An Introduction To Differential Geometry With Applications To Elasticity

AN INTRODUCTION TO DIFFERENTIAL GEOMETRY WITH APPLICATIONS TO ELASTICITY
Philippe G. Ciarlet City University of Hong Kong
Contents
Preface 1 Three-dimensional dierential geometry Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Curvilinear coordinates . . . . . . . . . . . . . . . . . . . . . . . 1.2 Metric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Volumes, areas, and lengths in curvilinear coordinates . . . . . . 1.4 Covariant derivatives of a vector eld . . . . . . . . . . . . . . . . 1.5 Necessary conditions satised by the metric tensor; the Riemann curvature tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Existence of an immersion dened on an open set in R3 with a prescribed metric tensor . . . . . . . . . . . . . . . . . . . . . . . 1.7 Uniqueness up to isometries of immersions with the same metric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.8 Continuity of an immersion as a function of its metric tensor . .
5 9 9 11 13 16 19 24 25 36 44
2 Dierential geometry of surfaces 59 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 2.1 Curvilinear coordinates on a surface . . . . . . . . . . . . . . . . 61 2.2 First fundamental form . . . . . . . . . . . . . . . . . . . . . . . 65 2.3 Areas and lengths on a surface . . . . . . . . . . . . . . . . . . . 67 2.4 Second fundamental form; curvature on a surface . . . . . . . . . 69 2.5 Principal curvatures; Gaussian curvature . . . . . . . . . . . . . . 73 2.6 Covariant derivatives of a vector eld dened on a surface; the Gau and Weingarten formulas . . . . . . . . . . . . . . . . . . . 79 2.7 Necessary conditions satised by the rst and second fundamental forms: the Gau and Codazzi-Mainardi equations; Gau Theorema Egregium . . . . . . . . . . . . . . . . . . . . . . . . . 82 2.8 Existence of a surface with prescribed rst and second fundamental forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 2.9 Uniqueness up to proper isometries of surfaces with the same fundamental forms . . . . . . . . . . . . . . . . . . . . . . . . . . 95 2.10 Continuity of a surface as a function of its fundamental forms . . 100 3
Contents
3 Applications to three-dimensional elasticity in curvilinear coordinates 109 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.1 The equations of nonlinear elasticity in Cartesian coordinates . . 112 3.2 Principle of virtual work in curvilinear coordinates . . . . . . . . 119 3.3 Equations of equilibrium in curvilinear coordinates; covariant derivatives of a tensor eld . . . . . . . . . . . . . . . . . . . . . 127 3.4 Constitutive equation in curvilinear coordinates . . . . . . . . . . 129 3.5 The equations of nonlinear elasticity in curvilinear coordinates . 130 3.6 The equations of linearized elasticity in curvilinear coordinates . 132 3.7 A fundamental lemma of J.L. Lions . . . . . . . . . . . . . . . . . 135 3.8 Korns inequalities in curvilinear coordinates . . . . . . . . . . . 137 3.9 Existence and uniqueness theorems in linearized elasticity in curvilinear coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 4 Applications to shell theory Introduction . . . . . . . . . . . . . . . . . . 4.1 The nonlinear Koiter shell equations . . . . 4.2 The linear Koiter shell equations . . . . . . 4.3 Korns inequalities on a surface . . . . . . . 4.4 Existence and uniqueness theorems for the equations; covariant derivatives of a tensor surface . . . . . . . . . . . . . . . . . . . . . 4.5 A brief review of linear shell theories . . . . References Index 153 153 155 164 172
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . linear Koiter shell eld dened on a . . . . . . . . . . . . . . . . . . . . . . . .
185 193 201 209
PREFACE
This book is based on lectures delivered over the years by the author at the Universit e Pierre et Marie Curie, Paris, at the University of Stuttgart, and at City University of Hong Kong. Its two-fold aim is to give thorough introductions to the basic theorems of dierential geometry and to elasticity theory in curvilinear coordinates. The treatment is essentially self-contained and proofs are complete. The prerequisites essentially consist in a working knowledge of basic notions of analysis and functional analysis, such as dierential calculus, integration theory and Sobolev spaces, and some familiarity with ordinary and partial dierential equations. In particular, no a priori knowledge of dierential geometry or of elasticity theory is assumed. In the rst chapter, we review the basic notions, such as the metric tensor and covariant derivatives, arising when a three-dimensional open set is equipped with curvilinear coordinates. We then prove that the vanishing of the Riemann curvature tensor is sucient for the existence of isometric immersions from a simply-connected open subset of Rn equipped with a Riemannian metric into a Euclidean space of the same dimension. We also prove the corresponding uniqueness theorem, also called rigidity theorem. In the second chapter, we study basic notions about surfaces, such as their two fundamental forms, the Gaussian curvature and covariant derivatives. We then prove the fundamental theorem of surface theory, which asserts that the Gau and Codazzi-Mainardi equations constitute sucient conditions for two matrix elds dened in a simply-connected open subset of R2 to be the two fundamental forms of a surface in a three-dimensional Euclidean space. We also prove the corresponding rigidity theorem. In addition to such classical theorems, which constitute special cases of the fundamental theorem of Riemannian geometry, we also include in both chapters recent results which have not yet appeared in book form, such as the continuity of a surface as a function of its fundamental forms. The third chapter, which heavily relies on Chapter 1, begins by a detailed derivation of the equations of nonlinear and linearized three-dimensional elasticity in terms of arbitrary curvilinear coordinates. This derivation is then followed by a detailed mathematical treatment of the existence, uniqueness, and regularity of solutions to the equations of linearized three-dimensional elasticity in 5
Preface
curvilinear coordinates. This treatment includes in particular a direct proof of the three-dimensional Korn inequality in curvilinear coordinates. The fourth and last chapter, which heavily relies on Chapter 2, begins by a detailed description of the nonlinear and linear equations proposed by W.T. Koiter for modeling thin elastic shells. These equations are two-dimensional, in the sense that they are expressed in terms of two curvilinear coordinates used for dening the middle surface of the shell. The existence, uniqueness, and regularity of solutions to the linear Koiter equations is then established, thanks this time to a fundamental Korn inequality on a surface and to an innitesimal rigid displacement lemma on a surface. This chapter also includes a brief introduction to other two-dimensional shell equations. Interestingly, notions that pertain to dierential geometry per se, such as covariant derivatives of tensor elds, are also introduced in Chapters 3 and 4, where they appear most naturally in the derivation of the basic boundary value problems of three-dimensional elasticity and shell theory. Occasionally, portions of the material covered here are adapted from excerpts from my book Mathematical Elasticity, Volume III: Theory of Shells, published in 2000 by North-Holland, Amsterdam; in this respect, I am indebted to Arjen Sevenster for his kind permission to rely on such excerpts. Otherwise, the bulk of this work was substantially supported by two grants from the Research Grants Council of Hong Kong Special Administrative Region, China [Project No. 9040869, CityU 100803 and Project No. 9040966, CityU 100604]. Last but not least, I am greatly indebted to Roger Fosdick for his kind suggestion some years ago to write such a book, for his permanent support since then, and for his many valuable suggestions after he carefully read the entire manuscript. Hong Kong, July 2005 Philippe G. Ciarlet Department of Mathematics and Liu Bie Ju Centre for Mathematical Sciences City University of Hong Kong
Chapter 1 THREE-DIMENSIONAL DIFFERENTIAL GEOMETRY
INTRODUCTION
Let be an open subset of R3 , let E3 denote a three-dimensional Euclidean space, and let : E3 be a smooth injective immersion. We begin by reviewing (Sections 1.1 to 1.3) basic denitions and properties arising when the three-dimensional open subset () of E3 is equipped with the coordinates of the points of as its curvilinear coordinates. Of fundamental importance is the metric tensor of the set (), whose covariant and contravariant components gij = gji : R and g ij = g ji : R are given by (Latin indices or exponents take their values in {1, 2, 3}):
j gij = g i g j and g ij = g i g j , where g i = i and g j g i = i .
The vector elds g i : R3 and g j : R3 respectively form the covariant, and contravariant, bases in the set (). It is shown in particular how volumes, areas, and lengths, in the set () are computed in terms of its curvilinear coordinates, by means of the functions gij and g ij (Theorem 1.3-1). We next introduce in Section 1.4 the fundamental notion of covariant derivatives vi j of a vector eld vi g i : R3 dened by means of its covariant components vi over the contravariant bases g i . Covariant derivatives constitute a generalization of the usual partial derivatives of vector elds dened by means of their Cartesian components. As illustrated by the equations of nonlinear and linearized elasticity studied in Chapter 3, covariant derivatives naturally appear when a system of partial dierential equations with a vector eld as the unknown (the displacement eld in elasticity) is expressed in terms of curvilinear coordinates. It is a basic fact that the symmetric and positive-denite matrix eld (gij ) dened on in this fashion cannot be arbitrary. More specically (Theorem 1.5-1), its components and some of their partial derivatives must satisfy necessary conditions that take the form of the following relations (meant to hold for 9
10
Three-dimensional dierential geometry
[Ch. 1
all i, j, k, q {1, 2, 3}): Let the functions ijq and p ij be dened by ijq = 1 pq pq 1 (j giq + i gjq q gij ) and p . ij = g ijq , where (g ) = (gij ) 2
Then, necessarily,
p j ikq k ijq + p ij kqp ik jqp = 0 in .
The functions ijq and p ij are the Christoel symbols of the rst, and second, kind and the functions
p Rqijk = j ikq k ijq + p ij kqp ik jqp
are the covariant components of the Riemann curvature tensor of the set (). We then focus our attention on the reciprocal questions: Given an open subset of R3 and a smooth enough symmetric and positivedenite matrix eld (gij ) dened on , when is it the metric tensor eld of an open set () E3 , i.e., when does there exist an immersion : E3 such that gij = i j in ? If such an immersion exists, to what extent is it unique? As shown in Theorems 1.6-1 and 1.7-1, the answers turn out to be remarkably simple to state (but not so simple to prove, especially the rst one!): Under the assumption that is simply-connected, the necessary conditions Rqijk = 0 in are also sucient for the existence of such an immersion . Besides, if is connected, this immersion is unique up to isometries of E3 . This means that, if : E3 is any other smooth immersion satisfying gij = i j in , there then exist a vector c E3 and an orthogonal matrix Q of order three such that (x) = c + Q(x) for all x . Together, the above existence and uniqueness theorems constitute an important special case of the fundamental theorem of Riemannian geometry and as such, constitute the core of Chapter 1. We conclude this chapter by showing (Theorem 1.8-5) that the equivalence class of , dened in this fashion modulo isometries of E3 , depends continuechet topologies. ously on the matrix eld (gij ) with respect to appropriate Fr
Sect. 1.1]
Curvilinear coordinates
11
1.1
CURVILINEAR COORDINATES
To begin with, we list some notations and conventions that will be consistently used throughout. All spaces, matrices, etc., considered here are real. Latin indices and exponents range in the set {1, 2, 3}, save when otherwise indicated, e.g., when they are used for indexing sequences, and the summation convention with respect to repeated indices or exponents is systematically used in conjunction with this rule. For instance, the relation g i (x) = gij (x)g j (x) means that
3
g i (x) =
j =1
gij (x)g j (x) for i = 1, 2, 3.
j Kroneckers symbols are designated by i , ij , or ij according to the context. Let E3 denote a three-dimensional Euclidean space, let a b and a b denote the of a, b E3 , and let |a| = Euclidean inner product and exterior product a a denote the Euclidean norm of a E3 . The space E3 is endowed with an orthonormal basis consisting of three vectors ei = ei . Let xi denote the Cartesian coordinates of a point x E3 and let i := / xi . In addition, let there be given a three-dimensional vector space in which three vectors ei = ei form a basis. This space will be identied with R3 . Let xi denote the coordinates of a point x R3 and let i := /xi , ij := 2 /xi xj , and ijk := 3 /xi xj xk . Let there be given an open subset of E3 and assume that there exist an open subset of R3 and an injective mapping : E3 such that () = . Then each point x can be unambiguously written as
x = (x), x , and the three coordinates xi of x are called the curvilinear coordinates of x (Figure 1.1-1). Naturally, there are innitely many ways of dening curvilinear coordinates in a given open set , depending on how the open set and the mapping are chosen! Examples of curvilinear coordinates include the well-known cylindrical and spherical coordinates (Figure 1.1-2). In a dierent, but equally important, approach, an open subset of R3 together with a mapping : E3 are instead a priori given. If C 0 (; E3 ) and is injective, the set := () is open by the invariance of domain theorem (for a proof, see, e.g., Nirenberg [1974, Corollary 2, p. 17] or Zeidler [1986, Section 16.4]), and curvilinear coordinates inside are unambiguously dened in this case.
12
[Ch. 1
x3 x e
3
g3 (x)
x e
3
g2 (x)
g1 (x)
e2 e1 x1
x2
E3
e 2
e 1
b E3 . The three Figure 1.1-1: Curvilinear coordinates and covariant bases in an open set b If the three coordinates x1 , x2 , x3 of x are the curvilinear coordinates of x b = (x) . vectors gi (x) = i (x) are linearly independent, they form the covariant basis at x b = (x) and they are tangent to the coordinate lines passing through x b.
x z E3
r
E3
Figure 1.1-2: Two familiar examples of curvilinear coordinates. Let the mapping be dened by : (, , z ) ( cos , sin , z ) E3 . Then (, , z ) are the cylindrical coordinates of x b = (, , z ). Note that ( + 2k, , z ) or ( + + 2k, , z ), k Z, are also cylindrical coordinates of the same point x b and that is not dened if x b is the origin of E3 . Let the mapping be dened by : (, , r ) (r cos cos , r cos sin , r sin ) E3 . Then (, , r ) are the spherical coordinates of x b = (, , r ). Note that ( + 2k, + 2 , r ) or ( + 2k, + + 2 , r ) are also spherical coordinates of the same point x b and that and are not dened if x b is the origin of E3 . In both cases, the covariant basis at x b and the coordinate lines are represented with self-explanatory notations.
Sect. 1.2]
Metric tensor
13
If C 1 (; E3 ) and the three vectors i (x) are linearly independent at all x , the set is again open (for a proof, see, e.g., Schwartz [1992] or Zeidler [1986, Section 16.4]), but curvilinear coordinates may be dened only locally in this case: Given x , all that can be asserted (by the local inversion theorem) is the existence of an open neighborhood V of x in such that the restriction of to V is a C 1 -dieomorphism, hence an injection, of V onto (V ).
1.2
METRIC TENSOR
= i ei : E 3
Let be an open subset of R3 and let
be a mapping that is dierentiable at a point x . If x is such that (x + x) , then (x + x) = (x) + (x)x + o(x), where the 3 3 matrix (x) and the column vector x are dened by x1 1 1 2 1 3 1 (x) := 1 2 2 2 3 2 (x) and x = x2 . 1 3 2 3 3 3 x3 Let the three vectors g i (x) R3 be dened by i 1 g i (x) := i (x) = i 2 (x), i 3 i.e., g i (x) is the i-th column vector of the matrix (x). Then the expansion of about x may be also written as (x + x) = (x) + xi g i (x) + o(x). If in particular x is of the form x = tei , where t R and ei is one of the basis vectors in R3 , this relation reduces to (x + tei ) = (x) + tg i (x) + o(t). A mapping : E3 is an immersion at x if it is dierentiable at x and the matrix (x) is invertible or, equivalently, if the three vectors g i (x) = i (x) are linearly independent. Assume from now on in this section that the mapping is an immersion at x. Then the three vectors g i (x) constitute the covariant basis at the point x = (x). In this case, the last relation thus shows that each vector g i (x) is tangent to the i-th coordinate line passing through x = (x), dened as the image by of the points of that lie on the line parallel to ei passing through x
14
[Ch. 1
(there exist t0 and t1 with t0 < 0 < t1 such that the i-th coordinate line is given by t ]t0 , t1 [ f i (t) := (x + tei ) in a neighborhood of x; hence f i (0) = i (x) = g i (x)); see Figures 1.1-1 and 1.1-2. Returning to a general increment x = xi ei , we also infer from the expansion of about x that (recall that we use the summation convention): |(x + x) (x)|2 = xT (x)T (x)x + o |x|2 = xi g i (x) g j (x)xj + o |x|2 . Note that, here and subsequently, we use standard notations from matrix algebra. For instance, xT stands for the transpose of the column vector x and (x)T designates the transpose of the matrix (x), the element at the i-th row and j -th column of a matrix A is noted (A)ij , etc. In other words, the principal part with respect to x of the length between the points (x + x) and (x) is {xi g i (x) g j (x)xj }1/2 . This observation suggests to dene a matrix (gij (x)) of order three, by letting gij (x) := g i (x) g j (x) = ((x)T (x))ij . The elements gij (x) of this symmetric matrix are called the covariant components of the metric tensor at x = (x). Note that the matrix (x) is invertible and that the matrix (gij (x)) is positive denite, since the vectors g i (x) are assumed to be linearly independent. The three vectors g i (x) being linearly independent, the nine relations
i g i (x) g j (x) = j
unambiguously dene three linearly independent vectors g i (x). To see this, let i . This gives a priori g i (x) = X ik (x)g k (x) in the relations g i (x) g j (x) = j ik i ik ik X (x)gkj (x) = j ; consequently, X (x) = g (x), where (g ij (x)) := (gij (x))1 . Hence g i (x) = g ik (x)g k (x). These relations in turn imply that g i (x) g j (x) = g ik (x)g k (x) g j (x)g (x)
j = g ij (x), = g ik (x)g j (x)gk (x) = g ik (x)k
and thus the vectors g i (x) are linearly independent since the matrix (g ij (x)) is positive denite. We would likewise establish that g i (x) = gij (x)g j (x). The three vectors g i (x) form the contravariant basis at the point x = (x) and the elements g ij (x) of the symmetric positive denite matrix (g ij (x)) are the contravariant components of the metric tensor at x = (x). Let us record for convenience the fundamental relations that exist between the vectors of the covariant and contravariant bases and the covariant and contravariant components of the metric tensor at a point x where the mapping is an immersion: gij (x) = g i (x) g j (x) g i (x) = gij (x)g j (x) and g ij (x) = g i (x) g j (x), and g i (x) = g ij (x)g j (x).
Sect. 1.2]
Metric tensor
15
A mapping : E3 is an immersion if it is an immersion at each point in , i.e., if is dierentiable in and the three vectors g i (x) = i (x) are linearly independent at each x . If : E3 is an immersion, the vector elds g i : R3 and g i : R3 respectively form the covariant, and contravariant bases. To conclude this section, we briey explain in what sense the components of the metric tensor may be covariant or contravariant. Let and be two domains in R3 and let : E3 and : E3 be two C 1 -dieomorphisms such that () = () and such that the vectors g i (x) := i (x) and g i (x) = i (x) of the covariant bases at the same point (x) = (x) E3 are linearly independent. Let g i (x) and g i (x) be the vectors of the corresponding contravariant bases at the same point x. A simple computation then shows that g i (x) = j i (x)g j (x) and g i (x) = (x)g j (x), xi xj
where = (j ) := C 1 (; ) (hence x = (x)) and = (i ) := 1 C 1 (; ). Let gij (x) and gij (x) be the covariant components, and let g ij (x) and g ij (x) be the contravariant components, of the metric tensor at the same point (x) = (x) E3 . Then a simple computation shows that gij (x) = k i j (x) (x)gk (x) and g ij (x) = (x) (x)g k (x). xi xj xk x
These formulas explain why the components gij (x) and g ij (x) are respectively called covariant and contravariant: Each index in gij (x) varies like that of the corresponding vector of the covariant basis under a change of curvilinear coordinates, while each exponent in g ij (x) varies like that of the corresponding vector of the contravariant basis. Remark. What is exactly the second-order tensor hidden behind its covariant components gij (x) or its contravariant exponents g ij (x) is beautifully explained in the gentle introduction to tensors given by Antman [1995, Chapter 11, Sections 1 to 3]; it is also shown in ibid. that the same tensor i (x), which turn out to be simply the Kronecker also has mixed components gj i symbols j . In fact, analogous justications apply as well to the components of all the other tensors that will be introduced later on. Thus, for instance, the covariant components vi (x) and vi (x), and the contravariant components v i (x) and v i (x) (both with self-explanatory notations), of a vector at the same point (x) = (x) satisfy (cf. Section 1.4) vi (x)g i (x) = vi (x)g i (x) = v i (x)g i (x) = v i (x)g i (x).
16
[Ch. 1
It is then easily veried that vi (x) = j i (x)vj (x) and v i (x) = (x)v j (x). xi xj
In other words, the components vi (x) vary like the vectors g i (x) of the covariant basis under a change of curvilinear coordinates, while the components v i (x) of a vector vary like the vectors g i (x) of the contravariant basis. This is why they are respectively called covariant and contravariant. A vector is an example of a rst-order tensor. Likewise, it is easily checked that each exponent in the contravariant components Aijk (x) of the three-dimensional elasticity tensor in curvilinear coordinates introduced in Section 3.4 again varies like that of the corresponding vector of the contravariant basis under a change of curvilinear coordinates. Remark. See again Antman [1995, Chapter 11, Sections 1 to 3] to decipher the fourth-order tensor hidden behind such contravariant components Aijk (x). Note, however, that we shall no longer provide such commentaries in the sequel. We leave it instead to the reader to verify in each instance that any index or exponent appearing in a component of a tensor indeed behaves according to its nature. The reader interested by such questions will nd exhaustive treatments of tensor analysis, particularly as regards its relevance to elasticity, in Boothby [1975], Marsden & Hughes [1983, Chapter 1], or Simmonds [1994].
1.3
VOLUMES, AREAS, AND LENGTHS IN CURVILINEAR COORDINATES
We now review fundamental formulas showing how volume, area, and length elements at a point x = (x) in the set = () can be expressed in terms of the matrices (x), (gij (x)), and matrix (g ij (x)). These formulas thus highlight the crucial r ole played by the matrix (gij (x)) for computing metric notions at the point x = (x). Indeed, the metric tensor well deserves its name! A domain in Rd , d 2, is a bounded, open, and connected subset D of Rd with a Lipschitz-continuous boundary, the set D being locally on one side of its boundary. All relevant details needed here about domains are found in Ne cas [1967] or Adams [1975]. Given a domain D R3 with boundary , we let dx denote the volume element in D, d denote the area element along , and n = ni ei denote the unit (|n| = 1) outer normal vector along (d is well dened and n is dened d-almost everywhere since is assumed to be Lipschitz-continuous). Note also that the assumptions made on the mapping in the next theorem guarantee that, if D is a domain in R3 such that D , then {D} ,
Sect. 1.3]
Volumes, areas, and lengths in curvilinear coordinates
17
{(D)} = (D), and the boundaries D of D and D of D are related by D = (D) (see, e.g., Ciarlet [1988, Theorem 1.2-8 and Example 1.7]). If A is a square matrix, Cof A denotes the cofactor matrix of A. Thus Cof A = (det A)AT if A is invertible. Theorem 1.3-1. Let be an open subset of R3 , let : E3 be an injective and smooth enough immersion, and let = (). (a) The volume element dx at x = (x) is given in terms of the volume element dx at x by dx = | det (x)| dx = g (x) dx, where g (x) := det(gij (x)).
(b) Let D be a domain in R3 such that D . The area element d(x) at x = (x) D is given in terms of the area element d(x) at x D by d(x) = | Cof (x)n(x)| d(x) = g (x) ni (x)g ij (x)nj (x) d(x),
where n(x) := ni (x)ei denotes the unit outer normal vector at x D. (c) The length element d (x) at x = (x) is given by d (x) = xT (x)T (x)x where x = xi ei . Proof. The relation dx = | det (x)| dx between the volume elements is well known. The second relation in (a) follows from the relation g (x) = | det (x)|2 , which itself follows from the relation (gij (x)) = (x)T (x). Indications about the proof of the relation between the area elements d(x) and d(x) given in (b) are found in Ciarlet [1988, Theorem 1.7-1] (in this formula, n(x) = ni (x)ei is identied with the column vector in R3 with ni (x) as its components). Using the relations Cof (AT ) = (Cof A)T and Cof(AB) = (Cof A)(Cof B), we next have: | Cof (x)n(x)|2 = n(x)T Cof (x)T (x) n(x) = g (x)ni (x)g ij (x)nj (x). Either expression of the length element given in (c) recalls that d (x) is by denition the principal part with respect to x = xi ei of the length |(x + x) (x)|, whose expression precisely led to the introduction of the matrix (gij (x)) in Section 1.2. The relations found in Theorem 1.3-1 are used in particular for computing volumes, areas, and lengths inside by means of integrals inside , i.e., in terms of the curvilinear coordinates used in the open set (Figure 1.3-1): Let D be a domain in R3 such that D , let D := (D), and let f L1 (D ) be given. Then
b D 1/2
= xi gij (x)xj
1/2
f (x) dx =
D
(f )(x)
g (x) dx.
18

n(x)
[Ch. 1
x) d( d(x)
C x+x x V
A
C d l( x)

V (x+x) (x) = x dx
dx
R3
E3
I t
Figure 1.3-1: Volume, area, and length elements in curvilinear coordinates. The elements b x b are expressed in terms of dx, d(x), and x at x by dx b, d( b), and d b(x b) at x b = (x) means of the covariant and contravariant components of the metric tensor; cf. Theorem 1.3-1. Given a domain D such that D and a d-measurable subset of D , the corresponding b = (D ) , b the area of b = () D b, relations are used for computing the volume of D b = (C ) , b where C = f (I ) and I is a compact interval of R. and the length of a curve C
In particular, the volume of D is given by vol D :=

b D
dx =
D
g (x) dx.
Next, let := D, let be a d-measurable subset of , let := () D, and let h L1 () be given. Then
b
h(x) d(x) =
(h )(x)
g (x)
ni (x)g ij (x)nj (x) d(x).
In particular, the area of is given by area :=

b
d(x) =
g (x)
ni (x)g ij (x)nj (x) d(x).
Finally, consider a curve C = f (I ) in , where I is a compact interval of R and f = f i ei : I is a smooth enough injective mapping. Then the length of the curve C := (C ) is given by length C :=
I
d ( f )(t) dt = dt
gij (f (t))
I
df i df j (t) (t) dt. dt dt
Sect. 1.4]
Covariant derivatives of a vector eld
19
This relation shows in particular that the lengths of curves inside the open set () are precisely those induced by the Euclidean metric of the space E3 . For this reason, the set () is said to be isometrically imbedded in E3 .
1.4
COVARIANT DERIVATIVES OF A VECTOR FIELD
Suppose that a vector eld is dened in an open subset of E3 by means of its Cartesian components vi : R, i.e., this eld is dened by its values vi (x)ei at each x , where the vectors ei constitute the orthonormal basis of E3 ; see Figure 1.4-1.

v3 ( x) x 3 x e 3 v1 ( x) v2 ( x) vi ( x) e i
E3
O e x 1
1
e 2
x 2
b the vector Figure 1.4-1: A vector eld in Cartesian coordinates. At each point x b , vi (x b b)b ei is dened by its Cartesian components v bi (x b) over an orthonormal basis of E3 formed by three vectors b ei . An example of a vector eld in Cartesian coordinates is provided by the displacement eld b } as its reference conguration; cf. Section 3.1. of an elastic body with {
Suppose now that the open set is equipped with curvilinear coordinates from an open subset of R3 , by means of an injective mapping : E3 satisfying () = . How does one dene appropriate components of the same vector eld, but this time in terms of these curvilinear coordinates? It turns out that the proper way to do so consists in dening three functions vi : R by requiring that (Figure 1.4-2) vi (x)g i (x) := vi (x)ei for all x = (x), x , where the three vectors g i (x) form the contravariant basis at x = (x) (Section i i and ei ej = j , we immediately nd 1.2). Using the relations g i (x) g j (x) = j
20
[Ch. 1
how the old and new components are related, viz., vj (x) = vi (x)g i (x) g j (x) = vi (x)ei g j (x), vi (x) = vj (x)ej ei = vj (x)g j (x) ei . The three components vi (x) are called the covariant components of the vector vi (x)g i (x) at x, and the three functions vi : R dened in this fashion are called the covariant components of the vector eld vi g i : E3 . Suppose next that we wish to compute a partial derivative j vi (x) at a point x = (x) in terms of the partial derivatives vk (x) and of the values vq (x) (which are also expected to appear by virtue of the chain rule). Such a task is required for example if we wish to write a system of partial dierential equations whose unknown is a vector eld (such as the equations of nonlinear or linearized elasticity) in terms of ad hoc curvilinear coordinates. As we now show, carrying out such a transformation naturally leads to a fundamental notion, that of covariant derivatives of a vector eld.
g3 (x) v3 (x) x3 x e3 e2 e1 x1
Figure 1.4-2: A vector eld in curvilinear coordinates. Let there be given a vector eld b by its Cartesian components b b) over the in Cartesian coordinates dened at each x b vi (x vectors b ei (Figure 1.4-1). In curvilinear coordinates, the same vector eld is dened at each x by its covariant components vi (x) over the contravariant basis vectors gi (x) in such a way that vi (x)gi (x) = v bi (x b)ei , x b = (x). An example of a vector eld in curvilinear coordinates is provided by the displacement b } = () as its reference conguration; cf. Section 3.2. eld of an elastic body with {
vi (x)g i (x) g2 (x) v2 (x) v1 (x) g1 (x)
x2
e 3
R3
e 1

e 2
E3
Theorem 1.4-1. Let be an open subset of R3 and let : E3 be an injective immersion that is also a C 2 -dieomorphism of onto := (). Given a vector eld vi ei : R3 in Cartesian coordinates with components vi C 1 (), let vi g i : R3 be the same eld in curvilinear coordinates, i.e., that dened by vi (x)ei = vi (x)g i (x) for all x = (x), x .
Sect. 1.4]
21
Then vi C 1 () and for all x , j vi (x) = vk [g k ]i [g ]j (x), x = (x), where vi and [g i (x)]k := g i (x) ek denotes the i-th component of g i (x) over the basis {e1 , e2 , e3 }. Proof. The following convention holds throughout this proof: The simultaneous appearance of x and x in an equality means that they are related by x = (x) and that the equality in question holds for all x . (i) Another expression of [g i (x)]k := g i (x) ek . Let (x) = k (x)ek and (x) = i (x)ei , where : E3 denotes the inverse mapping of : E3 . Since ((x)) = x for all x , the chain rule shows that the matrices (x) := (j k (x)) (the row index is k ) and (x) := (k i (x)) (the row index is i) satisfy (x)(x) = I, or equivalently, j 1 (x) i . k i (x)j k (x) = 1 i (x) 2 i (x) 3 i (x) j 2 (x) = j 3 j (x) The components of the above column vector being precisely those of the vector g j (x), the components of the above row vector must be those of the vector g i (x) since g i (x) is uniquely dened for each exponent i by the three i , j = 1, 2, 3. Hence the k -th component of g i (x) over relations g i (x) g j (x) = j the basis {e1 , e2 , e3 } can be also expressed in terms of the inverse mapping , as: [g i (x)]k = k i (x). (ii) The functions qk := g q g k C 0 (). We next compute the derivatives g q (x) (the elds g q = g qr g r are of class C on since is assumed to be of class C 2 ). These derivatives will be needed in (iii) for expressing the derivatives j ui (x) as functions of x (recall that ui (x) = uk (x)[g k (x)]i ). Recalling that the vectors g k (x) form a basis, we may write a priori g q (x) = qk (x)g k (x),
1 j p p := j vi p ij vp and ij := g i g j ,
22
[Ch. 1
thereby unambiguously dening functions qk : R. To nd their expressions in terms of the mappings and , we observe that
m = qm (x)g m (x) g k (x) = g q (x) g k (x). qk (x) = qm (x)k
Hence, noting that (g q (x) g k (x)) = 0 and [g q (x)]p = p q (x), we obtain qk (x) = g q (x) g k (x) = p q (x) k p (x) = q k (x). Since C 2 (; E3 ) and C 1 (; R3 ) by assumption, the last relations show that qk C 0 (). (iii) The partial derivatives i vi (x) of the Cartesian components of the vector eld vi ei C 1 (; R3 ) are given at each x = (x) by j vi (x) = vk (x)[g k (x)]i [g (x)]j , where vk (x) := vk (x) qk (x)vq (x), and [g k (x)]i and qk (x) are dened as in (i) and (ii). We compute the partial derivatives j vi (x) as functions of x by means of the relation vi (x) = vk (x)[g k (x)]i . To this end, we rst note that a dierentiable function w : R satises j w((x)) = w(x)j (x) = w(x)[g (x)]j , by the chain rule and by (i). In particular then, j vi (x) = j vk ((x))[g k (x)]i + vq (x)j [g q ((x))]i = vk (x)[g (x)]j [g k (x)]i + vq (x) [g q (x)]i [g (x)]j = ( vk (x) qk (x)vq (x)) [g k (x)]i [g (x)]j , since g q (x) = qk (x)g k (x) by (ii). The functions vi
j
= j vi p ij vp
dened in Theorem 1.4-1 are called the rst-order covariant derivatives of the vector eld vi g i : R3 . The functions p p ij = g i g j are called the Christoel symbols of the second kind (the Christoel symbols of the rst kind are introduced in the next section). The following result summarizes properties of covariant derivatives and Christoel symbols that are constantly used.
Sect. 1.4]
23
Theorem 1.4-2. Let the assumptions on the mapping : E3 be as in Theorem 1.4-1, and let there be given a vector eld vi g i : R3 with covariant components vi C 1 (). (a) The rst-order covariant derivatives vi j C 0 () of the vector eld vi g i : R3 , which are dened by vi
j p p := j vi p ij vp , where ij := g i g j ,
can be also dened by the relations j (vi g i ) = vi j g i vi

j
= j (vk g k ) g i .
p p 0 (b) The Christoel symbols p ij := g i g j = ji C () satisfy the relations j i i g p = p ij g and j g q = jq g i .
Proof. It remains to verify that the covariant derivatives vi j , dened in Theorem 1.4-1 by vi j = j vi p ij vp , may be equivalently dened by the relations j (vi g i ) = vi j g i . These relations unambiguously dene the functions vi j = {j (vk g k )} g i since the vectors g i are linearly independent at all points of by assumption. To this end, we simply note that, by denition, the Christoel symbols satisfy j i g p = p ij g (cf. part (ii) of the proof of Theorem 1.4-1); hence
k i j (vi g i ) = (j vi )g i + vi j g i = (j vi )g i vi i jk g = vi j g .
To establish the other relations j g q = i jq g i , we note that

