part II. Compact manifolds with boundaries. preprint, 1997. [114] BROOKS, J. and CHACON, R.: Continuity and compactness of measures. Advances in Math., 37, 16-26, 1980. [115] BROTHERS (ED.), J.: Some open problems in geometric measure theory and its applications suggested by partecipants of the 1984 AMS Summer Institute. Proc. Symp. Pure Math., 44, 441-464, 1986. [116] BURAGO, J. D. and MAZ'JA, V. G.: Some questions of potential theory and function theory for domains with non-regular boundaries. Zap. Nauchn. Sem. Leningr. Mat. Inst. Steklova, 3, 1-152, 1967. In Russian. English transl.: Sem. Math. Steklov Inst., Leningrad 3 (1969) 1-68. [117] BURAGO, Y. D. and ZALGALLER, V. A.: Geometric inequalities. Grundlehren math. Wiss. 285. Springer-Verlag, Berlin, 1988. [118] BURSTAL, F., LEMAIRE, L., and RAWNSLEY, J.: Harmonic maps bibliography.
http://www.bath. ac.uk/-masfeb/harmonic.html. [119] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones V. Arch. Math., 13, 512-526, 1962. [120] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones I-IV. Math. Ann., 151, 1-41, 1963. [121] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones IX. Math. Ann., 157, 219-230, 1964. [122] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones VI. J. London Math. Soc., 39, 307-319, 1964. [123] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones VII. Abh. Math. Sem. Univ. Hamburg, 27, 167-170, 1964.
[124] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Convexity on Grassmann Cones VIII. J. London Math. Soc., 39, 417-423, 1964. [125] BUSEMANN, H., EWALD, G., and SHEPHARD, G. C.: Convex bodies and Con-
vexity on Grassmann Cones X. Ann. Math. Pura e Appl., 70, 271-294, 1965. [126] BUSEMANN, H. and SHEPARD, G.: Convexity on nonconvex sets. In Proc. Coll. on Convexity, pp. 20-33. Copenhagen, 1965. [127] BUTTAZZO, G.: Semicontinuity, Relaxation and Integral Representation Prob-
lems in the Calculus of Variations. Pitman Res. Notes in Math. 207. Longman, Harlow, 1989.
[128] CACCIOPPOLI, R.: Sul carattere infinitesimale delle superfici quadrabili. Atti R. Accad. Lincei, 7, 901-905, 1928. [129] CACCIOPPOLI, R.: Misura e integrazione sugli insiemi dimensionalmente orientati. Rend. Acad. Naz. Lincei, 12, 3-11, 137-146, 1952. [130] CACCIOPPOLI, R. and ScoRZA DRAGONI, G.: Necessity della condizione di
Weierstrass per la semicontinuity di un integrale doppio sopra una data superficie. Memorie Acc. d'Italia, 9, 251-268, 1938. [131] CALDER6N, A. P. and ZYGMUND, A.: On the differentiability of functions which
are of bounded variation in Tonelli's sense. Revista Union Mat. Arg., 20, 102121, 1960.
BIBLIOGRAPHY
673
CALDERON, A. P. and ZYGMUND, A.: Local properties of solutions of elliptic partial differential equations. Studia Mat., 20, 171-225, 1961. [133] CALKIN, J. W.: Functions of several variables and absolute continuity I. Duke math. J., 6, 170-185, 1940. [134] CARATHEODORY, C.: Uber das lineare Mass von Punktmengen, eine Verallgemeinerung des Langenbegriffs. In Nach. Ges. Wiss. , pp. 404-426. Gottingen, [132]
1914.
[135] CARATHEODORY, C.: Vorlesungen fiber reelle Functionen.
Teubner, Leipzig,
1927.
[136] CARBOU, G.: Applications harmoniques a valeurs dans un cercle. C. R. Acad. Sci. Paris, 314, 359-362, 1992. [137] CASADIO TARABUSI, E.: An algebraic characterization of quasiconvex functions. [138] [139]
Ricerche di Mat., 42, 11-24, 1993. CASTAING, C. and VALADIER, M.: Convex analysis and measurable multifunctions. Lecture Notes in Math. 580. Springer, Berlin, 1977. CELADA, P. and DAL MASO, G.: Further remarks on the lower semicontinuity of polyconvex integrals. Ann. Institut H. Poincare Anal. Non Lineaire, 11, 661691, 1994.
[140]
CESARI, L.: Sulle funzioni a variazione limitata. Ann. Sc. Norm. Sup. Pisa, 5, 299-313, 1936.
[141] [142] [143]
CESARI, L.: Sulle funzioni assolutamente continue in due variabili. Ann. Sc. Norm. Pisa, 10, 91-101, 1941. CESARI, L.: Surface area. Annals of Math. Studies 35. Princeton Univ. Press, Princeton, 1956. CESARI, L.: Lower semicontinuity and lower closure theorems without seminormality condition. Ann. Mat. Pura e Appl., 98, 381-397, 1974.
[144] CHANDRESEKHAR, S.: Liquid crystals. Cambridge Univ. Press, Cambridge, 1977.
[145] CHANG, S. X.-D.: Two dimensional area minimizing integral currents are classical minimal surfaces. J. Amer. Math. Soc., 1, 699-778, 1988. [146] CHANILLO, A.: Sobolev inequalities involving divergence free maps. Comm. PDE, 16, 1969-1994, 1991. [147] CHANILLO, A. and Li, A.: Continuity of solutions of uniformly elliptic equations in R2. manuscripta math., 77, 415-433, 1992. [148] CHEN, N. K. and LIU, F. C.: Approximation of nonparametric surfaces of finite area. Chinese J. Math., 1981, 25-35, 1981. [149] CHEN, Y. and LIN, F. H.: Remarks on approximate harmonic maps. Comment. Math. Helvetici, 70, 161-169, 1995. [150] CHERN, S. S., ed.: Global differential geometry. The Mathematical Association [151] [152]
[153] [154] [155]
of America, 1989. CIARLET, P.: Mathematical elasticity. North Holland, Amsterdam, 1988.
CIARLET, P. and NECAS, J.: Injectivite presque partout, auto contact et non interpenetrabilite en elasticity non-lineaire tridimensionelle. C. R. Acad. Sci. Parts, 301, 621-624, 1985. CIARLET, P. and NECAS, J.: Unilateral problems in nonlinear three-dimensional elasticity. Arch. Rat. Mech. Anal., 87, 319-338, 1985. CIARLET, P. and NECAS, J.: Injectivity and selfcontact in nonlinear elasticity. Arch. Rat. Mech. Anal., 97, 171-188, 1987. COIFMAN, R., LIONS, P. L., MEYER, Y., and SEMMES, S.: Compacity par com-
pensation et espaces de Hardy. C. R. Acad. Sc. Paris, 311, 519-524, 1990. [156]
COIFMAN, R., LIONS, P. L., MEYER, Y., and SEMMES, S.: Compensated com-
pactness and Hardy spaces. J. Math. Pures Appl., 72, 247-286, 1993. [157] CORON, J. and GULLIVER, R.: Minimizing p-harmonic maps into spheres. J. reine angew. Math., 401, 82-100, 1989.
674
BIBLIOGRAPHY
[158] CORON, J. M., GHIDAGLIA, J. M., and HELEIN (ED.), F.: Nernatics. NATO series vol. 332. Kluwer academic Publ., Dodrecht, 1991. [159] COURANT, R.: On the problem of Plateau. Proc. Nat. Acad. Sci. USA, 22, 367-372, 1936.
[160] COURANT, R.: Plateau's problem and Dirichlet's Principle. Ann. Math., 38, 679-724, 1937.
[161] COURANT, R.: Dirichlet's principle, conformal mapping, and minimal surfaces. Interscience, New York, 1950. [162] COVENTRY, A.: Cartesian currents and weak convergence phenomena. Australian National University, BSc Honours Thesis, 1994. [163] CRONIN, J.: Fixed points and topological degree in nonlinear analysis. Amer. Math. Soc., Providence, 1964. [164] DACOROGNA, B.: Weak continuity and weak lower semicontinuity of non-linear functionals. Springer-Verlag, Berlin, 1982. [165] DACOROGNA, B.: Remarques sur les notions de polyconvexite, quasi-convexite et convexite de rang 1. J. Math. Pures Appl., 64, 403-438, 1985. [166] DACOROGNA, B.: Direct methods in the calculus of variations. Springer-Verlag, Berlin, 1989. [167] DACOROGNA, B. and MARCELLINI, P.: A counterexample in the vectorial calculus of variations. In Material Instabilities in Continuum Mechanics, edited by Ball, J. M., pp. 77-83. Clarendon Press, Oxford, 1988. [168] DACOROGNA, B. and MARCELLINI, P.: Semicontinuite pour des integrales polyconvexes sans continuity des determinants. C.R.A.S. Pans, 311, 393-396, 1990. [169] DACOROGNA, B. and MOSER, J.: On a partial differential equation involving the
Jacobian determinant. Ann. Institut H. Poincare Anal. Non Lineaire, 7, 1-26, 1990.
[170] DACOROGNA, B. and MURAT, F.: On the optimality of certain Sobolev exponents
for the weak continuity of determinants. J. Funct_ Anal., 105, 42-62, 1992. [171] DAL MASO, G.: Integral representation on BV(Q) of F-limits of variational integrals. manuscripta math., 30, 387-416, 1980. [172] DAL MASO, G.: An Introduction to T-convergence. Birkhauser, Boston, 1993. [173] DAL MASO, G. and SBORDONE, C.: Weak lower semicontinuity of polyconvex integrals: a borderline case. Math. Z., 218, 603-609, 1995. [174] DARST, R.: The Hausdorff dimension of the nondifferentiability set of the Cantor function is (log 2/ log 3)2. Proc. Am. Math. Soc., 119, 105-108, 1993. [175] DE GENNES, P.: The physics of liquid crystals. Oxford Univ. Press, Oxford, 1974.
[176] DE GIORGI, E.: Su una teoria generale dells misura (r - 1)-dimensionale in uno spazio ad r dimensioni. Ann. Mat. Pura Appl., 36, 191-213, 1954. [177] DE GIORGI, E.: Nuovi teoremi relativi alle misure (r - 1)-dimensionali in uno spazio ad r dimensioni. Ricerche di Matematica, 4, 95-113, 1955. [178] DE GIORGI, E.: Sulla differenziabilita e l'analiticita delle estremali degli integrali multipli regolari. Mem. Accad. Sci. Torino, 3, 25-43, 1957. [179] DE GIORGI, E.: Complementi alla teoria della misura (n - 1)-dimensionale in uno spazio n dimensionale. Seminario di Mat. della Scuola Normale Superiore di Pisa, 1960-61. [180] DE GIORGI, E.: Frontiere orientate di misura minima. Seminario di Mat. della Scuola Normale Superiore di Pisa, 1960-61. [181] DE GIORGI, E.: Una estensione del teorema di Bernstein. Ann. Sc. Norm. Sup. Pisa, 19, 79-85, 1965. [182] DE GIORGI, E.: Teoremi di Semicontinuita nel calcolo delle variazioni. Istit. Naz. Alta Mat., Roma, 1968-69.
BIBLIOGRAPHY
675
[183] DE GIORGI, E.: Sulla convergenza di alcune successioni di integrali del tipo dell'area. Rendic. Mat., 8, 277-294, 1975. [184] DE GIORGI, E. and AMBROSIO, L.: Un nuovo tipo di funzionale del calcolo delle
variazioni. Rend. Accad. Naz. Lincei, Cl. Sci. Fis. Mat. Natur., 82, 199-210, 1988.
[185] DE GIORGI, E., BUTTAZZO, G., and DAL MASO, G.: On the lower semicontinuity
of certain integral functionals. Atti Accad. Lincei, 74, 274-282, 1983. [186] DE GIORGI, E., COLOMBINI, F., and PICCININI, L. C.: Frontiere orientate di misura minima e questioni collegate. Scuola Normale Superiore, Pisa, 1972. [187] DE GIORGI, E. and STAMPACCHIA, G.: Sulle singolarita eliminabili delle ipersu-
perfici minimali. Atti Accad. Naz. Lincei, 38, 352-357, 1965. [188] DE RHAM, G.: Sur l'analysis situs des varietes a n-dimensions. J. Math. Pures et Appl., 10, 115-200, 1951. [189] DE RHAM, G.: Varietes differentiables, formes, courants, formes harrnoniques. Act. Sci. et Ind., 1222. Hermann, Paris, 1955. [190] DELLADIO, S.: Lowersemicontinuity and continuity of measures with respect to the strict convergence. Proc. Roy. Soc. Edinburgh, 119, 265-278, 1991. [191] DEMENGEL, F.: Une caracterization des applications de Wl,P(SN, S') qui peuvent etre approchees par des fonctions CO°. C. R. Acad. Sci. Paris, 330, 553-557, 1990. [192] DEMENGEL,
F. and HADIJI, R.: Relaxed energies for functionals on W1,2 (B2, S'). Nonlinear Analysis. Theory, Methods and Applications, 19, 625641, 1992.
[193] DENJOY, A.: Sur les fonctions derivee sommable. Bull. Soc. Mat. France, 43, 161-248, 1916. DIERKES, U., HILDEBRANDT, S., KUSTER, A., and WOHLRAB, 0.: Minimal surfaces I, 77. Grundlehren math. Wiss. 295, 296. Springer-Verlag, Berlin, 1992. [195] DOLD, A.: Lectures on algebraic topology. Springer, New York, 1972. [196] DOUGLAS, J.: The mapping theorem of Koebe and the Plateau problem. J. Math. Phys., 10, 106-130, 1930-31. [197] DOUGLAS, J.: Solution of the problem of Plateau. Trans. Am. Math. Soc., 33, 263-321, 1931. [198] DUZAAR, F. and KUWERT, E.: Minimization of conformally invariant energies [194]
in homotopy classes. Calc. Var., 1998. To appear. EELLS, J. and LEMAIRE, L.: A report on harmonic maps. Bull. London Math. Soc., 10, 1-68, 1978. [200] EELLS, J. and LEMAIRE, L.: Selected topics in harmonic maps. Published for the Conference Board of the Mathematical Sciences, Washington, DC; by the American Mathematical Society, Providence, RI, 1983. [201] EELLS, J. and LEMAIRE, L.: Examples of harmonic maps from disks to hemispheres. Math. Z., 185, 517-519, 1984. [202] EELLS, J. and LEMAIRE, L.: Another report on harmonic maps. Bull. London Math. Soc., 20, 385-524, 1988. [203] EELLS, J. and RATTO, A.: Harmonic maps between spheres and ellipsoids. Internat. J. Math., 1, 1-27, 1990. [204] EELLS, J. and SAMPSON, J.: Harmonic mappings of Riemannian manifolds. American J. Math., 86, 109-160, 1964. [205] EELLS, J. and WooD, J. C.: Restrictions on harmonic maps of surfaces. Topol-
[199]
ogy, 15, 263-266, 1976. [206] EILENBERG, S. and STEENROOD, N.: Foundations of algebraic topology. Prince[207]
ton Univ. Press, Princeton, NJ, 1952. EISEN, G.: A counterexample for some lower semicontinuity results. Math. Z., 162, 241-243, 1978.
676
BIBLIOGRAPHY
EISEN, G.: A selection lemma for sequences of measurable sets, and lower semicontinuity of multiple integrals. manuscripta math., 27, 73-79, 1979. [209] EKELAND, I. and TEMAM, R.: Convex analysis and variational problems. NorthHolland, Amsterdam, 1976. [210] ERICKSEN, J. L.: Equilibrium theory of liquid crystals. In Advances in Liquid [208]
Crystals, vol. 2, edited by Brown, G. H., pp. 233-299. Academic Press, New York, 1976.
ERICKSEN, J. L. and KINDERLEHRER (ED.), D.: Theory and Applications of Liquid Crystals. Springer, New York, 1987. [212] ESTEBAN, M. J.: A direct variational approach to Skyrme's model for mesons fields. Commun. Math. Phys., 105, 571-591, 1986. [213] ESTEBAN, M. J.: Existence of symmetric solutions for the Skyrme's problem. Ann. Mat. Pura. Appl., 147, 187-195, 1987. [214] ESTEBAN, M. J. and MULLER, S.: Sobolev maps with integer degree and applications to Skyrme's problem. Proc. R. Soc. London, 436, 197-201, 1992. [215] EVANS, L. C.: Weak convergence methods for nonlinear partial differential equations. CBMS Regional conference ser. in Math. n.74. AMS, Providence, 1990. [211]
[216] EVANS, L. C.: Partial regularity for stationary harmonic maps into spheres. Arch. Rat. Mech. Anal., 116, 101-, 1991. [217] EVANS, L. C. and GARIEPY, R. F.: Measure theory and fine properties of functions. CRC Press, Boca Raton, FL, 1992. [218] FALCONER, K.: The geometry of fractal sets. Cambridge Univ. Press, Cambridge, 1985.