p i p p 0 = j (g p g q ) = p ji g g q + g j g q = qj + g j g q .
Hence j g q = (j g q g p )g p = p qj g p .
Remark. The Christoel symbols p ij can be also dened solely in terms of the components of the metric tensor; see the proof of Theorem 1.5-1. If the ane space E3 is identied with R3 and (x) = x for all x , the relation j (vi g i )(x) = (vi j g i )(x) reduces to j (vi (x)ei ) = (j vi (x))ei . In this sense, a covariant derivative of the rst order constitutes a generalization of a partial derivative of the rst order in Cartesian coordinates.
24
[Ch. 1
1.5
NECESSARY CONDITIONS SATISFIED BY THE METRIC TENSOR; THE RIEMANN CURVATURE TENSOR
It is remarkable that the components gij = gji : R of the metric tensor of an open set () E3 (Section 1.2), dened by a smooth enough immersion : E3 , cannot be arbitrary functions. As shown in the next theorem, they must satisfy relations that take the form: p j ikq k ijq + p ij kqp ik jqp = 0 in , where the functions ijq and p ij have simple expressions in terms of the functions gij and of some of their partial derivatives (as shown in the next proof, it so happens that the functions p ij as dened in Theorem 1.5-1 coincide with the Christoel symbols introduced in the previous section; this explains why they are denoted by the same symbol). Note that, according to the rule governing Latin indices and exponents, these relations are meant to hold for all i, j, k, q {1, 2, 3}. Theorem 1.5-1. Let be an open set in R3 , let C 3 (; E3 ) be an immersion, and let gij := i j denote the covariant components of the metric tensor of the set (). Let the 1 functions ijq C 1 () and p ij C () be dened by ijq := 1 (j giq + i gjq q gij ), 2 pq pq 1 . p ij := g ijq where (g ) := (gij )
Then, necessarily,
p j ikq k ijq + p ij kqp ik jqp = 0 in .
Proof. Let g i = i . It is then immediately veried that the functions ijq are also given by ijq = i g j g q . For each x , let the three vectors g j (x) be dened by the relations g j (x) j g i (x) = j . Since we also have g j = g ij g i , the last relations imply that p ij = p i g j g . Therefore, i g j = p ij g p since i g j = (i g j g p )g p . Dierentiating the same relations yields k ijq = ik g j g q + i g j k g q , so that the above relations together give
p i g j k g q = p ij g p k g q = ij kqp .
Sect. 1.6] Existence of an immersion with a prescribed metric tensor Consequently, ik g j g q = k ijq p ij kqp . Since ik g j = ij g k , we also have ik g j g q = j ikq p ik jqp , and thus the required necessary conditions immediately follow.
25
Remark. The vectors g i and g j introduced above form the covariant and contravariant bases and the functions g ij are the contravariant components of the metric tensor (Section 1.2). As shown in the above proof, the necessary conditions Rqijk = 0 thus simply constitute a re-writing of the relations ik g j = ki g j in the form of the equivalent relations ik g j g q = ki g j g q . The functions ijq = and
p pq p p ij = g ijq = i g j g = ji
1 (j giq + i gjq q gij ) = i g j g q = jiq 2
are the Christoel symbols of the rst, and second, kinds. We saw in Section 1.4 that the Christoel symbols of the second kind also naturally appear in a dierent context (that of covariant dierentiation). Finally, the functions
p Rqijk := j ikq k ijq + p ij kqp ik jqp
are the covariant components of the Riemann curvature tensor of the set (). The relations Rqijk = 0 found in Theorem 1.4-1 thus express that the Riemann curvature tensor of the set () (equipped with the metric tensor with covariant components gij ) vanishes.
1.6
EXISTENCE OF AN IMMERSION DEFINED ON AN OPEN SET IN R3 WITH A PRESCRIBED METRIC TENSOR
Let M3 , S3 , and S3 > denote the sets of all square matrices of order three, of all symmetric matrices of order three, and of all symmetric positive denite matrices of order three. As in Section 1.2, the matrix representing the Fr echet derivative at x of a dierentiable mapping = ( ) : E3 is denoted (x) := (j (x)) M3 ,
26
[Ch. 1
where is the row index and j the column index (equivalently, (x) is the matrix of order three whose j -th column vector is j ). So far, we have considered that we are given an open set R3 and a smooth enough immersion : E3 , thus allowing us to dene a matrix eld C = (gij ) = T : S3 >, where gij : R are the covariant components of the metric tensor of the open set () E3 . We now turn to the reciprocal questions : Given an open subset of R3 and a smooth enough matrix eld C = (gij ) : 3 S3 > , when is C the metric tensor eld of an open set () E ? Equiva3 lently, when does there exist an immersion : E such that C = T in , or equivalently, such that gij = i j in ? If such an immersion exists, to what extent is it unique? The answers are remarkably simple: If is simply-connected, the necessary conditions p j ikq k ijq + p ij kqp ik jqp = 0 in found in Theorem 1.7-1 are also sucient for the existence of such an immersion. If is connected, this immersion is unique up to isometries in E3 . Whether the immersion found in this fashion is globally injective is a dierent issue, which accordingly should be resolved by dierent means. This result comprises two essentially distinct parts, a global existence result (Theorem 1.6-1) and a uniqueness result (Theorem 1.7-1). Note that these two results are established under dierent assumptions on the set and on the smoothness of the eld (gij ). In order to put these results in a wider perspective, let us make a brief incursion into Riemannian Geometry. For detailed treatments, see classic texts such as Choquet-Bruhat, de Witt-Morette & Dillard-Bleick [1977], Marsden & Hughes [1983], Berger [2003], or Gallot, Hulin & Lafontaine [2004]. Considered as a three-dimensional manifold, an open set R3 equipped with an immersion : E3 becomes an example of a Riemannian manifold (; (gij )), i.e., a manifold, the set , equipped with a Riemannian metric, the symmetric positive-denite matrix eld (gij ) : S3 > dened in this case by gij := i j in . More generally, a Riemannian metric on a manifold is a twice covariant, symmetric, positive-denite tensor eld acting on vectors in the tangent spaces to the manifold (these tangent spaces coincide with R3 in the present instance). This particular Riemannian manifold (; (gij )) possesses the remarkable property that its metric is the same as that of the surrounding space E3 . More specically, (; (gij )) is isometrically immersed in the Euclidean space E3 ,
Sect. 1.6] Existence of an immersion with a prescribed metric tensor
27
in the sense that there exists an immersion : E3 that satises the relations gij = i j . Equivalently, the length of any curve in the Riemannian manifold (; (gij )) is the same as the length of its image by in the Euclidean space E3 (see Theorem 1.3-1). The rst question above can thus be rephrased as follows: Given an open subset of R3 and a positive-denite matrix eld (gij ) : S3 > , when is the Riemannian manifold (; (gij )) at, in the sense that it can be locally isometrically immersed in a Euclidean space of the same dimension (three)? The answer to this question can then be rephrased as follows (compare with the statement of Theorem 1.6-1 below): Let be a simply-connected open subset of R3 . Then a Riemannian manifold (; (gij )) with a Riemannian metric (gij ) of class C 2 in is at if and only if its Riemannian curvature tensor vanishes in . Recast as such, this result becomes a special case of the fundamental theorem on at Riemannian manifolds, which holds for a general nitedimensional Riemannian manifold. The answer to the second question, viz., the issue of uniqueness, can be rephrased as follows (compare with the statement of Theorem 1.7-1 in the next section): Let be a connected open subset of R3 . Then the isometric immersions of a at Riemannian manifold (; (gij )) into a Euclidean space E3 are unique up to isometries of E3 . Recast as such, this result likewise becomes a special case of the so-called rigidity theorem; cf. Section 1.7. Recast as such, these two theorems together constitute a special case (that where the dimensions of the manifold and of the Euclidean space are both equal to three) of the fundamental theorem of Riemannian Geometry. This theorem addresses the same existence and uniqueness questions in the more general setting where is replaced by a p-dimensional manifold and E3 is replaced by a (p + q )-dimensional Euclidean space (the fundamental theorem of surface theory, established in Sections 2.8 and 2.9, constitutes another important special case). When the p-dimensional manifold is an open subset of Rp+q , an outline of a self-contained proof is given in Szopos [2005]. Another fascinating question (which will not be addressed here) is the following: Given again an open subset of R3 equipped with a symmetric, positivedenite matrix eld (gij ) : S3 , assume this time that the Riemannian manifold (; (gij )) is no longer at, i.e., its Riemannian curvature tensor no longer vanishes in . Can such a Riemannian manifold still be isometrically immersed, but this time in a higher-dimensional Euclidean space ? Equivalently, does there exist a Euclidean space Ed with d > 3 and does there exist an immersion : Ed such that gij = i j in ? The answer is yes, according to the following beautiful Nash theorem, so named after Nash [1954]: Any p-dimensional Riemannian manifold equipped with a continuous metric can be isometrically immersed in a Euclidean space of dimension 2p with an immersion of class C 1 ; it can also be isometrically immersed in a Euclidean space of dimension (2p + 1) with a globally injective immersion of class C 1 . Let us now humbly return to the question of existence raised at the beginning of this section, i.e., when the manifold is an open set in R3 .
28
[Ch. 1
Theorem 1.6-1. Let be a connected and simply-connected open set in R3 and let C = (gij ) C 2 (; S3 > ) be a matrix eld that satises
p Rqijk := j ikq k ijq + p ij kqp ik jqp = 0 in ,
where ijq := 1 (j giq + i gjq q gij ), 2 p ij := g pq ijq with (g pq ) := (gij )1 .
Then there exists an immersion C 3 (; E3 ) such that C = T in . Proof. The proof relies on a simple, yet crucial, observation. When a smooth enough immersion = ( ) : E3 is a priori given (as it was so far), its components satisfy the relations ij = p ij p , which are nothing but another way of writing the relations i g j = p ij g p (see the proof of Theorem 1.5-1). This observation thus suggests to begin by solving (see part (ii)) the system of partial dierential equations i F
j
= p ij F
in ,
whose solutions F j : R then constitute natural candidates for the partial derivatives j of the unknown immersion = ( ) : E3 (see part (iii)). To begin with, we establish in (i) relations that will in turn allow us to re-write the sucient conditions
p j ikq k ijq + p ij kqp ik jqp = 0 in
in a slightly dierent form, more appropriate for the existence result of part (ii). Note that the positive deniteness of the symmetric matrices (gij ) is not needed for this purpose. (i) Let be an open subset of R3 and let there be given a eld (gij ) pq C (; S3 ) of symmetric invertible matrices. The functions ijq , p ij , and g being dened by
2
ijq :=
1 (j giq + i gjq q gij ), 2
pq p ij := g ijq ,
(g pq ) := (gij )1 ,
dene the functions

p Rqijk := j ikq k ijq + p ij kqp ik jqp , p p p p p R ijk := j ik k ij + ik j ij k .
Then
p p pq R ijk = g Rqijk and Rqijk = gpq Rijk .
Sect. 1.6] Existence of an immersion with a prescribed metric tensor Using the relations jq +
jq
29
= j gq and ikq = gq ik ,
which themselves follow from the denitions of the functions ijq and p ij , and noting that (g pq j gq + gq j g pq ) = j (g pq gq ) = 0, we obtain
p pq g pq (j ikq r ik g pq (j gq ik jqr ) = j ik ikq j g jq )
= = Likewise,
j p ik j p ik
+ +
ik p j ik p j
ik (g j gq + gq j g pq )
pq
p p g pq (k ijq r ij kqr ) = k ij ij k , p pq and thus the relations R ijk = g Rqijk are established. The relations Rqijk = p gpq Rijk are clearly equivalent to these ones. We next establish the existence of solutions to the system
i F
= p ij F
in .
(ii) Let be a connected and simply-connected open subset of R3 and let p 1 there be given functions p ij = ji C () satisfying the relations
p p p j p ik k ij + ik j ij k = 0 in ,
which are equivalent to the relations

p j ikq k ijq + p ij kqp ik jqp = 0 in ,
by part (i). 3 Let a point x0 and a matrix (F 0 j ) M be given. Then there exists one, 2 3 and only one, eld (F j ) C (; M ) that satises i F j (x) = p ij (x)F p (x), x , F j (x0 ) = F 0 j. Let x1 be an arbitrary point in the set , distinct from x0 . Since is connected, there exists a path = ( i ) C 1 ([0, 1]; R3 ) joining x0 to x1 in ; this means that (0) = x0 , (1) = x1 , and (t) for all 0 t 1. Assume that a matrix eld (F j ) C 1 (; M3 ) satises i F j (x) = p ij (x)F p (x), x . Then, for each integer {1, 2, 3}, the three functions j C 1 ([0, 1]) dened by (for simplicity, the dependence on is dropped) j (t) := F j ( (t)), 0 t 1,
30
[Ch. 1
satisfy the following Cauchy problem for a linear system of three ordinary differential equations with respect to three unknowns : dj d i (t) = p (t)p (t), 0 t 1, ij ( (t)) dt dt 0 j (0) = j ,
0 where the initial values j are given by 0 := F 0 j j.
Note in passing that the three Cauchy problems obtained by letting = 1, 2, 0 . or 3 only dier by their initial values j It is well known that a Cauchy problem of the form (with self-explanatory notations) d (t) = A(t) (t), 0 t 1, dt (0) = 0 , has one and only one solution C 1 ([0, 1]; R3 ) if A C 0 ([0, 1]; M3 ) (see, e.g., Schwartz [1992, Theorem 4.3.1, p. 388]). Hence each one of the three Cauchy problems has one and only one solution. Incidentally, this result already shows that, if it exists, the unknown eld (F j ) is unique. In order that the three values j (1) found by solving the above Cauchy problem for a given integer {1, 2, 3} be acceptable candidates for the three unknown values F j (x1 ), they must be of course independent of the path chosen for joining x0 to x1 . So, let 0 C 1 ([0, 1]; R3 ) and 1 C 1 ([0, 1]; R3 ) be two paths joining x0 to x1 in . The open set being simply-connected, there exists a homotopy G = (Gi ) : [0, 1] [0, 1] R3 joining 0 to 1 in , i.e., such that G(, 0) = 0 , G(, 1) = 1 , G(t, ) for all 0 t 1, 0 1, G(0, ) = x0 and G(1, ) = x1 for all 0 1, and smooth enough in the sense that G C 1 ([0, 1] [0, 1]; R3 ) and G G = C 0 ([0, 1] [0, 1]; R3 ). t t
Let (, ) = (j (, )) C 1 ([0, 1]; R3 ) denote for each 0 1 the solution of the Cauchy problem corresponding to the path G(, ) joining x0 to x1 . We thus have j Gi (t, ) = p (t, )p (t, ) for all 0 t 1, 0 1, ij (G(t, )) t t 0 for all 0 1. j (0, ) = j
Sect. 1.6] Existence of an immersion with a prescribed metric tensor Our objective is to show that j (1, ) = 0 for all 0 1,
31
as this relation will imply that j (1, 0) = j (1, 1), as desired. For this purpose, a direct dierentiation shows that, for all 0 t 1, 0 1, j t where
p p = { q ij kq + k ij }p
Gk Gi Gi Gi + p + q q , ij p ij t t t
Gk j p , kj p p on the one hand (in the relations above and below, q ij , k ij , etc., stand for q p ij (G(, )), k ij (G(, )), etc.). On the other hand, a direct dierentiation of the equation dening the functions j shows that, for all 0 t 1, 0 1, j := j t But =
k j Gi Gi q q G p + i p + . + p p ij kj kj t t t t
Gi j = p p , so that we also have ij t t j t =
j Gi Gk Gi q p p + {i p + . + } p p ij kj kj iq t t t j j = Hence, subtracting the above relations and noting that t t Gi Gi = by assumption, we infer that and t t j Gk Gi Gi p q p q p + {i p q q = 0. ij kj k ij + kj iq ij kq }p t t t
p The assumed symmetries p ij = ji combined with the assumed relations p p p p j ik k ij + ik j ij k = 0 in show that p q p q p i p kj k ij + kj iq ij kq = 0,
on the one hand. On the other hand, j (0, ) = j Gk (0, ) p (0, ) = 0, kj (G(0, ))p (0, )
0 0 since j (0, ) = j and G(0, ) = x0 for all 0 1. Therefore, for any xed value of the parameter [0, 1], each function j (, ) satises a Cauchy problem for an ordinary dierential equation, viz.,
dj Gi (t, ) = q (t, )q (t, ), 0 t 1, ij (G(t, )) dt t j (0, ) = 0.
32
[Ch. 1
But the solution of such a Cauchy problem is unique; hence j (t, ) = 0 for all 0 t 1. In particular then, j (1, ) = j Gk (1, ) p (1, ) kj (G(1, ))p (1, ) = 0 for all 0 1,
and thus
j (1, ) = 0 for all 0 1, since G(1, ) = x1 for all 0 1.
For each integer , we may thus unambiguously dene a vector eld (F j ) : R3 by letting F j (x1 ) := j (1) for any x1 , where C 1 ([0, 1]; R3 ) is any path joining x0 to x1 in and the vector eld (j ) C 1 ([0, 1]) is the solution to the Cauchy problem d i dj ( ( t )) (t) = p (t)p (t), 0 t 1, ij dt dt 0 j (0) = j , corresponding to such a path. To establish that such a vector eld is indeed the -th row-vector eld of the 1 3 unknown matrix eld we are seeking, we need to show that (F j )3 j =1 C (; R ) p and that this eld does satisfy the partial dierential equations i F j = ij F p in corresponding to the xed integer used in the above Cauchy problem. Let x be an arbitrary point in and let the integer i {1, 2, 3} be xed in what follows. Then there exist x1 , a path C 1 ([0, 1]; R3 ) joining x0 to x1 , ]0, 1[, and an open interval I [0, 1] containing such that (t) = x + (t )ei for t I, where ei is the i-th basis vector in R3 . Since each function j is continuously difd i dj (t) = p (t)p (t) for all 0 t 1, ferentiable in [0, 1] and satises ij ( (t)) dt dt we have j (t) = j ( ) + (t ) dj ( ) + o(t ) dt = j ( ) + (t )p ij ( ( ))p ( ) + o(t )
for all t I . Equivalently, F j (x + (t )ei ) = F j (x) + (t )p ij (x)F p (x) + o(t x). This relation shows that each function F the set , given at each x by
j
possesses partial derivatives in
i F p (x) = p ij (x)F p (x).
33
Consequently, the matrix eld (F j ) is of class C 1 in (its partial derivatives are continuous in ) and it satises the partial dierential equations i F j = p ij F p in , as desired. Dierentiating these equations shows that the matrix eld (F j ) is in fact of class C 2 in . In order to conclude the proof of the theorem, it remains to adequately 0 choose the initial values F 0 j at x in step (ii). (iii) Let be a connected and simply-connected open subset of R3 and let (gij ) C 2 (; S3 > ) be a matrix eld satisfying
p j ikq k ijq + p ij kqp ik jqp = 0 in , pq being dened by the functions ijq , p ij , and g
ijq :=
1 (j giq + i gjq q gij ), 2
pq p ij := g ijq ,
(g pq ) := (gij )1 .
3 Given an arbitrary point x0 , let (F 0 j ) S> denote the square root of 0 0 3 the matrix (gij ) := (gij (x )) S> . Let (F j ) C 2 (; M3 ) denote the solution to the corresponding system
i F j (x) = p ij (x)F p (x), x , F j (x0 ) = F 0 j, which exists and is unique by parts (i) and (ii). Then there exists an immersion = ( ) C 3 (; E3 ) such that j = F
j
and gij = i j in .
To begin with, we show that the three vector elds dened by g j := (F j )3=1 C 2 (; R3 ) satisfy g i g j = gij in . To this end, we note that, by construction, these elds satisfy i g j = p ij g p in , g j (x0 ) = g 0 j,
0 3 where g 0 j is the j -th column vector of the matrix (F j ) S> . Hence the matrix eld (g i g j ) C 2 (; M3 ) satises m k (g i g j ) = m kj (g m g i ) + ki (g m g j ) in , 0 (g i g j )(x0 ) = gij .
The denitions of the functions ijq and p ij imply that k gij = ikj + jki and ijq = gpq p ij .
34
[Ch. 1
Hence the matrix eld (gij ) C 2 (; S3 > ) satises

m k gij = m kj gmi + ki gmj in , 0 gij (x0 ) = gij .
Viewed as a system of partial dierential equations, together with initial values at x0 , with respect to the matrix eld (gij ) : M3 , the above system can have at most one solution in the space C 2 (; M3 ). To see this, let x1 be distinct from x0 and let C 1 ([0, 1]; R3 ) be any path joining x0 to x1 in , as in part (ii). Then the nine functions gij ( (t)), 0 t 1, satisfy a Cauchy problem for a linear system of nine ordinary dierential equations and this system has at most one solution. An inspection of the two above systems therefore shows that their solutions are identical, i.e., that g i g j = gij . It remains to show that there exists an immersion C 3 (; E3 ) such that i = g i in , where g i := (F j )3=1 . p p 2 3 Since the functions p ij satisfy ij = ji , any solution (F j ) C (; M ) of the system i F j (x) = p ij (x)F p (x), x , F j (x0 ) = F 0 j satises i F
j
= j F i in .
The open set being simply-connected, Poincar es lemma (for a proof, see, e.g., Flanders [1989], Schwartz [1992, Vol. 2, Theorem 59 and Corollary 1, p. 234235], or Spivak [1999]) shows that, for each integer , there exists a function C 3 () such that i = F
i
in ,
or, equivalently, such that the mapping := ( ) C 3 (; E3 ) satises i = g i in . That is an immersion follows from the assumed invertibility of the matrices (gij ). The proof is thus complete. Remarks. (1) The assumptions
p p p j p ik k ij + ik j ij k = 0 in , p made in part (ii) on the functions p ij = ji are thus sucient conditions for p the equations i F j = ij F p in to have solutions. Conversely, a simple
35
computation shows that they are also necessary conditions, simply expressing that, if these equations have a solution, then necessarily ik F j = ki F j in . It is no surprise that these necessary conditions are of the same nature as those of Theorem 1.5-1, viz., ik g j = ij g k in . (2) The assumed positive deniteness of the matrices (gij ) is used only in part (iii), for dening ad hoc initial vectors g 0 i. The denitions of the functions p ij and ijq imply that the functions
satisfy, for all i, j, k, p, Rqijk = Rjkqi = Rqikj , Rqijk = 0 if j = k or q = i. These relations in turn imply that the eighty-one sucient conditions Rqijk = 0 in for all i, j, k, q {1, 2, 3}, are satised if and only if the six relations R1212 = R1213 = R1223 = R1313 = R1323 = R2323 = 0 in are satised (as is immediately veried, there are other sets of six relations that will suce as well, again owing to the relations satised by the functions Rqijk for all i, j, k, q ). To conclude, we briey review various extensions of the fundamental existence result of Theorem 1.6-1. First, a quick look at its proof reveals that it holds verbatim in any dimension d 2, i.e., with R3 replaced by Rd and E3 by a d-dimensional Euclidean space Ed . This extension only demands that Latin indices and exponents now range in the set {1, 2, . . . , d} and that the sets of mad d trices M3 , S3 , and S3 > be replaced by their d-dimensional counterparts M , S , d and S> . The regularity assumptions on the components gij of the symmetric positive denite matrix eld C = (gij ) made in Theorem 1.6-1, viz., that gij C 2 (), can be signicantly weakened. More specically, C. Mardare [2003] has shown that the existence theorem still holds if gij C 1 (), with a resulting mapping in the space C 2 (; Ed ). Then S. Mardare [2004] has shown that the existence 2, (), with a resulting mapping in the space theorem still holds if gij Wloc 2, d Wloc (; E ). As expected, the sucient conditions Rqijk = 0 in of Theorem 1.6-1 are then assumed to hold only in the sense of distributions, viz., as
p {ikq j + ijq k + p ij kqp ik jqp } dx = 0
for all D(). The existence result has also been extended up to the boundary of the set by Ciarlet & C. Mardare [2004a]. More specically, assume that the set
36
[Ch. 1
satises the geodesic property (in eect, a mild smoothness assumption on the boundary , satised in particular if is Lipschitz-continuous) and that the functions gij and their partial derivatives of order 2 can be extended by continuity to the closure , the symmetric matrix eld extended in this fashion remaining positive-denite over the set . Then the immersion and its partial derivatives of order 3 can be also extended by continuity to . Ciarlet & C. Mardare [2004a] have also shown that, if in addition the geodesic distance is equivalent to the Euclidean distance on (a property stronger than the geodesic property, but again satised if the boundary is Lipschitzcontinuous), then a matrix eld (gij ) C 2 (; Sn > ) with a Riemann curvature tensor vanishing in can be extended to a matrix eld (gij ) C 2 (; Sn > ) dened on a connected open set containing and whose Riemann curvature tensor still vanishes in . This result relies on the existence of continuous extensions to of the immersion and its partial derivatives of order 3 and on a deep extension theorem of Whitney [1934].
1.7
UNIQUENESS UP TO ISOMETRIES OF IMMERSIONS WITH THE SAME METRIC TENSOR
In Section 1.6, we have established the existence of an immersion : R3 E3 giving rise to a set () with a prescribed metric tensor, provided the given metric tensor eld satises ad hoc sucient conditions. We now turn to the question of uniqueness of such immersions. This uniqueness result is the object of the next theorem, aptly called a rigidity theorem in view of its geometrical interpretation: It asserts that, if two immersions C 1 (; E3 ) and C 1 (; E3 ) share the same metric tensor eld, then the set () is obtained by subjecting the set () either to a rotation (represented by an orthogonal matrix Q with det Q = 1), or to a symmetry with respect to a plane followed by a rotation (together represented by an orthogonal matrix Q with det Q = 1), then by subjecting the rotated set to a translation (represented by a vector c). The terminology rigidity theorem reects that such a geometric transformation indeed corresponds to the idea of a rigid transformation of the set () (provided a symmetry is included in this denition). Let O3 denote the set of all orthogonal matrices of order three. Theorem 1.7-1. Let be a connected open subset of R3 and let C 1 (; E3 ) and C 1 (; E3 ) be two immersions such that their associated metric tensors satisfy T = in . Then there exist a vector c E3 and an orthogonal matrix Q O3 such that (x) = c + Q(x) for all x .
T
Sect. 1.7]
Uniqueness of immersions with the same metric tensor
37
Proof. For convenience, the three-dimensional vector space R3 is identied throughout this proof with the Euclidean space E3 . In particular then, R3 inherits the inner product and norm of E3 . The spectral norm of a matrix A M3 is denoted |A| := sup{|Ab|; b R3 , |b| = 1}. To begin with, we consider the special case where : E3 = R3 is the identity mapping. The issue of uniqueness reduces in this case to nding C 1 (; E3 ) such that (x)T (x) = I for all x . Parts (i) to (iii) are devoted to solving these equations. (i) We rst establish that a mapping C 1 (; E3 ) that satises (x)T (x) = I for all x is locally an isometry: Given any point x0 , there exists an open neighborhood V of x0 contained in such that |(y ) (x)| = |y x| for all x, y V. Let B be an open ball centered at x0 and contained in . Since the set B is convex, the mean-value theorem (for a proof, see, e.g., Schwartz [1992]) can be applied. It shows that |(y ) (x)| sup |(z )||y x| for all x, y B.
z ]x,y [
Since the spectral norm of an orthogonal matrix is one, we thus have |(y ) (x)| |y x| for all x, y B. Since the matrix (x0 ) is invertible, the local inversion theorem (for a proof, see, e.g., Schwartz [1992]) shows that there exist an open neighborhood V of x0 contained in and an open neighborhood V of (x0 ) in E3 such that the restriction of to V is a C 1 -dieomorphism from V onto V . Besides, there is no loss of generality in assuming that V is contained in B and that V is convex (to see this, apply the local inversion theorem rst to the restriction of to B , thus producing a rst neighborhood V of x0 contained in B , then to the restriction of the inverse mapping obtained in this fashion to an open ball V centered at (x0 ) and contained in (V )). Let 1 : V V denote the inverse mapping of : V V . The chain rule applied to the relation 1 ((x)) = x for all x V then shows that 1 (x) = (x)1 for all x = (x), x V.
38
[Ch. 1
The matrix 1 (x) being thus orthogonal for all x V , the mean-value theorem applied in the convex set V shows that |1 (y ) 1 (x)| |y x| for all x, y V , or equivalently, that |y x| |(y ) (x)| for all x, y V. The restriction of the mapping to the open neighborhood V of x0 is thus an isometry. (ii) We next establish that, if a mapping C 1 (; E3 ) is locally an isometry, in the sense that, given any x0 , there exists an open neighborhood V of x0 contained in such that |(y ) (x)| = |y x| for all x, y V , then its derivative is locally constant, in the sense that (x) = (x0 ) for all x V. The set V being that found in (i), let the dierentiable function F : V V R be dened for all x = (xp ) V and all y = (yp ) V by F (x, y ) := ( (y ) (x))( (y ) (x)) (y x )(y x ). Then F (x, y ) = 0 for all x, y V by (i). Hence Gi (x, y ) := 1 F (x, y ) = (y )( (y ) (x)) i (y x ) = 0 2 yi yi
for all x, y V . For a xed y V , each function Gi (, y ) : V R is dierentiable and its derivative vanishes. Consequently, Gi (x, y ) = (y ) (x) + ij = 0 for all x, y V, xi yi xj or equivalently, in matrix form, (y )T (x) = I for all x, y V. Letting y = x0 in this relation shows that (x) = (x0 ) for all x V. (iii) By (ii), the mapping : M3 is dierentiable and its derivative vanishes in . Therefore the mapping : E3 is twice dierentiable and its second Fr echet derivative vanishes in . The open set being connected, a classical result from dierential calculus (see, e.g., Schwartz [1992, Theorem 3.7.10]) shows that the mapping is ane in , i.e., that there exists a vector
Sect. 1.7]
39
c E3 and a matrix Q M3 such that (the notation ox designates the column vector with components xi ) (x) = c + Qox for all x . Since Q = (x0 ) and (x0 )T (x0 ) = I by assumption, the matrix Q is orthogonal. (iv) We now consider the general equations gij = gij in , noting that they also read (x)T (x) = (x)T (x) for all x . Given any point x0 , let the neighborhoods V of x0 and V of (x0 ) and the mapping 1 : V V be dened as in part (i) (by assumption, the mapping is an immersion; hence the matrix (x0 ) is invertible). Consider the composite mapping := 1 : V E3 . Clearly, C 1 (V ; E3 ) and (x) = (x)1 (x) = (x)(x)1 for all x = (x), x V. Hence the assumed relations (x)T (x) = (x)T (x) for all x imply that (x)T (x) = I for all x V. By parts (i) to (iii), there thus exist a vector c R3 and a matrix Q O3 such that (x) = (x) = c + Q(x) for all x = (x), x V, hence such that (x) := (x)(x)1 = Q for all x V. The continuous mapping : V M3 dened in this fashion is thus locally constant in . As in part (iii), we conclude from the assumed connectedness of that the mapping is constant in . Thus the proof is complete. An isometry of E3 is a mapping J : E3 E3 of the form J(x) = c + Q ox for all x E3 , with c E3 and Q O3 (an analogous denition holds verbatim in any Euclidean space of dimension d 2). Clearly, an isometry preserves distances in the sense that |J(y ) J(x)| = |y x| for all x, y .
40
[Ch. 1
Remarkably, the converse is also true, according to the classical MazurUlam theorem, which asserts the following: Let be a connected subset in Rd , and let : Rd be a mapping that satises |(y ) (x)| = |y x| for all x, y . Then is an isometry of Rd . Parts (ii) and (iii) of the above proof thus provide a proof of this theorem under the additional assumption that the mapping is of class C 1 (the extension from R3 to Rd is trivial). In Theorem 1.7-1, the special case where is the identity mapping of R3 identied with E3 is the classical Liouville theorem. This theorem thus asserts that if a mapping C 1 (; E3 ) is such that (x) O3 for all x , where is an open connected subset of R3 , then is an isometry. Two mappings C 1 (; E3 ) and C 1 (; E3 ) are said to be isometrically equivalent if there exist c E3 and Q O3 such that = c + Q in , i.e., such that = J , where J is an isometry of E3 . Theorem 1.7-1 thus asserts that two immersions C 1 (; E3 ) and C 1 (; E3 ) share the same metric tensor eld over an open connected subset of R3 if and only if they are isometrically equivalent. Remark. In terms of covariant components gij of metric tensors, parts (i) to (iii) of the above proof provide the solution to the equations gij = ij in , while part (iv) provides the solution to the equations gij = i j in , where C 1 (; E3 ) is a given immersion. While the immersions found in Theorem 1.6-1 are thus only dened up to isometries in E3 , they become uniquely determined if they are required to satisfy ad hoc additional conditions, according to the following corollary to Theorems 1.6-1 and 1.7-1. Theorem 1.7-2. Let the assumptions on the set and on the matrix eld C be as in Theorem 1.6-1, let a point x0 be given, and let F0 M3 be any matrix that satises FT 0 F0 = C(x0 ). Then there exists one and only one immersion C 3 (; E3 ) that satises (x)T (x) = C(x) for all x , (x0 ) = 0 and (x0 ) = F0 . Proof. Given any immersion C 3 (; E3 ) that satises (x)T (x) = C(x) for all x (such immersions exist by Theorem 1.6-1), let the mapping : R3 be dened by (x) := F0 (x0 )1 ((x) (x0 )) for all x .
Sect. 1.7]
41
Then it is immediately veried that this mapping satises the announced properties. Besides, it is uniquely determined. To see this, let C 3 (; E3 ) and C 3 (; E3 ) be two immersions that satisfy (x)T (x) = (x)T (x) for all x . Hence there exist (by Theorem 1.7-1) c R3 and Q O3 such that (x) = c + Q(x) for all x , so that (x) = Q(x) for all x . The relation (x0 ) = (x0 ) then implies that Q = I and the relation (x0 ) = (x0 ) in turn implies that c = 0. Remark. One possible choice for the matrix F0 is the square root of the symmetric positive-denite matrix C(x0 ). Theorem 1.7-1 constitutes the classical rigidity theorem, in that both immersions and are assumed to be in the space C 1 (; E3 ). The next theorem is an extension, due to Ciarlet & C. Mardare [2003], that covers the case where one of the mappings belongs to the Sobolev space H 1 (; E3 ). The way the result in part (i) of the next proof is derived is due to Friesecke, James & M uller [2002]; the result of part (i) itself goes back to Reshetnyak [1967]. Let O3 + denote the set of all proper orthogonal matrices of order three, i.e., of all orthogonal matrices Q O3 with det Q = 1. Theorem 1.7-3. Let be a connected open subset of R3 , let C 1 (; E3 ) be a mapping that satises det > 0 in , and let H 1 (; E3 ) be a mapping that satises det > 0 a.e. in and T = a.e. in . Then there exist a vector c E3 and a matrix Q O3 + such that (x) = c + Q(x) for almost all x . Proof. The Euclidean space E3 is identied with the space R3 throughout the proof. (i) To begin with, consider the special case where (x) = x for all x . In other words, we are given a mapping H1 () that satises (x) O3 + for almost all x . Hence Cof (x) = (det (x))(x)T = (x)T for almost all x , on the one hand. Since, on the other hand, div Cof = 0 in (D (B ))3
T
42
[Ch. 1
in any open ball B such that B (to see this, combine the density of C 2 (B ) in H 1 (B ) with the classical Piola identity in the space C 2 (B ); for a proof of this identity, see, e.g., Ciarlet [1988, Theorem 1.7.1]), we conclude that = div Cof = 0 in (D (B ))3 . Hence = (j ) (C ())3 . For such mappings, the identity (i j i j ) = 2i j i (j ) + 2ik j ik j , together with the relations j = 0 and i j i j = 3 in , shows that ik j = 0 in . The assumed connectedness of then implies that there exist a vector 3 c E3 and a matrix Q O3 + (by assumption, (x) O+ for almost all x ) such that (x) = c + Q ox for almost all x . (ii) Consider next the general case. Let x0 be given. Since is an immersion, the local inversion theorem can be applied; there thus exist bounded open neighborhoods U of x0 and U of (x0 ) satisfying U and {U } (), such that the restriction U of to U can be extended to a C 1 -dieomorphism from U onto {U } . 1 : U U denote the inverse mapping of U , which therefore Let U 1 1 for all x = (x) U (the notation indicates satises U (x) = (x) that dierentiation is carried out with respect to the variable x U ). Dene the composite mapping
1 3 := U :U R . 1 1 Since H1 (U ) and U can be extended to a C -dieomorphism from {U } 1 3 onto U , it follows that H (U ; R ) and that 1 1 (x) = (x) U (x) = (x)(x)
for almost all x = (x) U (see, e.g., Adams [1975, Chapter 3]). Hence the assumptions det > 0 in , det > 0 a.e. in , and T = a.e. in , together imply that (x) O3 + for almost all x U . such that By (i), there thus exist c E3 and Q O3 + (x) = (x) = c + Q ox for almost all x = (x) U , or equivalently, such that (x) := (x)(x)1 = Q for almost all x U. Since the point x0 is arbitrary, this relation shows that L1 loc (). By a classical result from distribution theory (cf. Schwartz [1966, Section 2.6]),
T
Sect. 1.8]
43
we conclude from the assumed connectedness of that (x) = Q for almost all x , and consequently that (x) = c + Q(x) for almost all x .
Remarks. (1) The existence of H 1 (; E3 ) satisfying the assumptions of Theorem 1.7-3 thus implies that H 1 (; E3 ) and C 1 (; E3 ). (2) If C 1 (; E3 ), the assumptions det > 0 in and det > 0 in are no longer necessary; but then it can only be concluded that Q O3 : This is the classical rigidity theorem (Theorem 1.7-1), of which Liouvilles theorem is the special case corresponding to (x) = x for all x . (3) The result established in part (i) of the above proof asserts that, given a connected open subset of R3 , if a mapping H 1 (; E3 ) is such that 3 3 (x) O3 + for almost all x , then there exist c E and Q O+ such that (x) = c + Qox for almost all x . This result thus constitutes a generalization of Liouvilles theorem. (4) By contrast, if the mapping is assumed to be instead in the space H 1 (; E3 ) (as in Theorem 1.7-3), an assumption about the sign of det becomes necessary. To see this, let for instance be an open ball centered at the origin in R3 , let (x) = x, and let (x) = x if x1 0 and (x) = (x1 , x2 , x3 ) if x1 < 0. Then H 1 (; E3 ) and O3 a.e. in ; yet there does not exist any orthogonal matrix such that (x) = Q ox for all x , since () {x R3 ; x1 0} (this counter-example was kindly communicated to the author by Sorin Mardare). (5) Surprisingly, the assumption det > 0 in cannot be replaced by the weaker assumption det > 0 a.e. in . To see this, let for instance be an open ball centered at the origin in R3 , let (x) = (x1 x2 2 , x2 , x3 ) and let (x) = (x) if x2 0 and (x) = (x1 x2 , x , x ) if x < 0 (this counter2 3 2 2 example was kindly communicated to the author by Herv e Le Dret). (6) If a mapping C 1 (; E3 ) satises det > 0 in , then is an immersion. Conversely, if is a connected open set and C 1 (; E3 ) is an immersion, then either det > 0 in or det < 0 in . The assumption that det > 0 in made in Theorem 1.7-3 is simply intended to x ideas (a similar result clearly holds under the other assumption). (7) A little further ado shows that the conclusion of Theorem 1.7-3 is still 1 (; E3 ). valid if H 1 (; E3 ) is replaced by the weaker assumption Hloc Like the existence results of Section 1.6, the uniqueness theorems of this section hold verbatim in any dimension d 2, with R3 replaced by Rd and Ed by a d-dimensional Euclidean space.
44
[Ch. 1
1.8
CONTINUITY OF AN IMMERSION AS A FUNCTION OF ITS METRIC TENSOR
Let be a connected and simply-connected open subset of R3 . Together, Theorems 1.6-1 and 1.7-1 establish the existence of a mapping F that associates with any matrix eld C = (gij ) C 2 (; S3 > ) satisfying
p Rqijk := j ikq k ijq + p ij kqp ik jqp = 0 in ,
where the functions ijq and p ij are dened in terms of the functions gij as in Theorem 1.6-1, a well-dened element F (C) in the quotient set C 3 (; E3 )/R, where (, ) R means that there exist a vector a E3 and a matrix Q O3 such that (x) = a + Q(x) for all x . A natural question thus arises as to whether there exist natural topologies on the space C 2 (; S3 ) and on the quotient set C 3 (; E3 )/R such that the mapping F dened in this fashion is continuous. Equivalently, is an immersion a continuous function of its metric tensor? The object of this section, which is based on Ciarlet & Laurent [2003], is to provide an armative answer to this question (see Theorem 1.8-5). Note that such a question is not only clearly relevant to dierential geometry per se, but it also naturally arises in nonlinear three-dimensional elasticity. As we shall see more specically in Section 3.1, a smooth enough immersion = (i ) : E3 may be thought of in this context as a deformation of the set viewed as a reference conguration of a nonlinearly elastic body (although such an immersion should then be in addition injective and orientation-preserving in order to qualify for this denition; for details, see, e.g., Ciarlet [1988, Section 1.4] or Antman [1995, Section 12.1]). In this context, the associated matrix C(x) = (gij (x)) = (x)T (x), is called the (right) Cauchy-Green tensor at x and the matrix (x) = (j i (x)) M3 , representing the Fr echet derivative of the mapping at x, is called the deformation gradient at x. The Cauchy-Green tensor eld C = T : S3 > associated with a deformation : E3 plays a major role in the theory of nonlinear threedimensional elasticity, since the response function, or the stored energy function, of a frame-indierent elastic, or hyperelastic, material necessarily depends on the deformation gradient through the Cauchy-Green tensor (for a detailed description see, e.g., Ciarlet [1988, Chapters 3 and 4]). As already suggested by Antman [1976], the Cauchy-Green tensor eld of the unknown deformed conguration could thus also be regarded as the primary unknown rather than the deformation itself as is customary.
Sect. 1.8]
An immersion as a function of its metric tensor
45
To begin with, we list some specic notations that will be used in this section for addressing the question raised above. Given a matrix A M3 , we let (A) denote its spectral radius (i.e., the largest modulus of the eigenvalues of A) and we let |Ab| = {(AT A)}1/2 |A| := jsup |b| 3 b R
b=0
denote its spectral norm. Let be an open subset of R3 . The notation K means that K is , we dene the a compact subset of . If g C (; R), 0, and K semi-norms |g |
,K
= jsup | g (x)| and

x K ||=
,K
= jsup | g (x)|,
x K ||
where stands for the standard multi-index notation for partial derivatives. If C (; E3 ) or A C (; M3 ), 0, and K , we likewise set || |A|
,K
sup | (x)|
x K ||=
and
,K
sup | (x)|,
x K ||
,K
sup | A(x)|
x K ||=
and
,K
sup | A(x)|,
x K ||
where || denotes either the Euclidean vector norm or the matrix spectral norm. The next sequential continuity results (Theorems 1.8-1, 1.8-2, and 1.8-3) constitute key steps toward establishing the continuity of the mapping F (see n Theorem 1.8-5). Note that the functions Rqijk occurring in their statements are n in the same way that the funcmeant to be constructed from the functions gij tions Rqijk are constructed from the functions gij . To begin with, we establish the sequential continuity of the mapping F at C = I. Theorem 1.8-1. Let be a connected and simply-connected open subset of n n ) C 2 (; S3 R3 . Let Cn = (gij > ), n 0, be matrix elds satisfying Rqijk = 0 in , n 0, such that
n
lim
Cn I
2,K
= 0 for all K
Then there exist immersions n C 3 (; E3 ) satisfying (n )T n = Cn in , n 0, such that

n
lim
n id
3,K
= 0 for all K
where id denotes the identity mapping of the set , the space R3 being identied here with E3 (in other words, id(x) = x for all x ).
46
[Ch. 1
Proof. The proof is broken into four parts, numbered (i) to (iv). The rst part is a preliminary result about matrices (for convenience, it is stated here for matrices of order three, but it holds as well for matrices of arbitrary order). (i) Let matrices An M3 , n 0, satisfy
n
lim (An )T An = I.
Then there exist matrices Qn O3 , n 0, that satisfy

n
lim Qn An = I.
Since the set O3 is compact, there exist matrices Qn O3 , n 0, such that |Qn An I| = inf 3 |RAn I|.
RO
We assert that the matrices Q dened in this fashion satisfy limn Qn An = I. For otherwise, there would exist a subsequence (Qp )p0 of the sequence (Qn )n0 and > 0 such that |Qp Ap I| = inf 3 |RAp I| for all p 0.
RO
Since
p
lim |Ap | = lim
((Ap )T Ap ) =
(I) = 1,
the sequence (Ap )p0 is bounded. Therefore there exists a further subsequence (Aq )q0 that converges to a matrix S, which is orthogonal since ST S = lim (Aq )T Aq = I.
q
But then
q
lim ST Aq = ST S = I,
which contradicts inf RO3 |RAq I| for all q 0. This proves (i). In the remainder of this proof, the matrix elds Cn , n 0, are meant to be those appearing in the statement of Theorem 1.8-1 and the notation id stands for the identity mapping of the set . (ii) Let mappings n C 3 (; E3 ), n 0, satisfy (n )T n = Cn in (such mappings exist by Theorem 1.6-1). Then
n
lim |n id|
,K
= lim |n |
n
,K
= 0 for all K
and for
= 2, 3.
As usual, given any immersion C 3 (; E3 ), let g i = i , let gij = g i g j , q and let the vectors g q be dened by the relations g i g q = i . It is then immediately veried that ij = i g j = (i g j g q )g q = 1 (j giq + i gjq q gij )g q . 2
Sect. 1.8]
47
Applying this relation to the mappings n thus gives ij n = 1 n n n (j giq + i gjq q gij )(g q )n , n 0, 2
q where the vectors (g q )n are dened by means of the relations i n (g q )n = i . Let K denote an arbitrary compact subset of . On the one hand, n n n lim |j giq + i gjq q gij |0,K = 0,
n n |1,K = limn |gij ij |1,K = 0 by assumption. On the other since limn |gij q n hand, the norms |(g ) |0,K are bounded independently of n 0; to see this, observe that (g q )n is the q -th column vector of the matrix (n )1 , then that
|(n )1 |0,K = |{((n )T (n )1 )}1/2 |0,K

n 1 1/2 n 1 = |{((gij ) )} |0,K {|(gij ) |0,K }1/2 ,
and, nally, that

n n n 1 lim |(gij ) I|0,K = 0 = lim |(gij ) I|0,K = 0. n
Consequently,
n
lim |n id|2,K = lim |n |2,K = 0 for all K

n
Dierentiating the relations i g j g q = 1 2 (j giq + i gjq q gij ) yields ijp = ip g j = (ip g j g q )g q 1 (jp giq + ip gjq pq gij ) i g j p g q g q . = 2
n n Observing that limn |gij | ,K = limn |gij ij | ,K = 0 for = 1, 2 by q n assumption and recalling that the norms |(g ) |0,K are bounded independently of n 0, we likewise conclude that n
lim |n id|3,K = lim |n |3,K = 0 for all K

n n
.
n n
(iii) There exist mappings C 3 (; E3 ) that satisfy ( )T = Cn in , n 0, and

n
lim | id|1,K = 0 for all K
Let n C 3 (; E3 ) be mappings that satisfy ( n )T n = Cn in , n 0 (such mappings exist by Theorem 1.6-1), and let x0 denote a point in the set . Since limn n (x0 )T n (x0 ) = I by assumption, part (i) implies that there exist orthogonal matrices Qn (x0 ), n 0, such that
n
lim Qn (x0 ) n (x0 ) = I.
48

n
[Ch. 1
Then the mappings C 3 (; E3 ), n 0, dened by (x) := Qn (x0 ) n (x), x , satisfy

n n
( )T = Cn in , so that their gradients C 2 (; M3 ) satisfy

n
lim |i |0,K = lim | |2,K = 0 for all K

n
by part (ii). In addition,

n
lim (x0 ) = lim Qn n (x0 ) = I.

n
Hence a classical theorem about the dierentiability of the limit of a sequence of mappings that are continuously dierentiable on a connected open set and that take their values in a Banach space (see, e.g., Schwartz [1992, Theorem n 3.5.12]) shows that the mappings uniformly converge on every compact subset of toward a limit R C 1 (; M3 ) that satises i R(x) = lim i (x) = 0 for all x .
n n
This shows that R is a constant mapping since is connected. Consequently, n R = I since in particular R(x0 ) = limn (x0 ) = I. We have therefore established that
n
lim | id|1,K = lim | I|0,K = 0 for all K

n
(iv) There exist mappings n C 3 (; E3 ) satisfying (n )T n = Cn in , n 0, and

n
lim |n id|
,K
= 0 for all K
and for
= 0, 1.
The mappings n := { (x0 ) x0 } C 3 (; E3 ), n 0, clearly satisfy (n )T n = Cn in , n 0,

n n n
lim |n id|1,K = lim |n I|0,K = 0 for all K

n
n (x0 ) = x0 , n 0. Again applying the theorem about the dierentiability of the limit of a sequence of mappings used in part (iii), we conclude from the last two relations
Sect. 1.8]
49
that the mappings n uniformly converge on every compact subset of toward a limit C 1 (; E3 ) that satises (x) = lim n (x) = I for all x .
n
This shows that ( id) is a constant mapping since is connected. Consequently, = id since in particular (x0 ) = limn n (x0 ) = x0 . We have thus established that
n
lim |n id|0,K = 0 for all K
This completes the proof of Theorem 1.8-1. We next establish the sequential continuity of the mapping F at those matrix T elds C C 2 (; S3 > ) that can be written as C = with an injective 3 3 mapping C (; E ). Theorem 1.8-2. Let be a connected and simply-connected open subset of R3 . n n 2 3 Let C = (gij ) C 2 (; S3 > ) and C = (gij ) C (; S> ), n 0, be matrix elds n satisfying respectively Rqijk = 0 in and Rqijk = 0 in , n 0, such that
n
lim
Cn C
2,K
= 0 for all K
Assume that there exists an injective immersion C 3 (; E3 ) such that T = C in . Then there exist immersions n C 3 (; E3 ) satisfying (n )T n = Cn in , n 0, such that
n
lim
3,K
= 0 for all K
Proof. The assumptions made on the mapping : R3 E3 imply that the set := () E3 is open, connected, and simply-connected, and that the inverse mapping : E3 R3 belongs to the space C 3 (; R3 ). Dene n the matrix elds (gij ) C 2 (; S3 > ), n 0, by letting
n n (x)) := (x)T (gij (x))(x)1 for all x = (x) . (gij n Given any compact subset K of , let K := (K ). Since limn gij gij 2,K = 0 because K is a compact subset of , the denition of the functions n : R and the chain rule together imply that gij n
lim
n gij ij
b 2,K
= 0.
n Given x = (xi ) , let i = / xi . Let Rqijk denote the functions conn structed from the functions gij in the same way that the functions Rqijk are
50
[Ch. 1
constructed from the functions gij . Since it is easily veried that these funcn tions satisfy Rqijk = 0 in , Theorem 1.8-1 applied over the set shows that there exist mappings C 3 (; E3 ) satisfying
n n n n in , n 0, i j = gij
such that
n
lim
id
b 3,K
= 0 for all K
where id denotes the identity mapping of the set , the space E3 being identied here with R3 . Dene the mappings n C 3 (; S3 > ), n 0, by letting n (x) = (x) for all x = (x) . Given any compact subset K of , let K := (K ). Since limn n id 3,K b = 0, the denition of the mappings and the chain rule together imply that lim n 3,K = 0,
n n n
on the one hand. Since, on the other hand, (n )T n = Cn in , the proof is complete. We are now in a position to establish the sequential continuity of the mapping T F at any matrix eld C C 2 (; S3 > ) that can be written as C = with 3 3 C (; E ). Theorem 1.8-3. Let be a connected and simply-connected open subset of R3 . n n 2 3 Let C = (gij ) C 2 (; S3 > ) and C = (gij ) C (, S> ), n 0, be matrix elds n respectively satisfying Rqijk = 0 in and Rqijk = 0 in , n 0, such that
n
lim
Cn C
2,K
= 0 for all K
Let C 3 (; E3 ) be any immersion that satises T = C in (such immersions exist by Theorem 1.6-1). Then there exist immersions n C 3 (; E3 ) satisfying (n )T n = Cn in , n 0, such that
n
lim
3,K
= 0 for all K
Proof. The proof is broken into four parts. In what follows, C and Cn designate matrix elds possessing the properties listed in the statement of the theorem. (i) Let C 3 (; E3 ) be any mapping that satises T = C in . Then there exist a countable number of open balls Br , r 1, such that r = r=1 Br and such that, for each r 1, the set s=1 Bs is connected and the restriction of to Br is injective.
Sect. 1.8]
51
Given any x , there exists an open ball Vx such that the restriction of to Vx is injective. Since = x Vx can also be written as a countable union of compact subsets of , there already exist countably many such open balls, denoted Vr , r 1, such that = r=1 Vr . Let r1 := 1, B1 := Vr1 , and r2 := 2. If the set Br1 Vr2 is connected, let B2 := Vr2 and r3 := 3. Otherwise, there exists a path 1 in joining the centers of Vr1 and Vr2 since is connected. Then there exists a nite set I1 = {r1 (1), r1 (2), , r1 (N1 )} of integers, with N1 1 and 2 < r1 (1) < r1 (2) < < r1 (N1 ), such that 1 Vr1 Vr2
r I1
Vr .
Furthermore there exists a permutation 1 of {1, 2, . . . , N1 } such that the sets r N1 Vr1 ( s=1 V1 (s) ), 1 r N1 , and Vr1 ( s=1 V1 (s) ) Vr2 are connected. Let Br := V1 (r1) , 2 r N1 + 1,
N +2
BN1 +2 := Vr2 ,
r3 := min i {1 (1), . . . , 1 (N1 )}; i 3 .

1 If the set ( r=1 Br ) Vr3 is connected, let BN1 +3 := Vr3 . Otherwise, apply the same argument as above to a path 2 in joining the centers of Vr2 and Vr3 , and so forth. The iterative procedure thus produces a countable number of open balls Br , r 1, that possess the announced properties. In particular, = r=1 Br since, by construction, the integer ri appearing at the i-th stage satises ri i.
3 3 (ii) By Theorem 1.8-2, there exist mappings n 1 C (B1 ; E ) and 2 C 3 (B2 ; E3 ), n 0, that satisfy n T n (n 1 ) 1 = C in B1 n n (2 )T 2
and and
lim
n 1
3,K 3,K
= 0 for all K = 0 for all K
B1 , B2 ,
= Cn in B2
n lim 2 n
and by Theorem 1.7-1, there exist vectors cn E3 and matrices Qn O3 , n 0, such that n 2 (x) = cn + Qn n 1 (x) for all x B1 B2 . Then we assert that
n
lim cn = 0 and lim Qn = I.

n
Let (Qp )p0 be a subsequence of the sequence (Qn )n0 that converges to a (necessarily orthogonal) matrix Q and let x1 denote a point in the set B1 p p B2 . Since cp = 2 (x1 ) Qp 1 (x1 ) and limn 2 (x1 ) = limn p 1 (x1 ) = (x1 ), the subsequence (cp )p0 also converges. Let c := limp cp . Thus (x) = lim 2 (x)
p p p
= lim (cp + Qp p 1 (x)) = c + Q(x) for all x B1 B2 ,
52
[Ch. 1
on the one hand. On the other hand, the dierentiability of the mapping implies that (x) = (x1 ) + (x1 )(x x1 ) + o(|x x1 |) for all x B1 B2 . Note that (x1 ) is an invertible matrix, since (x1 )T (x1 ) = (gij (x1 )). Let b := (x1 ) and A := (x1 ). Together, the last two relations imply that b + A(x x1 ) = c + Qb + QA(x x1 ) + o(|x x1 |), and hence (letting x = x1 shows that b = c + Qb) that A(x x1 ) = QA(x x1 ) + o(|x x1 |) for all x B1 B2 . The invertibility of A thus implies that Q = I and therefore that c = bQb = 0. The uniqueness of these limits shows that the whole sequences (Qn )n0 and (cn )n0 converge.
3 3 (iii) Let the mappings n 2 C (B1 B2 ; E ), n 0, be dened by n n 2 (x) := 1 (x) for all x B1 , n T n n 2 (x) := (Q ) (2 (x) c ) for all x B2 . n
Then
n T n (n 2 ) 2 = C in B1 B2
(as is clear), and

n
lim
n 2
3,K
= 0 for all K
B1 B2 .
The plane containing the intersection of the boundaries of the open balls B1 and B2 is the common boundary of two closed half-spaces in R3 , H1 containing the center of B1 , and H2 containing that of B2 (by construction, the set B1 B2 is connected; see part (i)). Any compact subset K of B1 B2 may thus be written as K = K1 K2 , where K1 := (K H1 ) B1 and K2 := (K H2 ) B2 (that the open sets found in part (i) may be chosen as balls thus play an essential r ole here). Hence
n
lim
n 2
3,K1
= 0 and lim
n 2
3,K2
= 0,
the second relation following from the denition of the mapping n 2 on B2 K2 n and on the relations limn 2 3,K2 = 0 (part (ii)) and limn Qn = I and limn cn = 0 (part (iii)).
Sect. 1.8]
53
(iv) It remains to iterate the procedure described in parts (ii) and (iii). For r 3 3 some r 2, assume that mappings n r C ( s=1 Bs ; E ), n 0, have been found that satisfy
r n T n (n r ) r = C in s=1 r n
Bs , Bs .
s=1
lim
n r
2,K
= 0 for all K
Since the restriction of to Br+1 is injective (part (i)), Theorem 1.8-2 shows n that there exist mappings r+1 C 3 (Br+1 ; E3 ), n 0, that satisfy (r+1 )T r+1 = Cn in Br+1 ,
n n n
lim
r+1
3,K
= 0 for all K
Br+1 ,
+1 and since the set r s=1 Bs is connected (part (i)), Theorem 1.7-1 shows that n there exist vectors c E3 and matrices Qn O3 , n 0, such that
r+1 (x) = cn + Qn n r (x) for all x

s=1
Bs Br+1 .
Then an argument similar to that used in part (ii) shows that limn Qn = I and limn cn = 0, and an argument similar to that used in part (iii) (note that the ball Br+1 may intersect more than one of the balls Bs , 1 s r) r 3 3 shows that the mappings n r +1 C ( s=1 Bs ; E ), n 0, dened by
r n n r +1 (x) := r (x) for all x s=1 n T n n r +1 (x) := (Q ) (r (x) c ) for all x Br +1 , n
Bs ,
satisfy
r n
lim
n r +1
3,K
= 0 for all K
s=1
Bs .
Then the mappings n : E3 , n 0, dened by

r
n (x) := n r (x) for all x

s=1
Bs , r 1,
possess all the required properties: They are unambiguously dened since for all r n 3 s > r, n s (x) = r (x) for all x s=1 Bs by construction; they are of class C r n since the mappings r : s=1 Bs E3 are themselves of class C 3 ; they satisfy (n )T n = Cn in since the mappings n r satisfy the same relations r in s=1 Bs ; and nally, they satisfy limn n 3,K = 0 for all K r since any compact subset of is contained in s=1 Bs for r large enough. This completes the proof.
54
[Ch. 1
It is easily seen that the assumptions Rqijk = 0 in are in fact superuous in Theorem 1.8-3 (as shown in the next proof, these relations are consequences n of the assumptions Rqijk = 0 in , n 0, and limn Cn C 2,K = 0 for all K ). This observation gives rise to the following corollary to Theorem 1.8-3, in the form of another sequential continuity result, of interest by itself. The novelties are that the assumptions are now made on the immersions n , n 0, and that this result also provides the existence of a limit immersion . Theorem 1.8-4. Let be a connected and simply-connected open subset of R3 . Let there be given immersions n C 3 (; E3 ), n 0, and a matrix eld C C 2 (; S3 > ) such that
n
lim
(n )T n C
n
2,K
= 0 for all K
Then there exist immersions C 3 (; E3 ), n 0, of the form = cn + Qn n , with cn E3 and Qn O3 , which thus satisfy ( )T = (n )T n in for all n 0, and there exists an immersion C 3 (; E3 ) such that T = C in and lim
n n n n
3,K
= 0 for all K
n Proof. Let the functions Rqijk , n 0, and Rqijk be constructed from the n components gij and gij of the matrix elds Cn := (n )T n and C in the n usual way (see, e.g., Theorem 1.6-1). Then Rqijk = 0 in for all n 0, since these relations are simply the necessary conditions of Theorem 1.5-1. We now show that Rqijk = 0 in . To this end, let K be any compact subset of . The relations
Cn = C(I + C1 (Cn C)), n 0, together with the inequalities AB 2,K 4 A 2,K B 2,K valid for any matrix elds A, B C 2 (; M3 ), show that there exists n0 = n0 (K ) such that the matrix elds (I + C1 (Cn C))(x) are invertible at all x K for all n n0 . The same relations also show that there exists a constant M such that (Cn )1 2,K M for all n n0 . Hence the relations (Cn )1 C1 = C1 (C Cn )(Cn )1 , n n0 , together with the assumptions limn Cn C 2,K = 0, in turn imply that the components g ij,n , n n0 , and g ij of the matrix elds (Cn )1 and C1 satisfy
n
lim
g ij,n g ij
2,K
= 0.
With self-explanatory notations, it thus follows that

n
lim
n ijq ijq
1,K
= 0 and lim
p p,n ij ij
1,K
= 0,
Sect. 1.8]
55
n hence that limn Rqijk Rqijk 0,K = 0. This shows that Rqijk = 0 in K , hence that Rqijk = 0 in since K is an arbitrary compact subset of . By the fundamental existence theorem (Theorem 1.6-1), there thus exists a mapping C 3 (; E3 ) such that T = C in . Theorem 1.8-3 can now n be applied, showing that there exist mappings C 3 (; E3 ) such that
( )T = Cn in , n 0, and lim
3,K
for all K
Finally, the rigidity theorem (Theorem 1.7-1) shows that, for each n 0, n there exist cn E3 and Qn O3 such that = cn + Qn n in because n the mappings and n share the same metric tensor eld and the set is connected. It remains to show how the sequential continuity established in Theorem 1.8-3 implies the continuity of a deformation as a function of its metric tensor for ad hoc topologies. Let be an open subset of R3 . For any integers 0 and d 1, the space C (; Rd ) becomes a locally convex topological space when it is equipped with the Fr echet topology dened by the family of semi-norms ,K , K , dened earlier. Then a sequence (n )n0 converges to with respect to this topology if and only if
n
lim
,K
= 0 for all K
Furthermore, this topology is metrizable : Let (Ki )i0 be any sequence of subsets of that satisfy
Ki Then
n
and Ki int Ki+1 for all i 0, and =

i=0
Ki .
lim
,K
= 0 for all K
lim d (n , ) = 0,
n
where d ( , ) :=
i=0
1 ,Ki . 2i 1 + ,Ki
For details about Fr echet topologies, see, e.g., Yosida [1966, Chapter 1]. 3 (; E3 ) := C 3 (; E3 )/R denote the quotient set of C 3 (; E3 ) by the Let C equivalence relation R, where (, ) R means that and are isometrically equivalent (Section 1.7), i.e., that there exist a vector c E3 and a matrix Q O3 such that (x) = c + Q(x) for all x . Then it is easily veried 3 (; E3 ) becomes a metric space when it is equipped with the that the set C distance d3 dened by 3 ( ) = j inf d3 (, ) = j inf d3 (, c + Q ), , d
cE3 QO3
denotes the equivalence class of modulo R. where
56
[Ch. 1
We now show that the announced continuity of an immersion as a function of its metric tensor is a corollary to Theorem 1.8-1. If d is a metric dened on a set X , the associated metric space is denoted {X ; d}. Theorem 1.8-5. Let be a connected and simply-connected open subset of R3 . Let 2 2 3 (; S3 C0 > ) := {(gij ) C (; S> ); Rqijk = 0 in },
2 3 3 (; S3 and, given any matrix eld C = (gij ) C0 > ), let F (C) C (; E ) denote 3 3 the equivalence class modulo R of any C (; E ) that satises T = C in . Then the mapping 2 3 3 F : {C0 (; S3 > ); d2 } {C (; E ); d3 }
dened in this fashion is continuous.

2 3 3 Proof. Since {C0 (; S3 > ); d2 } and {C (; E ); d3 } are both metric spaces, it suces to show that convergent sequences are mapped through F into convergent sequences. 2 n 2 3 (; S3 Let then C C0 > ) and C C0 (; S> ), n 0, be such that n
lim d2 (Cn , C) = 0,
i.e., such that limn Cn C 2,K = 0 for all K . Given any F (C), Theorem 1.8-3 shows that there exist n F (Cn ), n 0, such that limn n 3,K = 0 for all K , i.e., such that
n
lim d3 (n , ) = 0.
Consequently,
n
3 (F (Cn ), F (C)) = 0. lim d
As shown by Ciarlet & C. Mardare [2004b], the above continuity result can be extended up to the boundary of the set , as follows. If is bounded and has a Lipschitz-continuous boundary, the mapping F of Theorem 1.8-5 can be extended to a mapping that is locally Lipschitz-continuous with respect to the topologies of the Banach spaces C 2 (; S3 ) for the continuous extensions of the symmetric matrix elds C, and C 3 (; E3 ) for the continuous extensions of the immersions (the existence of such continuous extensions is briey commented upon at the end of Section 1.6). Another extension, again motivated by nonlinear three-dimensional elasticity, is the following: Let be a bounded and connected subset of R3 , and let B be an elastic body with as its reference conguration. Thanks mostly to the landmark existence theory of Ball [1977], it is now customary in nonlinear three-dimensional elasticity to view any mapping H 1 (; E3 ) that is almosteverywhere injective and satises det > 0 a.e. in as a possible deformation
Sect. 1.8]
57
of B when B is subjected to ad hoc applied forces and boundary conditions. The almost-everywhere injectivity of (understood in the sense of Ciarlet & Ne cas [1987]) and the restriction on the sign of det mathematically express (in an arguably weak way) the non-interpenetrability and orientation-preserving conditions that any physically realistic deformation should satisfy. As mentioned earlier, the Cauchy-Green tensor eld T L1 (; S3 ) associated with a deformation H 1 (; E3 ) pervades the mathematical modeling of three-dimensional nonlinear elasticity. Conceivably, an alternative approach to the existence theory in three-dimensional elasticity could thus regard the Cauchy-Green tensor as the primary unknown, instead of the deformation itself as is usually the case. Clearly, the Cauchy-Green tensors depend continuously on the deformations, since the Cauchy-Schwarz inequality immediately shows that the mapping H 1 (; E3 ) T L1 (; S3 ) is continuous (irrespectively of whether the mappings are almost-everywhere injective and orientation-preserving). Then Ciarlet & C. Mardare [2004c] have shown that, under appropriate smoothness and orientation-preserving assumptions, the converse holds, i.e., the deformations depend continuously on their Cauchy-Green tensors, the topologies being those of the same spaces H 1 (; E3 ) and L1 (; S3 ) (by contrast with the orientation-preserving condition, the issue of non-interpenetrability turns out to be irrelevant to this issue). In fact, this continuity result holds in an arbitrary dimension d, at no extra cost in its proof; so it will be stated below in this more general setting. The notation Ed then denotes a d-dimensional Euclidean space and Sd denotes the space of all symmetric matrices of order d. This continuity result is itself a simple consequence of the following nonlinear Korn inequality, which constitutes the main result of ibid.: Let be a bounded and connected open subset of Rd with a Lipschitz-continuous boundary and let C 1 (; Ed ) be a mapping satisfying det > 0 in . Then there exists a constant C () with the following property: For each orientationpreserving mapping H 1 (; Ed ), there exist a proper orthogonal matrix R = R(, ) of order d (i.e., an orthogonal matrix of order d with a determinant equal to one) and a vector b = b(, ) in Ed such that (b + R)
H 1 (;Ed )
C () T T
1/2 . L1 (;Sd )
That a vector b and an orthogonal matrix R should appear in the lefthand side of such an inequality is of course reminiscent of the classical rigidity theorem (Theorem 1.7-1), which asserts that, if two mappings C 1 (; Ed ) and C 1 (; Ed ) satisfying det > 0 and det > 0 in an open connected subset of Rd have the same Cauchy-Green tensor eld, then the two mappings are isometrically equivalent, i.e., there exist a vector b in Ed and an orthogonal matrix R of order d such that (x) = b + R(x) for all x .
58
[Ch. 1
More generally, we shall say that two orientation-preserving mappings H 1 (; Ed ) and H 1 (; Ed ) are isometrically equivalent if there exist a vector b in Ed and an orthogonal matrix R of order d (a proper one in this case) such that (x) = b + R(x) for almost all x . One application of the above key inequality is the following sequential continuity property : Let k H 1 (; Ed ), k 1, and C 1 (; Ed ) be orientationpreserving mappings. Then there exist a constant C () and orientation-preserving mappings H 1 (; Ed ), k 1, that are isometrically equivalent to k such that
k H 1 (;Ed ) k k
C () (k )T k T
1/2 . L1 (;Sd )
1 d Hence the sequence ( ) k=1 converges to in H (; E ) as k if the k T k T 1 sequence (( ) )k=1 converges to in L (; Sd ) as k . Should the Cauchy-Green strain tensor be viewed as the primary unknown (as suggested above), such a sequential continuity could thus prove to be useful when considering inmizing sequences of the total energy, in particular for handling the part of the energy that takes into account the applied forces and the boundary conditions, which are both naturally expressed in terms of the deformation itself. They key inequality is rst established in the special case where is the identity mapping of the set , by making use in particular of a fundamental geometric rigidity lemma recently proved by Friesecke, James & M uller [2002]. It is then extended to an arbitrary mapping C 1 (; Rn ) satisfying det > 0 in , thanks in particular to a methodology that bears some similarity with that used in Theorems 1.8-2 and 1.8-3. Such results are to be compared with the earlier, pioneering estimates of John [1961], John [1972] and Kohn [1982], which implied continuity at rigid body deformations, i.e., at a mapping that is isometrically equivalent to the identity mapping of . The recent and noteworthy continuity result of Reshetnyak [2003] for quasi-isometric mappings is in a sense complementary to the above one (it also deals with Sobolev type norms).
Chapter 2 DIFFERENTIAL GEOMETRY OF SURFACES
INTRODUCTION
We saw in Chapter 1 that an open set () in E3 , where is an open set in R3 and : E3 is a smooth injective immersion, is unambiguously dened (up to isometries of E3 ) by a single tensor eld, the metric tensor eld, whose covariant components gij = gji : R are given by gij = i j . Consider instead a surface = ( ) in E3 , where is a two-dimensional open set in R2 and : E3 is a smooth injective immersion. Then by contrast, such a two-dimensional manifold equipped with the coordinates of the points of as its curvilinear coordinates, requires two tensor elds for its denition (this time up to proper isometries of E3 ), the rst and second fundamental forms of . Their covariant components a = a : R and b = b : R are respectively given by (Greek indices or exponents take their values in {1, 2}): a = a a and b = a3 a , a1 a2 . |a1 a2 | The vector elds ai : R3 dened in this fashion constitute the covariant bases along the surface , while the vector elds ai : R3 dened by the i constitute the contravariant bases along . relations ai aj = j These two fundamental forms are introduced and studied in Sections 2.1 to 2.5. In particular, it is shown how areas and lengths, i.e., metric notions , on the surface are computed in terms of its curvilinear coordinates by means of the components a of the rst fundamental form (Theorem 2.3-1). It is also shown how the curvature of a curve on can be similarly computed, this time by means of the components of both fundamental forms (Theorem 2.4-1). Other classical notions about curvature, such as the principal curvatures and the Gaussian curvature, are introduced and briey discussed in Section 2.5. We next introduce in Section 2.6 the fundamental notion of covariant derivatives i| of a vector eld i ai : R3 on , thus dened here by means of its covariant components i over the contravariant bases ai . We establish in this where a = and a3 = 59
60
Dierential geometry of surfaces
[Ch. 2
process the formulas of Gau and Weingarten (Theorem 2.6-1). As illustrated in Chapter 4, covariant derivatives of vector elds on a surface (typically, the unknown displacement vector eld of the middle surface of a shell) pervade the equations of shell theory. It is a basic fact that the symmetric and positive denite matrix eld (a ) and the symmetric matrix eld (b ) dened on in this fashion cannot be arbitrary. More specically, their components and some of their partial derivatives must satisfy necessary conditions taking the form of the following relations (meant to hold for all , , , {1, 2}), which respectively constitute the Gau, and Codazzi-Mainardi, equations (Theorem 2.7-1): Let the functions and be dened by = 1 1 ( a + a a ) and . = a , where (a ) := (a ) 2
Then, necessarily,
+ = b b b b in , b b + b b = 0 in .
The functions and are the Christoel symbols of the rst, and second, kind. We also establish in passing (Theorem 2.7-2) the celebrated Theorema Egregium of Gau : At each point of a surface, the Gaussian curvature is a given function (the same for any surface) of the components of the rst fundamental form and their partial derivatives of order 2 at the same point. We then turn to the reciprocal questions: Given an open subset of R2 and a smooth enough symmetric and positive denite matrix eld (a ) together with a smooth enough symmetric matrix eld (b ) dened over , when are they the rst and second fundamental forms of a surface ( ) E3 , i.e., when does there exist an immersion : E3 such that 1 2 a = and b = in ? |1 2 | If such an immersion exists, to what extent is it unique? As shown in Theorems 2.8-1 and 2.9-1 (like those of their three-dimensional counterparts in Sections 1.6 and 1.7, their proofs are by no means easy, especially that of the existence), the answers turn out to be remarkably simple: Under the assumption that is simply-connected, the necessary conditions expressed by the Gau and Codazzi-Mainardi equations are also sucient for the existence of such an immersion . Besides, if is connected, this immersion is unique up to proper isometries in E. This means that, if : E3 is any other smooth immersion satisfying a = and b = 1 2 |1 2 | in
Sect. 2.1]
Curvilinear coordinates on a surface
61
there then exist a vector c E3 and a proper orthogonal matrix Q of order three such that (y ) = c + Q(y ) for all y . Together, the above existence and uniqueness theorems constitute the fundamental theorem of surface theory, another important special case of the fundamental theorem of Riemannian geometry already alluded to in Chapter 1. As such, they constitute the core of Chapter 2. We conclude this chapter by showing (Theorem 2.10-3) that the equivalence class of , dened in this fashion modulo proper isometries of E3 , depends continuously on the matrix elds (a ) and (b ) with respect to appropriate Fr echet topologies.
2.1
CURVILINEAR COORDINATES ON A SURFACE
In addition to the rules governing Latin indices that we set in Section 1.1, we henceforth require that Greek indices and exponents vary in the set {1, 2} and that the summation convention be systematically used in conjunction with these rules. For instance, the relation
3 (i ai ) = ( | b 3 )a + (3| + b )a
means that, for = 1, 2,

3 2 2
i=1
i ai =
( | b 3 )a + 3| +
=1 =1
3 b a .
Kroneckers symbols are designated by , , or according to the context. Let there be given as in Section 1.1 a three-dimensional Euclidean space E3 , equipped with an orthonormal basis consisting of three vectors ei = ei , and let a b, |a|, and a b denote the Euclidean inner product, the Euclidean norm, and the vector product of vectors a, b in the space E3 . In addition, let there be given a two-dimensional vector space, in which two vectors e = e form a basis. This space will be identied with R2 . Let y denote the coordinates of a point y R2 and let := /y and := 2 /y y . Finally, let there be given an open subset of R2 and a smooth enough mapping : E3 (specic smoothness assumptions on will be made later, according to each context). The set
:= ( ) is called a surface in E3 .
62
[Ch. 2
If the mapping : E3 is injective, each point y can be unambiguously written as y = (y ), y , and the two coordinates y of y are called the curvilinear coordinates of y (Figure 2.1-1). Well-known examples of surfaces and of curvilinear coordinates and their corresponding coordinate lines (dened in Section 2.2) are given in Figures 2.1-2 and 2.1-3.
a2 (y ) (y ) a2 (y ) a1 (y ) = ( )
a1 (y )
E3
y2
R2
y1
Figure 2.1-1: Curvilinear coordinates on a surface and covariant and contravariant bases of the tangent plane. Let b = ( ) be a surface in E3 . The two coordinates y1 , y2 of y are the curvilinear coordinates of y b = (y ) b . If the two vectors a (y ) = (y ) are linearly independent, they are tangent to the coordinate lines passing through y b and they form the covariant basis of the tangent plane to b at y b = (y ). The two vectors a (y ) from this tangent form its contravariant basis. plane dened by a (y ) a (y ) =
Naturally, once a surface is dened as = ( ), there are innitely many other ways of dening curvilinear coordinates on , depending on how the domain and the mapping are chosen. For instance, a portion of a sphere may be represented by means of Cartesian coordinates, spherical coordinates, or stereographic coordinates (Figure 2.1-3). Incidentally, this example illustrates the variety of restrictions that have to be imposed on according to which kind of curvilinear coordinates it is equipped with!
Sect. 2.2]
Curvilinear coordinates on a surface
63
y y x
u
Figure 2.1-2: Several systems of curvilinear coordinates on a sphere. Let be a sphere of radius R. A portion of contained in the northern hemisphere can be represented by means of Cartesian coordinates, with a mapping of the form: : (x, y ) (x, y, {R2 (x2 + y 2 )}1/2 ) E3 . A portion of that excludes a neighborhood of both poles and of a meridian (to x ideas) can be represented by means of spherical coordinates, with a mapping of the form: : (, ) (R cos cos , R cos sin , R sin ) E3 . A portion of that excludes a neighborhood of the North pole can be represented by means of stereographic coordinates, with a mapping of the form: 2R2 u 2R2 v u2 + v2 R2 : (u, v) E3 . , 2 ,R 2 2 2 2 2 2 u +v +R u +v +R u + v2 + R2 The corresponding coordinate lines are represented in each case, with self-explanatory graphical conventions.
64
[Ch. 2
Figure 2.1-3: Two familiar examples of surfaces and curvilinear coordinates. A portion b of a circular cylinder of radius R can be represented by a mapping of the form : (, z ) (R cos , R sin , z ) E3 . A portion b of a torus can be represented by a mapping of the form : (, ) ((R + r cos ) cos , (R + r cos ) sin , r sin ) E3 , with R > r . The corresponding coordinate lines are represented in each case, with self-explanatory graphical conventions.
Sect. 2.2]
First fundamental form
65
2.2
FIRST FUNDAMENTAL FORM

= i ei : R2 ( ) = E3
Let be an open subset of R2 and let be a mapping that is dierentiable at a point y . If y is such that (y + y ) , then (y + y ) = (y ) + (y )y + o(y ), where the 3 2 matrix (y ) and the column vector y are dened by 1 1 2 1 y1 . (y ) := 1 2 2 2 (y ) and y = y2 1 3 2 3 Let the two vectors a (y ) R3 be dened by 1 a (y ) := (y ) = 2 (y ), 3 i.e., a (y ) is the -th column vector of the matrix (y ). Then the expansion of about y may be also written as (y + y ) = (y ) + y a (y ) + o(y ). If in particular y is of the form y = te , where t R and e is one of the basis vectors in R2 , this relation reduces to (y + te ) = (y ) + ta (y ) + o(t). A mapping : E3 is an immersion at y if it is dierentiable at y and the 3 2 matrix (y ) is of rank two, or equivalently if the two vectors a (y ) = (y ) are linearly independent. Assume from now on in this section that the mapping is an immersion at y . In this case, the last relation shows that each vector a (y ) is tangent to the -th coordinate line passing through y = (y ), dened as the image by of the points of that lie on a line parallel to e passing through y (there exist t0 and t1 with t0 < 0 < t1 such that the -th coordinate line is given by t ]t0 , t1 [ f (t) := (y + te ) in a neighborhood of y ; hence f (0) = (y ) = a (y )); see Figures 2.1-1, 2.1-2, and 2.1-3. The vectors a (y ), which thus span the tangent plane to the surface at y = (y ), form the covariant basis of the tangent plane to at y ; see Figure 2.1-1. Returning to a general increment y = y e , we also infer from the expansion of about y that (y T and (y )T respectively designate the transpose of the column vector y and the transpose of the matrix (y )) | (y + y ) (y )|2 = y T (y )T (y )y + o(|y |2 ) = y a (y ) a (y )y + o(|y |2 ).
66
[Ch. 2
In other words, the principal part with respect to y of the length between the points (y + y ) and (y ) is {y a (y ) a (y )y }1/2 . This observation suggests to dene a matrix (a (y )) of order two by letting a (y ) := a (y ) a (y ) = (y )T (y )
The elements a (y ) of this symmetric matrix are called the covariant components of the rst fundamental form, also called the metric tensor, of the surface at y = (y ). Note that the matrix (a (y )) is positive denite since the vectors a (y ) are assumed to be linearly independent. The two vectors a (y ) being thus dened, the four relations
a (y ) a (y ) =
unambiguously dene two linearly independent vectors a (y ) in the tangent plane. To see this, let a priori a (y ) = Y (y )a (y ) in the relations a (y ) a (y ) = . This gives Y (y )a (y ) = ; hence Y (y ) = a (y ), where (a (y )) := (a (y ))1 . Hence a (y ) = a (y )a (y ). These relations in turn imply that a (y ) a (y ) = a (y )a (y )a (y ) a (y )
= a (y )a (y )a (y ) = a (y ) = a (y ),
and thus the vectors a (y ) are linearly independent since the matrix (a (y )) is positive denite. We would likewise establish that a (y ) = a (y )a (y ). The two vectors a (y ) form the contravariant basis of the tangent plane to the surface at y = (y ) (Figure 2.1-1) and the elements a (y ) of the symmetric matrix (a (y )) are called the contravariant components of the rst fundamental form, or metric tensor, of the surface at y = (y ). Let us record for convenience the fundamental relations that exist between the vectors of the covariant and contravariant bases of the tangent plane and the covariant and contravariant components of the rst fundamental tensor: a (y ) = a (y ) a (y ) a (y ) = a (y )a (y ) and a (y ) = a (y ) a (y ), and a (y ) = a (y )a (y ).
A mapping : E3 is an immersion if it is an immersion at each point in , i.e., if is dierentiable in and the two vectors (y ) are linearly independent at each y . If : E3 is an immersion, the vector elds a : R3 and a : R3 respectively form the covariant, and contravariant, bases of the tangent planes.
Sect. 2.3]
Areas and lengths on a surface
67
A word of caution. The presentation in this section closely follows that of Section 1.2, the mapping : R2 E3 replacing the mapping : R3 E3 . There are indeed strong similarities between the two presentations, such as the way the metric tensor is dened in both cases, but there are also sharp dierences. In particular, the matrix (y ) is not a square matrix, while the matrix (x) is square!
2.3
AREAS AND LENGTHS ON A SURFACE
We now review fundamental formulas expressing area and length elements at a point y = (y ) of the surface = ( ) in terms of the matrix (a (y )); see Figure 2.3-1. These formulas highlight in particular the crucial r ole played by the matrix (a (y )) for computing metric notions at y = (y ). Indeed, the rst fundamental form well deserves metric tensor as its alias !