[219] FALCONER, K.: Fractal geometry. Mathematical foundations and applications. John Wiley & Sons, Ltd., Chichester, 1990. [220] FEDERER, H.: Surface area (I), (II). Trans. Am. Math. Soc., 55, 420-456, 1944. [221] FEDERER, H.: The Gauss-Green theorem. Trans. Am. Math. Soc., 58, 44-76, 1945.
[222] FEDERER, H.: The (0, k) rectifiable subsets of n space. Trans. Am. Math. Soc., 62, 114-192, 1947.
[223] FEDERER, H.: A note on the Gauss-Green theorem. Proc. Am. Math. Soc., 9, 447-451, 1958.
[224] FEDERER, H.: Curvature measures. Trans. Am. Math. Soc., 93, 418-491, 1959. [225] FEDERER, H.: Currents and area. Trans. Am. Math. Soc., 98, 204-233, 1961. [226] FEDERER, H.: Geometric measure theory. Grundlehren math. Wissen. 153. Springer-Verlag, Berlin, 1969. [227] FEDERER, H.: The singular set of area minimizing rectifiable currents with codi-
mension one and area minimizing flat chains modulo two with arbitrary codimension. Bull. Am. Math. Soc., 76, 767-771, 1970. [228] FEDERER, H.: Real flat chains, cochains and variational problems. Indiana Univ. Math. J., 24, 351-407, 1974. [229] FEDERER, H.: Colloquium lectures on geometric measure theory. Bull. Amer. Math. Soc., 84, 291-338, 1978. [230] FEDERER, H. and FLEMING, W. H.: Normal and integral currents. Ann. of Math., 72, 458-520, 1960. [231] FEDERER, H. and ZIEMER, W.: The Lebesgue set of a function whose distributional derivatives are p-th power summable. Indiana Univ. Math. J., 22, 139-158, 1972.
[232] FEFFERMAN, C.: Characterization of bounded mean oscillation. Bull. AMS, 77, 585-587, 1971. [233] FEFFERMAN, C. and STEIN, E.: H' spaces of several variables. Acta Math., 129, 137-193, 1972.
BIBLIOGRAPHY [234] [235] [236] [237] [238]
[239] [240] [241]
677
FINN, R.: Remarks relevant to minimal surfaces of prescribed mean curvature. J. Analyse Math., 14, 139-160, 1965. FINN, R.: Equilibrium capillary surfaces. Springer, New York, 1986. FISCHER-COLBRIE, D.: Some rigidity theorems for minimal submanifolds of sphere. Acta Math., 145, 29-46, 1980. FLEMING, W. H.: An example in the problem of least area. Proc. Amer. math. Soc., 7, 1063-1074, 1956. FLEMING, W. H.: Functions whose partial derivatives are measures. lll. J. Math., 4, 452-478, 1960. FLEMING, W. H.: On the oriented Plateau problem. Rend. Circ. Mat. Palermo, 11, 1-22, 1962. FLEMING, W. H.: Flat chains over a finite coefficient group. Trans. Am. Math. Soc., 121, 160-186, 1966. FLEMING, W. H. and RISHEL, R.: An integral formula for total gradient variation. Arch. Mat., 11, 218-222, 1960.
[242] FONSECA, I. and MARCELLINI, P.: Relaxation of multiple integrals in subcritical [243]
Sobolev spaces. preprint, 1994. FONSECA, I. and MULLER, S.: Quasiconvex integrands and lower semicontinuity
in L'. SIAM J. Math. Anal., 23, 1081-1098, 1992. S.: Relaxation of quasiconvex functionals in
[244] FONSECA, I. and MULLER, [245] [246] [247] [248]
[249] [250] [251]
[252] [253]
BV (12, RD) for integrands f (x, u, Du). Arch. Rat. Mech. Anal., 123, 1-49, 1993. FRANK, F. C.: On the theory of liquid crystals. Discuss. Faraday Soc., 28, 19-28, 1938. FREED, D. S. and UHLENBECK, K. K.: Instantons and four-manifolds. Springer, New York, 1984. FRIEDRICHS, K.: Differential forms on Riemannian manifolds. Comm. Pure Appl. Math., 8, 551-590, 1955. FUCHS, M.: p-harmonic obstacle problems. Part II: Extension of maps and applications. manuscripta math., 63, 381-419, 1989. FucHS, M.: p-harmonic obstacle problems. Part I: Partial regularity theory. Ann. Mat. Pura Appl., 156, 127-158, 1990. FUCHS, M.: p-harmonic obstacle problems. Part III: Boundary regularity. Ann. Mat. Pura Appl., 156, 159-180, 1990. FUCHS, M. and SEREGIN, G.: Holder continuity for weak extremals of some twodimensional variational problems related to nonlinear elasticity. Adv. Math. Sci. Appl., 7, 413-425, 1997.
Fusco, N. and HUTCHINSON, J.: Partial regularity in problems motivated by nonlinear elasticity. SIAM J. Math. Anal., 22, 1516-1551, 1991. Fusco, N. and HUTCHINSON, J.: Partial regularity and everywhere continuity for a model problem from nonlinear elasticity. J. Austral. Math. Soc., 57, 158169, 1994.
[254]
[255] [256]
Fusco, N. and HUTCHISON, J. E.: A direct proof for lower semicontinuity of polyconvex integrals. manuscripta math, 87, 35-50, 1995. FucIK, S., NECAS, J., SOUCEK, J., and SoucEK, V.: Spectral analysis of nonlinear operators. Lecture notes 346. Springer, Berlin, 1973. GAGLIARDO, E.: Caratterizzazione delle tracce sulla frontiera relative ad alcune classi di funzioni in n variabili. Rend. Sem. Mat. Univ. Padova, 27, 284-305, 1957.
[257] GANGBO, W.: On the weak lower semicontinuity of energies with polyconvex integrands. J. Math. Pures Appl., 73, 455-469, 1994. [258] GARDING, L.: Dirichlet problem for linear partial differential equations. Math. Scand., 1, 55-72, 1953.
678
BIBLIOGRAPHY
GASTEL, A.: Regularity theory for minimizing equivariant (p-)harmonic mappings. Calc. Var., 1998. To appear. [260] GEHRING, F.: The LP-integrability of the partial derivatives of a quasi conformal mapping. Acta Math., 130, 265-277, 1973. [261] GERHARDT, C.: On the regularity of solutions to variational problems in BV (f2). Math. Z., 149, 281-286, 1976. [262] GIAQUINTA, M.: On the Dirichlet problem for surfaces of prescribed mean curvature. manuscripta math., 12, 73-86, 1974. [263] GIAQUINTA, M.: Multiple integrals in the calculus of variations and nonlinear elliptic systems. Ann. of Math. Studies n.105. Princeton Univ. Press, Princeton, [259]
1983.
GIAQUINTA, M.: The problem of the regularity of minimizers. In Proceedings of the International Congress of Mathematicians, voll. II, pp. 1072-1083, 1986. [265] GIAQUINTA, M.: Quasiconvexity, growth conditions and partial regularity. In Partial Differential Equations and Calculus of Variations, edited by Hildebrandt, S. and Leis, R., Lecture notes 1357. Springer-Verlag, Berlin, 1988. [266] GIAQUINTA, M.: Problemi variazionali per applicazioni vettoriali. Aspetti geometrici ed analitici. Boll. UMI (7), 6-A, 1-34, 1992. [267] GIAQUINTA, M.: Introduction to regularity theory for nonlinear elliptic systems. Lectures in Math., ETH Zurich. Birkhauser, Basel, 1993. [268] GIAQUINTA, M.: Analytic and geometric aspects of variational problems for vector valued mappings. In Proceedings of the First European Congress of Mathematicians, Paris, July 1992. Birkhauser, Basel, 1994. [264]
GIAQUINTA, M. and GIUSTI, E.: Nonlinear elliptic systems with quadratic growth. manuscripta math., 24, 323-349, 1978. [270] GIAQUINTA, M. and GIUSTI, E.: On the regularity of the minima of variational integrals. Acta Math., 148, 31-46, 1982. [269]
[271] GIAQUINTA, M. and GIUSTI, E.: Differentiability of minima of nondifferentiable functionals. Invent. Math., 72, 285-298, 1983.
[272] GIAQUINTA, M. and GIUSTI, E.: The singular set of the minima of quadratic functionals. Ann. Sc. Norm. Sup. Pisa, 11, 45-55, 1984. [273] GIAQUINTA, M. and HILDEBRANDT, S.: Direct methods in the calculus of varia[274] [275] [276]
tions. In preparation. GIAQUINTA, M. and HILDEBRANDT, S.: A priori estimates for harmonic mappings. J. Reine Angew. Math., 336, 124-164, 1982. GIAQUINTA, M. and HILDEBRANDT, S.: Calculus of Variations, voll.2. Grundlehren math. Wiss. 310, 311. Springer, Berlin, 1996. GIAQUINTA, M. and MODICA, G.: Regularity results for some classes of higher order nonlinear elliptic systems. J. fur reine u. angew. Math., 311/312, 145-169, 1979.
[277] [278] [279]
[280]
GIAQUINTA, M. and MODICA, G.: Nonlinear systems of the type of the stationary
Navier- Stokes system. J. f2ir reine u. angew. Math., 330, 173-214, 1982. GIAQUINTA, M., MODICA, G., and SoucEK, J.: Functionals with linear growth in the calculus of variations. Comm. Math. Univ. Carolinae, 20, 143-171, 1979. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Cartesian currents and variational problems for mappings into spheres. Ann. Sc. Norm. Sup. Pisa, 16, 393-485, 1989. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Cartesian currents, weak dif-
feomorphisms and existence theorems in nonlinear elasticity. Arch. Rat. Mech. Anal., 106, 97-159, 1989. Erratum and addendum, Arch. Rat. Mech. Anal. 109 1990, 385-392. [281]
GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Partial regularity of Cartesian Currents which minimize certain variational integrals. In P.D.E. and the Calcu-
BIBLIOGRAPHY
[282]
679
his of Variations, Essays in Honor of Ennio De Giorgi, edited by Colombini, F., Marino, A., Modica, L., and Spagnolo, S., pp. 563-587. Birkauser, Boston, 1989. GIAQUINTA, M., MODICA, G., and SouOEx, J.: The Dirichlet energy of mappings with values into the sphere. manuscripta math., 65, 489-507, 1989.
[283] GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Liquid crystals: relaxed energies,
dipoles, singular lines and singular points. Ann. Sc. Norm. Sup. Pisa, 17, 415437, 1990. [284]
GIAQUINTA, M., MODICA, G., and SoucnK, J.: Cartesian Currents and Liquid Crystals: dipoles, singular lines and singular points. In Nematics, Mathematical and Physical Aspects, edited by Coron, J. M., Ghidaglia, J. M., and Helein, F., NATO ASI Series C, 332, pp. 113-127. Kluwer Academic Publishers, Dordrecht, 1991.
[285] [286]
GIAQUINTA, M., MODICA, G., and SoucEx, J.: Area and the area formula. Rend. Sem. Mat. Milano, 62, 53-87, 1992. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: The Dirichlet integral for map-
pings between manifolds: Cartesian currents and homology. Math. Ann., 294, [287] [288]
[289]
[290]
[291]
325-386, 1992. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: The gap phenomenon for vari-
ational integrals in Sobolev spaces. Proc. Roy. Soc. Edinburg, 120, 93-98, 1992. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Variational problems for the conformally invariant integral f Idul". In Progress in partial differential equations: calculus of variations, applications, edited by Bandle, C., Bemelmans, J., M., C., Gruter, M., and Paulin, J. S. J., Pittman Research Notes In Math. 267, pp. 27-47. Pittman, 1992. GIAQUINTA, M., MODICA, G., and SOUCEm, J.: Connectivity properties of the range of weak diffeomorphisms. Ann. Institut H. Poincard, Anal. Non Lineaire, 12, 61-73, 1993. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Graphs of finite mass which cannot be approximated in area by smooth graphs. manuscripta math., 78, 259-271, 1993. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Singular perturbations associ-
ated to the Dirichlet energy of maps with values in S2. J. Func. Anal., 118, [292]
188-197, 1993. GIAQUINTA, M., MODICA, G., and SouCEK, J.: Variational problems for maps
of bounded variation with values in S'. Calc. Var., 1, 87-121, 1993. [293] GIAQUINTA, M., MODICA, G., and SOUCEK, J.: A weak approach to finite elas-
ticity. Calc. Var., 2, 65-100, 1994. GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Remarks on the degree theory. J. F'unc. Analysis, 125, 172-200, 1994. [295] GIAQUINTA, M., MODICA, G., and SOUOEK, J.: Some remarks about the p[294]
Dirichlet integral. Comm. Mat. Univ. Carolinae, 35, 55-62, 1994. [296] GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Remarks on quasiconvexity and
lower semicontinuity. NoDEA, 2, 573-588, 1995. GIAQUINTA, M., MODICA, G., and SOUcEK, J.: Compositions of weak diffeomorphisms. Math. Z., 224, 385-402, 1997. [298] GIAQUINTA, M., MODICA, G., and SouCEK, J.: Energy minimizing maps from a domain of I[23 into S2. preprint, 1997. [299] GIAQUINTA, M., MODICA, G., and SOUCEK, J.: Some regularity properties of weakly invertible maps. In Nonlinear Analysis and Continuum Mechanics, edited by Galdi, P. and Pucci, P. Springer, New-York, to appear. Preprint 1993. [300] GIAQUINTA, M. and SOUCEK, J.: Esistenza per it problema dell'area e controesempio di Bernstein. Boll. UMI, 9, 807-817, 1974. [297]
680
BIBLIOGRAPHY
GIAQUINTA, M. and SoucEx, J.: Harmonic maps into a hemisphere. Ann. Scuola Norm. Sup. Pisa, 12, 81-90, 1985. [302] GILBARG, D. and TRUDINGER, N.: Elliptic partial differential equations of second order. Grundlehren mat. Wiss. Springer, Berlin, 1977. Second edition 1994. [301]
[303]
GILKEY, P. B.: Invariance theory, the heat equation, and the Atiyah-Singer Index theorem. Publish or Perish, Wilminton, Delaware, 1984.
[304]
GIUSTI, E.: Precisazione delle funzioni H 'P e singolarita delle soluzioni deboli di sistemi ellittici non lineari. Boll. UMI, 2, 71-76, 1969. GIUSTI, E.: Superfici cartesiane di area minima. Rend. Sem. Mat. Milano, 40,
[305]
1-21, 1970. [306]
[307] [308]
GIUSTI, E.: Boundary behavior of nonparametric minimal surfaces. Indiana Univ. Math. J., 22, 435-444, 1972. GIUSTI, E.: Boundary value problems for non-parametric surfaces of prescribed mean curvature. Ann. Sc. Norm. Sup. Pisa, 3, 501-548, 1976. GIUSTI, E.: Minimal surfaces and functions of bounded variation. Birkhauser, Basel, 1984.
[309]
GIUSTI, E.: Metodi diretti nel Calcolo delle Variazioni. Unione Matematica Italiana, Bologna, 1994.
[310]
GIUSTI, E. and MIRANDA, M.: Sulla regolarita delle soluzioni deboli di una classe
di sistemi ellittici quasilineari. Arch. Rat. Mech. Anal., 31, 173-184, 1968. [311] GOFFMAN, C.: Lower semicontinuity and area functionals. I. The nonparametric case. Rend. Circ. Mat. Palermo, 2, 203-235, 1953. [312] GOFFMAN, C.: Approximation of nonparametric surfaces of finite area. J. Math. Mech., 12, 737-745, 1963. [313] GOFFMAN, C.: A characterization of linearly continuous functions whose partial derivatives are measures. Acta Math., 117, 165-190, 1967. [314] GOFFMAN, C.: An example in surface area. J. Math. Mech., 18, 321-326, 1969. [315] GOFFMAN, C. and Liu, F. C.: Discontinuous mappings and surface area. Proc. London Math. Soc., 20, 237-248, 1970. [316] GOFFMAN, C. and Liu, F. C.: The area formula for Sobolev mappings. Indiana Univ. Math. J., 25, 871-876, 1976. [317] GOFFMAN, C. and Liu, F. C.: Derivatives measures. Proc. Amer. Math. Soc., 78, 218-220, 1980. [318] GOFFMAN, C., NEUGEBAUER, C. J., and NISHIURA, T.: Density topology and approximate continuity. Duke Math. J., 28, 497-505, 1961. [319] GOFFMAN, C. and WATERMAN, D.: Approximately continuous transformations. Proc. Am. Mat. Soc., 12, 116-121, 1961. [320] GOFFMANN, C. and SERRIN, J.: Sublinear functions of measures and variational integrals. Duke Math. J., 31, 159-178, 1964. [321] GOFFMANN, C. and ZIEMER, W. P.: Higher dimensional mappings for which the area formula holds. Ann, of Math., 92, 482-488, 1970. [322] GOLDBERG, S.: Curvature and homology. Dover Publications, 1962. [323]
GREEN, A. and ADKINS, J.: Large elastic deformations. Oxford Univ. Press,
[324]
London, 1970. GREEN, A. and ZERNA, W.: Theoretical elasticity. Oxford Univ. Press, London, 1968.