(y +y ) A ( dl y) (y ) = y E3 C
da ( y)
R2
C f t
y y +y dy A
R
Figure 2.3-1: Area and length elements on a surface. The elements db a(y b) and d b(y b) at y b = (y ) b are related to dy and y by means of the covariant components of the metric tensor of the surface b ; cf. Theorem 2.3-1. The corresponding relations are used for computing b = (A) b = (C ) the area of a surface A b and the length of a curve C b , where C = f (I ) and I is a compact interval of R.
68
[Ch. 2
Theorem 2.3-1. Let be an open subset of R2 , let : E3 be an injective and smooth enough immersion, and let = ( ). (a) The area element da(y ) at y = (y ) is given in terms of the area element dy at y by da(y ) = a(y ) dy, where a(y ) := det(a (y )).
(b) The length element d (y ) at y = (y ) is given by d (y ) = y a (y )y

1/2
Proof. The relation (a) between the area elements is well known. It can also be deduced directly from the relation between the area elements d(x) and d(x) given in Theorem 1.3-1 (b) by means of an ad hoc three-dimensional extension of the mapping . The expression of the length element in (b) recalls that d (y ) is by denition the principal part with respect to y = y e of the length | (y + y ) (y )|, whose expression precisely led to the introduction of the matrix (a (y )). The relations found in Theorem 2.3-1 are used for computing surface integrals and lengths on the surface by means of integrals inside , i.e., in terms of the curvilinear coordinates used for dening the surface (see again Figure 2.3-1). Let A be a domain in R2 such that A (a domain in R2 is a bounded, open, and connected subset of R2 with a Lipschitz-continuous boundary; cf. Section 1.3), let A := (A), and let f L1 (A) be given. Then f (y ) da(y ) =
A
b A
(f )(y )
a(y ) dy.
In particular, the area of A is given by area A := da(y ) =

A
b A
a(y ) dy.
Consider next a curve C = f (I ) in , where I is a compact interval of R and f = f e : I is a smooth enough injective mapping. Then the length of the curve C := (C ) is given by d ( f )(t) dt = dt df df (t) (t) dt. dt dt
length C :=
I
a (f (t))
I
The last relation shows in particular that the lengths of curves inside the surface ( ) are precisely those induced by the Euclidean metric of the space E3 . For this reason, the surface ( ) is said to be isometrically imbedded in E3 .
Sect. 2.4]
Second fundamental form; curvature on a surface
69
2.4
SECOND FUNDAMENTAL FORM; CURVATURE ON A SURFACE
While the image () E3 of a three-dimensional open set R3 by a smooth enough immersion : R3 E3 is well dened by its metric, uniquely up to isometries in E3 (provided ad hoc compatibility conditions are satised by the covariant components gij : R of its metric tensor ; cf. Theorems 1.6-1 and 1.7-1), a surface given as the image ( ) E3 of a two-dimensional open set R2 by a smooth enough immersion : R2 E3 cannot be dened by its metric alone. As intuitively suggested by Figure 2.4-1, the missing information is provided by the curvature of a surface. A natural way to give substance to this otherwise vague notion consists in specifying how the curvature of a curve on a surface can be computed. As shown in this section, solving this question relies on the knowledge of the second fundamental form of a surface, which naturally appears for this purpose through its covariant components (Theorem 2.4-1).
2
Figure 2.4-1: A metric alone does not dene a surface in E3 . A at surface b0 may be deformed into a portion b1 of a cylinder or a portion b2 of a cone without altering the length of any curve drawn on it (cylinders and cones are instances of developable surfaces; cf. Section 2.5). Yet it should be clear that in general b0 and b1 , or b0 and b2 , or b1 and b2 , are not identical surfaces modulo an isometry of E3 !
70
[Ch. 2
Consider as in Section 2.1 a surface = ( ) in E3 , where is an open subset of R2 and : R2 E3 is a smooth enough immersion. For each y , the vector a1 (y ) a2 (y ) a3 (y ) := |a1 (y ) a2 (y )| is thus well dened, has Euclidean norm one, and is normal to the surface at the point y = (y ). Remark. The denominator in the denition of a3 (y ) may be also written as |a1 (y ) a2 (y )| = a(y ),
where a(y ) := det(a (y )). This relation, which holds in fact even if a(y ) = 0, will be established in the course of the proof of Theorem 4.2-2. Fix y and consider a plane P normal to at y = (y ), i.e., a plane that contains the vector a3 (y ). The intersection C = P is thus a planar curve on . As shown in Theorem 2.4-1, it is remarkable that the curvature of C at y can be computed by means of the covariant components a (y ) of the rst fundamental form of the surface = ( ) introduced in Section 2.2, together with the covariant components b (y ) of the second fundamental form of . The denition of the curvature of a planar curve is recalled in Figure 2.4-2. 1 If the algebraic curvature of C at y is = 0, it can be written as , and R is R then called the algebraic radius of curvature of the curve C at y. This means that the center of curvature of the curve C at y is the point (y + Ra3 (y )); see Figure 2.4-3. While R is not intrinsically dened, as its sign changes in any system of curvilinear coordinates where the normal vector a3 (y ) is replaced by its opposite, the center of curvature is intrinsically dened. If the curvature of C at y is 0, the radius of curvature of the curve C at y is said to be innite ; for this reason, it is customary to still write the curvature 1 in this case. as R 1 Note that the real number is always well dened by the formula given in R the next theorem, since the symmetric matrix (a (y )) is positive denite. This implies in particular that the radius of curvature never vanishes along a curve on a surface ( ) dened by a mapping satisfying the assumptions of the next theorem, hence in particular of class C 2 on . It is intuitively clear that if R = 0, the mapping cannot be too smooth. Think of a surface made of two portions of planes intersecting along a segment, which thus constitutes a fold on the surface. Or think of a surface ( ) with 0 and (y1 , y2 ) = |y1 |1+ for some 0 < < 1, so that C 1 ( ; E3 ) but / C 2 ( ; E3 ): The radius of curvature of a curve corresponding to a constant y2 vanishes at y1 = 0.
Sect. 2.4]
Second fundamental form; curvature on a surface

(s) p(s) + R (s)
71
(s + s) (s)
(s + s)
p(s + s) (s) p(s)
Figure 2.4-2: Curvature of a planar curve. Let be a smooth enough planar curve, parametrized by its curvilinear abscissa s. Consider two points p(s) and p(s + s) with curvilinear abscissae s and s + s and let (s) be the algebraic angle between the two normals (s) and (s + s) (oriented in the usual way) to at those points. When s 0, (s) the ratio has a limit, called the curvature of at p(s). If this limit is non-zero, its s inverse R is called the algebraic radius of curvature of at p(s) (the sign of R depends on the orientation chosen on ). The point p(s) + R (s), which is intrinsically dened, is called the center of curvature of at p(s): It is the center of the osculating circle at p(s), i.e., the limit as s 0 of the circle tangent to at p(s) that passes through the point p(s + s). The center of curvature is also the limit as s 0 of the intersection of the normals (s) and (s + s). Consequently, the centers of curvature of lie on a curve (dashed on the gure), called la d evelopp ee in French, that is tangent to the normals to .
Theorem 2.4-1. Let be an open subset of R2 , let C 2 ( ; E3 ) be an injective immersion, and let y be xed. Consider a plane P normal to = ( ) at the point y = (y ). The intersection P is a curve C on , which is the image C = (C ) of a curve C in the set . Assume that, in a suciently small neighborhood of y , the restriction of C to this neighborhood is the image f (I ) of an open interval I R, where f = f e : I R is a smooth enough injective mapping that satises df (t) e = 0, where t I is such that y = f (t) (Figure 2.4-3). dt 1 of the planar curve C at y is given by the ratio Then the curvature R df df (t) (t) 1 dt dt = , R df df a (f (t)) (t) (t) dt dt b (f (t)) where a (y ) are the covariant components of the rst fundamental form of at y (Section 2.1) and b (y ) := a3 (y ) a (y ) = a3 (y ) a (y ) = b (y ).
72
[Ch. 2
a3(y ) y = (y ) C y + Ra3(y ) E3
e1
R2
e2
y = f (t)e
I t
df (t)e dt
f = f e
Figure 2.4-3: Curvature on a surface. Let P be a plane containing the vector a1 (y ) a2 (y ) , which is normal to the surface b = ( ). The algebraic curvature a3 (y ) = |a1 (y ) a2 (y )| 1 b =P of the planar curve C b = (C ) at y b = (y ) is given by the ratio R df df b (f (t)) (t) (t) 1 dt dt = , R df df a (f (t)) (t) (t) dt dt where a (y ) and b (y ) are the covariant components of the rst and second fundadf mental forms of the surface b at y b and (t) are the components of the vector tangent to dt 1 b = 0, the center of curvature of the curve C the curve C = f (I ) at y = f (t) = f (t)e . If R 3 at y b is the point (y b + Ra3 (y )), which is intrinsically dened in the Euclidean space E .
1 of Proof. (i) We rst establish a well-known formula giving the curvature R a planar curve. Using the notations of Figure 2.4-2, we note that sin (s) = (s) (s + s) = { (s + s) (s)} (s + s), so that 1 d (s) sin (s) := lim = lim = (s) (s). s0 s s0 R s ds
(ii) The curve ( f )(I ), which is a priori parametrized by t I , can be also parametrized by its curvilinear abscissa s in a neighborhood of the point y . I and a mapping p : J P , where J R is There thus exist an interval I an interval, such that s J. ( f )(t) = p(s) and (a3 f )(t) = (s) for all t I,
Sect. 2.5]
Principal curvatures; Gaussian curvature 1 of C is given by R d 1 = (s) (s), R ds
73
By (i), the curvature
where d d(a3 f ) dt df dt (s) = (t) = a3 (f (t)) (t) , ds dt ds dt ds dp d( f ) dt (s) = (s) = (t) ds dt ds df dt df dt = (f (t)) (t) (t) . = a (f (t)) dt ds dt ds Hence 1 df df dt 2 = a3 (f (t)) a (f (t)) (t) (t) . R dt dt ds 1 To obtain the announced expression for , it suces to note that R a3 (f (t)) a (f (t)) = b (f (t)), by denition of the functions b and that (Theorem 2.3-1 (b)) ds = y a (y )y
1/2
= a (f (t))
df df (t) (t) dt dt
1/2
dt.
The knowledge of the curvatures of curves contained in planes normal to suces for computing the curvature of any curve on . More specically, the radius of curvature R at y of any smooth enough curve C (planar or not) on the 1 cos = , where is the angle between the principal surface is given by R R 1 normal to C at y and a3 (y ) and is given in Theorem 2.4-1; see, e.g., Stoker R [1969, Chapter 4, Section 12]. The elements b (y ) of the symmetric matrix (b (y )) dened in Theorem 2.4-1 are called the covariant components of the second fundamental form of the surface = ( ) at y = (y ).
2.5
PRINCIPAL CURVATURES; GAUSSIAN CURVATURE
The analysis of the previous section suggests that precise information about the shape of a surface = ( ) in a neighborhood of one of its points y = (y ) can be gathered by letting the plane P turn around the normal vector a3 (y )
74
[Ch. 2
and by following in this process the variations of the curvatures at y of the corresponding planar curves P , as given in Theorem 2.4-1. As a rst step in this direction, we show that these curvatures span a compact interval of R. In particular then, they stay away from innity. Note that this compact interval contains 0 if, and only if, the radius of curvature of the curve P is innite for at least one such plane P . Theorem 2.5-1. (a) Let the assumptions and notations be as in Theorem 2.4-1. For a xed y , consider the set P of all planes P normal to the surface = ( ) at y = (y ). Then the set of curvatures of the associated planar 1 1 , curves P , P P , is a compact interval of R, denoted . R1 (y ) R2 (y ) (b) Let the matrix (b (y )), being the row index, be dened by
b (y ) := a (y )b (y ),
where (a (y )) = (a (y ))1 (Section 2.2) and the matrix (b (y )) is dened as in Theorem 2.4-1. Then 1 1 2 + = b1 1 (y ) + b2 (y ), R1 (y ) R2 (y ) det(b (y )) 1 2 2 1 = b1 . 1 (y )b2 (y ) b1 (y )b2 (y ) = R1 (y )R2 (y ) det(a (y )) 1 1 = , there is a unique pair of orthogonal planes P1 P R1 (y ) R2 (y ) and P2 P such that the curvatures of the associated planar curves P1 and 1 1 P2 are precisely and . R1 (y ) R2 (y ) (c) If Proof. (i) Let (P ) denote the intersection of P P with the tangent plane T to the surface at y , and let C (P ) denote the intersection of P with . Hence (P ) is tangent to C (P ) at y . In a suciently small neighborhood of y the restriction of the curve C (P ) to this neighborhood is given by C (P ) = ( f (P ))(I (P )), where I (P ) R is an open interval and f (P ) = f (P )e : I (P ) R2 is a smooth enough df (P ) (t)e = 0, where t I (P ) is such that injective mapping that satises dt y = f (P )(t). Hence the line (P ) is given by (P ) = y + where := d( f (P )) (t); R = {y + a (y ); R} , dt
df (P ) (t) and e = 0 by assumption. dt Since the line {y + e ; R} is tangent to the curve C (P ) := 1 (C (P )) at y (the mapping : R3 is injective by assumption) for each
Sect. 2.5]
Principal curvatures; Gaussian curvature
75
such parametrizing function f (P ) : I (P ) R2 and since the vectors a (y ) are linearly independent, there exists a bijection between the set of all lines (P ) T , P P , and the set of all lines supporting the nonzero tangent vectors to the curve C (P ). Hence Theorem 2.4-1 shows that when P varies in P , the curvature of the corresponding curves C = C (P ) at y takes the same values as does the ratio b (y ) when := 1 varies in R2 {0}. 2 a (y ) (ii) Let the symmetric matrices A and B of order two be dened by A := (a (y )) and B := (b (y )). Since A is positive denite, it has a (unique) square root C, i.e., a symmetric positive denite matrix C such that A = C2 . Hence the ratio b (y ) T C1 BC1 T B = , where = C , = T a (y ) T A is nothing but the Rayleigh quotient associated with the symmetric matrix C1 BC1 . When varies in R2 {0}, this Rayleigh quotient thus spans the compact interval of R whose end-points are the smallest and largest eigenvalue, 1 1 and , of the matrix C1 BC1 (for a proof, respectively denoted R1 (y ) R2 (y ) see, e.g., Ciarlet [1982, Theorem 1.3-1]). This proves (a). Furthermore, the relation C(C1 BC1 )C1 = BC2 = BA1 shows that the eigenvalues of the symmetric matrix C1 BC1 coincide with those of the (in general non-symmetric) matrix BA1 . Note that BA1 = 1 = (b (y )) with b (y ) := a (y )b (y ), being the row index, since A (a (y )). Hence the relations in (b) simply express that the sum and the product of the eigenvalues of the matrix BA1 are respectively equal to its trace and to its det(b (y )) determinant, which may be also written as since BA1 = (b (y )). det(a (y )) This proves (b). (iii) Let 1 = 2 =
1 1 2 1
= C 1 and 2 =
1 2 2 2
= C 2 , with 1 =
1 1 2 1
and
1 2 , be two orthogonal ( T 1 2 = 0) eigenvectors of the symmetric matrix 2 2 1 1 and , respectively. C1 BC1 , corresponding to the eigenvalues R1 (y ) R2 (y ) Hence T T T 0 = T 1 2 = 1 C C 2 = 1 A 2 = 0,
76
[Ch. 2
since CT = C. By (i), the corresponding lines (P1 ) and (P2 ) of the tangent a (y ) and 2 a (y ), which are orthogonal plane are parallel to the vectors 1 since 1 a (y ) 2 a (y ) = a (y )1 2 = T 1 A 2 . 1 1 = , the directions of the vectors 1 and 2 are uniquely R1 (y ) R2 (y ) determined and the lines (P1 ) and (P2 ) are likewise uniquely determined. This proves (c). If We are now in a position to state several fundamental denitions : The elements b (y ) of the (in general non-symmetric) matrix (b (y )) dened in Theorem 2.5-1 are called the mixed components of the second fundamental form of the surface = ( ) at y = (y ). 1 1 and (one or both possibly equal to 0) found The real numbers R1 (y ) R2 (y ) in Theorem 2.5-1 are called the principal curvatures of at y . 1 1 = , the curvatures of the planar curves P are the same in If R1 (y ) R2 (y ) 1 1 = = 0, the point y = (y ) is all directions, i.e., for all P P . If R1 (y ) R2 (y ) 1 1 called a planar point. If = = 0, y is called an umbilical point. R1 (y ) R2 (y ) It is remarkable that, if all the points of are planar, then is a portion of a plane. Likewise, if all the points of are umbilical, then is a portion of a sphere. For proofs, see, e.g., Stoker [1969, p. 87 and p. 99]. Let y = (y ) be a point that is neither planar nor umbilical; in other words, the principal curvatures at y are not equal. Then the two orthogonal lines tangent to the planar curves P1 and P2 (Theorem 2.5-1 (c)) are called the principal directions at y . A line of curvature is a curve on that is tangent to a principal direction at each one of its points. It can be shown that a point that is neither planar nor umbilical possesses a neighborhood where two orthogonal families of lines of curvature can be chosen as coordinate lines. See, e.g., Klingenberg [1973, Lemma 3.6.6]. 1 1 = 0 and = 0, the real numbers R1 (y ) and R2 (y ) are called If R1 (y ) R2 (y ) 1 = 0, the correspondthe principal radii of curvature of at y. If, e.g., R1 (y ) ing radius of curvature R1 (y ) is said to be innite, according to the convention made in Section 2.4. While the principal radii of curvature may simultaneously change their signs in another system of curvilinear coordinates, the associated centers of curvature are intrinsically dened. 1 1 1 1 + and , which are the principal The numbers 2 R1 (y ) R2 (y ) R1 (y )R2 (y ) invariants of the matrix (b (y )) (Theorem 2.5-1), are respectively called the mean curvature and the Gaussian, or total, curvature of the surface at y .
Sect. 2.5]
Principal curvatures; Gaussian curvature
77
A point on a surface is an elliptic, parabolic, or hyperbolic, point according as its Gaussian curvature is > 0, = 0 but it is not a planar point, or < 0; see Figure 2.5-1. An asymptotic line is a curve on a surface that is everywhere tangent to a direction along which the radius of curvature is innite; any point along an asymptotic line is thus either parabolic or hyperbolic. It can be shown that, if all the points of a surface are hyperbolic, any point possesses a neighborhood where two intersecting families of asymptotic lines can be chosen as coordinate lines. See, e.g., Klingenberg [1973, Lemma 3.6.12]. As intuitively suggested by Figure 2.4-1, a surface in R3 cannot be dened by its metric alone, i.e., through its rst fundamental form alone, since its curvature must be in addition specied through its second fundamental form. But quite surprisingly, the Gaussian curvature at a point can also be expressed solely in terms of the functions a and their derivatives! This is the celebrated Theorema Egregium (astonishing theorem) of Gau [1828]; see Theorem 2.6.2 in the next section. Another striking result involving the Gaussian curvature is the equally celebrated Gau-Bonnet theorem, so named after Gau [1828] and Bonnet [1848] (for a modern proof, see, e.g., Klingenberg [1973, Theorem 6.3-5] or do Carmo [1994, Chapter 6, Theorem 1]): Let S be a smooth enough, closed, orientable, and compact surface in R3 (a closed surface is one without boundary, such as a sphere or a torus; orientable surfaces, which exclude for instance Klein bottles, are dened in, e.g., Klingenberg [1973, Section 5.5]) and let K : S R denote its Gaussian curvature. Then K (y) da(y ) = 2 (2 2g (S )),
S
where the genus g (S ) is the number of holes of S (for instance, a sphere has genus zero, while a torus has genus one). The integer (S ) dened by (S ) := (2 2g (S )) is the Euler characteristic of . According to the denition of Stoker [1969, Chapter 5, Section 2], a developable surface is one whose Gaussian curvature vanishes everywhere. Developable surfaces are otherwise often dened as ruled surfaces whose Gaussian curvature vanishes everywhere, as in, e.g., Klingenberg [1973, Section 3.7]). A portion of a plane provides a rst example, the only one of a developable surface all points of which are planar. Any developable surface all points of which are parabolic can be likewise fully described: It is either a portion of a cylinder, or a portion of a cone, or a portion of a surface spanned by the tangents to a skewed curve. The description of a developable surface comprising both planar and parabolic points is more subtle (although the above examples are in a sense the only ones possible, at least locally; see Stoker [1969, Chapter 5, Sections 2 to 6]). The interest of developable surfaces is that they can be, at least locally, continuously rolled out, or developed (hence their name), onto a plane, without changing the metric of the intermediary surfaces in the process.
78
[Ch. 2
Figure 2.5-1: Dierent kinds of points on a surface. A point is elliptic if the Gaussian curvature is > 0 or equivalently, if the two principal radii of curvature are of the same sign; the surface is then locally on one side of its tangent plane. A point is parabolic if exactly one of the two principal radii of curvature is innite; the surface is again locally on one side of its tangent plane. A point is hyperbolic if the Gaussian curvature is < 0 or equivalently, if the two principal radii of curvature are of dierent signs; the surface then intersects its tangent plane along two curves.
Sect. 2.6] Covariant derivatives of a vector eld dened on a surface
79
More details about these various notions are found in classic texts such as Stoker [1969], Klingenberg [1973], do Carmo [1976], Berger & Gostiaux [1987], Spivak [1999], or K uhnel [2002].
2.6
COVARIANT DERIVATIVES OF A VECTOR FIELD DEFINED ON A SURFACE; THE GAUSS AND WEINGARTEN FORMULAS
As in Sections 2.2 and 2.4, consider a surface = ( ) in E3 , where : R2 E3 is a smooth enough injective immersion, and let a3 (y ) = a3 (y ) := a1 (y ) a2 (y ) , |a1 (y ) a2 (y )| y .
Then the vectors a (y ) (which form the covariant basis of the tangent plane to at y = (y ); see Figure 2.1-1) together with the vector a3 (y ) (which is normal to and has Euclidean norm one) form the covariant basis at y. Let the vectors a (y ) of the tangent plane to at y be dened by the rela tions a (y ) a (y ) = . Then the vectors a (y ) (which form the contravariant basis of the tangent plane at y; see again Figure 2.1-1) together with the vector a3 (y ) form the contravariant basis at y ; see Figure 2.6-1. Note that the vectors of the covariant and contravariant bases at y satisfy
i . ai (y ) aj (y ) = j
Suppose that a vector eld is dened on the surface . One way to dene such a eld in terms of the curvilinear coordinates used for dening the surface consists in writing it as i ai : R3 , i.e., in specifying its covariant components i : R over the vector elds ai formed by the contravariant bases. This means that i (y )ai (y ) is the vector at each point y = (y ) (Figure 2.6-1). Our objective in this section is to compute the partial derivatives (i ai ) of such a vector eld. These are found in the next theorem, as immediate consequences of two basic formulas, those of Gau and Weingarten. The Christoel symbols on a surface and the covariant derivatives of a vector eld dened on a surface are also naturally introduced in this process. A word of caution. The Christoel symbols on a surface introduced in this section and the next one, viz., and , are thus denoted by the same symbols as the three-dimensional Christoel symbols introduced in Sections 1.4 and 1.5. No confusion should arise, however. Theorem 2.6-1. Let be an open subset of R2 and let C 2 ( ; E3 ) be an immersion. (a) The derivatives of the vectors of the covariant and contravariant bases are given by
3 a = a + b a3 and a = a + b a , 3 a3 = a = b a = b a ,
80
[Ch. 2
i (y )ai (y ) 3 (y ) a3 (y ) a2 (y )
2 (y ) a1 (y ) 1 (y ) = ( )
E3
y2
R2
y1
Figure 2.6-1: Contravariant bases and vector elds along a surface. At each point y b = (y ) b = ( ), the three vectors ai (y ), where a (y ) form the contravariant basis of the a1 (y ) a2 (y ) tangent plane to b at y b (Figure 2.1-1) and a3 (y ) = , form the contravariant |a1 (y ) a2 (y )| basis at y b. An arbitrary vector eld dened on b may then be dened by its covariant components i : R. This means that i (y )ai (y ) is the vector at the point y b.
where the covariant and mixed components b and b of the second fundamental form of are dened in Theorems 2.4-1 and 2.5-1 and
:= a a .
(b) Let there be given a vector eld i ai : R3 with covariant components i C 1 ( ). Then i ai C 1 ( ) and the partial derivatives (i ai ) C 0 ( ) are given by
3 (i ai ) = ( b 3 )a + ( 3 + b )a 3 = ( | b 3 )a + (3| + b )a ,
where | := and 3| := 3 .
Sect. 2.6] Covariant derivatives of a vector eld dened on a surface
81
Proof. Since any vector c in the tangent plane can be expanded as c = 3 (c a )a = (c a )a , since a3 is in the tangent plane ( a3 a3 = 1 2 (a a3 ) = 0), and since a3 a = b (Theorem 2.4-1), it follows that a3 = ( a3 a )a = b a . This formula, together with the denition of the functions b (Theorem 2.5-1), implies in turn that a3 = ( a3 a )a = b (a a )a = b a a = b a . Any vector c can be expanded as c = (c ai )ai = (c aj )aj . In particular, a = ( a a )a + ( a a3 )a3 = a + b a3 , by denition of and b . Finally,
3 a = ( a a )a + ( a a3 )a3 = a + b a ,
since
a a3 = a a3 = b a a = b .
That i ai C 1 ( ) if i C 1 ( ) is clear since ai C 1 ( ) if C 2 ( ; E3 ). The formulas established supra immediately lead to the announced expression of (i ai ). The relations (found in Theorem 2.6-1)
3 a = a + b a3 and a = a + b a
and a3 = a3 = b a = b a , respectively constitute the formulas of Gau and Weingarten. The functions (also found in Theorem 2.6-1) | = and 3| = 3 are the rst-order covariant derivatives of the vector eld i ai : R3 , and the functions
:= a a = a a
are the Christoel symbols of the second kind (the Christoel symbols of the rst kind are introduced in the next section).
82
[Ch. 2
Remark. The Christoel symbols can be also dened solely in terms of the covariant components of the rst fundamental form; see the proof of Theorem 2.7-1 The denition of the covariant derivatives | = of a vector eld dened on a surface ( ) given in Theorem 2.6-1 is highly reminiscent of the denition of the covariant derivatives vi j = j vi p ij vp of a vector eld dened on an open set () given in Section 1.4. However, the former are more subtle to apprehend than the latter. To see this, recall that the covariant derivatives vi j = j vi p ij vp may be also dened by the relations (Theorem 1.4-2) vi j g j = j (vi g i ). By contrast, even if only tangential vector elds a on the surface ( ) are considered (i.e., vector elds i ai : R3 for which 3 = 0), their covariant derivatives | = satisfy only the relations | a = P { ( a )} , where P denotes the projection operator on the tangent plane in the direction of the normal vector (i.e., P(ci ai ) := c a ), since
3 ( a ) = | a + b a
for such tangential elds by Theorem 2.6-1. The reason is that a surface has in general a nonzero curvature, manifesting itself here by the extra term 3 b a . This term vanishes in if is a portion of a plane, since in this case b = b = 0. Note that, again in this case, the formula giving the partial derivatives in Theorem 2.9-1 (b) reduces to (i ai ) = (i| )ai .
2.7
NECESSARY CONDITIONS SATISFIED BY THE FIRST AND SECOND FUNDAMENTAL FORMS: THE GAUSS AND CODAZZI-MAINARDI EQUATIONS; GAUSS THEOREMA EGREGIUM
It is remarkable that the components a = a : R and b = b : R of the rst and second fundamental forms of a surface ( ), dened by a smooth enough immersion : E3 , cannot be arbitrary functions. As shown in the next theorem, they must satisfy relations that take the form:
+ = b b b b in , b b + b b = 0 in ,
where the functions and have simple expressions in terms of the functions a and of some of their partial derivatives (as shown in the next proof,
Sect. 2.7] Necessary conditions satised by the fundamental forms
83
it so happens that the functions as dened in Theorem 2.7-1 coincide with the Christoel symbols introduced in the previous section; this explains why they are denoted by the same symbol). These relations, which are meant to hold for all , , , {1, 2}, respectively constitute the Gau, and Codazzi-Mainardi, equations. Theorem 2.7-1. Let be an open subset of R2 , let C 3 ( ; E3 ) be an immersion, and let a := and b := 1 2 |1 2 |
denote the covariant components of the rst and second fundamental forms of 1 the surface ( ). Let the functions C 1 ( ) and C ( ) be dened by := Then, necessarily,
+ = b b b b in , b b + b b = 0 in .
1 ( a + a a ), 2 := a where (a ) := (a )1 .
Proof. Let a = . It is then immediately veried that the functions are also given by = a a . a1 a2 Let a3 = and, for each y , let the three vectors aj (y ) be dened |a1 a2 | j by the relations aj (y ) ai (y ) = i . Since we also have a = a a and a3 = a3 , the last relations imply that = a a , hence that a = a + b a3 , since a = ( a a )a + ( a a3 )a3 . Dierentiating the same relations yields = a a + a a , so that the above relations together give
a a = a a + b a3 a = + b b .
Consequently, a a = b b . Since a = a , we also have a a = b b .
84
[Ch. 2
Hence the Gau equations immediately follow. Since a3 = ( a3 a )a + ( a3 a3 )a3 and a3 a = b = a a3 , we have a3 = b a . Dierentiating the relations b = a a3 , we obtain b = a a3 + a a3 .
This relation and the relations a = a + b a3 and a3 = b a together imply that a a3 = b .
Consequently, a a3 = b + b . Since a = a , we also have a a3 = b + b . Hence the Codazzi-Mainardi equations immediately follow. Remark. The vectors a and a introduced above respectively form the covariant and contravariant bases of the tangent plane to the surface ( ), the unit vector a3 = a3 is normal to the surface, and the functions a are the contravariant components of the rst fundamental form (Sections 2.2 and 2.3). As shown in the above proof, the Gau and Codazzi-Mainardi equations thus simply constitute a re-writing of the relations a = a in the form of the equivalent relations a a = a a and a a3 = a a3 . The functions = and
= a = a a =
1 ( a + a a ) = a a = 2
are the Christoel symbols of the rst, and second, kind. We recall that the Christoel symbols of the second kind also naturally appeared in a dierent context (that of covariant dierentiation; cf. Section 2.6). Finally, the functions
S := +
are the covariant components of the Riemann curvature tensor of the surface ( ). The denitions of the functions and imply that the sixteen Gau equations are satised if and only if they are satised for = 1, = 2,
Sect. 2.8] Existence of a surface with prescribed fundamental forms
85
= 1, = 2 and that the Codazzi-Mainardi equations are satised if and only if they are satised for = 1, = 2, = 1 and = 1, = 2, = 2 (other choices of indices with the same properties are clearly possible). In other words, the Gau equations and the Codazzi-Mainardi equations in fact respectively reduce to one and two equations. Letting = 2, = 1, = 2, = 1 in the Gau equations gives in particular S1212 = det(b ). Consequently, the Gaussian curvature at each point (y ) of the surface ( ) can be written as 1 S1212 (y ) = , y , R1 (y )R2 (y ) det(a (y )) 1 det(b (y ) = (Theorem 2.5-1). By inspection of the function R1 (y )R2 (y ) det(a (y )) S1212 , we thus reach the astonishing conclusion that, at each point of the surface, a notion involving the curvature of the surface, viz., the Gaussian curvature, is entirely determined by the knowledge of the metric of the surface at the same point, viz., the components of the rst fundamental forms and their partial derivatives of order 2 at the same point! This startling conclusion naturally deserves a theorem: since Theorem 2.7-2. Let be an open subset of R2 , let C 3 ( ; E3 ) be an immersion, let a = denote the covariant components of the rst fundamental form of the surface ( ), and let the functions and S1212 be dened by := S1212 1 ( a + a a ), 2 1 := (212 a12 11 a22 22 a11 ) + a (12 12 11 22 ). 2
1 R1 (y )
Then, at each point (y ) of the surface ( ), the principal curvatures and R21 (y ) satisfy 1 S1212 (y ) = , y . R1 (y )R2 (y ) det(a (y ))
Theorem 2.7-2 constitutes the famed Theorema Egregium of Gau [1828], so named by Gau who had been himself astounded by his discovery.
2.8
EXISTENCE OF A SURFACE WITH PRESCRIBED FIRST AND SECOND FUNDAMENTAL FORMS
Let M2 , S2 , and S2 > denote the sets of all square matrices of order two, of all symmetric matrices of order two, and of all symmetric, positive denite matrices of order two.
86
[Ch. 2
So far, we have considered that we are given an open set R2 and a smooth enough immersion : E3 , thus allowing us to dene the elds 2 (a ) : S2 > and (b ) : S , where a : R and b : R are the covariant components of the rst and second fundamental forms of the surface ( ) E3 . Note that the immersion need not be injective in order that these matrix elds be well dened. We now turn to the reciprocal questions: Given an open subset of R2 and two smooth enough matrix elds (a ) : 2 S2 > and (b ) : S , when are they the rst and second fundamental forms of a surface ( ) E3 , i.e., when does there exist an immersion : E3 such that a = and b = 1 2 |1 2 | in ?
If such an immersion exists, to what extent is it unique? The answers to these questions turn out to be remarkably simple: If is simply-connected, the necessary conditions of Theorem 2.7-1, i.e., the Gau and Codazzi-Mainardi equations, are also sucient for the existence of such an immersion. If is connected, this immersion is unique up to isometries in E3 . Whether an immersion found in this fashion is injective is a dierent issue, which accordingly should be resolved by dierent means. Following Ciarlet & Larsonneur [2001], we now give a self-contained, complete, and essentially elementary, proof of this well-known result. This proof amounts to showing that it can be established as a simple corollary to the fundamental theorem on at Riemannian manifolds established in Theorems 1.6-1 and 1.7-1 when the manifold is an open set in R3 . This proof has also the merit to shed light on the analogies (which cannot remain unnoticed!) between the assumptions and conclusions of both existence results (compare Theorems 1.6-1 and 2.8-1) and both uniqueness results (compare Theorems 1.7-1 and 2.9-1). A direct proof of the fundamental theorem of surface theory is given in Klingenberg [1973, Theorem 3.8.8], where the global existence of the mapping is based on an existence theorem for ordinary dierential equations, analogous to that used in part (ii) of the proof of Theorem 1.6-1. A proof of the local version of this theorem, which constitutes Bonnets theorem, is found in, e.g., do Carmo [1976]. This result is another special case of the fundamental theorem of Riemannian geometry alluded to in Section 1.6. We recall that this theorem asserts that a simply-connected Riemannian manifold of dimension p can be isometrically immersed into a Euclidean space of dimension (p + q ) if and only if there exist tensors satisfying together generalized Gau, and Codazzi-Mainardi, equations and that the corresponding isometric immersions are unique up to isometries in the Euclidean space. A substantial literature has been devoted to this theorem and its various proofs, which usually rely on basic notions of Riemannian geometry, such as connections or normal bundles, and on the theory of dierential
87
forms. See in particular the earlier papers of Janet [1926] and Cartan [1927] and the more recent references of Szczarba [1970], Tenenblat [1971], Jacobowitz [1982], and Szopos [2005]. Like the fundamental theorem of three-dimensional dierential geometry, this theorem comprises two essentially distinct parts, a global existence result (Theorem 2.8-1) and a uniqueness result (Theorem 2.9-1), the latter being also called rigidity theorem. Note that these two results are established under different assumptions on the set and on the smoothness of the elds (a ) and (b ). These existence and uniqueness results together constitute the fundamental theorem of surface theory. Theorem 2.8-1. Let be a connected and simply-connected open subset of R2 2 2 and let (a ) C 2 ( ; S2 > ) and (b ) C ( ; S ) be two matrix elds that satisfy the Gau and Codazzi-Mainardi equations, viz.,
+ = b b b b in , b b + b b = 0 in ,
where := 1 ( a + a a ), 2 := a where (a ) := (a )1 .
Then there exists an immersion C 3 ( ; E3 ) such that a = and b = 1 2 |1 2 | in .
Proof. The proof of this theorem as a corollary to Theorem 1.6-1 relies on the following elementary observation: Given a smooth enough immersion : E3 and > 0, let the mapping : ], [ E3 be dened by (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) ], [ , where a3 := 1 2 , and let |1 2 | gij := i j . Then an immediate computation shows that g = a 2x3 b + x2 3 c and gi3 = i3 in ], [ , where a and b are the covariant components of the rst and second fundamental forms of the surface ( ) and c := a b b . Assume that the matrices (gij ) constructed in this fashion are invertible, hence positive denite, over the set ], [ (they need not be, of course; but
88
[Ch. 2
the resulting diculty is easily circumvented; see parts (i) and (viii) below). Then the eld (gij ) : ], [ S3 > becomes a natural candidate for applying the three-dimensional existence result of Theorem 1.6-1, provided of course that the three-dimensional sucient conditions of this theorem, viz.,
p j ikq k ijq + p ij kqp ik jqp = 0 in ,
can be shown to hold, as consequences of the two-dimensional Gau and Codazzi-Mainardi equations. That this is indeed the case is the essence of the present proof (see parts (i) to (vii)). By Theorem 1.6-1, there then exists an immersion : ], [ E3 that satises gij = i j in ], [. It thus remains to check that := (, 0) indeed satises (see part (ix)) a = and b = 1 2 |1 2 | in .
The actual implementation of this program essentially involves elementary, albeit sometimes lengthy, computations, which accordingly will be omitted for the most part; only the main intermediate results will be recorded. For clarity, the proof is broken into nine parts, numbered (i) to (ix). To avoid confusion with the three-dimensional Christoel symbols, those and C in this proof (and only in this on a surface will be denoted C proof).
2 2 (i) Given two matrix elds (a ) C 2 ( ; S2 > ) and (b ) C ( ; S ), let the 2 3 matrix eld (gij ) C ( R; S ) be dened by
g := a 2x3 b + x2 3 c and gi3 := i3 in R (the variable y is omitted; x3 designates the variable in R), where
c := b b and b := a b in .
Let 0 be an open subset of R2 such that 0 is a compact subset of . Then there exists 0 = 0 (0 ) > 0 such that the symmetric matrices (gij ) are positive denite at all points in 0 , where 0 := 0 ]0 , 0 [ . Besides, the elements of the inverse matrix (g pq ) are given in 0 by g = where
n 1 (B) := b and (B ) := b bn1 for n 2, i3 (n + 1)xn (Bn ) = i3 , 3a and g n0
i.e., (Bn ) designates for any n 0 the element at the -th row and -th column of the matrix Bn . The above series are absolutely convergent in the space C 2 (0 ).
89
Let a priori g = n0 xn 3 hn where hn are functions of y 0 only, so that the relations g g = read h 0 a + x3 (h1 a 2h0 b )
+
n2
xn 3 (hn a 2hn1 b + hn2 c ) = .
It is then easily veried that the functions h n are given by

n h n = (n + 1)a (B ) , n 0,
so that g =
1 (n + 1)xn b b 3a n1 . n0
It is clear that such a series is absolutely convergent in the space C 2 ( 0 [0 , 0 ]) if 0 > 0 is small enough.
being dened by (ii) The functions C := a C , C
where (a ) := (a )1 and C := dene the functions

b | := b + C b C b , b C b = b| . b | := b C
1 ( a + a a ), 2
Then
b | = a b| and b| = a b | .
Furthermore, the assumed Codazzi-Mainardi equations imply that

b | = b | and b| = b | .
The above relations follow from straightforward computations based on the denitions of the functions b | and b | . They are recorded here because they play a pervading r ole in the subsequent computations. (iii) The functions gij C 2 (0 ) and g ij C 2 (0 ) being dened as in part (i), 1 dene the functions ijq C 1 (0 ) and p ij C (0 ) by ijq := 1 pq (j giq + i gjq q gij ) and p ij := g ijq . 2
90
[Ch. 2
p Then the functions ijq = jiq and p ij = ji have the following expressions : 2 = C x3 (b | a + 2C b ) + x3 (b | b + C c ),
3 = 3 = b x3 c , 33 = 3 3 = 33q = 0,
= C n0 +1 xn b | (Bn ) , 3
3 = b x3 c , 3 =
n0 n+1 xn ) , 3 (B p 3 3 = 33 = 0, where the functions c , (Bn ) , and b | are dened as in parts (i) and (ii). All computations are straightforward. We simply point out that the assumed Codazzi-Mainardi equations are needed to conclude that the factor of x3 in the function is indeed that announced above. We also note that the computation of the factor of x2 3 in relies in particular on the easily established relations c = b | b + b | b + C c + C c . 1 (iv) The functions ijq C 1 (0 ) and p ij C (0 ) being dened as in part (iii), dene the functions Rqijk C 0 (0 ) by p Rqijk := j ikq k ijq + p ij kqp ik jqp .
Then, in order that the relations Rqijk = 0 in 0 hold, it is sucient that R1212 = 0, R2 3 = 0, R3 3 = 0 in 0 .
The above denition of the functions Rqijk and that of the functions ijq and p ij (part (iii)) together imply that, for all i, j, k, q , Rqijk = Rjkqi = Rqikj , Rqijk = 0 if j = k or q = i. Consequently, the relation R1212 = 0 implies that R = 0, the relations R2 3 = 0 imply that Rqijk = 0 if exactly one index is equal to 3, and nally, the relations R3 3 = 0 imply that Rqijk = 0 if exactly two indices are equal to 3. (v) The functions
p R3 3 := 33 3 3 + p 3 3p 33 p
Sect. 2.8] Existence of a surface with prescribed fundamental forms satisfy R3 3 = 0 in 0 .
91
These relations immediately follow from the expressions found in part (iii) for the functions ijq and p ij . Note that neither the Gau equations nor the Codazzi-Mainardi equations are needed here. (vi) The functions
p R2 3 := 23 3 2 + p 2 3p 23 p
satisfy R2 3 = 0 in 0 . The denitions of the functions g (part (i)) and ijq (part (iii)) show that 23 3 2 = (2 b b2 ) + x3 ( c2 2 c ). Then the expressions found in part (iii) show that
p p 2 3p 23 p = 3 2 23 = C b2 C2 b + x3 (b 2 | b b | b2 + C2 c C c2 ),
and the relations R3 3 = 0 follow by making use of the relations

c = b | b + b | b + C c + C c
together with the relations

2 b b2 + C b2 C2 b = 0,
which are special cases of the assumed Codazzi-Mainardi equations. (vii) The function
p R1212 := 1 221 2 211 + p 21 21p 22 11p
satises R1212 = 0 in 0 . The computations leading to this relation are fairly lengthy and they require some care. We simply record the main intermediary steps, which consist in evaluating separately the various terms occurring in the function R1212 rewritten as
R1212 = (1 221 2 211 ) + ( 12 12 11 22 ) + (123 123 113 223 ).
First, the expressions found in part (iii) for the functions 3 easily yield 123 123 113 223 = (b2 12 b11 b22 ) 2 + x3 (b11 c22 2b12 c12 + b22 c11 ) + x2 3 (c12 c11 c22 ).
92
[Ch. 2
Second, the expressions found in part (iii) for the functions and yield, after some manipulations:
12 12 11 22 = (C12 C12 C11 C22 )a + x3 {(C11 b2 |2 2C12 b1 |2 + C22 b1 |1 )a + 2(C11 C22 C12 C12 )b } + x2 3 {b1 |1 b2 |2 b1 |2 b1 |2 )a + (C11 b2 |2 2C12 b1 |2 + C22 b1 |1 )b + (C11 C22 C12 C12 )c }.
Third, after somewhat delicate computations, which in particular make use of the relations established in part (ii) about the functions b | and b | , it is found that 1 221 2 211 = 1 C221 2 C211
x3 {S1212 b + (C11 b2 |2 2C12 b1 |2 + C22 b1 |1 )a + 2(C11 C22 C12 C12 )b } + x2 3 {S 12 b1 b2 + (b1 |1 b2 |2 b1 |2 b1 |2 )a + (C11 b2 |2 2C12 b1 |2 + C22 b1 |1 )b + (C11 C22 C12 C12 )c },
where the functions