[325] GREENBERG, M.: Lectures on algebraic topology. W. A. Benjamin, New York, 1966.
GRIFFITS, P. and HARRIS, P.: Principles of Algebraic Geometry. J. Wiley & Sons, New York, 1978. [327] GROTOWSKI, J. F., SHEN, Y., and YAN, S.: On various classes of harmonic maps. Arch. Mat., 64, 353-358, 1995. [326]
BIBLIOGRAPHY
681
[328] GRUTER, M.: Regularity of weak H-surfaces. J. reine u. any. Math., 329, 1-15, 1981.
[329] GULLIVER, R.: Regularity of minimizing surfaces of prescribed mean curvature. Ann. Math., 97, 275-305, 1973. [330] GULLIVER, R. and LESLEY, F. D.: On boundary branch points of minimizing surfaces. Arch. Rat. Mech. Anal., 52, 20-25, 1973. [331] GULLIVER, R., OSSERMAN, R., and ROYDEN, H. L.: A theory of branched immersions of surfaces. Amer. J. Math., 95, 750-812, 1973. [332] GULLIVER, R. and SPRUCK, J.: On embedded minimal surfaces. Ann. Math., 103, 331-347, 1976. Correction in Ann. Math. 109 (1979) 407-412. [333] GUNTHER, M.: On the perturbation problem associated to isometric embeddings of Riemannian manifolds. Ann. Global. Anal. Geom., 7, 69-77, 1989. [334] GURTIN, M.: An introduction to continuum mechanics. Academic Press, New York, 1981.
[335] HAAR, A.: Uber das Plateausche Problem. Math. Ann., 97, 124-258, 1927. [336] HADAMARD, J.: Resolution d'un question relative aux determinants. Bull. Sci. Math., 17, 240-248, 1893. [337] HAJLASZ, P.: A note on weak approximation of minors. Ann. Inst. H. Poincare, Analyse Non Lineaire, 12, 415-424, 1995. [338] HALMOS, P.: Measure theory. Van Nostrand, New York, 1950. [339] HAMILTON, R.: Harmonic maps of manifolds with boundary. Lecture notes n.471. Springer, Berlin, 1975. [340] HARDT, R.: Singularities in some geometric variational problems. In Proceedings
of the International Congress of Mathematicians, vol. I, Berkeley, pp. 540-550, 1986.
[341] HARDT, R.: Singularities of harmonic maps. Bull. AMS, 34, 15-34, 1997. [342] HARDT, R. and KINDERLEHRER, D.: Mathematical questions of liquid crystal theory. In Theory and applications of liquid crystals, edited by Ericksen, J. L. and Kinderlehrer, D., pp. 151-184. Springer, New York, 1987. [343] HARDT, R., KINDERLEHRER, D., and LIN, F.: Stable defects of minimizers of constrained variational principles. Ann. Inst. H. Poincarg, Anal. Non Lineaire, 5, 297-322, 1988. [344] HARDT, R., KINDERLEHRER, D., and LIN, F. H.: Existence and partial regularity of static liquid crystals configurations. Comm. Math. Phys., 105, 547-570, 1986.
[345] HARDT, R., KINDERLEHRER, D., and LIN, F. H.: Existence and partial regularity of static liquid crystal configurations. Comm. Math. Phys., 105, 547-570, 1986.
[346] HARDT, R., KINDERLEHRER, D., and LIN, F. H.: The variety of configurations of liquid crystals. In Variational Methods, edited by Beresticky, H., Coron, J. M., and Ekeland, I., PNDE, pp. 115-131. Birkhauser, Basel, 1990. [347] HARDT, R., KINDERLEHRER, D., and LUSKIN, M.: Remarks about the mathe-
matical theory of liquid crystals. In Calculus of Variations and P.D.E., edited by Hildebrandt, S., Kinderlehrer, D., and Miranda, M., Springer Lecture Notes 1340, pp. 123-138. Springer, New York, 1988.
[348] HARDT, R., LAU, C. P., and LIN, F. H.: Non minimality of minimal graphs. Indiana Math. J., 36, 849-855, 1987. [349] HARDT, R. and LIN, F. H.: Tangential regularity near Cl-boundary. In Geometric Measure Theory and Calculus of Variations, edited by Allard, W. K. and Almgren, F. J. Proc. Symp. Pure Math. 44 AMS, Providence, 1985. [350] HARDT, R. and LIN, F. H.: A remark on Hl mappings. manuscripts math., 56, 1-10, 1986.
682
BIBLIOGRAPHY
[351] HARDT, R. and LIN, F. H.: Mappings minimizing the LP norm of the gradient. Comm. Pure Appl. Math., 40, 555-588, 1987. [352] HARDT, R. and LIN, F. H.: Stability of singularities of minimizing harmonic maps. J. Differential Geom., 29, 113-123, 1989. [353] HARDT, R. and LIN, F. H.: The singular set of an energy minimizing map from B4 to S2. Manuscripta Math., 69, 275-289, 1990. [354] HARDT, R., LIN, F. H., and POON, C. C.: Axially symmetric harmonic maps minimizing a relaxed energy. Comm. Pure Appl. Math., 45, 417-459, 1992. [355] HARDT, R. and PITTS, J.: Solving the Plateau's problem for hypersurfaces without the compactness theorem for integral currents. In Geometric Measure Theory and the Calculus of Variations, edited by Allard, W. K. and Almgren, F. J., Proc. Symp. Pure Math., 44, pp. 255-295. Am. Math. Soc., Providence, 1986. [356] HARDT, R. and SIMON, L.: Boundary regularity and embedded solutions for the oriented Plateau problem. Ann. of Math., 110, 439-486, 1979. [357] HARDT, R. and SIMON, L.: Seminar on geometric measure theory. Birkhauser Verlag, Basel-Boston, MA, 1986. [358] HARTMAN, P.: On homotopic harmonic maps. Can. J. Math., 19, 673-687, 1967. [359] HARTMAN, P. and STAMPACCHIA, G.: On some nonlinear eliptic differentialfunctional equations. Acta Math., 115, 271-310, 1966. [360] HARVEY, R. and LAWSON, H. B.: Extending minimal varieties. Inventiones Math., 28, 209-226, 1975. [361] HARVEY, R. and LAWSON, H. B.: Calibrated geometries. Acta math., 148, 47-157, 1982. [362] HAUSDORFF, F.: Dimension and ausseres Mass. Math. Ann., 79, 157-179, 1918. [363] HAYMAN, W. and KENNEDY, P.: Subharmonic functions. Academic Press, New York, 1976.
[364] HEBEY, E.: Sobolev spaces on Riemannian manifolds. Springer, Berlin, 1996. [365] HEINZ, E.: Uber die Losungen der Minimalflachengleichung. In Nachr. Akad. Wiss., pp. 51-56. Gottingen, 1952.
[366] HEINZ, E.: An elementary analytic theory of the degree of mappings in n[367]
dimensional spaces. J. Math. Mech., 8, 231-247, 1959. HEINZ, E.: Ein Regularitatssatz fur schwache Losungen nichtlinearer elliptische Systeme. Nachr. Akad. Wiss. Gottingen, Math.-Phys. Kl., 1, 1-13, 1975.
[368] HELEIN, F.: Minima de la fonctionelle energie libre des crystaux liquides. Comptes Rend. Acad. Sci. Paris, 305, 565-568, 1988. [369] HELEIN, F.: Approximation of Sobolev maps between an open set and an Euclidean sphere, boundary data, and singularites. Math. Ann., 285, 125-140, 1989.
[370] HELEIN, F.: Regularity des applications faiblement harmoniques entre une surface et une sphere. C. R. Acad. Sci. Paris Ser. I Math., 311, 519-524, 1990. [371] HELEIN, F.: Regularity des applications faiblement harmoniques entre une surface et une variety riemannienne. C. R. Acad. Sci. Paris Ser. I Math., 312, 591-596, 1991. [372] HELEIN, F.: Regularity of weakly harmonic maps from a surface into a manifold with symmetries. Manuscripta Math., 70, 203-218, 1991. [373] HELEIN, F.: Applications harmoniques, lois de conservation et reperes mobiles. Diderot editeur, Paris, 1996. [374] HERZ, C.: The Hardy-Littlewood maximal theorem. In Symposium on Harmonic Analysis. Univ. of Warwick, 1968. [375] HEWITT, E. and STROMBERG, K.: Real and abstract analysis. Springer, Berlin, 1965.
[376] HILDEBRANDT, S.: Boundary behaviour of minimal surfaces. Arch. Rat. Mech. Anal., 35, 47-82, 1969.
BIBLIOGRAPHY
683
[377] HILDEBRANDT, S.: Harmonic mappings of Riemannian manifolds. In Harmonic
mappings and minimal immersions, Lecture Notes in Math., 1161, pp. 1-117. Springer, Berlin-New York, 1985.
[378] HILDEBRANDT, S., JOST, J., and WIDMAN, K. 0.: Harmonic mappings and minimal submanifolds. Invent. Math., 62, 269-298, 1980. [379] HILDEBRANDT, S. and KAUL, H.: Two dimensional variational problems with obstructions, and Plateau's problem for H-surfaces in a Riemannian manifold. Comm. Pure Appl. Math., 25, 187-223, 1972. [380] HILDEBRANDT, S., KAUL, H., and WIDMAN, K.-O.: An existence theorem for harmonic mappings of Riemannian manifolds. Acta Math., 138, 1-16, 1977. [381] HODGE, W.: A Dirichlet problem for harmonic functions with applications to analytic varieties. Proc. London Math. Soc., 36, 257-303, 1934. [382] HODGE, W.: The theory and applications of harmonic integrals. Cambridge University Press, Cambridge, 1952. [383] HONG, M. C.: On the Jager-Kaul theorem concerning harmonic maps. preprint CMA, Canberra, 1996. [384] HoPF, E.: On S. Bernstein's theorem on surfaces z(x, y) of nonpositive curvature. Proc. Amer. Math. Soc., 1, 80-85, 1950. [385] HovE, L. V.: Sur 1'extension de la condition de Legendre du calcul des variations aux integrales multiples a plusieurs fonctions inconnues. Proc. Koninkl. Ned. Akad. Wetenschap, 50, 18-23, 1947.
[386] HovE, L. v.: Sur le signe de la variation seconde des integrales multiples a plusieurs fonctions inconnues. Koninkl. Belg. Acad. Klasse der Wetenschappen, 24, 65, 1949. [387] Hu, S. T.: Homotopy theory. Academic Press, New York, 1959. [388] HUGHES, T. and MARSDEN, J.: Mathematical foundations of elasticity. PrenticeHall, Englewood Cliffs, 1983. [389] Hucxs, R.: Functions of BVC type. Proc. Amar. Math. Soc., 12, 698-701, 1961. [390] HUSEMOLLER, D.: Fiber bundles. Springer, New York, 1993. [391] IOFFE, A. D.: On lower semicontinuity of integral functionals II. SIAM J. Cont. Optim., 15, 991-1000, 1977. [392] IOFFE, A. D.: On lower sermicontinuity of integral functionals I. SIAM J. Cont. Optim., 15, 521-538, 1977. [393] ISOBE, T.: Characterization of the strong closure of C' (B', S') in W"P (B', s2) (16/5 < p < 4). J. Math. Anal. Appl., 190, 361-372, 1995. [394] ISOBE, T.: Energy gap phenomenon and the existence of infinitely many weakly harmonic maps for the Dirichlet problem. J. Funct. Anal., 129, 243-267, 1995. [395] IsoBE, T.: Some new properties of Sobolev mappings: intersection theoretical approach. Proc. Roy. Soc. Edinburgh, 127A, 337-358, 1997. [396] JAGER, W. and KAUL, H.: Uniqueness and stability of harmonic maps and their Jacobi fields. Manuscripta Math., 28, 269-291, 1979. [397] JAGER, W. and KAUL, H.: Uniqueness of harmonic mappings and of solutions of elliptic equations on Riemannian manifolds. Math. Ann., 240, 231-250, 1979. [398] JAGER, W. and KAUL, H.: Rotationally symmetric harmonic maps from a ball into a sphere and the regularity problem for weak solutions of elliptic systems. J. Reine Angew. Math., 343, 146-161, 1983. [399] JENKINS, H. and SERRIN, J.: The Dirichlet problem for minimal surface equation in higher dimensions. J. reine u. angew. Math., 229, 170-187, 1968. [400] JOHN, F. and NIRENBERG, L.: On functions of bounded mean oscillation. Comm. Pure Appl. Math., 14, 415-426, 1961. [401] JORGENS, K.: Uber die Losungen der Differentialgleichung rt - s2 = 1. Math. Ann., 127, 130-134, 1954.
684
BIBLIOGRAPHY
[402] JOST, J.: A maximum principle for harmonic mappings which solve a Dirichlet problem. Manuscripta Math., 38, 129-130, 1982. [403] JOST, J.: Existence proofs for harmonic mappings with the help of a maximum principle. Math. Z., 184, 489-496, 1983. [404] JOST, J.: Harmonic mappings between Riemannian manifolds. Australian National University, Centre for Mathematical Analysis, Canberra, 1984. [405] JOST, J.: Harmonic maps between surfaces. Springer-Verlag, Berlin-New York, 1984.
[406] JOST, J.: The Dirichlet problem for harmonic maps from a surface onto a 2sphere with nonconstant boundary values. J. Dif. Geom., 19, 393-401, 1984. [407] JOST, J.: Lectures on harmonic maps (with applications to conformal mappings and minimal surfaces). In Harmonic mappings and minimal immersions, Lecture Notes in Math., 1161,, pp. 118-192. Springer, Berlin-New York, 1985. [408] JOST, J.: Two dimensional geometric variational problems. Wiley-Interscience, Chichester, 1991. [409] JOST, J.: Riemannian geometry and geometric analysis. Springer, Berlin, 1995. [410] JOST, J. and MEIER, M.: Boundary regularity for minima of certain quadratic functionals. Math. Ann., 262, 549-561, 1983. [411] KAPITANSKY, L. B. and LADYZHENSKAYA, 0. A.: On Coleman's principle con-
cerning the stationary points of invariant integrals. Zapiski. nauk. sem. LOMI, 127, 84-102, 1983. [412] KINDERLEHER, D. and PEDREGAL, P.: Gradient Young measures generated by sequences in Sobolev spaces. J. Geometrical Anal., 4, 59-90, 1994. [413] KINDERLERHER, D.: Recent developments in liquid crystal theory. In Frontiers in pure and applied mathematics, pp. 151-178. North-Holland, Amsterdam, 1991.
[414] KODAIRA, K.: Harmonic fields in Riemannian manifolds. Ann. of Math., 50, 587-665, 1949.
[415] KRICKEBERG, K.: Distributionen, Funktionen beschrankter Variation and Lebesguescher Inhalt nichtparametrischer Flachen. Ann. Mat. Pura Appl., 44, 105-134, 1957. [416] KUFNER, A. AND JOHN, O. AND FUCIK, S.: Function spaces. Academia, Praha, 1977.
[417] KUWERT, E.: Minimizing the energy of maps from a surface into a 2-sphere with
prescribed degree and boundary values. manuscripta math., 83, 31-38, 1994. [418] KUWERT, E.: Area-minimizing immersions of the disk type with boundary in a given homotopy class: a general existence theory, 1995. Habilitationsschrift, 1995, University of Bonn. [419] LADYZHENSKAYA, O. A.: On finding symmetrical solutions of field theory varia-
tional problems. In Proc. of Intern. Congr. of Mathematicians. Warszawa, 1983. [420] LADYZHENSKAYA, 0. A. and URAL'TSEVA, N. N.: Linear and quasilinear elliptic equations. Nauka, Moscow, 1964. (Engl. transl.: Academic Press, new York, 1968;
Second Russian edition: Nauka, 1973). [421] LAWSON, H. B.: Lectures on minimal surfaces. INPA, Rio de Janeiro, 1973. [422] LAWSON, H. B.: Minimal varieties. In Differential Geometry. Proc. Symp. Pure Math., AMS, Providence, 1975. [423] LAWSON, H. B. and OSSERMAN, R.: Non-existence, non-uniqueness and irregularity of solutions to the minimal surface system. Acta Math., 139, 1-17, 1977. [424] LEBESGUE, H.: Integral, longueur, aire. Ann. Mat. Pura Appl., 7, 231-359, 1902. [425] LEBESGUE, H.: Sur les integrales singulieres. Ann. Fac. Sc. Toulouse, pp. 25-117, 1909.