C C C S := C C + C
are precisely those appearing in the left-hand sides of the Gau equations. It is then easily seen that the above equations together yield R1212 = {S1212 (b11 b22 b12 b12 )} x3 {S1212 (b11 b22 b12 b12 )b }
+ x2 3 {S 12 b1 b2 + (c12 c12 c11 c22 )}.
Since
1 2 2 1 S 12 b 1 b2 = S1212 (b1 b2 b1 b2 ), 2 2 1 c12 c12 c11 c22 = (b11 b12 b11 b22 )(b1 1 b2 b1 b2 ),
it is nally found that the function R1212 has the following remarkable expression:
2 2 1 2 2 1 R1212 = {S1212 (b11 b22 b12 b12 )}{1 x3 (b1 1 + b2 ) + x3 (b1 b2 b1 b2 )}.
By the assumed Gau equations, S1212 = b11 b22 b12 b12 . Hence R1212 = 0 as announced.
93
(viii) Let be a connected and simply-connected open subset of R2 . Then there exist open subsets , 0, of R2 such that is a compact subset of for each 0 and = .
0
Furthermore, for each 0, there exists = ( ) > 0 such that the symmetric matrices (gij ) are positive denite at all points in , where := ] , [ . Finally, the open set :=
0
is connected and simply-connected. Let , 0, be open subsets of with compact closures such that = 0 . For each , a set := ] , [ can then be constructed in the same way that the set 0 was constructed in part (i). It is clear that the set := 0 is connected. To show that is simplyconnected, let be a loop in , i.e., a mapping C 0 ([0, 1]; R3 ) that satises (0) = (1) and (t) for all 0 t 1. Let the projection operator : be dened by (y, x3 ) = y for all (y, x3 ) , and let the mapping 0 : [0, 1] [0, 1] R3 be dened by 0 (t, ) := (1 ) (t) + ( (t)) for all 0 t 1, 0 1. Then 0 is a continuous mapping such that 0 ([0, 1][0, 1]) , by denition of the set . Furthermore, 0 (t, 0) = (t) and 0 (t, 1) = ( (t)) for all t [0, 1]. The mapping := C 0 ([0, 1]; R2 ) is a loop in since (0) = ( (0)) = ( (1)) = (1) and (t) for all 0 t 1. Since is simply connected, there exist a mapping 1 C 0 ([0, 1] [0, 1]; R2 ) and a point y 0 such that 1 (t, 1) = and 1 (t, 2) = y 0 for all 0 t 1 and 1 (t, ) for all 0 t 1, 1 2. Then the mapping C 0 ([0, 1] [0, 2]; R3 ) dened by (t, ) = 0 (t, ) (t, ) = 1 (t, ) for all 0 t 1, 0 1, for all 0 t 1, 1 2,
is a homotopy in that reduces the loop to the point (y 0 , 0) . Hence the set is simply-connected.
94
[Ch. 2
1 (ix) By parts (iv) to (viii), the functions ijq C 1 () and p ij C () constructed as in part (iii) satisfy p j ikq k ijq + p ij kqp ik jqp = 0
in the connected and simply-connected open set . By Theorem 1.6-1, there thus exists an immersion C 3 (; E3 ) such that gij = i j in , where the matrix eld (gij ) C 2 (; S3 > ) is dened by g = a 2x3 b + x2 3 c and gi3 = i3 in . Then the mapping C 3 ( ; E3 ) dened by (y ) = (y, 0) for all y , satises a = and b = 1 2 |1 2 | in .
Let g i := i . Then 33 = 3 g 3 = p 33 g p = 0; cf. part (iii). Hence there 1 3 3 exists a mapping C ( ; E ) such that (y, x3 ) = (y ) + x3 1 (y ) for all (y, x3 ) , and consequently, g = + x3 1 and g 3 = 1 . The relations gi3 = g i g 3 = i3 (cf. part (i)) then show that ( + x3 1 ) 1 = 0 and 1 1 = 1. These relations imply that 1 = 0. Hence either 1 = a3 or 1 = a3 in , where 1 2 . a3 := |1 2 | But 1 = a3 is ruled out since {1 2 } 1 = det(gij )|x3 =0 > 0. Noting that a3 = 0 implies a3 = a3 , we obtain, on the one hand, g = ( + x3 a3 ) ( + x3 a3 ) = 2x3 a3 + x2 3 a3 a3 in .
Sect. 2.9] Uniqueness of surfaces with the same fundamental forms Since, on the other hand, g = a 2x3 b + x2 3 c in by part (i), we conclude that a = and b = a3 in , as desired. This completes the proof.
95
Remarks. (1) The functions c = b b = a3 a3 introduced in part (i) are the covariant components of the third fundamental form of the surface ( ). (2) The series expansion g = n0 (n + 1)xn (Bn ) 3a found in part (i) is known; cf., e.g., Naghdi [1972]. (3) The functions b | and b | introduced in part (ii) will be identied later as covariant derivatives of the second fundamental form of the surface ( ); see Section 4.2. (4) The Gau equations are used only once in the above proof, for showing that R1212 = 0 in part (vii). The regularity assumptions made in Theorem 2.8-1 on the matrix elds (a ) and (b ) can be signicantly relaxed in several ways. First, Cristinel Mardare has shown by means of an ad hoc, but not trivial, modication of the proof given here, that the existence of an immersion C 3 ( ; E3 ) still holds under the weaker (but certainly more natural, in view of the regularity of the resulting immersion ) assumption that (b ) C 1 ( ; S2 ), all other assumptions of Theorem 2.8-1 holding verbatim. In fact, Hartman & Wintner [1950] had already shown the stronger result 0 2 that the existence theorem still holds if (a ) C 1 ( ; S2 > ) and (b ) C ( ; S ), 2 3 with a resulting mapping in the space C ( ; E ). Their result has been itself superseded by that of S. Mardare [2003b], who established that if (a ) 1, 2 Wloc ( ; S2 > ) and (b ) Lloc ( ; S ) are two matrix elds that satisfy the Gau and Codazzi-Mainardi equations in the sense of distributions, then there exists 2, ( ; E3 ) such that (a ) and (b ) are the fundamental a mapping Wloc forms of the surface ( ). As of June 2005, the last word in this direction seems to belong to S. Mardare [2005], who was able to further reduce these 1,p p 2 ( ; S2 regularities, to those of the spaces Wloc > ) and Lloc ( ; S ) for any p > 2, 2,p with a resulting mapping in the space Wloc ( ; E3 ).
2.9
UNIQUENESS UP TO PROPER ISOMETRIES OF SURFACES WITH THE SAME FUNDAMENTAL FORMS
In Section 2.8, we have established the existence of an immersion : R2 E3 giving rise to a surface ( ) with prescribed rst and second fundamental
96
[Ch. 2
forms, provided these forms satisfy ad hoc sucient conditions. We now turn to the question of uniqueness of such immersions. This is the object of the next theorem, which constitutes another rigidity theorem, called the rigidity theorem for surfaces. It asserts that, if two immersions C 2 ( ; E3 ) and C 2 ( ; E3 ) share the same fundamental forms, then the surface ( ) is obtained by subjecting the surface ( ) to a rotation (represented by an orthogonal matrix Q with det Q = 1), then by subjecting the rotated surface to a translation (represented by a vector c). Such a geometric transformation of the surface ( ) is sometimes called a rigid transformation , to remind that it corresponds to the idea of a rigid one in E3 . This observation motivates the terminology rigidity theorem. As shown by Ciarlet & Larsonneur [2001] (whose proof is adapted here), the issue of uniqueness can be resolved as a corollary to its three-dimensional counterpart, like the issue of existence. We recall that O3 denotes the set of all 3 orthogonal matrices of order three and that O3 + = {Q O ; det Q = 1} denotes the set of all proper orthogonal matrices of order three. Theorem 2.9-1. Let be a connected open subset of R2 and let C 2 ( ; E3 ) and C 2 ( ; E3 ) be two immersions such that their associated rst and second fundamental forms satisfy (with self-explanatory notations) a = a and b = b in . Then there exist a vector c E3 and a matrix Q O3 + such that (y ) = c + Q(y ) for all y . Proof. Arguments similar to those used in parts (i) and (viii) of the proof of Theorem 2.8-1 show that there exist open subsets of and real numbers > 0, 0, such that the symmetric matrices (gij ) dened by g := a 2x3 b + x2 3 c and gi3 = i3 , where c := a b b , are positive denite in the set :=
0
] , [ .
The two immersions C 1 (; E3 ) and C 1 (; E3 ) dened by (with self-explanatory notations) (y, x3 ) := (y ) + x3 a3 (y ) and (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) therefore satisfy gij = gij in .
Sect. 2.9] Uniqueness of surfaces with the same fundamental forms
97
By Theorem 1.7-1, there exist a vector c E3 and an orthogonal matrix Q such that (y, x3 ) = c + Q(y, x3 ) for all (y, x3 ) . Hence, on the one hand, det (y, x3 ) = det Q det (y, x3 ) for all (y, x3 ) . On the other hand, a simple computation shows that det (y, x3 ) =
2 2 1 2 2 1 det(a (y ){1 x3 (b1 1 + b2 )(y ) + x3 (b1 b2 b1 b2 )(y )}
for all (y, x3 ) , where

b (y ) := a (y )b (y ), y ,
so that det (y, x3 ) = det (y, x3 ) for all (y, x3 ) . Therefore det Q = 1, which shows that the orthogonal matrix Q is in fact proper. The conclusion then follows by letting x3 = 0 in the relation (y, x3 ) = c + Q(y, x3 ) for all (y, x3 ) .
A proper isometry of E3 is a mapping J+ : E3 E3 of the form J+ (x) = c + Qox for all x E3 , with c E3 and Q O3 + . Theorem 2.9-1 thus asserts that two immersions C 1 ( ; E3 ) and C 1 ( ; E3 ) share the same fundamental forms over an open connected subset of R3 if and only if = J+ , where J+ is a proper isometry of E3 . Remark. By contrast, the three-dimensional rigidity theorem (Theorem 1.7-1) involves isometries of E3 that may not be proper. Theorem 2.9-1 constitutes the classical rigidity theorem for surfaces, in the sense that both immersions and are assumed to be in the space C 2 ( ; E3 ). As a preparation to our next result, we note that the second fundamental form of the surface ( ) can still be dened under the weaker assumptions that a1 a2 C 1 ( ; E3 ), by means of the denition C 1 ( ; E3 ) and a3 = |a1 a2 | b := a a3 , which evidently coincides with the usual one when C 2 ( ; E3 ). Following Ciarlet & C. Mardare [2004a], we now show that a similar rea1 a2 sult holds under the assumptions that H 1 ( ; E3 ) and a3 := |a1 a2 | H 1 ( ; E3 ) (with self-explanatory notations). Naturally, our rst task will be to
98
[Ch. 2
verify that the vector eld a3 , which is not necessarily well dened a.e. in for an arbitrary mapping H 1 ( ; E3 ), is nevertheless well dened a.e. in for those mappings that satisfy the assumptions of the next theorem. This fact will in turn imply that the functions b := a a3 are likewise well dened a.e. in . Theorem 2.9-2. Let be a connected open subset of R2 and let C 1 ( ; E3 ) be an immersion that satises a3 C 1 ( ; E3 ). Assume that there exists a vector eld H 1 ( ; E3 ) that satises a = a a.e. in , a3 H 1 ( ; E3 ), and b = b a.e. in .
Then there exist a vector c E3 and a matrix Q O3 + such that (y ) = c + Q(y ) for almost all y . Proof. The proof essentially relies on the extension to a Sobolev space setting of the three-dimensional rigidity theorem established in Theorem 1.7-3. (i) To begin with, we record several technical preliminaries. First, we observe that the relations a = a a.e. in and the assumption that C 1 ( ; E3 ) is an immersion together imply that |a1 a2 | = det(a ) = det(a ) > 0 a.e. in .
Consequently, the vector eld a3 , and thus the functions b , are well dened a.e. in . Second, we establish that b = b in and b = b a.e. in , i.e., that a a3 = a a3 in and a a3 = a a3 a.e. in . To this end, we note that either the assumptions C 1 ( ; E3 ) and a3 C 1 ( ; E3 ) together, or the assumptions H 1 ( ; E3 ) and a3 H 1 ( ; E3 ) together, imply that a a3 = a3 L1 loc ( ), hence that a3 D ( ). Given any D( ), let U denote an open subset of R2 such that supp U and U is a compact subset of . Denoting by X , X the duality pairing between a topological vector space X and its dual X , we have
D ( )
a3 ,
D ( )
a3 dy (a3 ) dy

( ) a3 dy.
Observing that a3 = 0 a.e. in and that
(a3 ) dy
= =
(a3 ) dy ( ), a3
1 (U ; E 3 ) , H0
H 1 ( U ; E 3 )
Sect. 2.9] Uniqueness of surfaces with the same fundamental forms
99
we reach the conclusion that the expression D () a3 , D() is symmetric with respect to and since = in D (U ). Hence a3 = a3 in L1 loc ( ), and the announced symmetries are established. Third, let c := a3 a3 and c := a3 a3 . Then we claim that c = c a.e. in . To see this, we note that the matrix elds (a ) := (a )1 and (a ) := (a )1 are well dened and equal a.e. in since is an immersion and a = a a.e. in . The formula of Weingarten (Section 2.6) can thus be applied a.e. in , showing that c = a b b a.e. in . The assertion then follows from the assumptions b = b a.e. in . (ii) Starting from the set and the mapping (as given in the statement of Theorem 2.9-2), we next construct a set and a mapping that satisfy the assumptions of Theorem 1.7-2. More precisely, let (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) R. Then the mapping := R E3 dened in this fashion is clearly continuously dierentiable on R and det (y, x3 ) =
2 2 1 2 2 1 det(a (y )){1 x3 (b1 1 + b2 )(y ) + x3 (b1 b2 b1 b2 )(y )}
for all (y, x3 ) R, where

b (y ) := a (y )b (y ), y .
Let n , n 0, be open subsets of R2 such that n is a compact subset of and = n0 n . Then the continuity of the functions a , a , b and the assumption that is an immersion together imply that, for each n 0, there exists n > 0 such that det (y, x3 ) > 0 for all (y, x3 ) n [n , n ]. Besides, there is no loss of generality in assuming that n 1 (this property will be used in part (iii)). Let then (n ]n , n [). :=
n0
Then it is clear that is a connected open subset of R3 and that the mapping C 1 (; E3 ) satises det > 0 in . Finally, note that the covariant components gij C 0 () of the metric tensor eld associated with the mapping are given by (the symmetries b = b established in (i) are used here) g = a 2x3 b + x2 3 c , g3 = 0, g33 = 1.
100
[Ch. 2
(iii) Starting with the mapping (as given in the statement of Theorem 2.9-2), we construct a mapping that satises the assumptions of Theorem 1.7-2. To this end, we dene a mapping : E3 by letting (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) , where the set is dened as in (ii). Hence H 1 (; E3 ), since ]1, 1[. Besides, det = det a.e. in since the functions b := a b , which are well dened a.e. in , are equal, again a.e. in , to the functions b . Likewise, the components gij L1 () of the metric tensor eld associated with the mapping satisfy gij = gij a.e. in since a = a and b = b a.e. in by assumption and c = c a.e. in by part (i). (iv) By Theorem 1.7-2, there exist a vector c E3 and a matrix Q O3 + such that (y ) + x3 a3 (y ) = c + Q((y ) + x3 a3 (y )) for almost all (y, x3 ) . Dierentiating with respect to x3 in this equality between functions in H 1 (; E3 ) shows that a3 (y ) = Qa3 (y ) for almost all y . Hence (y ) = c + Q(y ) for almost all y as announced. Remarks. (1) The existence of H 1 ( ; E3 ) satisfying the assumptions of Theorem 2.9-2 implies that C 1 ( ; E3 ) and a3 C 1 ( ; E3 ), and that H 1 ( ; E3 ) and a3 H 1 ( ; E3 ). (2) It is easily seen that the conclusion of Theorem 2.9-2 is still valid if the assumptions H 1 ( ; E3 ) and a3 H 1 ( ; E3 ) are replaced by the weaker 1 1 assumptions Hloc ( ; E3 ) and a3 Hloc ( ; E3 ).
2.10
CONTINUITY OF A SURFACE AS A FUNCTION OF ITS FUNDAMENTAL FORMS
Let be a connected and simply-connected open subset of R2 . Together, Theorems 2.8-1 and 2.9-1 establish the existence of a mapping F that associates 2 2 to any pair of matrix elds (a ) C 2 ( ; S2 > ) and (b ) C ( ; S ) satisfying the Gau and Codazzi-Mainardi equations in a well-dened element F ((a ), (b )) in the quotient set C 3 ( ; E3 )/R, where ( , ) R means that there exists a vector c E3 and a matrix Q O3 + such that (y ) = c + Q (y ) for all y . A natural question thus arises as to whether there exist ad hoc topologies on the space C 2 ( ; S2 ) C 2 ( ; S2 ) and on the quotient set C 3 ( ; E3 )/R such that the mapping F dened in this fashion is continuous. Equivalently, is a surface a continuous function of its fundamental forms? The purpose of this section, which is based on Ciarlet [2003], is to provide an armative answer to the above question, through a proof that relies in an
Sect. 2.10]
A surface as a function of its fundamental forms
101
essential way on the solution to the analogous problem in dimension three given in Section 1.8. Such a question is not only relevant to surface theory, but it also nds its source in two-dimensional nonlinear shell theories, where the stored energy functions are often functions of the rst and second fundamental forms of the unknown deformed middle surface. For instance, the well-known stored energy function wK proposed by Koiter [1966, Equations (4.2), (8.1), and (8.3)] for modeling nonlinearly elastic shells made with a homogeneous and isotropic elastic material takes the form: wK = 3 a (a a )(a a ) + a (b b )(b b ), 2 6 4 a a + 2(a a +a a ), + 2
where 2 is the thickness of the shell, a :=
> 0 and > 0 are the two Lam e constants of the constituting material, a and b are the covariant components of the rst and second fundamental forms of the given undeformed middle surface, (a ) = (a )1 , and nally a and b are the covariant components of the rst and second fundamental forms of the unknown deformed middle surface (see Section 4.1 for a more detailed description of Koiters equations for a nonlinearly elastic shell). An inspection of the above stored energy functions thus suggests a tempting approach to shell theory, where the functions a and b would be regarded as the primary unknowns in lieu of the customary (Cartesian or curvilinear) components of the displacement. In such an approach, the unknown components a and b must naturally satisfy the classical Gau and Codazzi-Mainardi equations in order that they actually dene a surface. To begin with, we introduce the following two-dimensional analogs to the notations used in Section 1.8. Let be an open subset of R3 . The notation means that is a compact subset of . If f C ( ; R) or C ( ; E3 ), 0, and , we let f
,
:= jsup | f (y )| ,
y ||
:= jsup | (y )|,
y ||
where stands for the standard multi-index notation for partial derivatives and || denotes the Euclidean norm in the latter denition. If A C ( ; M3 ), 0, and , we likewise let A
,
= nsup | A(y )|,

y ||
where || denotes the matrix spectral norm. The next sequential continuity result constitutes the key step towards establishing the continuity of a surface as a function of its two fundamental forms in ad hoc metric spaces (see Theorem 2.10-2).
102
[Ch. 2
Theorem 2.10-1. Let be a connected and simply-connected open subset of R2 . 2 2 Let (a ) C 2 ( ; S2 > ) and (b ) C ( ; S ) be matrix elds satisfying the Gau 2 2 n and Codazzi-Mainardi equations in and let (an ) C ( ; S> ) and (b ) 2 2 C ( ; S ) be matrix elds satisfying for each n 0 the Gau and CodazziMainardi equations in . Assume that these matrix elds satisfy
n
lim
an a
2,
= 0 and lim
bn b
2,
= 0 for all
Let C 3 ( ; E3 ) be any immersion that satises a = and b = 1 2 |1 2 | in
(such immersions exist by Theorem 2.8-1). Then there exist immersions n C 3 ( ; E3 ) satisfying
n n n n an = and b =
1 n 2 n |1 n 2 n | .
in , n 0,
such that
n
lim
3,
= 0 for all
Proof. For clarity, the proof is broken into ve parts.

n (i) Let the matrix elds (gij ) C 2 ( R; S3 ) and (gij ) C 2 ( R; S3 ), n 0, be dened by
g := a 2x3 b + x2 3 c n n 2 n g := an 2 x b + x 3 3 c
and gi3 := i3 , n and gi 3 := i3 , n 0
(the variable y is omitted, x3 designates the variable in R), where cn

b (a ) := (a )1 , c := b b , := a b , ,n n ,n ,n n 1 := b b , b := a b , (a,n ) := (an , n 0. )
Let 0 be an open subset of R2 such that 0 0 (0 ) > 0 such that the symmetric matrices
. Then there exists 0 =
n C(y, x3 ) := (gij (y, x3 )) and Cn (y, x3 ) := (gij (y, x3 )), n 0,
are positive denite at all points (y, x3 ) 0 , where 0 := 0 ]0 , 0 [ . The matrices C(y, x3 ) S3 and Cn (y, x3 ) S3 are of the form (the notations are self-explanatory): C(y, x3 ) = C0 (y ) + x3 C1 (y ) + x2 3 C2 (y ),
n 2 n Cn (y, x3 ) = Cn 0 (y ) + x3 C1 (y ) + x3 C2 (y ), n 0.
Sect. 2.10]
103
First, it is easily deduced from the matrix identity B = A(I + A1 (BA)) n and the assumptions limn an a 0,0 = 0 and limn b b 0, 0 = 0 that there exists a constant M such that
1 (Cn 0) 0, 0
+ Cn 1
0, 0
+ Cn 2
0,0
M for all n 0.
This uniform bound and the relations C(y, x3 ) = C0 (y ){I + (C0 (y ))1 (2x3 C1 (y ) + x2 3 C2 (y ))},
n 1 2 n (2x3 Cn Cn (y, x3 ) = Cn 0 (y ){I + (C0 (y )) 1 (y ) + x3 C2 (y ))}, n 0,
together imply that there exists 0 = 0 (0 ) > 0 such that the matrices C(y, x3 ) and Cn (y, x3 ), n 0, are invertible for all (y, x3 ) 0 [0 , 0 ]. These matrices are positive denite for x3 = 0 by assumption. Hence they remain so for all x3 [0 , 0 ] since they are invertible. for each and (ii) Let , 0, be open subsets of R2 such that = 0 . By (i), there exist numbers = ( ) > 0, 0, such that the n symmetric matrices C(x) = (gij (x)) and Cn (x) = (gij (x)), n 0, dened for all x = (y, x3 ) R as in (i), are positive denite at all points x = (y, x3 ) , where := ] , [, hence at all points x = (y, x3 ) of the open set :=
0
which is connected and simply connected. Let the functions Rqijk C 0 () be dened from the matrix elds (gij ) C 2 (; S3 > ) by
where ijq := 1 pq pq 1 (j giq +i gjq q gij ) and p , ij := g ijq , with (g ) := (gij ) 2
n and let the functions Rqijk C 0 (), n 0 be similarly dened from the matrix n 2 3 elds (gij ) C (; S> ), n 0. Then n Rqijk = 0 in and Rqijk = 0 in for all n 0.
That is connected and simply-connected is established in part (viii) of the n proof of Theorem 2.8-1. That Rqijk = 0 in , and similarly that Rqijk = 0 in for all n 0, is established as in parts (iv) to (viii) of the same proof.
n n 2 3 (iii) The matrix elds C = (gij ) C 2 (; S3 > ) and C = (gij ) C (; S> ) dened in (ii) satisfy (the notations used here are those of Section 1.8)
lim
Cn C
2,K
= 0 for all K
104
[Ch. 2
Given any compact subset K of , there exists a nite set K of integers such that K K . Since by assumption,
n
lim
an a
2,
= 0 and lim
bn b
2,
= 0,
K ,
it follows that
n
lim
Cn p Cp
2,
= 0,
k , p = 0, 1, 2,
where the matrices Cp and Cn p , n 0, p = 0, 1, 2, are those dened in the proof of part (i). The denition of the norm 2, then implies that
n
lim
Cn C
2,
= 0,
K .
The conclusion then follows from the niteness of the set K . (iv) Conclusion. Given any mapping C 3 ( ; E3 ) that satises a = and b = let the mapping : E3 be dened by (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) , where a3 := 1 2 , and let |1 2 | gij := i j . Then an immediate computation shows that g = a 2x3 b + x2 3 c and gi3 = i3 in , where a and b are the covariant components of the rst and second fundamental forms of the surface ( ) and c = a b b . In other words, the matrices (gij ) constructed in this fashion coincide over the set with those dened in part (i). Since parts (ii) and (iii) of the above proof together show that all the assumptions of Theorem 1.8-3 are satised n n 2 3 by the elds C = (gij ) C 2 (; S3 > ) and C = (gij ) C (; S> ), there exist n n T n 3 3 n mappings C (; E ) satisfying ( ) = C in , n 0, such that lim n 3,K = 0 for all K .
n
1 2 |1 2 |
in ,
We now show that the mappings n () := n (, 0) C 3 ( ; E3 )
Sect. 2.10] indeed satisfy
105
n n n n an = and b =
1 n 2 n |1 n 2 n |
in .
Dropping the exponent n for notational convenience in this part of the proof, let g i := i . Then 33 = 3 g 3 = p 33 g p = 0, since it is easily veried that the functions p , constructed from the functions gij as indicated in part (ii), 33 vanish in . Hence there exists a mapping 1 C 3 ( ; E3 ) such that (y, x3 ) = (y ) + x3 1 (y ) for all (y, x3 ) . Consequently, g = + x3 1 and g 3 = 1 . The relations gi3 = g i g 3 = i3 then show that ( + x3 1 ) 1 = 0 and 1 1 = 1. These relations imply that 1 = 0. Hence either 1 = a3 or 1 = a3 in . But 1 = a3 is ruled out since we must have {1 2 } 1 = det(gij )|x3 =0 > 0. Noting that a3 = 0 implies a3 = a3 , we obtain, on the one hand, g = ( + x3 a3 ) ( + x3 a3 ) = 2x3 a3 + x2 3 a3 a3 in . Since, on the other hand, g = a 2x3 b + x2 3 c in , we conclude that a = and b = a3 in , as desired. It remains to verify that
n
lim
3,
= 0 for all
But these relations immediately follow from the relations

n
lim
3,K
= 0 for all K
combined with the observations that a compact subset of is also one of , that (, 0) = and n (, 0) = n , and nally, that n
3,
3, .
106
[Ch. 2
Remark. At rst glance, it seems that Theorem 2.10-1 could be established by a proof similar to that of its three-dimensional counterpart, viz., Theorem 1.8-3. A quick inspection reveals, however, that the proof of Theorem 1.8-2 does not carry over to the present situation. In fact, it is not necessary to assume in Theorem 2.10-1 that the limit matrix elds (a ) and (b ) satisfy the Gau and Codazzi-Mainardi equations (see the proof of the next theorem). More specically, another sequential continuity result can be derived from Theorem 2.10-1. Its interest is that the assumptions are now made on the immersions n that dene the surfaces n ( ) for all n 0; besides the existence of a limit surface ( ) is also established. Theorem 2.10-2. Let be a connected and simply-connected open subset of R2 . n For each n 0, let there be given immersions n C 3 ( ; E3 ), let an and b denote the covariant components of the rst and second fundamental forms of 2 the surface n ( ), and assume that bn C ( ). Let there be also given matrix 2 2 2 elds (a ) C ( ; S> ) and (b ) C ( ; S2 ) with the property that
n
lim
an a
2,
= 0 and lim
n
bn b
2,
= 0 for all
Then there exist immersions C 3 ( ; E3 ) of the form = cn + Qn n , with cn E3 and Qn O3 + (hence the rst and second fundamental forms of the surfaces ( ) and n ( ) are the same for all n 0), and there exists an immersion C 3 (, E3 ) such that a and b are the covariant components of the rst and second fundamental forms of the surface ( ). Besides,
n n n
lim
3,
= 0 for all
Proof. An argument similar to that used in the proof of Theorem 1.8-4 shows that passing to the limit as n is allowed in the Gau and Codazzi-Mainardi equations, which are satised in the spaces C 0 ( ) and C 1 ( ) respectively by the n functions an and b for each n 0 (as necessary conditions; cf. Theorem 2.7-1). Hence the limit functions a and b also satisfy the Gau and CodazziMainardi equations. By the fundamental existence theorem (Theorem 2.8-1), there thus exists an immersion C 3 ( ; E3 ) such that a = and b = 1 2 . |1 2 |
Theorem 2.10-1 can now be applied, showing that there exist mappings (now n denoted) C 3 ( ; E3 ) such that
n an = and b = n n n
1 2
n
n n
|1 2 |
in , n 0,
Sect. 2.10] and
107
lim
3,
= 0 for all
Finally, the rigidity theorem for surfaces (Theorem 2.9-1) shows that, for each n 0, there exist cn E3 and Qn O3 + such that = cn + Qn n in , since the surfaces ( ) and n ( ) share the same fundamental forms and the set is connected. It remains to show how the sequential continuity established in Theorem 2.10-1 implies the continuity of a surface as a function of its fundamental forms for ad hoc topologies. Let be an open subset of R2 . We recall (see Section 1.8) that, for any integers 0 and d 1, the space C ( ; Rd ) becomes a locally convex topological space when it is equipped with the Fr echet topology dened by the family of semi-norms , , . Then a sequence (n )n0 converges to with respect to this topology if and only if
n n n
lim
= 0 for all
Furthermore, this topology is metrizable : Let (i )i0 be any sequence of subsets of that satisfy
i Then
n
and i int i+1 for all i 0, and =

i=0
i .
lim
= 0 for all
lim d ( n , ) = 0,
n
where d ( , ) :=
i=0
1 ,i . 2i 1 + ,i
3 ( ; E3 ) := C 3 ( ; E3 )/R denote the quotient set of C 3 ( ; E3 ) by the Let C equivalence relation R, where (, ) R means that there exist a vector c E3 and a matrix Q O3 + such that (y ) = c + Q (y ) for all y . Then the 3 3 3 set C ( ; E ) becomes a metric space when it is equipped with the distance d dened by , ) := inf d3 (, ) = j inf d3 ( , c+Q ), 3 ( d j 3
c E QO3
denotes the equivalence class of modulo R. where The announced continuity of a surface as a function of its fundamental forms is then a corollary to Theorem 2.10-1. If d is a metric dened on a set X , the associated metric space is denoted {X ; d}.
108
[Ch. 2
Theorem 2.10-3. Let be connected and simply connected open subset of R2 . Let
2 2 2 2 2 2 C0 ( ; S2 > S ) := {((a ), (b )) C ( ; S> ) C ( ; S ); C C + C C C C = b b b b in , b C b = 0 in }. b b + C 2 2 Given any element ((a ), (b )) C0 ( ; S2 > S ), let F (((a ), (b ))) 3 3 ( ; E ) denote the equivalence class modulo R of any immersion C 3 ( ; E3 ) C that satises
a = and b = Then the mapping
1 2 |1 2 |
in .
2 2 3 3 3 F : {C0 ( ; S2 > S ); d2 } {C ( ; E ); d }
dened in this fashion is continuous.

2 3 3 3 Proof. Since {C0 ( ; S2 > S); d2 } and {C ( ; E ); d } are both metric spaces, it suces to show that convergent sequences are mapped through F into convergent sequences. 2 2 n n 2 2 ( ; S2 Let then ((a ), (b )) C0 > S ) and ((a ), (b )) C0 ( ; S> S2 ), n 0, be such that n n lim d2 (((an ), (b )), ((a ), (b ))) = 0,
i.e., such that

n
lim
an a
2,
= 0 and lim
bn b
2,
= 0 for all
Let there be given any F (((a ), (b ))). Then Theorem 2.10-1 shows n that there exist n F (((an ), (b ))), n 0, such that
n
lim
3,
= 0 for all
i.e., such that

n
lim d3 (n , ) = 0.
Consequently,
n
3 (F (((an ), (bn ))), F (((a ), (b )))) = 0, lim d
and the proof is complete. The above continuity results have been extended up to the boundary of the set by Ciarlet & C. Mardare [2005].
Chapter 3 APPLICATIONS TO THREE-DIMENSIONAL ELASTICITY IN CURVILINEAR COORDINATES
INTRODUCTION
The raison d etre of equations of three-dimensional elasticity directly expressed in curvilinear coordinates is twofold. First and foremost, they constitute an inevitable point of departure for the justication of most two-dimensional shell theories (such as those studied in the next chapter). Second, they are clearly more convenient than their Cartesian counterparts for modeling bodies with specic geometries, e.g., spherical or cylindrical. Consider a nonlinear elastic body, whose reference conguration is of the form (), where is a domain in R3 and : E3 is a smooth enough immersion. Let 0 1 denote a partition of the boundary such that area 0 > 0. The body is subjected to applied body forces in its interior (), to applied surface forces on the portion (1 ) of its boundary, and to a homogeneous boundary condition of place on the remaining portion (0 ) of its boundary (this means that the displacement vanishes there). We rst review in Section 3.1 the associated equations of nonlinear threedimensional elasticity in Cartesian coordinates, i.e., expressed in terms of the Cartesian coordinates of the set (). We then examine in Sections 3.2 to 3.5 how these equations are transformed when they are expressed in terms of curvilinear coordinates, i.e., in terms of the coordinates of the set . To this end, we put to use in particular the notion of covariant derivatives of a vector eld introduced in Chapter 1. In this fashion, we show (Theorem 3.3-1) that the variational equations of the principle of virtual work in curvilinear coordinates take the following form: ij (Ei j (u)v ) g dx = f i vi g dx + hi vi g d
for all v = (vi ) W(). In these equations, the functions ij = ji : R 109
110
Applications to three-dimensional elasticity
[Ch. 3
are the contravariant components of the second Piola-Kirchho stress tensor eld ; the functions Ei j (u) = 1 (ui 2
j
+ uj
+ g mn um i un j ) : R
are the covariant components of the Green-St Venant strain tensor eld associated with a displacement vector eld ui g i : R3 of the reference conguration (); the functions f i : R and hi : 1 R are the contravariant components of the applied body and surface forces; 0 1 denotes a partition of the boundary of ; nally, W() denotes a space of suciently smooth vector elds v = (vi ) : R3 that vanish on 0 . We also show (Theorem 3.3-1) that the above principle of virtual work is formally equivalent to the following equations of equilibrium in curvilinear coordinates : ( ij + kj g i u ( ij + kj g i u where functions such as tij
j pj iq = j tij + i + j pj t jq t , k) j k )nj
= f i in , = hi on 1 ,
which naturally appear in the derivation of these equations, provide instances of rst-order covariant derivatives of a tensor eld. These equations must be complemented by the constitutive equation of the elastic material, which in general takes the form ij (x) = Rij (x, (Em n (u)(x))) for all x for ad hoc functions Rij that characterize the elastic material constituting the body. In the important special case where the elastic material is homogeneous and isotropic and the reference conguration () is a natural state, the functions Rij are of the specic form Rij (x, E) = Aijk (x)Ek + o(E) for all x and E = (Ek ) S3 , where the functions Aijk = g ij g k + (g ik g j + g i g jk ) designate the contravariant components of the elasticity tensor of the elastic material and the constants and , which satisfy the inequalities 3 + 2 > 0 and > 0, are the Lam e constants of the material (Theorem 3.4-1). Together with boundary conditions such as ui = 0 on 0 , the equations of equilibrium and the constitutive equation constitute the boundary value problem of nonlinear three-dimensional elasticity in curvilinear coordinates (Section 3.5). Its unknown is the vector eld u = (ui ) : R3 ,
Ch. 3]
Introduction
111
where the functions ui : R are the covariant components of the unknown displacement eld ui g i : R3 of the reference conguration (). We then derive by means of a formal linearization procedure the equations that constitute the boundary value problem of linearized three-dimensional elasticity in curvilinear coordinates (Section 3.6). This problem is studied in detail in the rest of this chapter. The variational, or weak, formulation of this linear boundary value problem consists in seeking a vector eld u = (ui ) V() = {v = (vi ) H1 (); v = 0 on 0 } such that Aijk ek (u)ei j (v ) g dx = f i vi g dx + hi vi g d
for all v = (vi ) V(), where ui g i is now to be interpreted as a linearized approximation of the unknown displacement vector eld of the reference conguration. The functions ei j (v ) L2 (), which are dened for each v = (vi ) H1 () by 1 ei j (v ) = (vi j + vj i ) 2 are the covariant components of the linearized strain tensor in curvilinear coordinates. Equivalently, the vector eld u V() minimizes the functional J : V() R dened by J (v ) = 1 2 Aijk ek (v )ei j (v ) g dx f i vi g dx + hi vi g d
for all v H1 (). We then show how a fundamental lemma of J.L. Lions (Theorem 3.7-1) can be put to use for directly establishing a Korn inequality in curvilinear coordinates. When area 0 > 0, this key inequality asserts the existence of a constant C such that (Theorem 3.8-3) v
1,
C
i,j
ei j (v )
2 0,
1/2
for all v V().
Together with the uniform positive-deniteness of the elasticity tensor, which holds under the assumptions 3 + 2 > 0 and > 0 (Theorem 3.9-1), this inequality in turn yields (Theorem 3.9-2) the existence and uniqueness of a solution to the variational formulation of the equations of linearized threedimensional elasticity, again directly in curvilinear coordinates. In this chapter, expressions such as equations of nonlinear elasticity, Korns inequality, etc., are meant to be understood as equations of nonlinear three-dimensional elasticity, three-dimensional Korns inequality, etc.
112
[Ch. 3
3.1
THE EQUATIONS OF NONLINEAR ELASTICITY IN CARTESIAN COORDINATES
We briey review in this section the equations of nonlinear elasticity. All details needed about the various notions introduced in this section may be found in Ciarlet [1988], to which more specic references are also provided below. Additional references are provided at the end of this section. Latin indices or exponents range in the set {1, 2, 3}, except when they are used for indexing sequences. Let E3 denote a three-dimensional Euclidean space, let a b denote the Euclidean inner product of a, b E3 , and let |a| denote the Euclidean norm of a E3 . Let ei = ei denote the orthonormal basis vectors of E3 , let xi = xi denote the Cartesian coordinates of x = xi ei = xi ei E3 and let i := / xi . Let M3 and S3 denote the space of all matrices of order three and that of all symmetric matrices of order three, let A : B := tr AT B 3 denote the matrix inner product of A, B M , and let A := A : A denote the associated matrix norm of A M3 . We recall that a domain in E3 is a bounded, open, and connected subset of E3 with a Lipschitz-continuous boundary, the set being locally on the same side of its boundary (see Ne cas [1967] or Adams [1975]). Let be a domain in E3 , let dx denote the volume element in , let d denote the area element along the boundary of , and let n = ni ei denote the unit (|n| = 1) outer normal vector eld along . Remark. The reason we use in this section notations such as , , or n is to later aord a proper distinction between equations written in terms of the Cartesian coordinates xi of the points x of the set on the one hand as in this section, and equations written in terms of curvilinear coordinates xi on the other hand, the points x = (xi ) then varying in a domain with boundary and unit outer normal vector eld n (see Section 3.5). If v = (vi ) : {} R3 is a smooth enough vector eld, its gradient v : {} M3 is the matrix eld dened by 1 v1 2 v1 3 v1 v := 1 v2 2 v2 3 v2 . 1 v3 2 v3 3 v3 ij ) : {} M3 is a smooth enough matrix eld (the rst exponent If T = (t i is the row index), its divergence div T : {} M3 is the vector eld dened by 1j j t 2j div T := j t . 3j j t If W : {} M3 R is a smooth enough function, its partial derivative
Sect. 3.1]
Nonlinear elasticity in Cartesian coordinates
113
W (x, F) M3 at a point (x, F) = (x, (Fij )) {} M3 is the matrix that F satises W (x, F + G) = W (x, F) + Equivalently, W /F11 W (x, F) := W /F21 F W /F31 W /F12 W /F22 W /F32 W /F13 W /F23 (x, F). W /F33 W (x, F) : G + o( G ). F
Let there be given a body, which occupies the set {} in the absence of applied forces. The set {} is called the reference conguration of the body. Let = 0 1 be a d-measurable partition (0 1 = ) of the boundary of . The body is subjected to applied body forces in its interior , of density f = f i ei : R3 per unit volume, and to applied surface forces on the portion 1 of its boundary, of density h = hi ei : 1 R3 per unit area. We assume that these densities do not depend on the unknown, i.e., that the applied forces considered here are dead loads (cf. Ciarlet [1988, Section 2.7]). The unknown is the displacement eld u = ui ei = ui ei : {} E3 , where the three functions ui = ui : {} R are the Cartesian components of the displacement that the body undergoes when it is subjected to applied forces. This means that u(x) = ui (x)ei is the displacement of the point x {} (see Figure 1.4-1). It is assumed that the displacement eld vanishes on the set 0 , i.e., that it satises the (homogeneous) boundary condition of place u = 0 on 0 .
3 Let id{ b } denote the identity mapping of the space E . The mapping
: {} R3 dened by
:= id { b } + u, i.e., by (x) = ox + u(x) for all x {} , is called a deformation of the reference conguration {} and the set {} is called a deformed conguration. Since the approach in this section is for its most part formal, we assume throughout that the requirements that the deformation should satisfy in order to be physically admissible (orientation-preserving character and injectivity; cf. ibid., Section 1.4) are satised. Naturally, the deformation may be equivalently considered as the unknown instead of the displacement eld u.
114
[Ch. 3
The following equations of equilibrium in Cartesian coordinates (cf. ibid., Sections 2.5 and 2.6) are then satised in the reference conguration {} : There exists a matrix eld = ( ij ) : {} M3 , called the second PiolaKirchho stress tensor eld, such that div{(I + u)} = f in , (I + u)n = h on 1 , = in {} . Componentwise, the equations of equilibrium thus read: j ( ij + kj k ui ) = f i in , ( ij + kj k ui )nj = hi on 1 , ij = ji in {} . Note that the symmetry of the matrix eld is part of the equations of equilibrium. The components ij of the eld are called the second Piola-Kirchho stresses. The boundary conditions on the set 1 constitute a boundary condition of traction. ij ) : {} M3 , where The matrix eld T = (t T := (I + u) = ij = is called the rst Piola-Kirchho stress tensor eld. Its components t ij i kj ( + k u ) are called the rst Piola-Kirchho stresses. While the equations of equilibrium are thus expressed in a simpler manner in terms of the rst Piola-Kirchho stress tensor, it turns out that the constitutive equation of an elastic material (see below) is more naturally expressed in terms of the second Piola-Kirchho stress tensor. This is why the equations of equilibrium were directly written here in terms of the latter stress tensor. Let W() denote a space of suciently smooth vector elds v = vi ei = v i ei : {} E3 that vanish on 0 . The following Greens formula is then easily established: For any smooth enough matrix eld T : {} M3 , and vector eld v W(),
b T
div T v dx =
T : v dx +
b1
Tn v d.
If the unknown vector eld u = (ui ) and the elds f and h are smooth enough, it is immediately established, because of the above Greens formula, that the rst and second equations of equilibrium are formally equivalent to the principle of virtual work in Cartesian coordinates, which states that:
b
(I + u) : v dx =
f v dx +
b1
h v d for all v W().
Sect. 3.1]
115
Componentwise, the principle of virtual work thus reads: ( ij + kj k ui )j vi dx = f i vi dx + hi vi d
b1
for all vi ei W(). The principle of virtual work is thus nothing but the weak, or variational, form of the equations of equilibrium. The vector elds v W() that enter it are variations around the deformation = id { b } + u (cf. ibid., Section 2.6). Let the Green-St Venant strain tensor eld associated with an arbitrary displacement eld v = vi ei = v i ei : {} E3 of the reference conguration {} be dened by (cf. ibid., Section 1.8): E(v ) := 1 v T + v + v T v = (Eij (v )), 2
where the matrix eld v denotes the corresponding displacement gradient eld. The components Eij (v ) = 1 (i vj + j vi + i vm j v m ) 2
of the matrix eld E(v ) are called the Green-St Venant strains. Let i i := id { b } + v = i e = ei denote the associated deformation of the reference conguration {} and assume that the mapping : E3 is an injective immersion, so that the set () can be considered as being equipped with the Cartesian coordinates xi in E3 as its curvilinear coordinates. In this interpretation the covariant components of the metric tensor of the set () are thus given by (Section 1.2) ( )ij = i m j m = ij + 2Eij (v ) = (I + 2E(v ))ij . In the context of nonlinear three-dimensional elasticity, the matrix eld is called the Cauchy-Green tensor eld. Since the constant functions ij are the covariant components of the metric tensor of the set (which corresponds to the particular deformation id { b } ), the components Eij (v ) measure the dierences between the covariant components of the metric tensor of the deformed conguration and those of the reference conguration. This is why the eld E(v ) = 1 ( T I) 2
T T
is also aptly called the change of metric tensor eld associated with the displacement eld v . This also explains why strain means change of metric.
116
[Ch. 3
Remark. The linearized strain tensor e(v ) = (eij (v )) := where 1 ({v }T + v ), 2
1 (j vi + i vj ), 2 that naturally arises in linearized elasticity in Cartesian coordinates (see Section 3.6) is thus exactly the linear part with respect to v in the strain tensor E(v ). eij (v ) =
A material is elastic if, at each point x of the reference conguration {} , the stress tensor (x) is a known function (that characterizes the material) of the displacement gradient u(x) at the same point. The consideration of the fundamental principle of material frame-indierence further implies (cf. ibid., Theorem 3.6-2) that (x) is in fact a function of u(x) by means of the GreenSt Venant strain tensor E(u(x)) at x. Equivalently, there exists a response function R = (Rji ) : {} S3 S3 such that, at each point x of the reference conguration, the stress tensor (x) is given by the relation (x) = R(x, E(u)(x)), which is called the constitutive equation in Cartesian coordinates of the material. If the material is elastic, the unknown vector eld u = ui ei = ui ei : {} R3 should thus satisfy the following boundary value problem of nonlinear elasticity in Cartesian coordinates: div{(I + u)} = f in , u = 0 on 0 , (I + u)n = h on 1 , = R(, E(u)) in {} , 1 E(u) = (uT + u + uT u) in {} . 2 Componentwise, this boundary value problem reads: j ( ij + kj k ui ) = f i in , ui = 0 on 0 , ( ij + kj k ui )nj = hi on 1 , ij = Rij (, (Ek (u))) in {} , 1 Ek (u) = (k u + uk + k um um ) in {} . 2
Sect. 3.1]
117
If area 0 > 0 and area 1 > 0, the above boundary value problem is called a displacement-traction problem, by reference to the boundary conditions. For the same reason, it is called a pure displacement problem if 1 = , or a pure traction problem if 0 = . Thanks to the symmetry of the tensor eld and to the relations AB : C = B : AT C = BT : CT A valid for any A, B, C M3 , the integrand appearing in the left-hand side of the principle of virtual work can be re-written as (I + u) : v = : v + (u) : v T T 1 : v + : v + (u) : v + (u ) : v = 2 1 = : v + v T + uT v + v T u 2 = : E (u)v , where E (u)v = 1 v + v T + uT v + v T u 2
denotes the G ateaux derivative of the mapping E : v W() S3 (it is immediately veried that E (u)v is indeed the linear part with respect to v in the dierence E(u + v ) E(u)). Naturally, ad hoc topologies must be specied insuring that the mapping E is dierentiable (for instance, it is so if it is considered as a mapping from the space W1,4 () into the space L2 (; S3 )). Consequently, the left-hand of the principle of virtual work takes the form
b
(I + u) : v dx = =
b b
: (E (u)v ) dx R(, E(u)) : (E (u)v ) dx,
or, componentwise,
b
( ij + kj k ui )j vi dx = =
b b
ij (Eij (u)v ) dx Rij (x, E(u))(Eij (u)v ) dx.
This observation motivates the following denition: An elastic material is hyperelastic if there exists a stored energy function W : {} S3 R such that W R(x, E) = (x, E) for all (x, E) {} S3 , E or equivalently, such that Rij (x, E) = W Eij (x, E ) for all (x, E) = (x, (Eij )) {} S3 .
118
[Ch. 3
If the elastic material is hyperelastic, the principle of virtual work thus takes the form W (, E(u)) : (E (u)v ) dx = E f v dx + h v d
b1
for all v W(), so that it becomes formally equivalent to the equations J (u)v = 0 for all v W(), ateaux derivative at u of the energy J : W() R where J (u)v denotes the G dened for all v W() by J (v ) = W (, E(v )) dx f v dx + h v d .
b1
Naturally, an ad hoc topology in the space W() must be again specied, so that the energy J is dierentiable on that space. To sum up, if the elastic material is hyperelastic, nding the unknown displacement eld u amounts, at least formally, to nding the stationary points (hence in particular, the minimizers) of the energy J over an ad hoc space W() of vector elds v satisfying the boundary conditions v = 0 on 0 , i.e., to nding those vector elds u W() such that the Fr echet derivative J (u) vanishes. In other words, the boundary value problem of three-dimensional nonlinear elasticity becomes, at least formally, equivalent to a problem in the calculus of variations if the material is hyperelastic. This observation was put to a beautiful use when Ball [1977] established in a landmark paper the existence of minimizers of the energy for hyperelastic materials whose stored energy is polyconvex and satises ad hoc growth conditions (the notion of polyconvexity, which is due to John Ball, plays a fundamental role in the calculus of variations). This theory accommodates non-smooth boundaries and boundary conditions such as those considered here and is not restricted to small enough forces. However, it does not provide the existence of a solution to the corresponding variational problem (let alone to the original boundary value problem), because the energy is not dierentiable in the spaces where the minimizers are found (a detailed account of John Balls theory is also found in Ciarlet [1988, Chapter 7]). If the elastic material is homogeneous and isotropic and the reference conguration {} is a natural state, i.e., is stress-free (these notions are dened in ibid., Chapter 3), the response function takes a remarkably simple form for small deformations (ibid., Theorem 3.8-1): There exist two constants and , called the Lam e constants of the material, such that R(E) = (tr E)I + 2E + o(E) for all E S3
Sect. 3.2]
Principle of virtual work in curvilinear coordinates
119
(the response function no longer depends on x by virtue of the assumption of homogeneity). Equivalently, Rij (E) = Aijk Ek + o(E) for all E = (Ek ) S3 , where the constants Aijk = ij k + ( ik j + i jk ) are called the Cartesian components of the elasticity tensor (characterizing the elastic body under consideration). Experimental evidence shows that the Lam e constants of actual elastic materials satisfy 3 + 2 > 0 and > 0. For such materials, an existence theory is available that relies on the implicit function theorem. It is, however, restricted to smooth boundaries, to small enough applied forces, and to special classes of boundary conditions, which do not include those of the displacement-traction problems considered here (save exceptional cases). See Ciarlet [1988, Section 6.7] and Valent [1988]. Detailed expositions of the modeling of three-dimensional nonlinear elasticity are found in Truesdell & Noll [1965], Germain [1972], Wang & Truesdell [1973], Germain [1981], Marsden & Hughes [1983, Chapters 15], and Ciarlet [1988, Chapters 15]. Its mathematical theory is exposed in Ball [1977], Marsden & Hughes [1983], Valent [1988], and Ciarlet [1988, Chapters 6 and 7].
3.2
PRINCIPLE OF VIRTUAL WORK IN CURVILINEAR COORDINATES
Our rst objective is to transform the principle of virtual work expressed in Cartesian coordinates (Section 3.1) into similar equations, but now expressed in arbitrary curvilinear coordinates. Remark. Our point of departure could be as well the formally equivalent equations of equilibrium in Cartesian coordinates. It has been instead preferred here to derive the equations of equilibrium in curvilinear coordinates as natural corollaries to the principle of virtual work in curvilinear coordinates; see Section 3.3. Our point of departure thus consists of the variational equations
b
ij (Eij (u)v ) dx =
f i vi dx +
b1
hi vi d,
which are satised for all v = vi ei W(). We recall that is a domain in E3 with its boundary partitioned as = 0 1 , W() is a space of sufciently smooth vector elds v = (vi ) that vanish on 0 , u = (ui ) W() is the displacement eld of the reference conguration {} , the functions ij = ji : {} R are the second Piola-Kirchho stresses, the functions Eij (u)v = 1 (j vi + i vj + i um j v m + j um i v m ) 2
120
[Ch. 3
are the G ateaux derivatives of the Green-St Venant strains Eij : W() R dened by 1 Eij (v ) = (i vj + j vi + i vm j v m ), 2 and nally, (f i ) : R3 and (hi ) : 1 R3 are the densities of the applied forces. The above equations are expressed in terms of the Cartesian coordinates xi of the points x = (xi ) {} and of the Cartesian components of the functions ij , ui , vi , f i , and hi . Assume that we are also given a domain in R3 and a smooth enough injective immersion : E3 such that () = {} . Our objective consists in expressing the equations of the principle of virtual work in terms of the curvilinear coordinates xi of the points x = (x) {} , where x = (xi ) . In other words, we wish to carry out a change of variables, from the old variables xi to the new variables xi , in each one of the integrals appearing in the above variational equations, which we thus wish to write as dx = dx and d = d,
b1
where 1 is the subset of the boundary of that satises (1 ) = 1 . As expected, we shall make an extensive use of notions introduced in Chapter 1 in this process. A word of caution. From now on, we shall freely extend without notice all the denitions given (for instance, that of an immersion), or properties established, in Chapter 1 on open sets to their analogs on closures of domains. In particular, the denition of m-th order dierentiability, m 1, can be extended as follows for functions dened over the closure of a domain. Let be an open subset of Rn , n 2. For any integer m 1, dene the space C m () as the subspace of the space C m () consisting of all functions f C m () that, together with all their partial derivatives of order m, possess continuous extensions to the closure . Because, in particular, of a deep extension theorem of Whitney [1934], one can then show that, if the boundary of is Lipschitz-continuous, the space C m () can be dened equivalently as C m () = {f | ; f C m (Rn )} (irrespective of whether is bounded); for a proof, see, e.g., Ciarlet & C. Mardare [2004a, Theorem 4.2]. For further results in this direction, see Stein [1970]. Because the old unknowns ui : {} R are the components of a vector eld, some care evidently must be exercised in the denition of the new unknowns, which must reect the physical invariance of the displacement vector ui (x)ei at each point x {} . Accordingly, we proceed as in Section 1.4.
Sect. 3.2]
121
Since the mapping : E3 considered above is assumed to be an immersion, the three vectors g i (x) := i (x), which are linearly independent at all points x , thus form the covariant basis at x = (x) {} (Section 1.2). We may thus unambiguously dene three new unknowns ui : R by requiring that ui (x)g i (x) := ui (x)ei for all x = (x), x , where the three vectors g i (x) form the contravariant basis at x = (x) {} i i (see Figure 1.4-2). Using the relations g i (x) g j (x) = j and ei ej = j , we note (again as in Section 1.4) that the old and new unknowns are related by uj (x) = ui (x)ei g j (x) and ui (x) = uj (x)g j (x) ei . Let [g j (x)]i := g j (x) ei and [g j (x)]i := g j (x) ei , i.e., [g j (x)]i denotes the i-th component of the vector g j (x), and [g j (x)]i denotes the i-th component of the vector g j (x), over the basis {e1 , e2 , e3 } = {e1 , e2 , e3 } of the Euclidean space E3 . In terms of these notations, the preceding relations thus become uj (x) = ui (x)[g j (x)]i and ui (x) = uj (x)[g j (x)]i for all x = (x), x . The three components ui (x) are called the covariant components of the displacement vector ui (x)g i (x) at x, and the three functions ui : R dened in this fashion are called the covariant components of the displacement eld ui g i : R3 . A word of caution. While the old unknown vector (ui (x)) R3 can be, and will henceforth be, justiably identied with the displacement vector u(x) = ui (x)ei since the basis {e1 , e2 , e3 } is xed in E3 , this is no longer true for the new unknown vector (ui (x)) R3 , since its components ui (x) now represent the components of the displacement vector over the basis {g 1 (x), g 2 (x), g 3 (x)}, which varies with x . In the same vein, the vector elds u = (ui ) and u := ui g i , which are both dened on , must be carefully distinguished ! While the latter has an intrinsic character, the former has not; it only provides a means of recovering the eld u via its covariant components ui . We likewise associate new functions vi : R with the old functions vi : {} R appearing in the equations of the principle of virtual work by letting vi (x)g i (x) := vi (x)ei for all x = (x), x .
122
[Ch. 3
We begin the change of variables by considering the integrals found in the right-hand side of the variational equations of the principle of virtual work, i.e., those corresponding to the applied forces. With the Cartesian components f i : = () R of the applied body force density, let there be associated its contravariant components f i : R, dened by f i (x)g i (x) := f i (x)ei for all x = (x), x . This denition shows that f i (x) = f j (x)[g i (x)]j , and consequently that f i (x)vi (x) = (f i (x)ei ) (vj (x)ej ) = (f i (x)g i (x)) (vj (x)g j (x)) = f i (x)vi (x) for all x = (x), x . Hence f i (x)vi (x) dx = f i (x)vi (x) since dx = g (x) dx for all x = (x), x ,
g (x) dx (Theorem 1.3-1 (a)), and thus f i vi dx = f i vi g dx.
Remark. What has just been proved is in eect the invariance of the number f i (x)vi (x) with respect to changes of curvilinear coordinates, provided one vector (here f i (x)g i (x)) appears by means of its contravariant components (i.e., over the covariant basis) and the other vector (here vi (x)g i (x)) appears by means of its covariant components (i.e., over the contravariant basis). Naturally, this number is nothing but the Euclidean inner product of the two vectors. With the Cartesian components hi : 1 = (1 ) R of the applied surface force density, let there likewise be associated its contravariant components hi : 1 R, dened by hi (x)g i (x) g (x) d(x) := hi (x)ei d(x) for all x = (x), x 1 ,
where the area elements d(x) at x = (x) 1 and d(x) at x 1 are related by d(x) = g (x) nk (x)g k (x)n (x) d(x), where the functions nk : 1 R denote the components of the unit outer normal vector eld along the boundary of (Theorem 1.3-1 (b)). This denition shows that hi (x)ei = nk (x)g k (x)n (x)
1/2
hi (x)g i (x) for all x = (x), x 1 ,
Sect. 3.2] and that
123
hi (x) = The factor
nk (x)g k (x)n (x) hj (x)[g i (x)]j .
g introduced in the denition of the functions hi implies that hi vi d = hi vi g d.
b1
As a result, the same factor g appears in both integrals f i vi g dx and i h vi g d. This common factor in turn gives rise to a more natural 1 boundary condition on 1 when the variational equations of the principle of virtual work are used to derive the equilibrium equations in curvilinear coordinates (Theorem 3.3-1). Since our treatment in this section is essentially formal (as in Section 3.1), the functions ui , vi , ij , f i , and hi appearing in the next theorem are again assumed to be smooth enough, so as to insure that all the computations involved make sense. Transforming the integrals appearing in the left-hand side of the variational equations of the principle of virtual work relies in particular on the fundamental notion of covariant dierentiation of a vector eld, introduced in Section 1.4. Theorem 3.2-1. Let be a domain in E3 , let be a domain in R3 , and let : E3 be an injective immersion that is also a C 2 -dieomorphism from onto {} = (). Dene the vector elds u = (ui ) : R3 and v = (vi ) : R3 and the symmetric matrix elds ( ij ) : S3 and (Eij (v )) : S3 by ui (x) := uk (x)[g i (x)]k and vi (x) := vk (x)[g i (x)]k , for all x = (x), x , ij (x) := k (x)[g i (x)]k [g j (x)] for all x = (x), x , Eij (v )(x) := Ek (v )(x) [g i (x)]k [g j (x)] for all x = (x), x . Then the functions Eij = Eji : W() R are also given by Eij (v ) := 1 (vi 2
j
+ vj
+ g mn vm i vn j ) for all v = (vi ) : R3 ,
p p where the functions vi j := j vi p ij vp , where ij := g i g j , are the covariant i 3 derivatives of the vector eld vi g : R . The G ateaux derivatives Eij (u)v are then related to the G ateaux derivatives
Eij (u)v := of the functions Ei

j
1 vi 2
+ vj
+ g mn {um i vn
+ un j vm i }
: W() R by the relations
Eij (u)v (x) = Ek (u)v [g k ]i [g ]j (x) at all x (x), x .
124
[Ch. 3
Let W() denote a space of suciently smooth vector elds v = (vi ) : R3 that vanish on 0 . Then the above matrix and vector elds satisfy the following variational equations: ij Eij (u)v g dx =
f i vi g dx +
hi vi g d
for all v = (vi ) W(), where 0 := 1 (0 ) and 1 := 1 (1 ). Proof. The following conventions hold throughout this proof: The simultaneous appearance of x and x in an equality means that they are related by x = (x) and that the equality in question holds for all x . The appearance of x alone in a relation means that this relation holds for all x . (i) Expression of the matrix eld ( ij ) : {} S3 in terms of the matrix eld ( ij ) : S3 as dened in the statement of the theorem. In part (i) of the proof of Theorem 1.4-1, it was shown that [g i (x)]k = k i (x), where = i ei : {} R3 denotes the inverse mapping of = k ek : E3 . Since j (x) = [g j (x)] , the relation (x)(x) = I shows that i . [g p (x)]k [g p (x)]i = k We thus have
i j ij (x) = k (x)k
= k (x)[g p (x)]k [g p (x)]i [g q (x)] [g q (x)]j = pq (x)[g p (x)]i [g q (x)]j , since, by denition of the functions pq , k (x)[g p (x)]k [g q (x)] = pq (x). (ii) Expressions of the functions Eij (v ) : {} R in terms of the functions Eij (v ) : R as dened in the statement of the theorem. We likewise have
p q Eij (v )(x) = Epq (v )(x)i j
= Epq (v )(x)[g k (x)]p [g (x)]q [g k (x)]i [g (x)]j = Ek (v )[g k ]i [g ]j (x), since, by denition of the functions Ek , Epq (v )(x) [g k (x)]p [g (x)]q = Ek (v )(x).
Sect. 3.2]
125
(iii) Expressions of the functions Eij (v ) in terms of the elds v = (vi ) : R3 as dened in the statement of the theorem. In Theorem 1.4-1, it was established that j vi (x) = vk [g k ]i [g ]j (x). Then this relation and the relation [g k (x)]m [g (x)]m = g k (x) g (x) = g k (x) together imply that Eij (v )(x) = 1 i vj + j vi + i vm j v m (x) 2 1 (vk + v k + g mn vm k vn )[g k ]i [g ]j (x). = 2
Because of the equivalence (with self-explanatory notations) Aij (x) = Ak [g k ]i [g ]j (x) Ak (x) = Apq (x) [g k ]p [g ]q (x), we thus conclude from part (ii) that the functions Eij (v ) are also given by Eij (v ) = 1 (vi 2
j
+ vj
+ g mn vm i vn j ) for all v = (vi ) : R3 .
ateaux (iv) Expression of the G ateaux derivatives Eij (u)v in terms of the G derivatives Eij (u)(v ). We likewise have Eij (u)v (x) = = = 1 (vk 2 1 (j vi + i vj + i um j v m + j um i v m ) (x) 2
k
+v
+ g mn {um k vn
+ un vm k })[g k ]i [g ]j (x)
(Ek (u)v )[g k ]i [g ]j (x),
since it is immediately veried that Eij (u)v is indeed given by (as the linear part with respect to v in the dierence {Eij (u + v ) Eij (u)}): Eij (u)v = 1 vi 2
j
+ vj
+ g mn {um i vn
+ un j vm i } .
(v) Conclusions: Parts (i) and (iv) together show that ij (Eij (u)v ) (x) = pq (Ek (u)v )[g p ]i [g q ]j [g k ]i [g ]j (x) = ij (Eij (u)v ) (x),
126 since
[Ch. 3
k [g p (x)]i [g k (x)]i = g p (x) g k (x) = p and [g q (x)]j [g (x)]j = g q (x) g (x) = q .
Finally, dx =
g (x) dx (Theorem 1.3-1 (a)). Therefore, ij Eij (u)v dx = ij (Eij (u)v ) g dx,
on the one hand. At the beginning of this section, it was also shown that the denition of the functions f i and hi implies that
b
f i vi dx =
f i vi g dx and
b1
hi vi d =
hi vi g d,
on the other hand. Thus the proof is complete. Naturally, if E3 is identied with R3 and = id R3 , each vector g i (x) is equal to ei and thus g i (x) = ei , g (x) = 1, g ij (x) = ij , and p ij (x) = 0 for all x . Consequently, the elds ( ij ) and ( ij ), (ui ) and (ui ), (f i ) and (f i ), and (hi ) and (hi ) coincide in this case. The variational problem found in Theorem 3.2-1 constitutes the principle of virtual work in curvilinear coordinates and the functions ij : R are the contravariant components of the second Piola-Kirchho stress tensor eld. They are also called the second Piola-Kirchho stresses in curvilinear coordinates. The functions Eij (v ) : R are the covariant components of the Green-St Venant strain tensor eld associated with an arbitrary displacement eld vi g i : R of the reference conguration (). They are also called the Green-St Venant strains in curvilinear coordinates. We saw in Section 3.1 that the Green-St Venant strain tensor in Cartesian coordinates is indeed a change of metric tensor, since it may also be written as 1 T {(id{ E(v ) = b } +v )} (id{ b } +v ) I . 2 Naturally, this interpretation holds as well in curvilinear coordinates. Indeed, a straightforward computation shows that the Green-St Venant strains in curvilinear coordinates may also be written as Eij (v ) = where gij (v ) := i ( + vk g k ) j ( + v g ) and gij = i j respectively denote the covariant components of the metric tensors of the deformed conguration ( + vk g k )() associated with a displacement eld vk g k , and of the reference conguration (). 1 (gij (v ) gij ), 2
Sect. 3.3]
Equations of equilibrium in curvilinear coordinates
127
3.3
EQUATIONS OF EQUILIBRIUM IN CURVILINEAR COORDINATES; COVARIANT DERIVATIVES OF A TENSOR FIELD
While deriving the equations of equilibrium, i.e., the boundary value problem that is formally equivalent to the variational equations of the principle of virtual work, simply amounts in Cartesian coordinates to applying the fundamental Green formula, doing so in curvilinear coordinates is more subtle. As we next show, it relies in particular on the notion of covariant dierentiation of a tensor eld. As in Sections 3.1 and 3.2, our treatment is formal, in that ad hoc smoothness is implicitly assumed throughout. We recall that the space W() denotes a space of suciently smooth vector elds v : R3 that vanish on 0 . Theorem 3.3-1. Let be a domain in R3 , let = 0 1 be a partition of the boundary of , let ni denote the components of the unit outer normal vector eld along , and let f i : R and hi : 1 R be given functions. Then a symmetric matrix eld ( ij ) : S3 and a vector eld u = (ui ) : R3 satisfy the principle of virtual work in curvilinear coordinates found in Theorem 3.2-1, viz., ij (Eij (u)v ) g dx = f i vi g dx + hi vi g d
for all v = (vi ) W() if and only if these elds satisfy the following boundary value problem: ( ij + kj g i u ( ij + kj g i u
k) j k )nj
= f i in , = hi on 1 ,
where, for an arbitrary tensor eld with smooth enough contravariant components tij : R, pj iq tij j := j tij + i + j pj t jq t . Proof. (i) We rst establish the relations j g = g q qj . To this end, we recall that g = | det |, that the column vectors of the matrix are g 1 , g 2 , g 3 (in this order), and that the vectors g i are linearly independent at all points in . Assume for instance that det > 0, so that g = det = det(g 1 , g 2 , g 3 ) in . Then j g = det(j g 1 , g 2 , g 3 ) + det(g 1 , j g 2 , g 3 ) + det(g 1 , g 2 , j g 3 ) p p = p 1j det(g p , g 2 , g 3 ) + 2j det(g 1 , g p , g 3 ) + 3j det(g 1 , g 2 , g p )
128
[Ch. 3
since j g q = p qj g p (Theorem 1.4-2 (a)); hence q 2 3 j g = 1 1j + 2j + 3j det(g 1 , g 2 , g 3 ) = qj g. The proof is similar if g = det in .
(ii) Because of the symmetries ij = ji , the integrand in the left-hand side of the principle of virtual work (Theorem 3.2-1) can be re-written as ij (Eij (u)v ) = ij (vi
j
+ g mn um i vn j ).
Taking into account the relations j g = g q qj (part (i)) and using the fundamental Green formula (j v )w dx = vj w dx + vwnj d,
we obtain: ij vi
j
g dx =
ij g j vi dx
= = =
j g ij vi dx +
pj iq g j ij + i + j vi dx + pj jq ( ij
j )vi
ij g nj vi d
ij p g ij vp dx
pj i g pj vi dx
g dx +
ij nj vi g d
ij g nj vi d
for all vector elds v = (vi ) : R3 . We likewise obtain: ij g mn um i vn

j
g dx = +
g ( kj g i u
k) j
vi dx
kj i g( g u
k )nj vi d
for all vector elds v = (vi ) : R3 . Hence the variational equations of the principle of virtual work imply that g ( ij + kj g i u
k) j
+ f i vi dx
k )nj
=
1
g ( ij + kj g i u
hi vi d
for all (vi ) W(). Letting (vi ) vary rst in (D())3 , then in W(), yields the announced boundary value problem. The converse is established by means of the same integration by parts formulas.
Sect. 3.4]
Constitutive equation in curvilinear coordinates
129
The equations ( ij + kj g i u ( ij + kj g i u
k) j k )nj ij
= f i in , = hi on 1 = ji in ,
constitute the equations of equilibrium in curvilinear coordinates. The functions tij = ij + kj g i u k are the contravariant components of the rst Piola-Kirchho stress tensor. Finally, the functions tij
j pj iq := j tij + i + j pj t jq t
are instances of rst-order covariant derivatives of a tensor eld, dened here by means of its contravariant components tij : R (for a more systematic derivation, see, e.g., Lebedev & Cloud [2003, Chapter 4]). We emphasize that, like their Cartesian special cases, the principle of virtual work and the equations of equilibrium are valid for any continuum (see, e.g., Ciarlet [1988, Chapter 2]), i.e., regardless of the nature of the continuum. The object of the next section is precisely to take this last aspect into account.
3.4
CONSTITUTIVE EQUATION IN CURVILINEAR COORDINATES
Because of Theorem 3.2-1, the constitutive equation of an elastic material (cf. Section 3.1) can be easily converted in terms of curvilinear coordinates. Theorem 3.4-1. Let the notations and assumptions be as in Theorem 3.2-1. Let there be given functions Rij : {} S3 R such that ij (x) = Rij (x, (Ek (u)(x))) for all x {} is the constitutive equation of an elastic material with {} as its reference conguration. Then there exist functions Rij : S3 R, depending only on the functions Rij and on the mapping : E3 , such that the second Piola-Kirchho stresses in curvilinear coordinates ij : R are given in terms of the covariant components of the Green-St Venant strain tensor as ij (x) = Rij (x, (Ek (u)(x))) for all x . Assume that, in addition, the elastic material is homogeneous and isotropic and that its reference conguration {} is a natural state, in which case the functions Rij satisfy Rij (E) = Aijk Ek + o(E) for all E = (Ek ) S3 ,
130 with
[Ch. 3
Aijk = ij k + ( ik j + i jk ). Then, in this case, the functions Rij satisfy Rij (x, E) = Aijk (x)Ek + o(E) for all E = (Ek ) S3 at each x , where Aijk := g ij g k + (g ik g j + g i g jk ) = Ajik = Ak
ij
Proof. Let = i ei : {} R3 denote the inverse mapping of the C -dieomorphism : E3 . Theorem 3.2-1 and its proof show that, for all x = (x) ,
2
ij (x) = Rk (x, (Epq (u)(x)))[g i (x)]k [g j (x)] , Epq (u)(x) = (Emn (v )[g m ]p [g n ]q )(x), [g i (x)]k = k i (x). The announced conclusions immediately follow from these relations. Given an elastic material with () as its reference conguration, the relations ij (x) = Rij (x, (Ek (u)(x))) for all x form its constitutive equation in curvilinear coordinates, the matrix eld (Rij ) : S3 S3 being its response function in curvilinear coordinates. The functions Aijk = g ij g k + (g ik g j + g i g jk ) are the contravariant components of the three-dimensional elasticity tensor in curvilinear coordinates (characterizing a specic elastic body).
3.5
THE EQUATIONS OF NONLINEAR ELASTICITY IN CURVILINEAR COORDINATES
Assembling the various relations found in Sections 3.2, 3.3, and 3.4, we are in a position to describe the basic boundary value problem of nonlinear elasticity in curvilinear coordinates. We are given: a domain in R3 whose boundary is partitioned as = 0 1 and an immersion : E3 that is also a C 2 -dieomorphism from onto its image; a matrix eld (Rij ) : S3 S3 , which is the response function of an elastic material with the set () E3 as its reference conguration ; a vector eld (f i ) : R3 and a vector eld (hi ) : 1 R3 , whose components are the contravariant components of the applied body and surface force densities.
Sect. 3.5]
Nonlinear elasticity in curvilinear coordinates
131
We are seeking a vector eld u = (ui ) : R3 , whose components are the covariant components of the displacement eld ui g i of the set (), that satises the following boundary value problem: ( ij + kj g i u ( ij + kj g i u
k) j
= f i in , ui = 0 on 0 , = hi on 1 , = Rij (, (Ek (u))) in ,
k )nj ij
where tij u
j k j iq pj := j tij + i pj t + jq t ,
:= k u p ij up , 1 Ek (u) := (uk + u k + g mn um k un ). 2 If in addition the elastic material is homogeneous and isotropic and its reference conguration () is a natural state, the functions Rij satisfy Rij (x, E) = Aijk (x)Ek + o(E) for all E = (Ek ) S3 at each x , where Aijk = g ij g k + (g ik g j + g i g jk ). The above nonlinear boundary value problem is a pure displacement problem if 0 = , a pure traction problem if 1 = , and a displacement-traction problem if area 0 > 0 and area 1 > 0. Let W() denote a space of suciently smooth vector elds v = (vi ) : R3 that vanish on 0 . Then the unknown vector eld u W() satises the above boundary value problem if and only if it satises, at least formally, the following variational equations : Rij (, (Ek (u))) Eij (u)v g dx =
f i vi g dx +
hi vi g d
for all v = (vi ) W(), where Eij (u)v := 1 vi 2