[426] LEMAIRE, L.: Applications harmoniques de surfaces riemanniennes. J. Differential Geom., 13, 51-78, 1978.
BIBLIOGRAPHY
685
[427] LEMAIRE, L.: Boundary value problems for harmonic and minimal maps of surfaces into manifolds. Ann. Scuola Norm. Sup. Pisa, 9, 91-103, 1982. [428] LEONETTI, F.: Maximum principle for functionals depending on minors of the Jacobian matrix of vector-valued mappings. Tech. rep., Centre for Mathematical Analysis, Australian National University, 1990. [429] LERAY, J.: Discussion d'un probleme de Dirichlet. J. Math. Pure Appl., 18, 249-284, 1939. [430]
LESLIE, F. C.: Theory of flow phenomena in liquid crystals. In Advances in Liquid Crystals, vol. 2, edited by Brown, G. H., pp. 1-81. Academic Press, New York, 1976.
[431] LEvY, P.: Theorie de l'addition des variables aleatoires. Gauthier-Villars, Paris, 1954. [432] [433] [434] [435]
LIEB, E. H.: Remarks on Skyrme model, ????
LIN, F. H.: A remark on the map x/jxj. C. R. Acad. Sci. Paris Ser. I Math., 305, 529-531, 1987. LIN, S. Y.: Numerical analysis for liquid crystal problems. Thesis, 1987. LIONS, P. L.: The concentration-compactness principle in the calculus of varia-
tions. The locally compact case. I. Ann. Inst. H. Poincare Anal. Non Lineaire, [436]
[437]
[438] [439]
1, 109-145, 1984. LIONS, P. L.: The concentration-compactness principle in the calculus of variations. The locally compact case. II. Ann. Inst. H. Poincare Anal. Non Lineaire, 1, 223-283, 1984. LIONS, P. L.: The concentration-compactness principle in the calculus of variations. The limit case. II. Rev. Mat_ Iberoamericana, 1, 45-121, 1985. LIONS, P. L.: The concentration-compactness principle in the calculus of variations. The limit case. I. Rev. Mat. Iberoamericana, 1, 145-201, 1985. LIu, F. C.: Essential multiplicity function for a.e. approximately differentiable mapping. Proc. Acad. Sinica, 1976.
[440]
LIu, F. C.: A Lusin type property of Sobolev functions. Indiana Univ. Math.
[441]
J., 26, 645-651, 1977. LIu, F. C.: Approximately differentiable mappings and surface area. In "Studies and essays in Commemoration of the Golden Jubilee of Academia Sinica" Taipei, 1978.
LIu, F. C. and TAI, W. S.: Approximate Taylor polynomials and differentiation of functions. preprint, 1991. [443] LIu, F. C. and TAI, W. S.: Maximal mean steepness and Lusin type properties. preprint, 1991. [444] LuCKHAUS, S.: Partial Holder continuity for minima of certain energies among maps into a Riemannian manifold. Indiana Math. J., 37, 349-367, 1988. [445] LuCKHAUS, S.: Convergence of minimizers for the p-Dirichlet integral. Math. Z., 213, 449-456, 1993. [446] LUKES, J., MALY, J., and ZAJIbEK, L.: Fine Topology Methods in Real Analysis and Potential Theory. Lecture notes 1189. Springer-Verlag, Berlin, 1986. [447] MALY, J.: LP-approximation of Jacobians. Comment. Math. Univ. Carolinas, 32, 659-666, 1991. [448] MALY, J.: Weak lower semicontinuity of polyconvex integrals. Proc. Roy. Soc. Edinburgh, 123A, 681-691, 1993. [449] MALY, J.: Lower semicontinuity of quasiconvex integrals. manuscripta math., 85, 419-428, 1994. [450] MALY, J. and MARTIO, 0.: Lusin's condition (N) and mappings of the class Wl'". J. Reine Angew. Math., 458, 19-36, 1995. [451] MANTON, N. S. and RUBECK, P. J.: Skyrmions in flat space and curved space. Phys. Lett., B181, 137-140, 1986. [442]
686
BIBLIOGRAPHY
[452] MARCELLINI, P.: Quasiconvex quadratic forms in two dimensions. Appl. Math. Optim., 11, 183-189, 1984. [453] MARCELLINI, P.: Approximation of quasiconvex functions and lower semicontinuity of multiple integrals. manuscripta math., 51, 1-28, 1985. [454] MARCELLINI, P.: On the definition and the lower semicontinuity of certain quasiconvex integrals. Ann. Inst. H. Poincare, Anal. Non Lineaire, 3, 391-409, 1986.
[455] MARCELLINI, P.: The stored-energy for some discontinuous deformations in non-
linear elasticity. In PDE and the Calculus of Variations, edited by Colombini, F., Marino, A., Modica, L., and Spagnolo, S., pp. 767-786. Birkhauser, Boston, 1989.
[456] MARCELLINI, P. and SBORDONE, C.: Semicontinuity problems in the calculus of
variations. Nonlinear Anal., Theory, Meth. and Appl., 4, 241-257, 1980. [457] MARCUS, M. and MIZEL, V.: Transformations by functions in Sobolev spaces and lower semicontinuity for parametric variational problems. Bull. Amer. Math. Soc., 79, 790-795, 1973. [458] MARCUS, M. and MIZEL, V. J.: Absolute continuity on tracks and mappings of Sobolev spaces. Arch. Rat. Mech. Anal., 45, 294-320, 1972. [459] MASSARI, U.: Esistenza e regolarita delle ipersuperfici di curvatura media assegnata in R'. Arch. Rat. Mech. Anal., 55, 357-382, 1974.
[460] MASSARI, U.: Frontiere orientate di curvatura media assegnata in L. Rend. Sem. Mat. Padova, 53, 37-52, 1975. [461] MASSARI, U. and MIRANDA, M.: Minimal surfaces of codimension one. NorthHolland, Amsterdam, 1984. [462] MATTILA, P.: Geometry of sets and measures in Euclidean spaces. Cambridge Univ. Press, 1995. [463] MAZ'JA, V. G.: Sobolev spaces. Springer-Verlag, Berlin, 1985.
[464] MCSHANE, B. J.: On the necessary condition of Weierstrass in the multiple integral problem of the calculus of variations. Ann. Math., 32, 578-590, 1931. [465] MEEKS, W. H. and YAu, S.-T.: The classical Plateau problem and the topology of three-dimensional manifolds. Topology, 21, 409-440, 1982. [466] MEYERS, N. G.: Quasiconvexity and lower semicontinuity of multiple variational integrals of any order. Trans. A.M.S., 119, 125-149, 1965.
[467] MEYERS, N. G. and ZIEMER, W. P.: Integral inequalities of Poincare and Wirtinger type for BV functions. Amer. J. Math., 99, 1345-1360, 1977. [468] MICALLEF, M. and WHITE, B.: The structure of branch points in minimal surfaces and to pseudoholomorphic curves. Ann. of Math., 141, 35-85, 1995. [469] MICHAEL, E.: Continuous selections. Ann. Math., 63, 361-381, 1956. [470] MICHAEL, J. H.: The equivalence of two areas for nonparametric discontinuous surfaces. Illinois J. Math., 7, 59-78, 1963. [471] MICKLE, E. J.: A remark on a theorem of Serge Bernstein. Proc. Amer. Math. Soc., 1, 86-89, 1950. [472] MILNOR, J.: On manifolds homeomorphic to the 7-sphere. Ann. of Math., 64, 399-405, 1956. [473] MILNOR, J.: Topology from the differentiable viewpoint. Univ. Press of Virginia, 1965.
[474] MIRANDA, M.: Distribuzioni aventi derivate misure. Insiemi di perimetro finito. Ann. Sc. Norm. Sup. Pisa, 18, 27-56, 1964. [475] MIRANDA, M.: Superfici cartesiane generalizzate ed insiemi di perimetro finito
sui prodotti cartesiani. Ann. Sc. Norm. Sup. Pisa, 18, 515-542, 1964. [476] MIRANDA, M.: Sul minimo dell'integrale del gradiente di una funzione. Ann. Sc. Norm. Sup. Pisa, 19, 627-665, 1965.
BIBLIOGRAPHY
687
[477] MIRANDA, M.: Un teorema di esistenza e unicity per it problema dell'area in n variabili. Ann. Sc. Norm. Sup. Pisa, 19, 233-249, 1965. [478] MIRANDA, M.: Comportamento delle successioni convergenti di frontiere minimali. Rend. Sem. Mat. Univ. Padova, 38, 238-257, 1967. [479] MIRANDA, M.: Un principio di massimo forte per le frontiere minimali e sua applicazione alle superfici di area minima. Rend. Sem. Mat. Padova, 45, 355366, 1971.
[480] MIRANDA, M.: Grafici minimi completi. Ann. Univ. Ferrara, 23, 269-272, 1977. [481] MIRANDA, M.: Sulle singolarity eliminabili delle soluzioni dell'equazione delle superfici minime. Ann. Sc. Norm. Sup. Pisa, 4, 129-132, 1977. [482] MIRANDA, M.: Superfici minime illimitate. Ann. Sc. Norm. Sup. Pisa, 4, 313322, 1977.
[483] MORGAN, F.: Area-minimizing currents bounded by higher multiples of curves. Rend. Circ. Mat. Palermo, 33, 37-46, 1984. [484] MORGAN, F.: Area-minimizing surfaces, faces of Grassmannians, and calibrations. Am. Math. Monthly, 95, 813-822, 1988. [485] MORGAN, F.: Geometric measure theory. A beginner's guide. Academic Press, Inc., Boston, MA, 1988. second edition, 1995. [486] MORREY, C. B.: On the solutions of quasi-linear elliptic partial differential equations. Trans. Am. Math. Soc., 43, 126-166, 1938. [487] MORREY, C. B.: Multiple integral problems in the calculus of variations and related topics. Univ. of California Publ. in Math., 1, 1-130, 1943. [488] MORREY, C. B.: The problem of Plateau on a Riemannian manifold. Ann. of Math., 49, 807-851, 1948. [489] MORREY, C. B.: Quasi-convexity and the lower semicontinuity of multiple integrals. Pacific J. Math., 2, 25-53, 1952. [490] MORREY, C. B.: Multiple integrals in the calculus of variations. Grundlehren math. Wissenschaften 130. Springer-Verlag, Berlin, 1966. [491] MORREY, C. B.: Partial regularity results for nonlinear elliptic systems. Jour. Math. and Mech., 17, 649-670, 1968. [492] MosER, J.: A new proof od De Giorgi's theorem concerning the regularity problem for elliptic differential eqiuations. Comm. Pure Appl. Math., 13, 457-468, 1960.
[493] MOSER, J.: On Harnack's theorem for elliptic differential equations. Comm. Pure Appl. Math., 14, 577-591, 1961. [494] Mou, L.: Harmonic maps with prescribed finite singularities. Comm. in PDE, 14, 1509-1540, 1989.
[495] MUCci, D.: Graphs of finite mass which cannot be approximated by smooth graphs with equibounded area. J. Func. Anal., to appear, ??, ??, ??
[496] Mucci, D.: Approximation in area of graphs with isolated singularities. manuscripta math., 88, 135-146, 1995. [497] Mucci, D.: Approssimazione in area di grafici continui in dimensione e codimensione qualunque. Ph.D. thesis, University di Firenze, 1996. [498] Mucci, D.: Approximation in area of continuous graphs. Calc. Var., 4, 525-557, 1996.
[499] MULLER, S.: Weak continuity of determinants and nonlinear elasticity. C. R. Acad. Sci. Paris, 307, 501-506, 1988. [500] MULLER, S.: Det=det. A remark on the distributional determinant. C. R. Acad. Sci Paris, 311, 13-17, 1990. [501] MULLER, S.: Higher integrability of determinants and weak convergence in V. J. Reine Angew. Math., 412, 20-34, 1990. [502] MULLER, S.: On quasiconvex functions which are homogeneous of degree 1. Indiana Univ. Math. J., 41, 295-301, 1992.
688
BIBLIOGRAPHY
[503] MULLER, S.: On the Singular Support of the Distributional Determinant. Ann. Inst. H. Poincare Anal. Non Lineaire, 10, 657-696, 1993. [504] MULLER, S., TANG, Q., and YAN, B.: On a new class of elastic deformations not allowing for cavitation. Ann. Inst. H. Poincare Anal. Non Lineaire, 11, 217-243, 1994. [505] MUNROE, M.: Measure and integration. Addison-Wesley, Reading, 1971. Second edition.
[506] MURAT, F.: Compacite par compensation. Ann. Sc. Norm. Sup. Pisa, 5, 489507, 1978.
[507] MURAT, F.: Compacite par compensation, II. In Proceedings of the International Meeting on Recent Methods in Nonlinear Analysis, edited by De Giorgi, E., Magenes, E., and Mosco, U. Pitagora, Bologna, 1979. [508] MURAT, F.: Compacite par compensation: condition necessaire et suffisante de
continuite faible sous une hypothese de rang constant. Ann. Sc. Norm. Sup. Pisa, 8, 69-102, 1981. [509] NAGUMO, M.: Degree of mappings in convex linear topological spaces. Amer. J. Math., 73, 497-511, 1951. [510] NASH, J.: The imbedding problem for riemannian manifolds. Ann. of Math., 63, 20-63, 1956.
[511] NASH, J.: Continuity of solutions of parabolic and elliptic equations. Amer. J. Math., 8, 931-954, 1958. [512] NATANSON, I. P.: Theory of functions of real variable 1,11. Frederick ungar Publ., New York, 1964. [513] NIRENBERG, L.: Topics in nonlinear functional analysis. Courant Institute, New York University, New York, 1974. [514] NISHIURA, T. and BRECKENRIDGE, J. C.: Differentiation, Integration, and
Lebesgue Area. Imdiana Univ. Math. J., 26, 515-536, 1977. [515] NITSCHE, J. C. C.: Elementary proof of Bernstein's theorem on minimal surfaces. Ann. of Math., 66, 543-544, 1957. [516] NITSCHE, J. C. C.: On new results in the theory of minimal surfaces. Bull. AMS, 71, 195-270, 1965.
[517] NITSCHE, J. C. C.: Uber ein verallgemeinertes Dirichletsches Problem fur die Minimalflachengleichung and hebbare Unstetigkeiten ihrer Losungen. Math. Ann., 158, 203-214, 1965. [518] NITSCHE, J. C. C.: The boundary behaviour of minimal surfaces. Kellog's theorem and branch points on the boundary. Invent. Math., 8, 313-333, 1969. Addendum, Invent. Math. 9 (1970) 270. [519] NITSCHE, J. C. C.: Vorlesungen fiber Minimaflachen. Grundlehren 199. Springer-Verlag, Berlin, 1975. [520] NITSCHE, J. C. C.: Lectures on minimal surfaces. Cambridge Univ. Press, 1989. [521] NOBELING, G.: Uber die Flachenmasse im Euclidischen Raum. Math. Ann., 118, 687-701, 1943.
[522] OLECH, C.: Existence theory in optimal control problems, the underlying ideas. In International Conference on Differential Equations , pp. 612-629. Academic Press, New York, 1975. [523] OLECH, C.: A characterization of L'-weak lower semicontinuity of integral functionals. Bull. Acad. Pol. Sci., 25, 135-142, 1977. [524] OSEEN, C. W.: The theory of liquid crystals. Trans. Faraday Soc., 29, 883-889, 1933.
[525] OSSERMAN, R.: A survey of minimal surfaces. Van Nostrand, New York, 1969. (Dover Edition, 1986). [526] OSSERMAN, R.: Minimal varieties. Bull. AMS, 75, 1092-1120, 1969.
BIBLIOGRAPHY
689
[527] OSSERMAN, R.: A proof of the regularity everywhere of the classical solution to Plateau problem. Ann. Math., 91, 550-569, 1970. [528] OSSERMAN, R.: Some properties of solutions to the minimal surfaces system for arbitrary codimension. In Global Analysis. Proc. Symp. Math. 15, AMS, Providence, 1970. [529] OSSERMAN, R.: On Bers' Theorem on isolated singularities. Indiana Univ. Math. J., 23, 337-342, 1973. [530] PEDREGAL, P.: Parametrczed measures and variational principles. Birkauser, 1997.
[531] PERELOMOV, A. M.: Instanton-like solutions in chiral models. Physica 4D, pp. 1-25, 1981. [532] PODIO-GUIDUGLI, P. and VERGARA CAFFARELLI, G.: Discontinuous energy minimizers in nonlinear elastostatics: an example of J. Ball revisited. J. Elasticity, 15, 75-96, 1986. [533] POHOZAEV, S. I.: Eigenfunctions of the equation du + .X f (u) = 0. Soviet Math. Doklady, 6, 1408-1411, 1965. (Transl. from the Russian Dokl. Akad. Nauk SSSR 165, 1965, 33-36). [534] PONOMAREV, S. P.: Property N of homeomorphisms of class W''. Sib. Math. J., 28, 291-298, 1987.
[535] POON, C. C.: Some new harmonic maps from B3 to S2. J. Dif. Geom., 34, 165-168, 1991.
PREISS, D.: Geometry of measures in R': distribution, rectifiability, and densities. Ann. of Math. (2), 125, 537-643, 1987. [537] QING, J.: Remark on the Dirichlet problem for harmonic maps from the disc into the 2-sphere. Proc. Roy. Soc. Edinburgh, 122A, 63-67, 1992. [538] RADEMACHER, H.: Uber partielle and totale Differentierbarkeit I. Math. Ann., 79, 340-359, 1919. [539] RAD6, T.: Geometrische Betrachtungen Uber zweidimensionale reguldre Variationsprobeme. Acta Litt. Sci. Szeged, 2, 228-253, 1926. [540] RADO, T.: Zu einem Satze von S. Bernstein Uber Minimalflachen in Grossen. Math. Z., 26, 559-565, 1927. [541] RAD6, T.: Uber zweidimensionale regulase Variationsprobleme. Math. Ann., 101, 620-632, 1929. [542] RADO, T.: On Plateau's problem. Ann. Math., 32, 457-469, 1930. [543] RADO, T.: The problem of least area and the problem of Plateau. Math. Z., 32,
[536]
763-796, 1930.
[544] RADO, T.: On the problem of Plateau. Springer, Berlin, 1933. [545] RADO, T.: Length and area. Colloquium Publications 30. Amer. Math. Soc., New York, 1948.
[546] RADO, T. and REICHELDERFER, P.: Continuous transformations in analysis. Springer, Berlin, 1955. [547] REIFENBERG, R. E.: Solution of the Plateau problem for m-dimensional surfaces of varying topological type. Acta Math., 104, 1-92, 1960. [548] REIFENBERG, R. E.: An epiperimetric inequality related to the analyticity of minimal surfaces. Ann. of Math., 80, 1-14, 1964. [549] REIFENBERG, R. E.: On the analyticity of minimal surfaces. Ann. of Math., 80, 15-21, 1964.
[550] RESHETNYAK, Y. G.: On the stability of conformal mappings in multidimensional spaces. Sib. Math. Zhu., 8, 91-114, 1967. In Russian. English transl.: Siberian Math. J. 8 (1967) 69-85. [551] RESHETNYAK, Y. G.: Stability theorems for mappings with bounded excursion. Sib. Math. Zhu., 9, 667-684, 1968. In Russian. English transl.: Siberian Math. J. 9 (1968) 499-512.
690
BIBLIOGRAPHY
[552] RESHETNYAK, Y. G.: Weak convergence of completely additive vector functions
on a set. Sib. Math. J., 9, 1039-1045, 1968. [553] RESHETNYAK, Y. G.: General theorems on semicontinuity and on convergence with a functional. Sib. Math. J., 9, 801-816, 1969. [554] RESHETNYAK, Y. G.: Space mappings with bounded distorsion. Transl. Math. Monographs. Amer. Math. Soc., Providence, 1989. [555] RICKMAN, S.: On the number of omitted values of entire quasiregular mappings. J. Analyse Math., 37, 100-117, 1980. [556] RIEsz, F. and Sz.-NAGY, B.: Legons d'analyse fonctionnelle, 1952. 4th edition, Gauthier-Villars, Paris, 1965. [557] RIVIERE, T.: Applications harmoniques de B3 dans S2 ayant une ligne de sin-
gularites. C. R. Acad. Sci. Paris Ser. I Math., 313, 583-587, 1991. RIVIERE, T.: Applications harmoniques de B3 dans S2 partout discontinues. C. R. Acad. Sci. Paris Ser. I Math., 314, 719-723, 1992. [559] RIVIERE, T.: Applications harmoniques entre varietes. These de Doctorat, 1993, Universite de Paris VI. [560] RIVIERE, T.: Harmonic maps from B3 into S2 having a line of singularities. In Applications harmoniques entre varietes, These de Doctorat, 1993, Universite de [558]
Paris VI. [561]
RIVIERE, T.: Everywhere discontinuous maps into spheres. Acta Math., 175, 197-226, 1995.
RIVIERE, T.: Minimizing fibrations and p-harmonic maps in homotopy classes from S3 into S2. Preprint, 1996. [563] ROCKAFELLAR, R.: Convex analysis. Princeton Univ. Press, Princeton, 1970. [564] ROGERS, C.: Hausdorff measure. Cambridge Univ. Press, Cambridge, 1970. [565] ROUBICEK, T.: Relaxation in optimization theory and variational calculus. Walter de Gruyter, Berlin, 1997. [566] RUDIN, W.: Real and complex analysis. McGraw-Hill, New York, 1966. [567] RUH, E. and VILMS, J.: The tensor field of the Gauss map. Trans. Amer. Math. Soc., 149, 569-573, 1970. [568] SACKS, J. and UHLENBECK, K.: The existence of minimal immersions of 2spheres. Ann. of Math., 113, 1-24, 1981. [569] SAKS, S.: On the surfaces without tangent planes. Ann. of Math., 34, 114-124, [562]
1933.
SAKS, S.: Theorie de l'integrale. Monografje Matematyczne, Warsawa, 1933. [571] SCHOEN, R.: Analytic aspects of the harmonic map problem. In Seminar on nonlinear partial differential equations, Math. Sci. Res. Inst. Publ., 2,, pp. 321[570]
358. Springer, New York-Berlin, 1984. [572] SCHOEN, R.: The effect of curvature on the behaviour of harmonic functions and mappings. In Nonlinear Partial Differential Equations in Differential Geometry, edited by Hardt, R. and Wolf, M., pp. 127-183. Amer. Math. Soc., Providence, RI, 1996. [573] SCHOEN, R. and SIMON, L.: A new proof of the regularity theorem for rectifiable currents which minimize parametric elliptic functionals. Indiana Univ. Math. J., 31, 415-434, 1982. [574] SCHOEN, R., SIMON, L., and ALMGREN, F. J.: Regularity and singularity estimates on hypersurfaces minimizing parametric elliptic variational integrals. Acta Math., 139, 217-265, 1977. [575] SCHOEN, R., SIMON, L., and YAU, S. T.: Curvature estimates for minimal hypersurfaces. Acta Math., 134, 276-288, 1975. [576] SCHOEN, R. and UHLENBECK, K.: A regularity theory for harmonic maps. J. Differential Geom., 17, 307-335, 1982.
BIBLIOGRAPHY
691
[577] SCHOEN, R. and UHLENBECK, K.: Boundary regularity and the Dirichlet problem for harmonic maps. J. Differential Geom., 18, 253-268, 1983. [578] SCHOEN, R. and UHLENBECK, K.: Correction to: "A regularity theory for harmonic maps". J. Differential Geom., 18, 1983. [579] SCHOEN, R. and UHLENBECK, K.: Regularity of minimizing harmonic maps into the sphere. Invent. Math., 78, 89-100, 1984. [580] SCHOEN, R. and YAU, S.: Lectures on harmonic maps. International Press. To appear.
[581] SCHOEN, R. and YAU, S.: Existence of incompressible minimal surfaces and the topology of three dimensional manifolds with non-negative scalar curvature. Ann. of Math., 110, 127-142, 1979. [582] SCHWARZ, G.: Hodge decomposition. A method for solving boundary value problems. Lecture Notes 1607. Springer, Berlin, 1995.
[583] SEMMES, S.: A primer on Hardy spaces, and some remarks on a theorem of Evans and Muller. Comm. Part. Diff. Eq., 19, 277-319, 1994. [584] SERRE, D.: Formes quadratiques et calcul des variations. J. Math. Pures Appl., 62, 117-196, 1983. [585] SERRIN, J.: On the differentiability of functions of several variables. Arch. Rat. Mech. Anal., 7, 359-372, 1961. [586] SERRIN, J.: The problem of Dirichlet for quasilinear elliptic differential equations with many independent variables. Phil. Trans. Roy. Soc. London, 264, 413-496, 1969.
[587] SERRIN, J. and VARBERG, D. E.: A general chain rule for derivatives and the change of variables formula for the Lebesgue integral. Amer. Math. Monthly, 76, 514-520, 1969.
SIMON, L.: Boundary regularity for solutions of the non-parametric least area problem. Ann. of Math., 103, 429-455, 1976. [589] SIMON, L.: Remarks on curvature estimates for minimal hypersurfaces. Duke Math. J., 43, 545-553, 1976. [590] SIMON, L.: On a theorem of De Giorgi and Stampacchia. Math. Z., 115, 199-204, [588]
1977.
SIMON, L.: Asymptotics for a class of non-linear evolution equations, with applications to geometric problems. Ann. of Math., 118, 525-572, 1983. [592] SIMON, L.: Lectures on geometric measure theory. The Centre for mathematical Analysis, Canberra, 1983. [593] SIMON, L.: Survey lectures on minimal submanifolds. In Seminar on minimal submanifolds, edited by Bombieri, E. Princeton Univ. Press., Princeton, 1983. [594] SIMON, L.: Isolated singularities of extrema of geometric variational problems. In Harmonic Mappings and Minimal Immersions, Lecture notes in Math. n.1161. Springer, Berlin, 1985. [595] SIMON, L.: Rectifiability of the singular set of energy minimizing maps. Calc. Var., 3, 1-65, 1995. [596] SIMON, L.: Theorems on regularity and singularity of energy minimizing maps. ETH Lectures on Mathematics. Birkauser, Basel, 1995. [597] SIMON, L.: Proof of the basic regularity theorem for harmonic maps. In Nonlinear Partial Differential Equations in Differential Geometry, edited by Hardt, R. and Wolf, M., pp. 225-256. Amer. Math. Soc., Providence, RI, 1996. [598] SIMON, L.: Singularities of geometric variational problems. In Nonlinear Partial Differential Equations in Differential Geometry, edited by Hardt, R. and Wolf, M., pp. 185-224. Amer. Math. Soc., Providence, RI, 1996. [599] SIMONS, J.: Minimal varieties in riemannian manifolds. Ann. of Math., 88, 62[591]
105, 1968.
692
BIBLIOGRAPHY
[600] SINGER, I. M. and THORPE, J. A.: Lectures Notes on elementary topology and geometry. Scott, Foresman and C., Genview, Illinois, 1967. [601] SIVALOGANATHAN, J.: Uniqueness of regular and singular equilibria for spherically symmetric problems of nonlinear elasticity. Arch. Rat. Mech. Anal., 96, 97-136, 1986.
[602] SKYRME, T. H. R.: A nonlinear filed theory. Proc. Roy. Soc., A260, 127-138, 1961.
[603] SKYRME, T. H. R.: A unified field theory of mesons and baryons. Nuc. Phys., 31, 556-559, 1962. [604] SOLOMON, B.: A new proof of the closure theorem for integral currents. Indiana Univ. Math. J., 33, 393-418, 1984. [605] SOUCEK, J.: Spaces of functions on the domain 17 whose k-th derivatives are measures defined in (1. Cas. Pestovani Mat., 97, 10-46, 1972. [606] SOYEUR, A.: The Dirichlet problem for harmonic maps from the disc into the 2-sphere. Proc. Roy. Soc. Edinburgh, 113A, 229-234, 1989. [607] SPANIER, E. H.: Algebraic topology. McGraw-Hill, New York, 1966. [608] STAMPACCHIA, G.: On some regular multiple integral problems in the calculus of variations. Comm. Pure Appl. Math., 16, 383-421, 1963. [609] STAMPACCHIA, G.: Equations elliptiques du second ordre d coefficients discontinus. Les Presses de l'Universite de Montreal, Montreal, 1966. [610] STEFFEN, K. and WENTE, H.: The non-existence of branch points in a solution to certain classes of Plateau type variational problems. Math. Z., 163, 211-238, 1978.
[611] STEIN, E. M.: Singular integrals and differentiability properties of functions. Princeton Univ. Press, Princeton, 1970. [612] STEIN, E. M.: Harmonic analysis. Princeton Univ. Press, Princeton, 1993. [613] STEIN, E. M. and WEiss, G.: Introduction to Fourier analysis on Euclidean spaces. Princeton Univ. Press, Princeton, 1971. [614] STEPANOFF, W.: Sur les conditions de 1'existence de la differentielle totale. Rec. Math. Soc. Moscou, 32, 511-526, 1925. [615] STRUwE, M.: Plateau's problem and the Calculus of Variations. Mathemathical Notes, 35. Princeton Univ. Press, Princeton, 1988. [616] STRUWE, M.: Variational methods. Springer-Verlag, Berlin, 1990. Second edition 1996.
[617] STUART, C. A.: Radially symmetric cavitation for hyperelastic materials. Ann IHP, Analyse Non Lineaire, 2, 33-66, 1985. [618] SVERAK, V.: Regularity properties of deformations with finite energy. Arch. Rat. Mech. Anal., 100, 105-127, 1988. [619] SVERAK, V.: Quasiconvex functions with subquadratic growth. Proc. Roy. Soc. London, A 433, 723-725, 1991. [620] SVERAK, V.: New examples of quasiconvex functions. Arch. Rat. Mech. Anal., 119, 293-300, 1992. [621] SVERAK, V.: Rank-one convexity does not imply quasiconvexity. Proc. Roy. Soc. Edinburgh, 120A, 185-189, 1992. [622] SYCHEV, M. A.: Characterization of weak-strong convergence property of integral functionals in terms of integrands. preprint, 1994. [623] SYCHEV, M. A.: A criterion for continuity of an integral functional on a sequence of functions. Sib. Math. J., 36, 146-156, 1995. [624] TAMANINI, I.: Boundaries of Caccioppoli sets with Holder-continuous normal vector. J. reine u. angew. Math., 334, 27-39, 1982. [625] TANG, Q.: Almost everywhere injectivity in non linear elasticity. Proc. Roy. Soc. Edinburg, 109 A, 79-95, 1988.
BIBLIOGRAPHY
693
[626] TARTAR, L.: The compensated compactness method applied to systems of conser-
vation laws. In Systems of nonlinear partial differential equations. Proc. NATO Advanced Study Inst., Oxford 1982, edited by Ball, J. M. Reidel, Dordrecht, 1983. [627] TARTAR, L.: Weak limits of semilinear hyperbolic systems with oscillating data. In Macroscopic modelling of turbulent flows, vol. 230 of Lecture Notes in Physics. Springer, Berlin, 1985. [628] TAUBES, C.: The existence of a non-minimal solution to the SU(2) Yang-MillsHiggs equations on R3. Comm. Math. Phys., 86, 257-320, 1982.
[629] TERPSTRA, F. J.: Die Darstellung biquadratischer Formen als Summen von Quadraten mit Anwendung auf die Variationsrechnung. Math. Ann., 116, 166180, 1938.
[630] TOMI, F. and TROMBA, A. J.: Extreme curves bound an embedded minimal surface of disk type. Math. Z., 158, 137-145, 1978. [631] TONELLI, L.: Fondamenti di Calcolo delle Variazioni, voll. 2. Zanichelli, Bologna, 1921-23.
[632] TONELLI, L.: Sul problema di Plateau, I-II. Rend. R. Accad. Lincei, 24, 333-339, 393-398, 1936. (See also Opere scelte, vol.3, 328-341). [633] TONELLI, L.: Opere scelte I-IV. Edizioni Cremonese, Roma, 1960-63. [634] TORCHINSKY, A.: Real-Variable Methods in Harmonic Analysis. Academic Press, New York, 1986. [635] TRISCARI, D.: Sul'esistenza di cilindri con frontiere di misura minima. Ann. Sc. Norm. Sup. Pisa, 17, 387-399, 1963. [636] TRISCARI, D.: Sulle singolarita delle frontiere minime orientate di misura min-
ima. Ann. Sc. Norm. Sup. Pisa, 17, 349-371, 1963. [637] TRISCARI, D.: Sulle singolarita delle frontiere orientate di misura minima nello spazio euclideo a 4 dimensioni. Le Matematiche, 18, 139-163, 1963. [638] TRUDINGER, N. S.: A new proof of the interior gradient bound for the minimal surface equation in n variables. Proc. Nat. Acad. Sci. USA, 69, 821-823, 1972. [639] TRUESDELL, C.: The elements of continuum mechanics. Springer Verlag, Berlin, 1966.
[640] TRUESDELL, C.: A first course in rational continuum mechanics.
Academic Press, New York, 1977. [641] TRUESDELL, C. and NOLL, W.: The non-linear field theories of mechanics. In Handbuch der Physik, Vol III/3. Springer Verlag, Berlin, 1965. [642] UHLENBECK, K.: Harmonic maps: a direct method in the calculus of variations. Bull. AMS, 76, 1082-1087, 1970. [643] UHLENBECK, K.: Minimal spheres and other conformal variational problems.