j
+ vj
+ g mn {um i vn
+ un j vm i }
are the G ateaux derivatives of the functions Eij : W() R. If the elastic material is hyperelastic, there exists a stored energy function W : S3 R such that Rij (x, E) = W (x, E) for all (x, E) = (x, (Eij )) S3 . Eij
132
[Ch. 3
In this case, the above variational equations thus become formally equivalent to the equations J (u)v = 0 for all v W(), where the energy in curvilinear coordinates J : W() R is dened for all v W() by J (v ) =
W (, (Ek (v ))) g dx
f i vi g dx +
hi vi g d .
Finding the unknown vector eld u in this case thus amounts, at least formally, to nding the stationary points, and in particular the minimizers, of the energy J over an ad hoc space W().
3.6
THE EQUATIONS OF LINEARIZED ELASTICITY IN CURVILINEAR COORDINATES
Formally, the boundary value problem of nonlinear elasticity in curvilinear coordinates (Section 3.5) consists in nding a vector eld u = (ui ) such that u W() and (A(u), B (u)) = (f , h) in F() H(1 ), where W(), F(), and H(1 ) are spaces of smooth enough vector elds respectively dened on and vanishing on 0 , dened in , and dened on 1 , A : W() F() and B : W() H(1 ) are nonlinear operators dened by (A(u))i = ( ij + kj g i u where ij = Rij (, Ek (u)) and the elds f = (f i ) F() and h = (hi ) H(1 ) are given. Assume henceforth that the elastic material is homogeneous and isotropic and that the reference conguration is a natural state. This last assumption thus implies that (A(0), B(0)) = (0, 0), so that u = 0 is a particular solution corresponding to f = 0 and h = 0. Assume in addition that W(), F(), and H(1 ) are normed vector spaces and that the nonlinear operator (A, B) : W() F() H(1 ) is Fr echet dierentiable at the origin 0 W(). Then, by denition, the boundary value problem of linearized elasticity consists in nding u = (ui ) such that u W() and (A (0)u, B (0)u) = (f , h) in F() H(1 ). In other words, this linear boundary value problem is the tangent at the origin of the nonlinear boundary value problem. As such, its solution can be
k) j
and (B (u))i = ( ij + kj g i u
k )nj ,
Sect. 3.6]
Linearized elasticity in curvilinear coordinates
133
expected to be a good approximation, at least for small enough forces, of the displacement eld that satises the original nonlinear problem. Indeed, this raison d etre can be rigorously justied in ad hoc function spaces, but only in some carefully circumscribed situations (see, e.g., Ciarlet [1988, Theorem 6.8-1]). Because it is linear, and thus incomparably simpler to solve than the original nonlinear problem, this linear problem also provides an extraordinarily ecient and versatile means of numerically approximating displacement and stress elds in innumerable elastic structures arising in engineering. For this reason, devising ecient approximation schemes for this problem has pervaded the numerical analysis and computational mechanics literature during the past decades and continues to do so to a very large extent, even to this day. In order to compute the components of the vector elds A (0)u and B (0)u, it clearly suces, once derivability is assumed, to compute the terms that are linear with respect to v in the dierences A(v ) A(0) = A(v ) and B(v ) B(0) = B(v ). Doing so shows that solving the boundary value problem of linearized elasticity consists in nding a vector eld u = (ui ) : R3 that satises ij
ij j
= f i in ,
ui = 0 on 0 , nj = hi on 1 , ij = Aijk ek (u) in , where ek (u) := Recall that ij v

j pj iq := j ij + i + j pj jq ,
1 (uk 2
+u
k ).
Aijk := g ij g k + (g ik g j + g i g jk ),
k
:= k v p ij vp .
The functions ek (v ) := 1 2 (vk +v k ) are called the covariant components of the linearized strain tensor, or the linearized strains in curvilinear coordinates, associated with a displacement vector eld vi g i of the reference conguration (). By denition, they satisfy eij (v ) = [Eij (v )]lin , where [ ]lin denotes the linear part with respect to v = (vi ) in the expression [ ]. The linearized constitutive equation ij = Aijk ek (u) is called Hookes law in curvilinear coordinates and the functions ij that appear in this equation are called the linearized stresses in curvilinear coordinates.
134
[Ch. 3
The above linear boundary value problem is a pure displacement problem if 0 = , a pure traction problem if 1 = , and a displacementtraction problem if area 0 > 0 and area 1 > 0. Writing the above boundary value problem as a variational problem naturally relies on the following linearized version of Theorem 3.3-1. As before, ad hoc smoothness is implicitly assumed throughout. Theorem 3.6-1. A symmetric matrix eld ( ij ) : S3 satises ij
j
= f i in ,
ij nj = hi on 1 , if and only if it satises the variational equations ij eij (v ) g dx = f i vi g dx + hi vi g d
for all v = (vi ) W(), where W() is a space of smooth enough vector elds that vanish on 0 . Proof. The proof relies on the integration by part formula ij vi
j
g dx =
( ij
j )vi
g dx +
ij nj vi g d,
valid for all vector elds (vi ) : R3 , established in the proof of Theorem 3.3-1. The variational equations derived in Theorem 3.6-1 constitute the linearized principle of virtual work in curvilinear coordinates. Together with the assumed symmetry of the matrix eld ( ij ), the equations ij j = f i in and ij nj = hi constitute the linearized equations of equilibrium in curvilinear coordinates. Because of Theorem 3.6-1, it is immediately seen that the boundary value problem of linearized elasticity is formally equivalent to nding a vector eld u W() that satises the following variational equations :
Aijk ek (u)eij (v ) g dx =
f i vi g dx +
hi vi g d
for all v = (vi ) W(). Thanks to the symmetries Aijk = Ak ij , nding the solutions to these variational equations also amounts to nding the stationary points of the functional J : W() R dened by J (v ) = 1 2
Aijk ek (v )eij (v ) g dx
f i vi g dx +
hi vi g d
for all v W().
Sect. 3.7]
A fundamental lemma of J.L. Lions
135
Our objective in the next sections is to establish the existence and uniqueness of the solution to these problems in ad hoc function spaces. More specically, dene the space V() := {v = (vi ) H 1 (; R3 ); v = 0 on 0 }, dene the symmetric bilinear form B : V() V() R by B (u, v ) :=
Aijk ek (u)eij (v ) g dx
for all (u, v ) V() V(), and dene the linear form L : V() R by L(v ) :=
f i vi g dx
hi vi g d
for all v V(). Clearly, both forms B and L are continuous on the space V() if the data are smooth enough. Our main result will then consist in establishing that, if area 0 > 0, the bilinear form B is V()-elliptic, i.e., that there exists a constant > 0 such that v 2 1, B (v , v ) for all v V(), where 1, denotes the norm of the Hilbert space H 1 (; R3 ). This inequality holds, because of a fundamental Korn inequality in curvilinear coordinates (Theorem 3.8-3), which asserts the existence of a constant C such that v
1,
C
i,j
eij (v )
2 0,
1/2
for all v V(),
in addition to the uniform positiveness of the elasticity tensor (Theorem 3.9-1), meaning that there exists a constant Ce such that |tij |2 Ce Aijk (x)tk tij
i,j
for all x and all symmetric matrices (tij ). The existence and uniqueness of a solution to the above variational equations then follow from the well-known Lax-Milgram lemma and the symmetry of the bilinear form further implies that this solution is also the unique minimizer of the functional J over the space V().
3.7
A FUNDAMENTAL LEMMA OF J.L. LIONS
We rst review some essential denitions and notations, together with a fundamental lemma of J.L. Lions (Theorem 3.7-1). This lemma plays a key r ole in the proof of the Korn inequality in curvilinear coordinates.
136
[Ch. 3
Recall that a domain in Rd is an open, bounded, connected subset of Rd with a Lipschitz-continuous boundary , the set being locally on one side of . As is Lipschitz-continuous, a measure d can be dened along and a unit outer normal vector = (i )d i=1 (unit means that its Euclidean norm is one) exists d-almost everywhere along . m () denote Let be a domain in Rd . For each integer m 1, H m () and H0 the usual Sobolev spaces. In particular, H 1 () := {v L2 (); i v L2 (), 1 i d}, H 2 () := {v H 1 (); ij v L2 (), 1 i, j d}, where i v and ij v denote partial derivatives in the sense of distributions, and
1 () := {v H 1 (); v = 0 on }, H0
where the relation v = 0 on is to be understood in the sense of trace. Boldface letters denote vector-valued or matrix-valued functions, also called vector elds or matrix elds, and their associated function spaces. The norm in L2 () or L2 () is noted 0, and the norm in H m () or Hm (), m 1, is noted m, . In particular then, v v
0,
:=
d
|v |2 dx |vi |2 0,
2 0,
1/2
if v L2 (),
2 if v = (vi )d i=1 L (), 1/2
0,
:=
i=1
1/2
1,
:=
v
d
+
i=1 2 1, d
i v
1/2
2 0,
if v H 1 (),
1,
:=
i=1
vi v
2 0,
1 if v = (vi )d i=1 H (), d 1/2
2,
+
i=1
i v
2 0,
+
i,j =1
ij v
2 0,
if v H 2 (), etc.
Detailed treatments of Sobolev spaces are found in Ne cas [1967], Lions & Magenes [1968], Dautray & Lions [1984, Chapters 13], Adams [1975]. An excellent introduction is given in Brezis [1983]. In this section, we also consider the Sobolev space
1 H 1 () := dual space of H0 (). 1 () being Another possible denition of the space H0 1 H0 () = closure of D() with respect to
1,
where D() denotes the space of innitely dierentiable real-valued functions dened over whose support is a compact subset of , it is clear that v L2 () = v H 1 () and i v H 1 (), 1 i n,
Sect. 3.8]
Korns inequalities in curvilinear coordinates
137
since (the duality between the spaces D() and D () is denoted by , ): | v, | = | i v, | = | v, i | =
v dx v
0,
1, ,
vi dx v
0,
1,
for all D(). It is remarkable (but also remarkably dicult to prove!) that the converse implication holds: Theorem 3.7-1. Let be a domain in Rd and let v be a distribution on . Then {v H 1 () and i v H 1 (), 1 i d} = v L2 (). This implication was rst proved by J.L. Lions, as stated in Magenes & Stampacchia [1958, p. 320, Note (27 )]; for this reason, it will be henceforth referred to as the lemma of J.L. Lions. Its rst published proof for domains with smooth boundaries appeared in Duvaut & Lions [1972, p. 111]; another proof was also given by Tartar [1978]. Various extensions to genuine domains, i.e., with Lipschitz-continuous boundaries, are given in Bolley & Camus [1976], Geymonat & Suquet [1986], and Borchers & Sohr [1990]; Amrouche & Girault [1994, Proposition 2.10] even proved that the more general implication {v D () and i v H m (), 1 i n} = v H m+1 () holds for arbitrary integers m Z. Note that some minimal regularity of the boundary is anyway required: Geymonat & Gilardi [1998] have shown that the lemma of J.L. Lions does not hold if the open set satises only the segment property (this means that, for every x , there exists an open set Ux containing x and a vector ax = 0 such that (y + tax ) for all y Ux and all 0 < t < 1). Remark. Although Theorem 3.7-1 shall be referred to as the lemma of J.L. Lions in this volume, there are other results of his that bear the same name in the literature, such as his compactness lemmas (Lions [1961, Proposition 4.1, p. 59] or Lions [1969, Section 5.2, p. 57]) or his singular perturbation lemma (Lions [1973, Lemma 5.1, p. 126]).
3.8
KORNS INEQUALITIES IN CURVILINEAR COORDINATES
As already noted in Section 3.6, the existence and uniqueness of a solution to the variational equations of three-dimensional linearized elasticity found in Section 3.6 will essentially follow from the ellipticity of the associated bilinear form. To this end, an essential step consists in establishing a three-dimensional Korn inequality in curvilinear coordinates (Theorem 3.8-3), due to Ciarlet [1993]
138
[Ch. 3
(see also Chen & Jost [2002], who have shown that, in fact, such an inequality holds in the more general context of Riemannian geometry). In essence, this inequality asserts that, given a domain R3 with boundary and a subset 0 of with area 0 > 0, the L2 ()-norm of the matrix elds (eij (v )) is equivalent to the H1 ()-norm of the vector elds v for all vector elds v H1 () that vanish on 0 (the ellipticity of the bilinear form also relies on the positive deniteness of the three-dimensional elasticity tensor; cf. Theorem 3.9-1). Recall that the functions eij (v ) = 1 2 (vi j + vj i ) are the covariant components of the linearized strain tensor associated with a displacement eld vi g i of the reference conguration (). As a rst step towards proving such an inequality, we establish in Theorem 3.8-1 a Korns inequality in curvilinear coordinates without boundary conditions. This means that this inequality is valid for all vector elds v = (vi ) H1 (), i.e., that do not satisfy any specic boundary condition on . This inequality is truly remarkable, since only six dierent combinations of rst-order partial derivatives, viz., 1 2 (j vi + i vj ), occur in its right-hand side, while all nine partial derivatives i vj occur in its left-hand side! A similarly striking observation applies to part (ii) of the next proof, which entirely rests on the crucial lemma of J.L. Lions recalled in the previous section. Theorem 3.8-1. Let be a domain in R3 and let be a C 2 -dieomorphism of onto its image () E3 . Given a vector eld v = (vi ) H1 (), let eij (v ) := 1 2 (j vi + i vj ) p ij vp L () 2
denote the covariant components of the linearized change of metric tensor associated with the displacement eld vi g i . Then there exists a constant C0 = C0 (, ) such that v
1,
C0
i
vi
2 0,
+
i,j
eij (v )
2 0,
1/2
for all v = (vi ) H1 ().
Proof. The proof is essentially an extension of that given in Duvaut & Lions [1972, p. 110] for proving Korns inequality without boundary conditions in Cartesian coordinates. (i) Dene the space W() := {v = (vi ) L2 (); eij (v ) L2 ()}. Then, equipped with the norm v
W() W()
dened by +
i,j
:=
i
vi
2 0,
eij (v )
2 0,
1/2
the space W() is a Hilbert space.
Sect. 3.8]
139
The relations eij (v ) L2 () appearing in the denition of the space W() are naturally to be understood in the sense of distributions. This means that there exist functions in L2 (), denoted eij (v ), such that eij (v ) dx = 1 (vi j + vj i ) + p ij vp dx 2 for all D().
k k Let there be given a Cauchy sequence (v k ) k=1 with elements v = (vi ) W(). The denition of the norm W() shows that there exist functions vi L2 () and eij L2 () such that k vi vi in L2 () and eij (v k ) eij in L2 () as k ,
since the space L2 () is complete. Given a function D(), letting k in the relations eij (v k ) dx = 1 k k k (v j + vj i ) + p ij vp dx, k 1, 2 i
shows that eij = eij (v ). (ii) The spaces W() and H1 () coincide. Clearly, H1 () W(). To establish the other inclusion, let v = (vi ) W(). Then sij (v ) := 1 2 (j vi + i vj ) = {eij (v ) + p ij vp } L (), 2
0 2 since eij (v ) L2 (), p ij C (), and vp L (). We thus have
k vi H 1 (), j (k vi ) = {j sik (v ) + k sij (v ) i sjk (v )} H 1 (), since w L2 () implies k w H 1 (). Hence k vi L2 () by the lemma of J.L. Lions (Theorem 3.7-1) and thus v H1 (). (iii) Korns inequality without boundary conditions. The identity mapping from the space H1 () equipped with 1, into the space W() equipped with W() is injective, continuous (there clearly exists a constant c such that v W() c v 1, for all v H1 ()), and surjective by (ii). Since both spaces are complete (cf. (i)), the closed graph theorem (see, e.g., Brezis [1983, p. 19] for a proof) then shows that the inverse mapping 1 is also continuous; this continuity is exactly what is expressed by Korns inequality without boundary conditions.
140
[Ch. 3
Our next objective is to dispose of the norms vi 0, in the right-hand side of the Korn inequality established in Theorem 3.8-1 when the vector elds v = (vi ) H1 () are subjected to the boundary condition v = 0 on a subset 0 of the boundary that satises area 0 > 0. To this end, we rst establish in the next theorem the weaker property that the semi-norm v
i,j
eij (v )
2 0,
1/2
becomes a norm for such vector elds. Part (a) in the next theorem constitutes the innitesimal rigid displacement lemma in curvilinear coordinates without boundary conditions, while part (b) constitutes the innitesimal rigid displacement lemma in curvilinear coordinates, with boundary conditions. The adjective innitesimal reminds that if eij (v ) = 0 in , i.e., if only the linearized part of the full change of metric tensor Eij (v ) vanishes in (cf. Section 3.6), then the corresponding displacement eld vi g i is in a specic sense only the linearized part of a genuine rigid displacement. More precisely, let an innitesimal rigid displacement vi g i of the set () be dened as one whose associated vector eld v = (vi ) H1 () satises eij (v ) = 0 in . Then one can show that the space of such innitesimal rigid displacements coincides with the tangent space at the origin to the manifold of rigid displacements of the set (). Such rigid displacements are dened as those whose associated deformed conguration ( + vi g i )() is obtained by means of a rigid transformation of the reference conguration () (cf. Section 1.7); for details, see Ciarlet & C. Mardare [2003]. Theorem 3.8-2. Let be a domain in R3 and let be a C 2 -dieomorphism onto its image () E3 . (a) Let a vector eld v = (vi ) H1 () be such that eij (v ) = 0 in . Then there exist two vectors a, b R3 such that the associated vector eld vi g i is of the form vi (x)g i (x) = a + b (x) for all x . (b) Let 0 be a d-measurable subset of the boundary that satises area 0 > 0, and let a vector eld v = (vi ) H1 () be such that eij (v ) = 0 in and v = 0 on 0 . Then v = 0 in . Proof. An argument similar to that used in part (ii) of the proof of Theorem 3.2-1 shows that eij (v )(x) = ek (v )[g k ]i [g ]j (x) for all x = (x), x ,
Sect. 3.8]
141
1 where eij (v ) = 1 2 (j vi + i vj ) and the vector elds v = (vi ) H () and v = (vi ) H1 () are related by
vi (x)ei = vi (x)g i (x) for all x = (x), x . Hence eij (v ) = 0 in implies eij (v ) = 0 in , and the identity (actually, the same as in the proof of Theorem 3.8-1) j (k vi ) = j eik (v ) + k eij (v ) i ejk (v ) in D () further shows that eij (v ) = 0 in implies j (k vi ) = 0 in D (). By a classical result from distribution theory (Schwartz [1966, p. 60]), each function vi is therefore a polynomial of degree 1 in the variables xj , since the set is connected. There thus exist constants ai and bij such that vi (x) = ai + bij xj for all x = (xi ) . But eij (v ) = 0 also implies that bij = bji . Hence there exist two vectors a, b R3 such that (ox denotes the column vector with components xi ) v i (x)ei = a + b ox for all x , or equivalently, such that vi (x)g i (x) = a + b (x) for all x . Since the set where such a vector eld vi ei vanishes is always of zero area unless a = b = 0 (as is easily proved; see, e.g., Ciarlet [1988, Theorem 6.3-4]), the assumption area 0 > 0 implies that v = 0. Remark. Since the elds g i are of class C 1 on by assumption, the components vi of a eld v = (vi ) H1 () satisfying eij (v ) = 0 in are thus automatically in C 1 () since vi = (vj g j ) g i . Remarkably, the eld vi g i = a + b inherits in this case even more regularity, as it is of class C 2 on ! We are now in a position to prove the announced Korn inequality in curvilinear coordinates with boundary conditions. Theorem 3.8-3. Let be a domain in R3 , let be a C 2 -dieomorphism of onto its image () E3 , let 0 be a d-measurable subset of the boundary that satises area 0 > 0, and let the space V() be dened by V() := {v = (vi ) H1 (); v = 0 on 0 }. Then there exists a constant C = C (, 0 , ) such that v
1, 1/2
C
i,j
eij (v )
2 0,
for all v V().
142
[Ch. 3
Proof. If the announced inequality is false, there exists a sequence (v k ) k=1 of elements v k V() such that vk
1,
= 1 for all k and lim
eij (v k )
0,
= 0.
1 Since the sequence (v k ) k=1 is bounded in H (), a subsequence (v ) =1 2 converges in L () by the Rellich-Kondra sov theorem. Furthermore, since
lim eij (v )
0,
= 0,
2 each sequence (eij (v )) =1 also converges in L () (to 0, but this information is not used at this stage). The subsequence (v ) =1 is thus a Cauchy sequence with respect to the norm
v = (vi )
i
vi
2 0,
+
i,j
eij (v )
2 0,
1/2
hence with respect to the norm 1, by Korns inequality without boundary conditions (Theorem 3.8-1). The space V() being complete as a closed subspace of H1 (), there exists v V() such that lim v = v in H1 (),
and the limit v satises eij (v ) 0, = lim eij (v ) 0, = 0; hence v = 0 by Theorem 3.8-2. But this contradicts the relations v 1, = 1 for all 1, and the proof is complete. Identifying E3 with R3 and letting (x) = x for all x shows that Theorems 3.8-1 to 3.8-3 contain as special cases the Korn inequalities and the innitesimal rigid displacement lemma in Cartesian coordinates (see, e.g., Duvaut & Lions [1972, p. 110]). Let Rig() := {r H1 (); eij (r ) = 0 in } denote the space of vector elds v = (vi ) H1 () whose associated displacement elds vi g i are innitesimal rigid displacements of the reference conguration (). We conclude our study of Korns inequalities in curvilinear coordinates by showing that the Korn inequality without boundary conditions (Theorem 3.8-1) is equivalent to yet another Korns inequality in curvilinear coordinates, over the quotient space H1 ()/ Rig(). As we shall see later (Theorem 3.9-3), this inequality is the basis of the existence theorem for the pure traction problem. designates the equivalence class of an In the next theorem, the notation v element v H1 () in the quotient space H1 ()/ Rig(). In other words, := {w H1 (); (w v ) Rig()}. v
Sect. 3.8]
143
Theorem 3.8-4. Let be a domain in R3 and let be a C 2 -dieomorphism of onto its image () E3 . Let 1, designate the quotient norm 1, over the Hilbert space H1 ()/ Rig(), dened by v
1,
:=
r Rig()
inf
v+r
1,
H1 ()/ Rig(). for all v
=C (, ) such that Then there exists a constant C v

1,
C
i,j
) eij (v
2 0,
1/2
H1 ()/ Rig(). for all v
Moreover, this Korn inequality over the quotient space H1 ()/ Rig() is equivalent to the Korn inequality without boundary condition of Theorem 3.8-1. Proof. (i) To begin with, we show that the Korn inequality without boundary conditions implies the announced Korn inequality over the quotient space H1 ()/ Rig(). By Theorem 3.8-2, a vector eld r = (ri ) H1 () satises eij (r ) = 0 in if and only if there exist two vectors a, b R3 such that vi (x)g i (x) = a + b (x) for all x . This shows that the space Rig() is nite-dimensional, of dimension six. By the Hahn-Banach theorem, there thus exist six continuous linear forms 1 on H (), 1 6 with the following property: An element r Rig() is equal to 0 if and only if (r ) = 0, 1 6. We then claim that it suces to establish the existence of a constant D such that v
1,
D
i,j
eij (v )
2 0,
1/2
+
=1
| (v )| for all v H1 (),
since this inequality will in turn imply the desired inequality: Given any v H1 (), let r(v ) Rig() be such that (v + r (v )) = 0, 1 6; then v
1,
r Rig()
inf
v+r eij (v )
1, 2 0,
v + r (v )
1/2
1,
D
i,j
=D
i,j
) eij (v
2 0,
1/2
To establish the existence of such a constant D, assume the contrary. Then there exist v k H1 (), k 1, such that vk and eij (v k )
i,j 2 0, 1/2 1,
= 1 for all k 1,
6
+
=1
| (v k )|
0.
144
[Ch. 3
By Rellich theorem, there exists a subsequence (v ) =1 that converges in L2 (). Since each sequence (eij (v )) also converges in L2 (), the subse=1 quence (v ) is a Cauchy sequence with respect to the norm =1 v v
0,
+
i,j
eij (v )
2 0,
}1/2 ,
hence also with respect to the norm 1, , by Korns inequality without boundary conditions (Theorem 3.8-1). Consequently, there exists v H1 () such that v v 1, 0. But then v = 0 since eij (v ) = 0 and (v ) = 0, 1 6, in contradiction with the relations v
1,
= 1 for all
1.
(ii) We next show that, conversely, Korns inequality in the quotient space H1 ()/ Rig() implies Korns inequality without boundary conditions. Assume the contrary. Then there exist v k H1 (), k 1, such that vk
1,
= 1 for all k 1 and
vk
0,
+ e(v k )
0,
0.
Let rk Rig() denote for each k 1 the projection of v k on Rig() with respect to the inner-product of H1 (), which thus satises: vk rk
1,
r Rig()
inf
vk r
1,
and vk
2 1,
= vk rk
2 1,
+ rk
2 1, .
The space Rig() being nite-dimensional, the inequalities rk 1, 1 for 1 all k 1 imply the existence of a subsequence (r ) =1 that converges in H () to an element r Rig(). Besides, Korns inequality in the quotient space H1 ()/ Rig() implies that v r 1, 0, so that v r 1, 0. Hence v r
0,
0 , which forces r to be 0, since v

1,
0,
0 on the
other hand. We thus reach the conclusion that v
0, a contradiction.
3.9
EXISTENCE AND UNIQUENESS THEOREMS IN LINEARIZED ELASTICITY IN CURVILINEAR COORDINATES

V() = {v H1 (); v = 0 on 0 },
Let the space V() be dened as before by
where 0 is a subset of the boundary satisfying area 0 > 0. Our objective consists in showing that the bilinear form B : V() V() R dened by B (u, v ) :=
Sect. 3.9]
Existence theorems in curvilinear coordinates
145
for all (u, v ) V() V() is V()-elliptic. As a preliminary, we establish the uniform positive-deniteness of the elasticity tensor (uniform means with respect to points in and to symmetric matrices of order three). Recall that the assumptions made in the next theorem about the Lam e constants and also reect experimental evidence. Theorem 3.9-1. Let be a domain in R3 and let be a C 2 -dieomorphism of onto () E3 , let the contravariant components Aijk : R of the elasticity tensor be dened by Aijk = g ij g k + (g ik g j + g i g jk ), and assume that 3 + 2 > 0 and > 0. Then there exists a constant Ce = Ce (, , , ) > 0 such that |tij |2 Ce Aijk (x)tk tij
i,j
for all x and all symmetric matrices (tij ). Proof. We recall that Md and Sd respectively designate the set of all real matrices of order d and the set of all real symmetric matrices of order d. The elegant proof of part (ii) given below is due to Cristinel Mardare. (i) To begin with, we establish a crucial inequality. Let d 2 be an integer and let and be two constants satisfying d + 2 > 0 and > 0. Then there exists a constant = (d, , ) > 0 such that tr(BT B) (tr B)2 + 2 tr(BT B) for all B Md . If 0 and > 0, this inequality holds with = 2. It thus remains to 2 < < 0 and > 0. Given any matrix B Md , consider the case where d dene the matrix C Md by C = AB := (tr B)I + 2B. The linear mapping A : Md Md dened in this fashion can be easily inverted if d + 2 = 0 and = 0, as B = A 1 C = 1 (tr C)I + C. 2(d + 2) 2
Noting that the bilinear mapping (B, C) Md Md B : C := tr BT C denes an inner product over the space Md , we thus obtain (tr B)2 + 2 tr(BT B) = (AB) : B = C : A1 C 1 1 (tr C)2 + tr(CT C) C:C = 2(d + 2) 2 2
146
[Ch. 3
for any B = A1 C Md if
2 < < 0 and > 0. Since there clearly exists d a constant = (d, , ) > 0 such that B : B C : C for all B = A1 C Md , the announced inequality also holds if 2 < < 0 and > 0, with = d
(2 )1 in this case.
(ii) We next show that, for any x and any nonzero symmetric matrix (tij ), Aijk (x)tk tij g ik (x)g j (x)tk tij > 0, where > 0 is the constant of (i) corresponding to d = 3. Given any x and any symmetric matrix (tij ), let G(x) := (g ij (x)) and T = (tij ). Then it is easily veried that Aijk (x)tk tij = tr(G(x)T) + 2 tr G(x)TG(x)T . In order to render this expression similar to that appearing in the righthand side of the inequality of (i), let H(x) S3 be the unique square root of G(x) S3 (i.e., the unique positive-denite symmetric matrix that satises (H(x))2 = G(x); for details about such square roots, see, e.g., Ciarlet [1988, Theorem 3.2-1]), and let B(x) := H(x)TH(x) S3 . Because tr(BC) = tr(CB) for any B, C M3 , we may then also write Aijk (x)tk tij = tr(B(x))
2
+ 2 tr B(x)T B(x) .
By (i), there thus exists a constant > 0 such that Aijk (x)tk tij tr B(x)T B(x) . Since B(x) = H(x)TH(x) = 0 only if T = 0, it thus follows that, for any x and any nonzero symmetric matrix (tij ), tr B(x)T B(x) = tr G(x)TG(x)T = g ik (x)g j (x)tk tij > 0. (iii) Conclusion: Since the mapping (x, (tij )) K := (tij ) S3 ;
i,j
|tij |2 = 1 g ik (x)g j (x)tk tij
Sect. 3.9]
147
is continuous and its domain of denition is compact, we infer that = (; ) := Hence

i,j (x,(tij ))K
inf
g ik (x)g j (x)tk tij > 0.
|tij |2 g ik (x)g j (x)tk tij ,
and thus
i,j
|tij |2 Ce Aijk (x)tk tij
for all x and all symmetric matrices (tij ) with Ce := ( )1 . Remark. Letting the matrices B in the inequality of (i) be equal to the identity matrix and to any nonzero matrix with a vanishing trace shows that the inequalities d + 2 > 0 and > 0 are also necessary for the validity of this inequality. With a little further ado, it can likewise be shown that the inequalities 3 + 2 > 0 and > 0 necessarily hold if the elasticity tensor is uniformly positive denite. Combined with the Korn inequality with boundary conditions (Theorem 3.8-3), the positive-deniteness of the elasticity tensor leads to the existence and uniqueness of a weak solution, i.e., a solution to the variational equations of three-dimensional linearized elasticity in curvilinear coordinates. Theorem 3.9-2. Let be a domain in R3 , let 0 be a d-measurable subset of = that satises area 0 > 0, and let be a C 2 -dieomorphism of onto its image () E3 . Finally, let there be given constants and satisfying 3 + 2 > 0 and > 0 and functions f i L6/5 () and hi L4/3 (1 ), where 1 := 0 . Then there is one and only one solution u = (ui ) to the variational problem: u V() := {v = (vi ) H1 (); v = 0 on 0 },
f i vi g dx +
hi vi g d
for all v = (vi ) V(), where Aijk = g ij g k + g ik g j + g i g jk , 1 p eij (v ) = (j vi + i vj ) p p ij vp , ij = g i g j . 2 The eld u V() is also the unique solution to the minimization problem: J (u) =
v V()
inf
J (v ),
148 where J (v ) := 1 2
[Ch. 3
Aijk ek (v )eij (v ) g dx
f i vi g dx +
hi vi g d .
Proof. As a closed subspace of H1 (), the space V() is a Hilbert space. The assumptions made on the mapping ensure in particular that the functions Aijk , p ij , and g are continuous on the compact set . Hence the bilinear form B : (u, v ) H1 () H1 () Aijk ek (u)eij (v ) g dx
is continuous. The continuous imbedding H 1 () L6 () and the continuity of the trace operator tr : H 1 () L4 () imply that the linear form L : v H1 () f i vi g dx + hi vi g d
is continuous. Since the symmetric matrix (gij (x)) is positive-denite for all x , there exists a constant g0 such that 0 < g0 g (x) = det(gij (x)) for all x . Finally, the Korn inequality with boundary conditions (Theorem 3.8-3) and the uniform positive-deniteness of the elasticity tensor (Theorem 3.9-1) together imply that
1 2 C g0 v Ce 2 1,
Aijk ek (v )eij (v ) g dx for all v V().
Hence the bilinear form B is V()-elliptic. The bilinear form being also symmetric since Aijk = Ak ij , all the assumptions of the Lax-Milgram lemma in its symmetric version are satised. Therefore, the variational problem has one and only one solution, which may be equivalently characterized as the solution of the minimization problem stated in the theorem (for a proof of the Lax-Milgram lemma in its symmetric form used here, see, e.g., Ciarlet [1988, Theorem 6.3-2]). An immediate corollary with a more intrinsic avor to Theorem 3.9-2 is the existence and uniqueness of a displacement eld ui g i , whose covariant components ui H 1 () are thus obtained by nding the solution u = (ui ) to the variational problem. Since the vector elds g i formed by the contravariant bases belong to the space C 1 () by assumption, the displacement eld ui g i also belongs to the space H1 (). Naturally, the existence and uniqueness result of Theorem 3.9-2 holds a fortiori in Cartesian coordinates (to see this, identify E3 with R3 and let = id ).
Sect. 3.9]
149
Remark. Combining the relation

b
Aijk ek (u)eij (v ) dx =
with the three-dimensional Korn inequality in Cartesian coordinates (see, e.g., Duvaut & Lions [1972, p. 110]) and with a classical result about composite mappings in Sobolev spaces (see, Ne cas [1967, Chapter 2, Lemma 3.2] or Adams [1975, Theorem 3.35]), one can also show directly that the bilinear form B : (u, v ) V() V() Aijk ek (u)eij (v ) g dx
is V()-elliptic, thus providing another proof to Theorem 3.9-2. The above existence and uniqueness result applies to the linearized pure displacement and displacement-traction problems, i.e., those that correspond to area 0 > 0. We now consider the linearized pure traction problem, i.e., corresponding to 1 = 0 . In this case, we seek a vector eld u = (ui ) H1 () such that
f i vi g dx +
hi vi g d
for all v H1 (). Clearly, such variational equations can have a solution only if their right-hand side vanishes for any vector eld r = (ri ) H1 () that satises eij (r) = 0 in , since replacing v by (v + r) with any such eld r does not aect their left-hand side. We now show that this necessary condition is in fact also sucient for the existence of solutions, thanks in this case to the Korn inequality on the quotient space H1 ()/ Rig( (Theorem 3.8-4). Evidently, the uniqueness of solutions can then hold only up to the addition of vector elds satisfying eij (r ) = 0, which implies that the solution is now sought in the same quotient space H1 ()/ Rig(). Theorem 3.9-3. Let be a domain in R3 and let be a C 2 -dieomorphism of onto its image () E3 . Let there be given constants and satisfying 3 + 2 > 0 and > 0 and functions f i L6/5 () and hi L4/3 (). Dene the space Rig() := {r H1 (); eij (r) = 0 in }, and assume that the functions f i and hi are such that f i ri g dx + hi ri g d = 0 for all r = (ri ) Rig().
Finally, let the functions Aijk be dened as in Theorem 3.9-2. H1 ()/ Rig() to the variaThen there is one and only one solution u tional equations )eij (v ) g dx = Aijk ek (u f iv i g dx + hi v i g d
150
[Ch. 3
= (v for all v i ) H1 ()/ Rig(). H1 ()/ Rig() is also the unique solution to the The equivalence class u minimization problem ) = J (u where ) := J (v 1 2 )eij (v ) g dx Aijk ek (v f iv i g dx + hi v i g d .
H1 ()/ Rig() v
inf
), J (v
Proof. The proof is analogous to that of Theorem 3.9-2 and for this reason is omitted. When 0 = and the boundary is smooth enough, a regularity result shows that the weak solution obtained in Theorem 3.9-2 is also a classical solution, i.e., a solution of the corresponding pure displacement boundary value problem, according to the following result. Theorem 3.9-4. Let be a domain in R3 with a boundary of class C 2 and let be a C 2 -dieomorphism of onto its image () E3 . If 0 = and f := (fi ) Lp (), p 6 5 , the weak solution u V() = 1 H0 () found in Theorem 3.9-2 is in the space W2,p () and satises the equations Aijk ek (u) j = f i in Lp (). Let m 1 be an integer. If the boundary is of class C m+2 , if is a C m+2 dieomorphism of onto its image, and if f Wm,p (), the weak solution m+2,p u H1 (). 0 () is in the space W Proof. We very briey sketch the main steps of the proof, which is otherwise long and delicate. As in Section 3.6, we let A (0) denote the linear operator dened by A (0) : v (Aijk ek (v ) j ) for any smooth enough vector elds v = (vi ) : R3 . (i) Because the system associated with operator A (0) is strongly elliptic, the regularity result f L2 () = u H2 () H1 0 () cas [1967, p. 260]). Hence the anholds if the boundary is of class C 2 (Ne nounced regularity holds for m = 0, p = 2. (ii) Because the linearized pure displacement problem is uniformly elliptic and satises the supplementary and complementing conditions, according to the
Sect. 3.9]
151
denitions of Agmon, Douglis & Nirenberg [1964], it follows from Geymonat [1965, Theorem 3.5] that, considered as acting from the space Vp () := {v W2,p (); v = 0 on } into the space Lp (), the mapping A (0) has an index ind A (0) that is independent of p ]1, [. Recall that ind A (0) = dim Ker A (0) dim Coker A (0), where Coker A (0) is the quotient space of the space Lp () by the space Im A (0) (the index is well dened only if both spaces Ker A (0) and Coker A (0) are nite-dimensional). In the present case, we know by (i) that ind A (0) = 0 for p = 2 since A (0) is a bijection in this case (Ker A (0) = {0} if and only if A (0) is injective, and Coker A (0) = {0} if and only if A (0) is surjective). Since the space Vp () is continuously imbedded in the space H1 0 () for p p p 6 , the mapping A ( 0 ) : V () L () is injective for these values of p 5 (if f L6/5 (), the weak solution is unique in the space H1 (); cf. Theorem 0 3.9-2); hence dim Ker A (0) = 0. Since ind A (0) = 0 on the other hand, the mapping A (0) is also surjective in this case. Hence the regularity result holds for m = 0, p 6 5. (iii) The weak solution u W2,p () H1 0 () satises the variational equations
f i vi g dx for all v = (vi ) D().
Hence we can apply the same integration by parts formula as in Theorem 3.6-1. This gives ) g dx = Aijk ek (u)eij (v (Aijk ek (u))vi g dx
for all v = (vi ) D(), and the conclusion follows since {D()} = Lp (). (iv) Once the regularity result f Wm,p () = u Wm+2,p () has been established for m = 0, it follows from Agmon, Douglis & Nirenberg [1964] and Geymonat [1965] that it also holds for higher values of the integer m if the boundary is of class C m+2 . The regularity results of Theorem 3.9-4 can be extended to linearized displacement-traction problems, but only if the closures of the sets 0 and 1 do not intersect. They also apply to linearized pure traction problems, provided the functions hi also possess ad hoc regularity. For instance, for m = 2, the functions hi are assumed to belong to the space W 1(1/p),p () (for details about such trace spaces, see, e.g., Adams [1975, Chapter 7]).
Chapter 4 APPLICATIONS TO SHELL THEORY
INTRODUCTION
Consider a nonlinearly elastic shell with middle surface S = ( ) and thickness 2 > 0, where is a domain in R2 and : E3 is a smooth enough injective immersion. The material constituting the shell is homogeneous and isotropic and the reference conguration is a natural state ; hence the material is characterized by two Lam e constants and satisfying 3 + 2 > 0 and > 0. The shell is subjected to a homogeneous boundary condition of place along a portion of its lateral face with (0 ) as its middle curve, where 0 is a portion of the boundary that satises length 0 > 0. Finally, the shell is subjected to applied body forces in its interior and to applied surface forces on its upper and lower faces. Let pi : R denote the contravariant components of the resultant (after integration across the thickness) of these forces, and let a = 4 a a + 2(a a + a a ) + 2
denote the contravariant components of the shell elasticity tensor. Then Koiters equations for a nonlinearly elastic shell, which are described in detail in Section 4.1, take the following form when they are expressed as a minimization problem : The unknown vector eld = (i ), where the functions i : R are the covariant components of the displacement eld i ai of the middle surface S , should be a stationary point (in particular a minimizer) of the functional j dened by j ( ) := 1 2 1 + 2 a (a ( ) a )(a ( ) a ) a dy 4 3 a (b ( ) b )(b ( ) b ) a dy 3
pi i a dy,
over an appropriate set of vector elds = (i ) satisfying ad hoc boundary conditions on 0 . For each such eld = (i ), the functions a ( ) and b ( ) respectively denote the covariant components of the rst and second fundamental forms 153
154
Applications to shell theory
[Ch. 4
of the deformed surface ( + i ai )( ), and the functions 1 2 (a ( ) a ) and (b ( ) b ) are the covariant components of the change of metric, and change of curvature, tensor elds associated with the displacement eld i ai of the middle surface S . Such equations provide instances of two-dimensional shell equations. Twodimensional means that such equations are expressed in terms of curvilinear coordinates (those that describe the middle surface of the shell) that vary in a two-dimensional domain . The rest of this chapter is then devoted to a mathematical analysis of another set of two-dimensional shell equations, viz., those obtained from the nonlinear Koiter equations by a formal linearization, a procedure detailed in Section 4.2. The resulting Koiter equations for a linearly elastic shell take the following weak, or variational, form, i.e., when they are expressed as a variational problem : The unknown = (i ), where i ai is now to be interpreted as a linearized approximation of the unknown displacement eld of the middle surface S , satises: = (i ) V( ) = { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on 0 }, a ( ) ( ) +
3 a ( ) ( ) a dy 3 pi i a dy for all = (i ) V( ),
where, for each = (i ) V( ), the functions ( ) L2 ( ) and ( ) L2 ( ) are the components of the linearized change of metric, and linearized change of curvature, tensors associated with a displacement eld = i ai of S , respectively given by ( ) = 1 ( a + a ) 2 ( ) = ( ) a3 .
Equivalently, the unknown vector eld V( ) minimizes the functional j : V( ) R dened by j ( ) = 1 2 a ( ) ( )
3 a ( ) ( ) a dy 3
pi i a dy
for all V( ). As shown in Sections 4.3 and 4.4, the existence and uniqueness of a solution to these equations essentially rely on a fundamental Korn inequality on a surface (Theorem 4.3-4), itself a consequence of the same crucial lemma of J.L. Lions as in Chapter 3, and on the uniform positive-deniteness of the shell elasticity tensor, which holds under the assumptions 3 + 2 > 0 and > 0 (Theorem 4.4-1).
Sect. 4.1]
The nonlinear Koiter shell equations
155
The Korn inequality on a surface asserts that, given any subset 0 of satisfying length 0 > 0, there exists a constant c such that
2 1,
+ 3
2 2,
1/2
c
,
( )
2 0,
+
,
( )
2 0,
1/2
for all = (i ) V( ). We also derive (Theorem 4.4-4) the boundary value problem that is formally equivalent to the above variational equations. This problem takes the form
b n = p3 in , m | b b m (n + b m )| b (m | ) = p in ,
i = 3 = 0 on 0 , m = 0 on 1 , (m | ) + (m ) = 0 on 1 ,
(n + 2b m ) = 0 on 1 ,
where 1 = 0 , n = a ( ), and such functions as