In Seminar on minimal submanifolds, Ann. of Math. Stud., 103,, pp. 169-176. Princeton Univ. Press, Princeton, NJ, 1983. [644] VALENT, T.: Boundary value problems of finite elasticity. Springer, Heidelberg, 1988.
[645] VICK, J. W.: Homology theory. Academic Press, New York, 1973. [646] VIRGA, E.: Variational theories for liquid crystals. Chapmans & Hall, London, 1994.
VISINTIN, A.: Strong convergence results related to strict convexity. Comm. P.D.E., 9, 439-466, 1984. [648] VITALI, G.: Sulle funzioni integrali. Atti Acad. Sci. Torino, 40, 3-16, 1905. [649] VITALI, G.: Opere. Edizioni Cremonese, Bologna, 1984. [650] VODOPYANOV, S. K. and GOLDSTEIN, V. M.: Quasiconformal mappings and spaces of functions with generalized first derivatives. Siberian Math. J., 17, [647]
515-531, 1977.
[651] VOL'PERT, A. I.: The space BV and quasi-linear equations. Mat. Sb., 73, 255302, 1967. In Russian. English transl.: Math. USSR-Sb. 2 (1967) 225-267.
694
BIBLIOGRAPHY
[652] VOL'PERT, A. I. and HUDJAEV, S. I.: Analysis in classes of discontinuous functions and equations of mathematical physics. Nijhoff, Dordrecht, 1985. [653] VOLTERRA, V.: Legons sur l'integration des equations differentielles aux derivees
partielles. Almgvist&Sviksell, Upsala, 1906. Second edition: Hermann, Paris, 1912.
[654] VOLTERRA, V.: Sur 1'equilibre des corps elastiques multiplement connexes. Ann. Sc. Norm. Paris, 24, 401-518, 1907. [655] VOLTERRA, V.: Opere Matematiche, voll_5. Accademia Nazionale dei Lincei, Roma, 1954-62.
[656] VOLTERRA, V. and VOLTERRA, E.: Sur les distorsion des corps elastiques. Theorie et applications. Gauthier-Villars, Paris, 1960. [657]
SILHAVY, M.: The mechanics and thermodynamics of continuous media. Springer, Berlin, 1996.
[658] WENTE, H.: An existence theorem for surfaces of constant mean curvature. J. Math. Anal. Appl., 26, 318-344, 1969. [659] WENTE, H.: The Dirichlet problem with a volume constaint. manuscripta math., 11, 141-157, 1974.
[660] WENTE, H.: The differential equation Ax = 2Hxu,Ax with vanishing values. Proc. AMS, 50, 59-77, 1975. [661] WEYL, H.: The method of orthogonal projection in potential theory. Duke Math. J., 7, 411-440, 1940. [662] WHEEDEN, R. L. and ZYGMUND, A.: Measure and integral. M. Dekker, New York, ????
[663] WHITE, B.: Existence of least area mappings of N-dimensional domains. Ann. Math., 113, 179-185, 1983. [664] WHITE, B.: Regularity of area-minimizing hypersurfaces at boundaries with multiplicity. In Seminar on minimal submanifolds, edited by Bombieri, E. Princeton Univ. Press, Princeton, 1983. [665] WHITE, B.: Mappings that minimize area in their homotopy classes. J. Diff. Geom., 20, 433-446, 1984.
[666] WHITE, B.: The last area bounded by multiples of a curve. Proc. Am. Math. Soc., 90, 230-232, 1984.
[667] WHITE, B.: Infima of energy functionals in homotopy classes of mappings. J. Diff. Geom., 23, 127-142, 1986.
[668] WHITE, B.: Homotopy classes in Sobolev spaces and the existence of energy minimizing maps. Acta Math., 160, 1-17, 1988. [669] WHITE, B.: A new proof of the compactness theorem for integral currents. Comment. Math. Helv., 64, 207-220, 1989. [670] WHITE, B.: , 1993. Private communication. [671] WHITEHEAD, G. W.: Homotopy Theory. MIT Press, Cambridge, 1966. [672] WHITNEY, H.: Analytic extensions of differentiable functions defined in closed sets. Trans. Am. Math. Soc., 36, 63-89, 1934. [673] WHITNEY, H.: On totally differentiable and smooth functions. Pacific J. Math., 1, 143-159, 1951. [674] WHITNEY, H.: Geometric integration theory. Princeton Univ. Press, Princeton, 1957.
[675] WIENHOLTZ, D.: A method to exclude branch points of minimal surfaces. Calc. Var., to appear, 1997. [676] YAU, S. T.: Survey on partial differential equations in differential geometry. In
Seminar on differential geometry, edited by Yau, S. T. Princeton Univ. Press, Princeton, 1982.
[677] YAU (ED.), S. T.: Seminar on differential geometry. Princeton univ. Press, Princeton, 1982.
BIBLIOGRAPHY
695
[678] YOUNG, L. C.: Generalized curves and the existence of an attained absolute minimum in the calculus of variations. Comptes Rendus Soc. Sci. et Lettres Varsovie, 30, 212-234, 1937. [679] YOUNG, L. C.: Generalized surfaces in the Calculus of Variations. Ann. of Math., 43, 84-103, 1942.
[680] YOUNG, L. C.: Generalized surfaces in the Calculus of Variations II. Ann. of Math., 43, 530-544, 1942. [681] YOUNG, L. C.: Some extremal questions for simplicial compexes V : The relative
area of a Klein bottle. Rend. Circ. Mat. Palermo, 12, 257-274, 1963. [682] YOUNG, L. C.: Lectures on the Calculus of Variations and optimal Control Theory. Saunders, 1969. Reprinted by Chelsea 1980. [683] ZHANG, D.: The existence of nonminimal regular harmonic maps from B3 to S2. Ann. Sc. Norm. Sup. Pisa, 16, 355-365, 1989. [684] ZHANG, K.: Biting theorems for Jacobians and their applications. Ann. Institut H. Poincare Anal. Non Lineaire, 7, 581-621, 1990. [685] ZHOU, Y.: On the density of smooth maps in Sobolev spaces between two manifolds. Ph.D. thesis, Columbia University, 1993. [686] ZIEMER, W. P.: Weakly differentiable functions. Springer-Verlag, New York, 1989.
Index
a.e.
- for BV-functions, 1-331
- continuous, 11-245 - homeomorphism, 11-245 - open, 11-244, 11-245 absence of fractures, 11-167 Acerbi-Fnsco l.s.c. theorem, 11-119, 11-120 adjoints, 1-238 admissible - charts, 1-549 - deformations, 11-143 Allard boundary regularity theorem, 11-576 - regularity theorem, 11-576
- for Cart', 1-307 - for Cartesian maps, 1-296 - for radial currents, 1-448
Almgren (n - 2) partial regularity theorem, 11-570 anchorage - strong, II-84, 11-262, 11-431 - weak, II-262, II-431, 11-437 A'-maps, 1-229, 1-642 AP-maps, 1-293 Ap,q maps, 1-255 approximate - continuity, 1-33, 1-211 - differentiability, 1-215, 1-640 in the LP-sense, 1-197
-- of BV-functions, I-369 - of Sobolev maps, 1-216 - dipole of length 2, 11-405 - limit, 1-33, 1-210, 1-214 - lower limit, 1-214 - tangent space, 1-92, 1-226
-- to measures, 1-94 - upper limit, 1-214 approximation - p-strong, 1-308 - p--strong, 1-308
- for cart(fl x S'), 11-609 - for cart °'(f2 x S2), 11-365 - for cart2°'(f2 x SZ), 11-419
- for cart','(f? x S2), 11-412
- for Sobolev maps, 1-632
- strong for i.m. rectifiable currents, 1-160, 1-504
- strong for normal currents, 1-161, 11-102
- weak polyhedral, 1-159 area - formula, 1-71, 1-74, 1-220, 1-221
- of rectifiable graphs, I-225 - parametric polyconvex extension, 11-47 - relaxed functional, 11-582, II-610, 11-640
area minimizers - boundary regularity, 11-576 - interior regularity, 11-570 - minimal surfaces, 11-292 axially symmetric - Cartesian currents, 11-459, 11-461 - maps, 11-459 Baker-Ericksen inequality, 11-175 Ball - theorem on global invertibility, 11-242
- theorem on Young measures, 1-63 Banach indicatrix, 1-75, 1-218, 1-220 Banach-Saks theorem, 1-45, 11-13 Banach-Steinhaus theorem, 1-36 Beppo Levi theorem, 1-7 Bernstein - problem, 11-589 - theorem, 11-589 Besicovitch covering theorem, 1-30
Besicovitch-Federer structure theorem, 1-98
Bethuel - approximation theorem, 1-637, 11-396 - density theorem, 1-636
698
Index
- theorem, 11-349 Bethuel-Zhang - theorem, 11-393, 11-413, 11-420 Binet formula, 1-118, 11-252
biting - convergence, 1-64 - lemma, 1-65 BMO-space, 1-260 Bombieri-De Giorgi-Giusti theorem, 11-589
Bombieri-De Giorgi-Miranda theorem, II-581
Bore' - function, I-10 - measure, I-10 - or-algebra, I-10 branch points, 11-623 Brezis-Coron theorem, 11-378 Brezis-Coron-Lieb theorem, 11-345, 11-402 bubbling off - of circles, 1-412, 11-592 - of cylinders, 1-417, 11-593
- of general vertical parts, 1-418 - of given degree, 11-360 - of spheres, 1-414, 11-356 - of vertical parts of different dimensions, 1-420
BV-functions, 1-327, 1-330 - approximate differentiability, 1-369
- approximation, I-331 - Cantor part, 1-357 - coarea formula, 1-337 - compactness, 1-336 - compositions and products, 1-486 - in one variable, 1-479 - jump part, 1-357 - slicing, 1-484 - Sobolev-Poincare inequality, 1-335, 1-366
- special, 1-487 - structure theorem, 1-371 - traces, I-482 - variational characterization, 1-335 Caccioppoli inequality, 11-313 Caccioppoli sets, 1-328, 1-340, 11-564
- approximation, 1-342 - compactness, 1-342 - measure theoretic boundary, 1-329, 1-352
- of minimal perimeter, 11-565 - perimeter, 1-328 - reduced boundary, 1-328, 1-341
Calderon-Zygmund - decomposition argument, 1-188 - theorem, 1-197 calibration, 11-580 Cantor set, 1-15 Cantor-Vitali function, 1-19, 1-408 Caratheodory - criterion, 1-10 - theorem, 11-27 Cartesian currents, 1-385
- approximation in (1 x Si, 11-609 - axially symmetric, 11-459, 11-461 - closure theorem, 1-387, 1-645 - codimension one, 1-329 - homological part, 11-507
- in ,fl x S', 11-600 - in manifolds, 1-644 - reduced, 11-502 - structure theorem, 1-391, 1-645 - weak convergence, 1-387 - with Cantor mass on the minors, 1-423 - with fractures, 11-476 Cartesian maps, 1-233, 1-286 - closure property, 1-292 - compactness property, 1-292 - join, 1-313 p-Cartesian maps, 1-293, 1-297 - closure property, 1-294 - compactness property, 1-294 u-Cartesian maps, 1-296 Cauchy stress tensor, 11-155, 11-267 Cauchy-Binet formula, 1-73, 1-109 cavitation, 11-268 change of variable formula, 1-72, 1-75, 1-222
chiral model, 11-376 Christoffel symbols
- of the first kind, 11-284 - of the second kind, 11-284 closure
- for cart2°' (,fl x y), 1-647 - for Cartesian currents, 1-387, 1-645 - for graph currents, 1-387 - for radial currents, 1-448 - homology classes, 1-585, 1-588, 1-589, 1-599, 1-610, 1-618, 1-619
- of cart' (Q, I ' ), 1-292 - of carte(Q,RN), 1-294
- of dif P,9 (.fl, 0), 11-186
- of dif','(Q, dl), 11-187 of dif P'4(Q, IE8"`), 11-196
of dif',I (S?, EV), 11-197
Index
- sequential weak, 1-302 coarea formula, 1-72, 1-82, 1-337 coboundary operator, 1-567 codifferential operator, 1-538
-- in dif P'4, II-185
cohomology, 1-547
-
- cochains, 1-613 - long sequence, 1-568, I-613 - map, 1-649, 1-658 collar theorem, 1-592 comass, 11-39, 11-41
- of a k-covector, I-110 - of a k-vector, I-110 compactness
- for Cartesian currents, 1-387 - for currents of finite mass, 1-127 - for i.m. rectifiable currents, 1-141 - in cart 1(42,IRN), 1-292 - in cartP4Q,lR' ), 1-294
- in carte' (9 x Y), 1-647 - in dif P'9(Q,12), 11-186 - in dif 1'1(42, fl), 11-187 - in if P'9 (0, II811-196
- in dif 1'1(9, k), II-197 - in BV, 1-336 compositions, 11-253, 11-257
concentration-compactness lemma, 1-56 cone over a current, 1-135 conformal
- invariance of the Dirichlet integral, 11-287
- invariant integrals, 11-548 - maps, 11-371 - metric, 11-286 conformality relations, 11-290 connection, 1-559
- Levi-Civita, 1-561 constancy theorem, 1-130, 1-452, 1-456, 1-573, 1-574
constitutive condition, 11-150, 11-151, II-175
continuous selection theorem, 11-19 convergence
- ACP of minors, 1-399 - in mass norm, 1-36 - in measure, 1-7 - in the biting sense, 1-64 - in the sense ACP, 1-66 - quasi, 1-37 - weak
-- for Cartesian currents, 1-387 -- in A1, 1-232 -- in AP(Q,RN), 1-293
-
699
in dif P'4, 11-194
-- in L' (X; IL), 1-7 in LP, 1-39 in the sense of measure or as
measures, 1-39
-- of currents, 1-123 - of measures, 1-36 -- with the mass, 1-36 - weak' in L°°, 1-44 convex
- envelop, 11-27 - hull, 11-27 count ably measurable - functions, 1-9
- sets, 1-9 counting - function, 1-75, 1-218, 1-220 - measure, 1-14 Courant-Lebesgue lemma, 11-236, 11-334 covering theorem - Besicovitch, 1-30 - Vitali, 1-30 critical point of D(u), 11-284, 11-291 currents, 1-123 - boundary, 1-123
- carried by M, 1-137 - carried by smooth graphs, 1-137 - Cartesian product, 1-124 - cone over, 1-135 - deformation theorem, 1-157 - homotopy formula, 1-134 - i.m. rectifiable, 1-140
-
closure, 1-141
- compactness, 1-141 -- decomposition, 1-459
-
image, 1-149 push forward, 1-149
-- strong approximation, 1-160 -- weak polyhedral approximation, 1-159
- image, 1-132, 1-133 - integer multiplicity rectifiable, 1-140
- integral, 1-513 - intersection, 1-599, 1-600 - Kronecker index, 1-605, 1-607 - lift, 1-581 - mass, 1-125 - mollifying, 1-505 - normal, 1-129 -- of maximal dimension, characterization, 1-454
-- representation, 1-510
700
Index
-- strong approximation, 1-161 - of finite mass -- compactness-closure, 1-127 - Mk(U), 1-126 - of type r(M, 0, C), 1-140 - on submanifolds, 1-572 - push forward, 1-132 - radial, 1-442 - approximation, 1-448 -- closure theorem, 1-448 - rectifiable, 1-140 - restriction, 1-127 - slice, 1-154, 1-156
- support, 1-124 curvature tensor, 1-563 D-field, 11-387, 11-388, 11-474, 11-487, 11-488
D-map, 1-649 d-open, 11-218 do-connected, 11-218 De Giorgi - decay estimate, 11-567 - l.s.c. theorem, 11-132 - rectifiability theorem, 1-328, -346
- regularity theorem for elliptic equations, 11-579
- regularity theorem for minimal sets, 11-566
de Rham - cohomology, 1-547 - duality, 1-570, 1-584, 1-598
- on manifolds with boundary, 1-598 on manifolds without boundary, 1-584 duality on Lipschitz category, 1-613 - relative cohomology, 1-558 theorem, 1-584, 1-585, 1-598, 1-614 decomposition theorem, 1-491
- for i.m. rectifiable currents of maximal dimension, 1-459 - Hodge-Kodaira-Morrey, 1-546 - Hodge-Morrey, 1-557 deformation, 11-141 - elastic, 11-168
- infinitesimal, 11-143 - local, 11-164 - theorem, 1-498, 1-503 - with fractures, 11-264 degree, 1-227, 1-460, 1-462, 1-656
- mapping, 1-461, 1-656 density, 1-33
- lower, 1-34 - topology, 11-217
- upper, 1-34 determinant - distributional, 1-251 - of the product of two matrices, 11-252 - of the sum of two matrices, 1-313 Dif P"' (0, b)-maps, 11-213 differentiable functions, 1-205 dipole - approximate, 11-405 - for liquid crystals, 11-472 - irrotational, 11-480 - of length $, 11-403 - problem, 11-400, 11-449, 11-489, 11-492 - Riviere, 11-459, 11-464
- solenoidal, II-481 Dirichlet conditions - strong anchorage, 1-270 - weak anchorage, 1-270 Dirichlet growth theorem of Morrey, 11-298
Dirichlet integral, 1-540, 11-282
- existence among Cartesian currents, 11-430
- minimizers among Cartesian currents, 11-370, 11-430
- parametric polyconvex extension, 11-49, 11-51, 11-57, 860, 11-61
- relaxed, 11-425 - singular perturbations, 11-456 disintegration of measures, 1-35 distribution function, 1-180 distributional - determinant, 1-251 - divergence, 1-248 divergence theorem, I-100 double Lusin condition, 11-245 Douglas-Rado theorem, 11-621 dyadic decomposition, 1-188 Fells-Sampson theorem, 11-337 Egoroff theorem, 1-6 elastic deformation, 11-168, 11-199 energy density, 11-282 energy minimizing - elastic deformations, 11-262 - weak harmonic maps, 11-294 energy-momentum tensor, 11-266, 11-288 envelop - convex, 11-27 - convex l.s.c., 11-28 - parametric polyconvex l.s.c., 11-38, 11-43
- polyconvex, 11-35
Index
- polyconvex l.s.c., 11-35 equi-integrable, 1-39 Ericksen-Leslie theory, 11-467 essentially connected, 11-220 Euler - variation, 11-153
- cochain, 1-526, 1-613
- variation in the deformed state, 11-153
flatness theorem, 1-522, 1-573 Fleming theorem, 11-623 Fleming-Rishel theorem, 1-338 forms - exterior derivative, 1-119, 1-529
Evans theorem, 11-346 example - Hardt-Lin, 11-406
- Neumann, 1-303 - Riviere, 11-463 - stored energies, 11-179 excess, 11-567
existence of minimizers - in An-1,n/(n-1), 11-263 - in classes of Cartesian currents, 11-85, II-539
- in classes of Cartesian maps, 1-292, 11-84
- in classes of weak diffeomorphisms, 11-262
- in homology classes, 1-585, 1-589, 1-599, 1-611, 1-618, 1-619
extension - for Sobolev maps, 11-547 - lemma, 11-340, 11-439
Fatou lemma, 1-7 Federer - closure theorem for homology classes, 1-588, 1-589, 1-610
flatness theorem, 1-522 - semicontinuity theorem, 11-129 strong approximation theorem, 11-96,
-- representation theorem, 1-526 - integral distance, 1-513 - integral norm, 1-512 integrand, 11-97
- norm, 1-514
- harmonic, 1-539 - integration along the normal, 1-578 - mollifying, 1-505 - normal part, 1-549 - null to a submanifold, 1-530 - periods, 1-585, 1-598 - pull-back, 1-120 - relatively closed, 1-558 - relatively exact, 1-558 - Sobolev, 1-537 - tangential part, 1-549 formula - area, 1-71, 1-74, 1-220,1-221 - Binet, 1-118, 11-252 - Cauchy-Binet, 1-73 - change of variable, 1-72, 1-222 - coarea, 1-72, 1-82, 1-337 - determinant of the sum, 1-313 - homotopy for currents, 1-134 - homotopy for forms, 1-135 - Kiinneth, 1-652 - Laplace, 1-238 - monotonicity, 11-320, 11-435, 11-566, 11-576
II-97
- Wietzenbock, 1-564
support theorem, 1-525 theorem on approximate differentiabil-
fracture, 11-167, 11-168, 11-476 frame indifference, 11-169
ity, 1-217
Fubini theorem, 1-9 function - integrable, 1-5 - measurable, 1-5 - of bounded variation, 1-329 - polyconvex, 11-150 - quasiconvex, 11-173 - rank-one convex, 11-112
- theorem on the dimension of singular set, 11-569
Federer-Fleming closure theorem, 1-141 -- for homology classes, 1-618, 1-619 - deformation theorem, 1-157, 1-503 Fefferman theorem, 1-261 five lemma, 1-597 flat - chain, 1-515, 1-521
- - integral, I-512 - mollifying, I-522 push-forward, 1-520
- - representation, 1-523
701
Gaffney theorem, 1-543, 1-553 gamma function, 1-13 gap phenomenon, 11-405, 11-460, 11-556, II-559 Garding inequality, 11-115 Gauss-Green
702
Index
- formula, 1-328, 1-340, 1-346
- theorem, 1-272 Giaquinta-Giusti theorem, 11-312 globally invertible maps, 11-192, 11-238
gradient relative to M, 1-102 graph currents, 1-230, 1-385 - closure theorem, 1-387 graph or rectifiable graph, 1-224 Helein theorem, 11-346
Hadamard determinant theorem, 11-64 Hahn-Banach separation theorem, 11-27 Hamilton energy-momentum tensor, 11-155, 11-266, 11-288
Hamilton theorem, 11-337 Hardt-Kinderlehrer-Lin theorem, 11-343 Hardt-Lin theorem, 11-352, 11-406
Hardt-Simon boundary regularity theorem, 11-576 Hardy space, 1-260 Hardy-Littlewood - maximal function, 1-180 - maximal theorem, 1-181 harmonic form, 1-539, 1-544 harmonic map, 11-283 - energy minimizing, 11-307, 11-308, 11-319, 11-327
- stationary, 11-291 - weak, 11-459 Harnack theorem, 11-587 Hausdorff - dimension, 1-14 - measure, 1-13 - measure and connected sets, 11-491 higher integrability - of the determinant, 1-258 - of the gradient, 11-314, 11-342 Hildebrandt regularity theorem, 11-622 Hildebrandt-Jost-Widman theorem, II-631
Hildebrandt-Kaul-Widman theorem, 11-308
Hodge
- operator, 1-535 - theorem, 1-547 Hodge-Kodaira-Morrey decomposition theorem, 1-546 Hodge-Morrey theorem, 1-556, 1-557 homological - closed sets, 1-474 - connected components, 1-478 - disconnected sets, 1-474 homology
- closure of classes, 1-585, 1-588, 1-589, 1-599, 1-610
- duality isomorphism, 1-575, 1-582 - integral, 1-616 - integral relative, 1-619 - intersection of cycles, 1-604 - long sequence, 1-591, 1-609 - map, 1-650, 1-661 - of Lipschitz chains, 1-586 - relative, 1-590 - simplicial, 1-586, 1-616 homotopy formula - for currents, 1-134 - for flat chains, 1-520 - for forms, 1-135 homotopy smoothing operator, 1-507 Hopf - degree, 11-558 - differential, 11-293 - map, 11-560, 11-630 Hurewitz isomorphism theorem, 11-633 impenetrability of matter, 11-142, 11-157 index - intersection, 1-607 - Kronecker, 1-605, 1-607 inequality - Baker-Ericksen, 11-175 - Caccioppoli, 11-313 - Garding, 11-115 - isodiametric, 1-15 - isoperimetric, 1-158, 1-344,1-587, 1-610, 1-617, 1-619, 11-49, 11-65, 11-68
- Jensen, 11-16 - monotonicity, 11-345 - Morrey, 1-193 - Poincare, 1-327 - Sobolev, 1-327 - Sobolev-Poincare, 1-336, 1-366 - Wirtinger, 11-570 inner - extremal, 11-288 - strong extremal, 11-290 - variation, 11-288 instanton, 11-376 integrable function, 1-6 integrals - coercive, 11-3 - conformally invariant, 11-548 - lower semicontinuous, 11-3 - parametric, 11-24, 11-44 - parametric extension, 11-45 - regular, II-1, 11-23, 11-42
Index
- relaxed, II-6, 11-7 integrand - X-quasiconvex, 11-127 - flat, 11-97 - parametric, 11-75 - quasiconvex, 11-108, II-111 - rank-one convex, 11-112 interior multiplications, 1-533 intersection - currents, 1-600 - index, 1-571, 1-607 - map, 1-604 - of cycles, 1-604 Ioffe l.s.c. theorem, 11-132 irrotational dipole, 11-480 isodiametric inequality, 1-15 isoperimetric inequality, 1-158, 1-344, I-587, I-610, I-617, I-619, II-49, 865, 11-68
- for the determinant, 1-253 isotropic material, 11-170
Jager-Kaul theorem, 11-308 Jacobian, 1-71, 1-72, 1-102, 11-214 Jensen inequality, 11-16 John-Nirenberg space, 1-260 join
- of Cartesian maps, 1-313 - of diffeomorphisms, 11-255 Jost theorem, 11-308, 11-378 jump points, 1-329
Kirszbraun theorem, 1-202 Kronecker index, 1-605, 1-607 Kunneth formula, 1-652 Laplace formulas, 1-238 Laplace-Beltrami operator, 1-539, 1-540, 11-284
Lebesgue - area, 11-649
- decomposition theorem, 1-24 - differentiation theorem, 1-183 - dominated convergence theorem, 1-7 - extension, 11-7 - outer measure, 1-5 - representative, 1-185 - set, 1-185 - value, 1-185 Lebesgue-Besicovitch differentiation theorem, 1-32 Legendre-Fenchel transform, 11-29 Legendre-Hadamard condition, 11-113 Lemaire theorem, 11-337, 11-371, 11-373
703
lemma - biting, 1-65 - extension, 11-340, 11-439 - five lemma, 1-597 Levi-Civita connection, 1-561 Lipschitz functions, 1-202 liquid crystals - D-field, 11-474 - cholesteric, 11-469
- existence among Cartesian currents, 11-471
- existence in Sobolev spaces, 11-470 - general dipole problem, 11-484 - nernatic, 11-468 - parametric extension, 11-72, 11-470 - regularity in Sobolev spaces, 11-470
- relaxed energy among Cartesian currents, 11-471 - relaxed energy in Sobolev spaces,
II-475 Liu theorem, 1-207 L log L-space, I-191 long sequence - cohomology, 1-568,1-613 - homology, I-591 Lusin - double (N) condition, 11-245
- property (N), I-11 - representative, 1-222 - theorem, 1-12 Malt' - approximation theorem, 1-303 - l.s.c. theorem, 11-122 mass, 1-28, 1-125
- Euclidean, 1-28, 1-125 - lower semicontinuity, 1-127 - of a k-vector, 11-39 material body - part, 11-156 maximal function, 1-180 mean curvature, I-100, 1-101, 11-628 - generalized, 11-575 measurable - functions, 1-5 - sets, 1-3 measure - absolutely continuous, 1-24 - Borel, 1-10 - Borel regular, I-10 - Caratheodory construction, 1-22 - countably additive, 1-2 - counting, 1-14
704
Index
- disintegration, 1-35 - Hausdorff, 1-13 - image, 1-8 - Lebesgue, 1-5 - outer, 1-3 - positive and negative parts, 1-25 - product, 1-9 - Radon, I-10 - regular, 1-4 - restriction, 1-4 - signed, 1-25 - spherical, 1-14 - total variation, 1-25, 1-26 - vector valued, 1-25 - Young, 1-59 measure-theoretic - boundary, 1-32, 1-329, 1-352 - exterior, 1-32 - interior, 1-32 measures mutually singular, 1-24 Meeks-Yau theorem, 11-624 metric tensor, 1-81 Meyers-Serrin theorem, 1-632 minimal cone, 11-568 minimal connection - of the singularities, 11-415
- of the singularities allowing connections to 9s2, 1221 minimal graphs - existence in X x S1, 11-614 - existence in homology classes of
XxS1,II-615 - existence with prescribed homological singularities in X x Si, 1417 minimal integral connection - allowing connections to 351, 11-420 - in 51, 11-420 - in 12, 11-413, 11-459 minimal perimeter, 11-565 minimal real connection - in 51, 11-421 - in 12, 11-413, 11-459 minimal surface, 11-292
- BV approach, 11-582 - Dirichlet problem, 11-581, 11-582, II-586 - equation, 11-579 - in higher codimension, 11-630 - Liouville type theorem, 11-587 - non parametric, 11-579 - parametric, 11-292 - regularity, 11-586 - removable singularities, 11-587 - system, 11-628
minors, 1-104
- approximation, 1-303 - weak continuity, 1-288, 1-643 mollifying
- currents, 1-505 - flat chains, 1-522 - forms, 1-505 monotonicity - for minimal Caccioppoli sets, 11-566 - formula, 11-320, 11-435, 11-566, 11-576
- inequality, 11-345 Morrey - e-conformality theorem, 11-360, 11-620 - inequality, 1-193 - l.s.c. theorem, 11-108, II-111, 11-117 - regularity theorem, 11-304 Morrey-Sobolev theorem, 1-200 Moser-Harnack theorem, 11-587 Mucci theorem, 11-650 Miiller
- theorem on higher integrability of the determinant, 1-258
- theorem on the distributional determinant, 1-252 Miiller-Tang-Yan theorem, 1-256 multiplicity function, 1-75 multivectors, 1-104 - simple, 1-107 Neumann example, 1-303 Noll condition, 11-172 normal currents of maximal dimension
- characterization, 1-454 null Lagrangian, 11-112
ordinal numbers, 1-299 Oseen-Frank energy, 11-68, 11-467 - parametric polyconvex extension, 11-72 outer measure, 1-3 p-Dirichlet integral, 11-552 - parametric extension, 11-66 pairing - de Rham, 1-570, 1-584, 1-598
- de Rham-Federer, 1-614 - Poincare, 1-570 - Poincare-Lefschetz, 1-569 parametric extension - p-Dirichlet integral, 11-66 - area, 11-47 - Dirichlet integral, 11-51, 11-53, 11-57, 11-61
- liquid crystals, 11-72, 11-470 - Skyrme like integrals, 11-68
Index
- total variation, 11-48 parametric integrals - elliptic, 11-127
- Diff-, 11-174 - X-, 11-127
- lower semicontinuity, 11-77, 11-82 parametric integrand, 11-75 part of a body, 11-156
Rado theorem, 11-625 Rademacher theorem, 1-203 radial currents, 1-442
partial regularity for minimizers of the
- approximation, 1-448 - closure theorem, 1-448 Rado-Reichelderfer theorem, 1-223 Radon measure, I-10 Radon-Nikodym - derivative, 1-24 - theorem, I-24, 1-31 Radon-Riesz theorem, 1-47
relaxed energy, 11-434 perimeter, 1-328, 1-340 Piola identities, 1-244 Piola-Kirchhoff stress tensor, 11-152, 11-154, 11-267
Plateau problem, 11-290 Poincare dual form, 1-583
-- on manifolds with boundary, 1-597 -- on open sets, 1-576 duality isomorphism, 1-575, 1-582
- - on compact manifolds without boundary, 1-582
-- on open sets, 1-574, 1-576 duality theorem, 1-566 - inequality, 1-327 - lemma, 1-136 - pairing, 1-570 duality isomorphism, 1-583 Poincare-Lefschetz
- duality theorem, 1-569 - isomorphism theorem, 1-592 - pairing, 1-569 polar coordinates, 1-89 polar decomposition, 1-70 polar function, 11-29 polyconvex - envelop, 11-35 - function, 11-34, 11-150 - parametric l.s.c. envelop, 11-38, 11-43 polyhedral chains - integral, 1-160 - with real coefficients, 1-160 Preiss theorem, 1-98 preservation of orientation, 11-168 principle of transfinite induction, 1-299
property - p-strong approximation, 1-308 - Lusin, I-11 - tension extension, 11-175 - Vitali symmetric, 1-29 quasiconvex functions, 11-108, II-111 quasiconvexity - Carte, 11-127 - carte, 11-127
705
recession function, 11-583
rectifiability - boundary, 1-163 - for currents, 1-166 - for measures, 1-94 - for sets, 1-98 rectifiable sets, 1-91 - countably, 1-90 - image, 1-221 - orientation, 1-122 reduced - boundary, 1-328, 1-341 - Cartesian currents, 11-502 - Sobolev classes, 11-553 regular ball, 11-308 relative homology, 1-589 relaxed energy, 11-6, 11-7
Reshetnyak - approximation theorem, 11-90, 11-92 - l.s.c. theorem, 11-22, 11-80 - theorem, 1-281, 1-282 reverse Holder inequalities with increasing supports, 11-299 Riemann-Lebesgue lemma, 1-46 Riesz theorem, 1-27 Riviere - dipole, 11-459, 11-464 - example, 11-459, 11-463 Sacks-Uhlenbeck theorem, 11-337, 11-339
Sard - theorem, 1-89 - type theorem, 1-89 SB V-functions, 1-487 Schoen-Uhlenbeck - boundary regularity, 11-333 - density theorem, 1-634 - regularity theorem, 11-319 Schoen-Yau theorem, 11-336
706
Index
a-algebra, I-1 v-measure, 1-2
- approximation for cart 2'1(l2 x 52),
simple vectors, 1-107 - characterization, 1-115, 11-57 - convex hull, 11-33 - in V,,,k, 11-57, 11-65 simplicial homology, 1-586, 1-616 singular values of a matrix, 11-170 Skyrme - model, 11-560
- approximation for
- parametric extension, 11-68 slice
- of BV-functions, 1-484 - of a current, 1-154, 1-156 Sobolev
- inequality, 1-327 - spaces, 1-179 Sobolev-Poincare inequality, 1-336, 1-366 solenoidal dipole, II-481 spherical cycles, 11-497 spherical measure, 1-14 stationary point, 11-291 stereographic projection, 1-414 - modified, 11-358 Stokes theorem, 1-272 stored energy density, 11-147 stress, 11-152
strong approximation - for i.m. rectifiable currents, 1-160, 1-504 - for normal currents, 1-161 structure theorem - for cartf x S1), 11-606, 11-608 - for cart '1(Q x S2), 11-363, 11-409, 11-494
- for cart2'1(Q x y), 1-646 - for cart2'1(,f2 x y), 11-498 - for BV-functions, 1-371 - for Cartesian currents, 1-391, 1-645 - for weak diffeomorphisms, 11-194 subgraph, 1-371 summable function, 1-6 support theorem, 1-525
tangent n-vector to graphs, 1-113 theorem - De Giorgi decay estimate, 11-567 - absolute continuity of integral, 1-7 - Acerbi-Fusco l.s.c., 11-119, 11-120
- Allard boundary regularity, 11-576 - Allard regularity, 11-576
- Almgren on (n - 2) partial regularity, 11-570
- approximation for cart(,f2 x S'), 11-609
11-365
_ x S2),
11-412
- approximation for BV-functions, 1-331 - approximation for radial currents, 1-448 - approximation in cart 2'1(Q x S2)111-419
- Ball global invertibility, 11-242 - Ball on Young measures, 1-63 - Banach-Saks, 1-45, 11-13 - Banach-Steinhaus, 1-36 - Beppo Levi, 1-7 - Bernstein, 11-589 - Besicovitch covering, I-30 - Besicovitch-Federer structure, 1-98 - Bethuel, 11-349 - Bethuel approximation, 1-637, 11-396 - Bethuel density, 1-636 - Bethuel-Zhang, 11-393, 11-413, 11-420 - Bombieri-De Giorgi-Giusti, 11-589 - Bombieri-De Giorgi-Miranda, 11-581 - boundary rectifiability, 1-163 - Brezis-Coron, 11-378 - Brezis-Coron-Lieb, 11-345, 11-402 - Calderon-Zygmund, 1-197 - Carathdodory, 11-27
- characterization of Cart', 1-307 - closure for Cartesian currents, 1-387, 1-645
- closure for graph currents, 1-387 - closure for radial currents, 1-448 - collar, 1-592 - compactness for i.m. rectifiable currents, I-141 - compactness for varifolds, 11-575 - compactness-closure for currents of finite mass, 1-127 - concentration-compactness, 1-56 - constancy, 1-130, 1-452, 1-456, 1-573, 1-574
- continuous selection, 11-19 - Courant-Lebesgue, 11-236, 11-334 - De Giorgi l.s.c., 11-132
- De Giorgi on the regularity for elliptic equations, 11-579
- De Giorgi on the regularity for minimal sets, 11-566
- De Giorgi rectifiability, 1-328, 1-346 - de Rham, 1-584, 1-585, 1-598 - de Rham-Federer, 1-614 - decomposition, 1-491
Index
- decomposition for i.m. rectifiable current of maximal dimension, 459 - deformation, 1-498, 1-503 - Dirichlet growth, 11-298 - divergence, I-100 - Douglas-Radd, 11-621 - Eells-Sampson, 11-337 - Egoroff, 1-6 - Evans, 11-346
- Fatou, 1-7 - Federer on approximate differentiability, 1-217
- Federer on the dimension of singular set, 11-569
- Federer semicontinuity, 11-129 - Federer strong approximation, 11-96, 11-97
- Federer-Fleming closure, 1-141 - Federer-Fleming deformation, 1-157 - Fefferman, 1-261 - flatness, 1-522, 1-573 - Fleming, 11-623 - Fleming-Rishel, 1-338 - Fubini, 1-9 - Gaffney, 1-543, 1-553 - Gauss-Green, 1-272 - Giaquinta-Giusti, II-312 - Helein, 11-346 - Hadamard determinant, 11-64 - Hahn-Banach, 11-27 - Hamilton, 11-337 - Hardt-Kinderlehrer-Lin, 11-343 - Hardt-Lin, 11-352, 11-406
- Hardt-Simon boundary regularity, 11-576
- Hardy-Littlewood maximal, 1-181 - Harnack, 11-587
- higher integrability of the gradient, II-314
- Hildebrandt regularity, 11-622 - Hildebrandt-Jost-Widman, 11-631 - Hildebrandt-Kaul-Widman, 11-308 - Hodge, 1-547 - Hodge-Kodaira-Morrey decomposition, 1-546
- Hodge-Morrey, 1-556 - Hodge-Morrey decomposition, 1-557 - Hurewitz isomorphism, 11-633 - Ioffe l.s.c., 11-132 - Jager-Kaul, 11-308 - Jost, 11-308, 11-378 - Kirszbraun, 1-202 - Lebesgue decomposition, 1-24
707
- Lebesgue differentiation, 1-183 - Lebesgue dominated convergence, 1-7 - Lebesgue-Besicovitch differentiation, 1-32
- Lemaire, 11-337, 11-371, 11-373
- Liu, 1-207 - Lusin, 1-12 - Miiller on higher integrability of the determinant, 1-258 - Muller on the distributional determinant, 1-252 - Miiller-Tang-Yan, 1-256
- Mali approximation, 1-303 - Mali l.s.c., 11-122 - Meeks-Yau, 11-624 - Meyers-Serrin, 1-632 - monotonicity formula, 11-320 - Morrey e-conformality, 11-360, 11-620 - Morrey l.s.c., 11-108, II-111, 11-117 - Morrey regularity, 11-304 - Morrey-Sobolev, 1-200 - Mucci, 11-650 - Poincare, 1-136 - Poincare duality, 1-566 - Poincare-Lefschetz duality, 1-569 - Poincare-Lefschetz isomorphism, 1-592
- Preiss, 1-98 - Rado, 11-625 - Rademacher, 1-203 - Rado-Reichelderfer, 1-223 - Radon-Nikodym, 1-24, 1-31 - Radon-Riesz, 1-47 - rectifiability for currents, 1-166 - rectifiability for measures, 1-94 - rectifiability for sets, 1-98 - rectifiability for varifolds, 11-575 - representation of cochains, 1-526 - representation of flat chains, 1-523 - Reshetnyak, I-281, 1-282 - Reshetnyak approximation, 11-90, 11-92 - Reshetnyak l.s.c., 11-22, 11-80 - Riemann-Lebesgue, 1-46 - Riesz, 1-27 - Sacks-Uhlenbeck, 11-337, 11-339
- Said, 1-89 - Sard type, 1-89 - Schoen-Uhlenbeck, 11-333 - Schoen-Uhlenbeck density, 1-634 - Schoen-Uhlenbeck regularity, 11-319 - Schoen-Yau, 11-336 - semicontinuity, 11-12, 11-18, 11-21, 11-22 - Stokes, 1-272 - strong approximation, 1-504
708
Index
-- for i.m. rectifiable currents, 1-160 - for normal currents, I-161 - structure for cart(i7 x S'), 11-606, 11-608
structure for cart2"1(S? x S2), 11-409
structure for cart2"(fl x y), 11-498 - structure for BV-functions, 1-371 - structure for Cartesian currents, 1-391, 1-645
structure for weak diffeomorphisms, II-194
- support, 1-525 - Tietze, 1-12 - Tonelli, I-10 - Tonelli-Morrey semicontinuity, 11-15 - Vitali, 1-29 - Vitali covering, 1-30 - Vitali-Caratheodory, 1-12 - Vodopianov-Goldstein, 11-236 - weak continuity of minors, 1-643 - weak polyhedral approximation, 1-159 - White, 11-633, 11-635, 11-637, 11-638
- White on mapping area, 11-633 - Whitney, 1-206 Tietze extension theorem, 1-12 Tonelli theorem, I-10 Tonelli-Morrey semicontinuity theorem, 11-15
torsion groups, 1-620
total variation - parametric polyconvex extension, 11-48 traces - of BV-functions, 1-482 - of Sobolev maps, 11-547 transfinite induction, 1-300
- rectifiability theorem, 11-575 - stationary, 11-572, 11-574 Vitali - covering lemma, 1-30 - symmetric property, 1-29 - theorem, 1-29 Vitali-Caratheodory theorem, 1-12 Vodopianov-Goldstein theorem, 11-236 volume preserving - diffeomorphisms, 11-216 - transformations, 11-215 weak - convergence
-
implies strong, 1-48, 1-55, 11-92, 11-95, 11-97
-- of compositions, 11-258, 11-259 -- of inverse maps, 11-202 - deformation, 11-166 - diffeomorphisms, 11-184 -- discontinuous, 11-189 - with variable range, 11-194 - inverse, 11-158 - invertible maps, 11-158 - local diffeomorphisms, 11-229 - one-to-one transformation, 11-158
- polyhedral approximation theorem, 1-159
- topology, 1-302 wedge product, 1-105 White - theorem, 11-633, 11-635, 11-637, 11-638
- theorem on mapping area, 11-633 Whitney theorem, 1-206 Wietzenbock formula, 1-564 Wirtinger inequality, 11-570
unrectifiable sets, 1-97 Young measure, 1-59
variation - Euler, 11-153 - in the deformed state, 11-153 - inner, 11-288 - of the deformed state, 11-153 - of the domain, 11-153 - of the reference state, 11-153 - strong inner, 11-290 varifold, 11-570, 11-573
- Allard theorem, 11-576 - compactness theorem, 11-575 - first variation, 11-574 - i.m. rectifiable, 11-570 - image, 11-571, 11-573
- mean curvature, 11-572
- associated to u, 1-58 Zygmund space, 1-191
Symbols
0 2cc T, 1-135 [w](o,l), 1-124
, 1-154, 1-156 < [T]rey [w]rel >b, 1-598 < [T], [w] >, 1-571, 1-584, 1-598
< [T], [w] >', 1-598
nx, no , n' I-604 ,
x, 1-113, 1-313, 1-444 * operator, 1-535 adj Gad, 1-238
A'(M RN), 1-273 Al (.f2, JRN ), 1-229, 1-642
AP(Q,][8n), 1-293
apDu(x), 1-215 aplimy.. f (y), 1-33 aplimy_,, f (y), 1-210, 1-214 apliminfy-,, f (y), 1-33, 1-214 aplimsupy_,, f (y), 1-33, 1-214 AP,q(Q,R'), I-255 BMO, 1-260 Be(X), I-23 (aG,,)(k), 1-239 8T, 1-123 0E, 1-354 8* E, 1-329 8µE, 1-329, 1-352 a- E, 1-328, 1-341 BV(fl), 1-327, 1-330
cart(fl x Y), 1-646 cart (f2 x RN), 1-385, 1-389 cart' (.fl, y), 1-646
cart 1(Q,RN), 1-233, 1-286, 1-643 cartn(.f2 x y), 11-548, 11-549 cart2'1(,f2 x y), 1-646 cart2,1(d2 x S2), 11-423
cart ,1(i x S2), 11-412, 11-419 cart(X x Y), 1-657 CARTP (fl, RN), 1-298 cartAS (fl, S2), 11-461 Carte (,fl, RN), 1-297 cartP (,fl, RN), 1-293 carte(,f2 x RN), 1-391 cart, (1?, an), 11-229 Carte (.fl, RN), 1-308
Cf(x), 11-27 Cf2'1(Q,S2), 11-476 deg (T, A, y), 1-462 deg (T, R, y), 1-461 deg (u, A, y), 1-227, 1-460 8, I-538
Det, 1-251 V(T,,fl x S2), 11-424 Diff (X, Y), 11-142 dif (fl x ,(2), 11-205
dif (,fl x (2), 11-205 Diff (X, Y), II-142 dif (,fl x ][8n), 11-209 dif P,9 (S2, ,fl), II-184 dif p,q (12, Rn), 11-194 dif (1) (17, .fl), 11-216 P
dif (1) (.fl, ]E8n), 11-216 Div, 1-248
Symbols
710
divMX, I-100 Dk(U), 1-120, 1-122 Dk(U), 1-123 DN,v(x), D,,v(x), DNv(x), 1-31 D(n, G), 11-387, 11-487 DP, 1-531 Dp(u), 11-552 Dp, 1-531 D(u), 11-388, 11-488 D(u), 11-388, 11-412, 11-474, 11-488 (Du)(c), 1-357 (Du)(i), 1-357 dx", 1-119 Ek(U), 1-120 Ek(U), 1-134 y, II-355 Em y, II-368
E,
, ,
E(T, !2 x S 2 ), 11-470 -c-K, 1-512
FK, I-514
.,
,
K, 1-512
,7m , cpt, 1-512
FIn I-521 Fm,cpt, 1-515
f#T, 1-132, 1-149 Fn (P), 11-421 Fo (P), 11-413
FCf, 11-28 FFf, 11-35 1(T)-f, 11-6
Hk(A, B; G), 1-620 Hk(A, Z), 1-616 Hk (X, IR), I-570 Hk (X, aX, IR), 1-589 7{s, 1-13
H' (A), I-13 7{00, 1-35
Z,,,,,K, 1-513
ix(S,T), 1-607 Jf(x), 1-74, 1-82
Jf (x), 1-102 J,,,(dM f (x)), I-103
k(S,T), 1-605 A k L, 1-117
AkV , 1-106 Al, I-Ao, A+, 11-31 AkL, 1-116 AkV, 1-105 L log L, 1-191 Ll (X; p), 1-6 LP(X; µ), 1-39 LT, 11-364, 11-409, 11-494 L(T), 11-498 L(T, a), 11-497 L,, (T), 11-498 L(u), 11-415, 11-475 Au, 1-185 µ L A, 1-4
I'(kl, k2, k3), 11-471 graph(.f2 x R N), 1-385 Gu, 1-137, 1-230 Gu,an, 1-273
f.G L f , 1-24
Gu(k), 1-231
Mu (v), 1-28 M, 1-125 M(G), 1-114 M(k) (G), 1-115 M(GR), 1-116, 11-32 mi (T), 11-102 mi(P), 11-413 m°,n(P), 11-420
gJu,n, 1-137, 1-224
H2ph (y, Z), 11-497 7-11(IR'' ), 1-260
H pt (X \ aX ), 1-593 HHR(X), 1-547 HHR(X, aX), 1-558
QM 1, 1-136
Ma (G), 1-104, 1-114
M(T), I-125
Symbols
Mk(U), 1-126 M+(X), M(X), M(X,]EYm), 1-25 mr(r), 11-102 mr(P), 11-413
M(u)(x),I-180 VM f , I-99 VM f (x), 1-102 VXC 1-559 N(f, A, y), 1-75 N(u, A, y), 1-220 Nk(U), 1-129 NA(T), 1-129
SBV, 1-487 set (T), 1-140 SG,,, 11-597 Sg,,,Q, I-371 E++ 11-143 E, i I-Eo, E+, 11-31 ST,sing, II-499, II-549 strong-c1AP (C' n AP (f?, RN)), 1-298 succw-limx C, 1-301 sw-clx C, 1-302 sw-limx C, 1-301 sw-limAP (C' n AP (Q, RN)) , 1-296
11 u IIA11 1-229
I-140
IITHHD, 11-549 11w1HD, 11-549
9(E, x), 1-33 0''(µ, A, x), 1-34
II U IldifP,9(.a,h)> 11-184 11611 f, 11-41 11 T 111, 1-381 11 T 11°, 1-381
B* (µ, A, x), 1-34 9* (E, x), 1-33 Bn* (µ, A, x), 1-34 B* (E, x), 1-33
T2, 11-507
P(E, (1), 1-328, 1-340
P f,
Too, 1-381
11-35
it*, 1-581
u(A)r, 1-221
7r*, 1-578
Pk(R'n), 1-160
Pr(Rn), 1-160 Pb, 1-591 P, PT, 1-597 PT, 1-597 PT, 1-576, 1-583
Poinc<, >, 1-566 P(u), 11-412 Q(w), 11-355, 11-377
RP(Xn, ym), 1-637 RAs, 11-460 red-cart2"1(1? x y), 11-502
R°(B3,S2), 11-396 C)O
R2 (B3, S2), 11-396
R- (B3, S2), 11-396 Rp (Xn, y"n), 1-637 Rk (U), 1-140 RW1"P(Q, Y), 11-553
711
Tk D T, 11-549 Tk -k T, 1-123 W1"P(Q), 1-179