n | = n + + , n n | ) m | = (m | ) + (m
m =
3 a ( ), 3
which naturally appear in the course of this derivation, provide instances of rst-order, and second-order, covariant derivatives of tensor elds dened on a surface. This chapter also includes, in Sections 4.1 and 4.5, brief introductions to other nonlinear and linear shell equations that are also two-dimensional, indicating in particular why Koiter shell equations may be regarded as those of an all-purpose shell theory. Their choice here was motivated by this observation.
4.1
THE NONLINEAR KOITER SHELL EQUATIONS
To begin with, we briey recapitulate some important notions already introduced and studied at length in Chapter 2. Note in this respect that we shall extend without further notice all the denitions given, or properties studied, on arbitrary open subsets of R2 in Chapter 2 to their analogs on domains in R2 (a similar extension, this time from open subsets to domains in R3 , was carried out in Chapter 3). We recall that a domain U in Rd is an open, bounded, connected subset of Rd , whose boundary is Lipschitz-continuous, the set U being locally on one side of its boundary.
156
[Ch. 4
Greek indices and exponents (except in the notation ) range in the set {1, 2}, Latin indices and exponents range in the set {1, 2, 3} (save when they are used for indexing sequences), and the summation convention with respect to repeated indices and exponents is systematically used. Let E3 denote a threedimensional Euclidean space. The Euclidean scalar product and the exterior product of a, b E3 are noted a b and a b and the Euclidean norm of a E3 is noted |a|. Let be a domain in R2 . Let y = (y ) denote a generic point in the set , and let := /y . Let there be given an immersion C 3 ( ; E3 ), i.e., a mapping such that the two vectors a (y ) := (y ) are linearly independent at all points y . These two vectors thus span the tangent plane to the surface S := ( ) at the point (y ), and the unit vector a3 (y ) := a1 (y ) a2 (y ) |a1 (y ) a2 (y )|
is normal to S at the point (y ). The three vectors ai (y ) constitute the covariant basis at the point (y ), while the three vectors ai (y ) dened by the relations
i ai (y ) aj (y ) = j , i is the Kronecker symbol, constitute the contravariant basis at the point where j (y ) S (recall that a3 (y ) = a3 (y ) and that the vectors a (y ) are also in the tangent plane to S at (y )). The covariant and contravariant components a and a of the rst fundamental form of S , the Christoel symbols , and the covariant and mixed of the second fundamental form of S are then dened components b and b by letting:
a := a a ,
a := a a ,
:= a a ,
b := a3 a , b := a b . The area element along S is a dy , where
Note that a = |a1 a2 |. Let := ], [, let x = (xi ) denote a generic point in the set (hence x = y ), and let i := /xi . Consider an elastic shell with middle surface S = ( ) and thickness 2 > 0, i.e., an elastic body whose reference conguration is the set ( [, ]), where (cf. Figure 4.1-1) (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) [, ] .
a := det(a ).
Sect. 4.1]

a3 (y ) x
157
S
x3 x
) ( =
= ( ) E3
y
= ] , [ R3
Figure 4.1-1: The reference conguration of an elastic shell. Let be a domain in R2 , let = ], [ > 0, let C 3 ( ; E3 ) be an immersion, and let the mapping : E3 be dened by (y, x3 ) = (y ) + x3 a3 (y ) for all (y, x3 ) . Then the mapping is globally injective on if the immersion is globally injective on and > 0 is small enough (Theorem 4.1-1). In this case, the set () may be viewed as the reference conguration of an elastic shell with thickness 2 and middle surface S = ( ). The coordinates (y1 , y2 , x3 ) of an arbitrary point x are then viewed as curvilinear coordinates of the point x b = (x) of the reference conguration of the shell.
Naturally, this denition makes sense physically only if the mapping is globally injective on the set . Following Ciarlet [2000a, Theorem 3.1-1], we now show that this is indeed the case if the immersion is itself globally injective on the set and is small enough. Theorem 4.1-1. Let be a domain in R2 and let C 3 (; E3 ) be an injective immersion. Then there exists > 0 such that the mapping : E3 dened by (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) , where := ], [, is a C 2 -dieomorphism from onto () and det(g 1 , g 2 , g 3 ) > 0 in , where g i := i . Proof. The assumed regularity on implies that C 2 ( [, ] ; E3 ) for any > 0. The relations g = = a + x3 a3 and g 3 = 3 = a3 imply that det(g 1 , g 2 , g 3 )|x3 =0 = det(a1 , a2 , a3 ) > 0 in .
158
[Ch. 4
Hence det(g 1 , g 2 , g 3 ) > 0 on [, ] if > 0 is small enough. Therefore, the implicit function theorem can be applied if is small enough: It shows that, locally, the mapping is a C 2 -dieomorphism: Given any y , there exist a neighborhood U (y ) of y in and (y ) > 0 such that is a C 2 dieomorphism from the set U (y ) [(y ), (y )] onto (U (y ) [(y ), (y )]). See, e.g., Schwartz [1992, Chapter 3] (the proof of the implicit function theorem, which is almost invariably given for functions dened over open sets, can be easily extended to functions dened over closures of domains, such as the sets [, ]; see, e.g., Stein [1970]). To establish that the mapping : [, ] E3 is injective provided > 0 is small enough, we proceed by contradiction: If this property is false, n n there exist n > 0, (y n , xn 3 ), and (y , x3 ), n 0, such that n 0 as n , y n , y n , |xn 3 | n , |xn 3 | n ,
n n n n n n (y n , xn 3 ) = (y , x3 ) and (y , x3 ) = (y , x3 ).
Since the set is compact, there exist y and y , and there exists a subsequence, still indexed by n for convenience, such that y n y, Hence
n n (y ) = lim (y n , xn 3 ) = lim (y , x3 ) = (y ), n n
y n y,
xn 3 0,
xn 3 0 as n .
by the continuity of the mapping and thus y = y since the mapping is injective by assumption. But these properties contradict the local injectivity (noted above) of the mapping . Hence there exists > 0 such that is injective on the set = [, ]. In what follows, we assume that > 0 is small enough so that the conclusions of Theorem 4.1-1 hold. In particular then, (y1 , y2 , x3 ) constitutes a bona de system of curvilinear coordinates for describing the reference conguration () of the shell. Let 0 be a measurable subset of the boundary := . If length 0 > 0, we assume that the shell is subjected to a homogeneous boundary condition of place along the portion (0 [, ]) of its lateral face ( [, ]), which means that its displacement eld vanishes on the set (0 [, ]). The shell is subjected to applied body forces in its interior () and to applied surface forces on its upper and lower faces (+ ) and ( ), where := {}. The applied forces are given by the contravariant components (i.e., over the covariant bases g i = i ) f i L2 () and hi L2 (+ ) of their densities per unit volume and per unit area, respectively. We then dene functions pi L2 ( ) by letting pi :=

f i x3 + hi (, +) + hi (, ).
Sect. 4.1]
159
Finally, the elastic material constituting the shell is assumed to be homogeneous and isotropic and the reference conguration () of the shell is assumed to be a natural state. Hence the material is characterized by two Lam e constants and satisfying 3 + 2 > 0 and > 0. Such a shell, endowed with its natural curvilinear coordinates, namely the coordinates (y1 , y2 , x3 ) , can thus be modeled as a three-dimensional problem. According to Chapter 3, the corresponding unknowns are thus the three covariant components ui : R of the displacement eld ui g i : R of the points of the reference conguration (), where the vector elds g i i denote the contravariant bases (i.e., dened by the relations g i g j = j , where g j = j ; that these vector elds are well dened also follows from Theorem 4.1-1); cf. Figure 4.1-2. These unknowns then satisfy the equations of elasticity in curvilinear coordinates, as described in Sections 3.5 and 3.6.
g 3 (x)
ui (x)g i(x) g 2 (x)
g 1 (x)
(0)
S
x3 x y
Figure 4.1-2: An elastic shell modeled as a three-dimensional problem. Let = ], [. The set (), where (y, x3 ) = (y ) + x3 a3 (y ) for all x = (y, x3 ) , is the reference conguration of a shell, with thickness 2 and middle surface S = ( ) (Figure 4.1-1), which is subjected to a boundary condition of place along the portion (0 ) of its lateral face (i.e., the displacement vanishes on (0 )), where 0 = 0 [, ] and 0 = . The shell is subjected to applied body forces in its interior () and to applied surface forces on its upper and lower faces (+ ) and ( ) where = {}. Under the inuence of these forces, a point (x) undergoes a displacement ui (x)gi (x), where the three vectors gi (x) form the contravariant basis at the point (x). The unknowns of the problem are the three covariant components ui : R of the displacement eld ui gi : R3 of the points of (), which thus satisfy the boundary conditions ui = 0 on 0 . The objective consists in nding ad hoc conditions aording the replacement of this three-dimensional problem by a two-dimensional problem posed over the middle surface S if is small enough; see Figure 4.1-3. Note that, for the sake of visual clarity, the thickness is overly exaggerated.
160
[Ch. 4
In a two-dimensional approach , the above three-dimensional problem is replaced by a presumably much simpler two-dimensional problem, this time posed over the middle surface S of the shell . This means that the new unknowns should be now the three covariant components i : R of the displacement eld i ai : E3 of the points of the middle surface S = ( ); cf. Figure 4.1-3. During the past decades, considerable progress has been made towards a rigorous justication of such a replacement. The central idea is that of asymptotic analysis: It consists in showing that, if the data are of ad hoc orders of magnitude, the three-dimensional displacement vector eld (once properly scaled) converges in an appropriate function space as 0 to a limit vector eld that can be entirely computed by solving a two-dimensional problem. In this direction, see Ciarlet [2000a, Part A] for a thorough overview in the linear case and the key contributions of Le Dret & Raoult [1996] and Friesecke, James, Mora & M uller [2003] in the nonlinear case.
i(y )ai(y ) a3(y ) a2(y ) a1(y )
( 0)
y 0
Figure 4.1-3: An elastic shell modeled as a two-dimensional problem. For > 0 small enough and data of ad hoc orders of magnitude, the three-dimensional shell problem (Figure 4.1-2) is replaced by a two-dimensional shell problem. This means that the new unknowns are the three covariant components i : R of the displacement eld i ai : R3 of the points of the middle surface S = ( ). In this process, the three-dimensional boundary conditions on 0 need to be replaced by ad hoc two-dimensional boundary conditions on 0 . For instance, the boundary conditions of clamping i = 3 = 0 on 0 (used in Koiters linear equations; cf. Section 4.2) mean that the points of, and the tangent spaces to, the deformed and undeformed middle surfaces coincide along the set (0 ).
Sect. 4.1]
161
We now describe the nonlinear Koiter shell equations, so named after Koiter [1966], and since then a nonlinear model of choice in computational mechanics (its relation to an asymptotic analysis as 0 is briey discussed at the end of this section). Given an arbitrary displacement eld i ai : R3 of the surface S with smooth enough components i : R, dene the vector eld := (i ) : R3 and let a ( ) := a ( ) a ( ), where a ( ) := ( + i ai ), denote the covariant components of the rst fundamental form of the deformed surface ( + i ai )( ). Then the functions G ( ) := 1 (a ( ) a ) 2
denote the covariant components of the change of metric tensor associated with the displacement eld i ai of S . Remark. An easy computation, which simply relies on the formulas of Gau and Weingarten (Theorem 2.6-1), shows that G ( ) = where 1 ( 2
+ amn m n ),
a3 = a3 = 0 and a33 = 1
(otherwise the functions a denote as usual the contravariant components of the rst fundamental form of S ), and
:= b 3
and
:= 3 + b .
If the two vectors a ( ) are linearly independent at all points of , let b ( ) := where a( ) := det(a ( )), denote the covariant components of the second fundamental form of the deformed surface ( + i ai )( ). Then the functions R ( ) := b ( ) b denote the covariant components of the change of curvature tensor eld associated with the displacement eld i ai of S . Note that a( ) = |a1 ( ) a2 ( )|. 1 a( ) ( + i ai ) {a1 ( ) a2 ( )},
162
[Ch. 4
The nonlinear two-dimensional equations proposed by Koiter [1966] for modeling an elastic shell are derived from those of nonlinear three-dimensional elasticity on the basis of two a priori assumptions: One assumption, of a geometrical nature, is the Kirchho-Love assumption. It asserts that any point situated on a normal to the middle surface remains on the normal to the deformed middle surface after the deformation has taken place and that, in addition, the distance between such a point and the middle surface remains constant. The other assumption, of a mechanical nature, asserts that the state of stress inside the shell is planar and parallel to the middle surface (this second assumption is itself based on delicate a priori estimates due to John [1965, 1971]). Taking these a priori assumptions into account, W.T. Koiter then reached the conclusion that the unknown vector eld = (i ) should be a stationary point, in particular a minimizer, over a set of smooth enough vector elds = (i ) : R3 satisfying ad hoc boundary conditions on 0 , of the functional j dened by (cf. Koiter [1966, eqs. (4.2), (8.1), and (8.3)]): j ( ) = 1 2
a G ( )G ( ) +
3 a R ( )R ( ) a dy 3
pi i a dy,
where the functions a := 4 a a + 2(a a + a a ) + 2
denote the contravariant components of the shell elasticity tensor. The above functional j is called Koiters energy for a nonlinear elastic shell. Remark. The specic form of the functions a can be fully justied, in both the linear and nonlinear cases, by means of an asymptotic analysis of the solution of the three-dimensional equations as the thickness 2 approaches zero; see Ciarlet [2000a], Le Dret & Raoult [1996] and Friesecke, James, Mora & M uller [2003]. The stored energy function wK found in Koiters energy j is thus dened by wK ( ) = 3 a G ( )G ( ) + a R ( )R ( ) 2 6
for ad hoc vector elds . This expression is the sum of the membrane part wM ( ) = and of the exural part wF ( ) = 3 a R ( )R ( ). 6 a G ( )G ( ) 2
Sect. 4.2]
163
As hinted at earlier, the long-standing question of how to rigorously identify and justify the nonlinear two-dimensional equations of elastic shells from threedimensional elasticity was nally settled in two key contributions, one by Le Dret & Raoult [1996] and one by Friesecke, James, Mora & M uller [2003], who respectively justied the equations of a nonlinearly elastic membrane shell and those of a nonlinearly elastic exural shell by means of -convergence theory (a nonlinearly elastic shell is a membrane shell if there are no nonzero admissible displacements of its middle surface S that preserve the metric of S ; it is a exural shell otherwise). The stored energy function wM of a nonlinearly elastic membrane shell is an ad hoc quasiconvex envelope, which turns out to be only a function of the covariant components a ( ) of the rst fundamental form of the unknown deformed middle surface (the notion of quasiconvexity, which plays a central role in the calculus of variations, is due to Morrey [1952]; an excellent introduction to this notion is provided in Dacorogna [1989, Chapter 5]). The function wM reduces to the above membrane part wM in Koiters stored energy function wK only for a restricted class of displacement elds i ai of the middle surface. By contrast, the stored energy function of a nonlinearly elastic exural shell is always equal to the above exural part wF in Koiters stored energy function wK . Remark. Interestingly, a formal asymptotic analysis of the three-dimensional equations is only capable of delivering the above restricted expression wM ( ), but otherwise fails to provide the general expression, i.e., valid for all types of displacements, found by Le Dret & Raoult [1996]. By contrast, the same formal approach yields the correct expression wF ( ). For details, see Miara [1998], Lods & Miara [1998], and Ciarlet [2000a, Part B]. Another closely related set of nonlinear shell equations of Koiters type has been proposed by Ciarlet [2000b]. In these equations, the denominator a ( ) that appears in the functions R ( ) = b ( ) b is simply replaced by a, thereby avoiding the possibility of a vanishing denominator in the expression wK ( ). Then Ciarlet & Roquefort [2001] have shown that the leading term of a formal asymptotic expansion of a solution to this two-dimensional model, with the thickness 2 as the small parameter, coincides with that found by a formal asymptotic analysis of the three-dimensional equations. This result thus raises hopes that a rigorous justication, again by means of -convergence theory, of either types of nonlinear Koiters models might be possible.
164
[Ch. 4
4.2
THE LINEAR KOITER SHELL EQUATIONS
Consider the Koiter energy j for a nonlinearly elastic shell, dened by (cf. Section 4.1) j ( ) = 1 2
a G ( )G ( ) +
3 a R ( )R ( ) a dy 3
pi i a dy,
for smooth enough vector elds = (i ) : R3 . One of its virtues is that the integrands of the rst two integrals are quadratic expressions in terms of the covariant components G ( ) and R ( ) of the change of metric, and change of curvature, tensors associated with a displacement eld i ai of the middle surface S = ( ) of the shell. In order to obtain the energy corresponding to the linear equations of Koiter [1970], it thus suces to replace the covariant components 1 G ( ) = (a ( ) a ) and R ( ) = b ( ) b , 2 of these tensors by their linear parts with respect to = (i ), respectively denoted ( ) and ( ) below. Accordingly, our rst task consists in nding explicit expressions of such linearized tensors. To begin with, we compute the components ( ). A word of caution. The vector elds = (i ) and := i ai , which are both dened on , must be carefully distinguished! While the latter has an intrinsic character, the former has not; it only provides a means of recovering the eld via its covariant components i . Theorem 4.2-1. Let be a domain in R2 and let C 2 ( ; E3 ) be an immersion. Given a displacement eld := i ai of the surface S = ( ) with smooth enough covariant components i : R, let the function ( ) : R be dened by 1 ( ) := [a ( ) a ]lin , 2 where a and a ( ) are the covariant components of the rst fundamental form of the surfaces ( ) and ( + i ai )( ), and [ ]lin denotes the linear part with respect to = (i ) in the expression [ ]. Then ( ) = 1 ( a + a ) = ( ) 2 1 = (| + | ) b 3 2 1 = ( + ) b 3 , 2
Sect. 4.2]
The linear Koiter shell equations
165
where the covariant derivatives | are dened by | = (Theorem 2.6-1). In particular then, H 1 ( ) and 3 L2 ( ) ( ) L2 ( ). Proof. The covariant components a ( ) of the metric tensor of the surface ( + i ai )( ) are by denition given by a ( ) = ( + ) ( + ). Note that both surfaces ( ) and ( + i ai )( ) are thus equipped with the same curvilinear coordinates y . The relations ( + ) = a + then show that a ( ) = (a + ) (a + ) = a + a + a + , hence that ( ) = 1 1 [a ( ) a ]lin = ( a + a ). 2 2
The other expressions of ( ) immediately follow from the expression of = (i ai ) given in Theorem 2.6-1. The functions ( ) are called the covariant components of the linearized change of metric tensor associated with a displacement i ai of the surface S . We next compute the components ( ). Theorem 4.2-2. Let be a domain in R2 and let C 3 ( ; E3 ) be an immersion. Given a displacement eld := i ai of the surface S = ( ) with smooth enough and small enough covariant components i : R, let the functions ( ) : R be dened by ( ) := [b ( ) b ]lin , where b and b ( ) are the covariant components of the second fundamental form of the surfaces ( ) and ( + i ai )( ), and [ ]lin denotes the linear part with respect to = (i ) in the expression [ ]. Then ( ) = ( ) a3 = ( )
= 3| b b 3 + b | + b | + b | = 3 3 b b 3 +b ( ) + b ( ) +( b + b b ) ,
166
[Ch. 4
where the covariant derivatives | are dened by | = (Theorem 2.6-1) and

3| := 3 3 and b | := b + b b .
In particular then, H 1 ( ) and 3 H 2 ( ) ( ) L2 ( ). The functions b | satisfy the symmetry relations

b | = b | .
Proof. For convenience, the proof is divided into ve parts. In parts (i) and (ii), we establish elementary relations satised by the vectors ai and ai of the covariant and contravariant bases along S . (i) The two vectors a = satisfy |a1 a2 | = a, where a = det(a ). Let A denote the matrix of order three with a1 , a2 , a3 as its column vectors. Consequently, det A = (a1 a2 ) a3 = (a1 a2 ) Besides, a1 a2 = |a1 a2 |. |a1 a2 |
(det A)2 = det(AT A) = det(a ) = a, since a a = a and a a3 = 3 . Hence |a1 a2 | = a. 2 (ii)1The vectors ai and a are related by a1 a3 = aa and a3 a2 = aa . To prove that two vectors c and d coincide, it suces to prove that c ai = d ai for i {1, 2, 3}. In the present case, (a1 a3 ) a1 = 0 and (a1 a3 ) a3 = 0, (a1 a3 ) a2 = (a1 a2 ) a3 = a,
aa3 = a1 a2 by (i), on the one hand; on the other hand, aa2 a1 = aa2 a3 = 0 and aa2 a2 = a, i . Hence a1 a3 = aa2 . The other relation is similarly since ai aj = j established. since (iii) The covariant components b ( ) satisfy b ( ) = b + ( ) a3 + h.o.t., where h.o.t. stands for higher-order terms, i.e., terms of order higher than linear with respect to = (i ). Consequently, ( ) := [b ( ) b ]lin = a3 = ( ).
Sect. 4.2]
167
Since the vectors a = are linearly independent in and the elds = (i ) are smooth enough by assumption, the vectors ( + i ai ) are also linearly independent in provided the elds are small enough, e.g., with respect to the norm of the space C 1 ( ; R3 ). The following computations are therefore licit as they apply to a linearization around = 0. Let a ( ) := ( + ) = a + and a3 ( ) := where a( ) := det(a ( )) and a ( ) := a ( ) a ( ). Then b ( ) = a ( ) a3 ( ) = = + 1 a( ) ( a + ) (a1 a2 + a1 2 + 1 a2 + h.o.t.) a1 ( ) a2 ( ) a( ) ,
1 a(b + a3 ) a( ) 1 a( ) ( a + b a3 ) (a1 2 + 1 a2 ) + h.o.t. ,
since b = a a3 and a = a + b a3 by the formula of Gau (Theorem 2.6-1 (a)). Next, ( a + b a3 ) (a1 2 ) = 2 a2 (a1 2 ) b 2 (a1 a3 ) 2 , = a 2 2 a3 + b 2 a since, (ii), a2 (a1 2 ) = 2 (a1 a2 ) = a2 a3 and a1 a3 = by aa2 ; likewise, 1 ( a(1 a + b a3 ) (1 a2 ) = 1 a3 + b 1 a ). Consequently, b ( ) = a a ) + ( b (1 + ) a3 + h.o.t. . a( )
There remains to nd the linear term with respect to = (i ) in the expan1 1 = (1 + ). To this end, we note that sion a a( ) det(A + H) = (det A)(1 + tr A1 H + o(H)),
168
[Ch. 4
with A := (a ) and A + H := (a ( )). Hence H = ( a + a + h.o.t.) , since [a ( ) a ]lin = a + a (Theorem 4.2-1). Therefore, a( ) = det(a ( )) = det(a )(1 + 2 a + h.o.t.), since A1 = (a ); consequently, 1 a( ) 1 = (1 a + h.o.t.). a
Noting that there are no linear terms with respect to = (i ) in the product (1 a )(1 + a ), we nd the announced expansion, viz., b ( ) = b + ( ) a3 + h.o.t. (iv) The components ( ) can be also written as
( ) = 3| b b 3 + b | + b | + b | ,
where the functions 3| and b | are dened as in the statement of the theorem. By Theorem 2.6-1 (b),
3 = ( b 3 )a + ( 3 + b )a .
Hence
a3 = ( 3 + b ), i since ai a3 = 3 . Again by Theorem 2.6-1 (b), a3 = ( b 3 )a 3 +( 3 + b )a } a3 = ( b 3 ) a a3 3 +( 3 + ( b ) + b )a a3 3 +( 3 + b ) a a3 = b ( ) b b 3 + 3 +( b ) + b ,
since
3 a a3 = ( a + b a ) a3 = b ,
a3 a3 = b a a3 = 0,
Sect. 4.2]
169
by the formulas of Gau and Weingarten (Theorem 2.6-1 (a)). We thus obtain ( ) = ( ) a3
= b ( ) b b 3 + 3 + ( b ) + b ( 3 + b ).
While this relation seemingly involves only the covariant derivatives 3| and | , it may be easily rewritten so as to involve in addition the functions | and b | . The stratagem simply consists in using the relation b b = 0! This gives
( ) = ( 3 3 ) b b 3 +b ( ) + b ( ) +( b + b b b ) .
(v) The functions b | are symmetric with respect to the indices and . Again, because of the formulas of Gau and Weingarten, we can write
3 3 0 = a a = a + b a a + b a 3 3 = ( )a + a b a + ( b )a b b a 3 3 +( )a a + b a ( b )a + b b a .
Consequently,
0 = ( a a ) a3 = b b + b b ,
on the one hand. On the other hand, we immediately infer from the denition of the functions b | that we also have
b | b | = b b + b b ,
and thus the proof is complete. The functions ( ) are called the covariant components of the linearized change of curvature tensor associated with a displacement i ai of the surface S . The functions
3| = 3 3 and b | = b + b b
respectively represent a second-order covariant derivative of the vector eld i ai and a rst-order covariant derivative of the second fundamental form of S , dened here by means of its mixed components b . Remarks. (1) Covariant derivatives b | can be likewise dened. More specically, each function
b | := b b b
170
[Ch. 4
represents a rst-order covariant derivatives of the second fundamental form, dened here by means of its covariant components b . By a proof analogous to that given in Theorem 4.2-2 for establishing the symmetry relations b | = b | , one can then show that these covariant derivatives likewise satisfy the symmetry relations b | = b| , which are themselves equivalent to the relations
b b + b b = 0,
i.e., the familiar Codazzi-Mainardi equations (Theorem 2.7-1)! (2) The functions c := b b = c appearing in the expression of ( ) are the covariant components of the third fundamental form of S . For details, see, e.g., Stoker [1969, p. 98] or Klingenberg [1973, p. 48]. (3) The functions b ( ) are not always well dened (in order that they be, the vectors a ( ) must be linearly independent in ), but the functions ( ) are always well dened. (4) The symmetry ( ) = ( ) follows immediately by inspection of the expression ( ) = ( ) a3 found there. By contrast, deriving the same symmetry from the other expression of ( ) requires proving rst that the covariant derivatives b | are themselves symmetric with respect to the indices and (cf. part (v) of the proof of Theorem 4.2-1). While the expression of the components ( ) in terms of the covariant components i of the displacement eld is fairly complicated but well known (see, e.g., Koiter [1970]), that in terms of = i ai is remarkably simple but seems to have been mostly ignored, although it already appeared in Bamberger [1981]. Together with the expression of the components ( ) in terms of (Theorem 4.2-1), this simpler expression was eciently put to use by Blouza & Le Dret [1999], who showed that their principal merit is to aord the denition of the components ( ) and ( ) under substantially weaker regularity assumptions on the mapping . More specically, we were led to assume that C 3 ( ; E3 ) in Theorem 4.2-1 in order to insure that ( ) L2 ( ) if H 1 ( ) H 1 ( ) H 2 ( ). The culprits responsible for this regularity are the functions b | appearing in the functions ( ). Otherwise Blouza & Le Dret [1999] have shown how this regularity assumption on can be weakened if only the expressions of ( ) and ( ) in terms of the eld are considered. We shall return to such aspects in Section 4.3. We are now in a position to describe the linear Koiter shell equations. Let 0 be a measurable subset of = that satises length 0 > 0, let denote the outer normal derivative operator along , and let the space V( ) be dened by V( ) := = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on 0 . Then the unknown vector eld = (i ) : R3 , where the functions i are the covariant components of the displacement eld i ai of the middle surface
Sect. 4.2]
171
S = ( ) of the shell, should be a stationary point over the space V( ) of the functional j dened by j ( ) = 1 2
a ( ) ( ) +
3 a ( ) ( ) a dy 3
pi i a dy
for all V( ). This functional j is called Koiters energy for a linearly elastic shell. Equivalently, the vector eld V( ) should satisfy the variational equations 3 a ( ) ( ) + a ( ) ( ) a dy 3 = pi i a dy for all = (i ) V( ).
We recall that the functions a := 4 a a + 2(a a + a a ) + 2
denote the contravariant components of the shell elasticity tensor ( and are the Lam e constants of the elastic material constituting the shell), ( ) and ( ) denote the covariant components of the linearized change of metric, and change of curvature, tensors associated with a displacement eld i ai of S , and the given functions pi L2 ( ) account for the applied forces. Finally, the boundary conditions i = 3 = 0 on 0 express that the shell is clamped along the portion (0 ) of its middle surface (see Figure 4.1-3). The choice of the function spaces H 1 ( ) and H 2 ( ) for the tangential components and normal components 3 of the displacement elds i ai is guided by the natural requirement that the functions ( ) and ( ) be both in L2 ( ), so that the energy is in turn well dened for = (i ) V( ). Otherwise these choices can be weakened to accommodate shells whose middle surfaces have little regularity (cf. Section 4.3). Remark. A justication of Koiters linear equations (by means of an asymptotic analysis of the three-dimensional equations as 0) is provided in Section 4.5. Our objective in the next sections is to study the existence and uniqueness of the solution to the above variational equations. To this end, we shall establish (Theorem 4.4-1) that, under the assumptions 3 + 2 > 0 and > 0, there exists a constant ce > 0 such that |t |2 ce a (y )t t
,
172
[Ch. 4
for all y and all symmetric matrices (t ). When length 0 > 0, the existence and uniqueness of a solution to this variational problem by means of the LaxMilgram lemma will then be a consequence of the existence of a constant c such that
2 1,
+ 3
2 2,
1/2
c
,
( )
2 0,
+
,
( )
2 0,
1/2
for all V( ).
The objective of the next section precisely consists in showing that such a fundamental Korns inequality on a surface indeed holds (Theorem 4.3-4).
4.3
KORNS INEQUALITIES ON A SURFACE
In Section 3.8, we established three-dimensional Korn inequalities, rst without boundary conditions (Theorem 3.8-1), then with boundary conditions (Theorem 3.8-3), the latter depending on an innitesimal rigid displacement lemma (Theorem 3.8-2). Both Korn inequalities involved the covariant components eij (v ) of the three-dimensional linearized change of metric tensor. But while this tensor is the only one that is attached to a displacement vi g i of the three-dimensional set () in E3 , we saw in the previous section that two tensors, the linearized change of metric and the linearized change of curvature tensors, are attached to a displacement eld i ai of a surface in E3 . It is thus natural to seek to likewise establish Korns inequalities on a surface, rst without boundary conditions (Theorem 4.3-1), then with boundary conditions (Theorem 4.3-4), the latter again depending on an innitesimal rigid displacement lemma on a surface (Theorem 4.3-3). As expected, such inequalities will now involve the covariant components ( ) and ( ) of the linearized change of metric tensor and linearized change of curvature tensor dened in the previous section. The innitesimal rigid displacement lemma and the Korn inequality with boundary conditions were rst established by Bernadou & Ciarlet [1976]. A simpler proof, which we follow here, was then proposed by Ciarlet & Miara [1992] (see also Bernadou, Ciarlet & Miara [1994]). Its rst stage consists in establishing a Korns inequality on a surface, without boundary conditions, again as a consequence of the same lemma of J.L. Lions as in dimension three (cf. Theorem 3.8-1). Note that such an inequality holds as well in the more general context of Riemannian geometry; cf. Chen & Jost [2002]. Recall that the notations 0, and m, respectively designate the norms in L2 ( ) and H m ( ), m 1; cf. Section 3.6.
Sect. 4.3]
Korns inequalities on a surface
173
Theorem 4.3-1. Let be a domain in R2 and let C 3 (; E3 ) be an injective immersion. Given = (i ) H 1 ( ) H 1 ( ) H 2 ( ), let ( ) := 1 ( a + a ) L2 ( ), 2
2 ( ) := ( ) a3 L ( )
denote the covariant components of the linearized change of metric, and linearized change of curvature, tensors associated with the displacement eld := i ai of the surface S = (). Then there exists a constant c0 = c0 (, ) such that
2 1,
+ 3
1
2 2, 2 0,
1/2
c0
+ 3
1
2 1,
+
, 2
( )
2 0,
+
,
( )
2 0,
1/2
for all = (i ) H ( ) H ( ) H ( ). Proof. The fully explicit expressions of the functions ( ) and ( ), as found in Theorems 4.2-1 and 4.2-2, are used in this proof, simply because they are more convenient for its purposes. (i) Dene the space W( ) := = (i ) L2 ( ) L2 ( ) H 1 ( ); ( ) L2 ( ), ( ) L2 ( ) . Then, equipped with the norm
W ( ) W ( ) 2 1,
dened by ( )
, 2 0,
:=
2 0,
+ 3
+
,
( )
2 0,
1/2
the space W( ) is a Hilbert space. The relations ( ) L2 ( ) and ( ) L2 ( ) appearing in the denition of the space W( ) are to be understood in the sense of distributions. They mean that a vector eld L2 ( ) L2 ( ) H 1 ( ) belongs to W( ) if there exist functions in L2 ( ), denoted ( ) and ( ), such that for all D( ), ( ) dy =

1 ( + ) + + b 3 dy, 2
3 + 3 + b b 3 + (b ) + b + (b ) + b b + b b dy.
( ) dy =

174
[Ch. 4
k k Let there be given a Cauchy sequence ( k ) k=1 with elements = (i ) W( ). The denition of the norm W() shows that there exist L2 ( ), 3 H 1 ( ), L2 ( ), and L2 ( ) such that k in L2 ( ), ( k ) in L2 ( ), k 3 3 in H 1 ( ), ( k ) in L2 ( )
as k . Given a function D( ), letting k in the relations ( k ) d = . . . and ( k ) d = . . . then shows that = ( ) and = ( ). (ii) The spaces W( ) and H 1 ( ) H 1 ( ) H 2 ( ) coincide. Clearly, H 1 ( ) H 1 ( ) H 2 ( ) W( ). To prove the other inclusion, let = (i ) W( ). The relations s ( ) := 1 ( + ) = ( ) + + b 3 2
then imply that e ( ) L2 ( ) since the functions and b are continuous on . Therefore, H 1 ( ), ( ) = { s ( ) + s ( ) s ( )} H 1 ( ), since L2 ( ) implies H 1 ( ). Hence L2 ( ) by the lemma of J.L. Lions (Theorem 3.7-1) and thus H 1 ( ). The denition of the functions ( ), the continuity over of the func 2 tions , b , b , and b , and the relations ( ) L ( ) then imply that 2 2 3 L ( ), hence that 3 H ( ). (iii) Korns inequality without boundary conditions. The identity mapping from the space H 1 ( )H 1 ( )H 2 ( ) equipped with 2 1/2 its product norm = (i ) { 2 into the space W( ) 1, + 3 2, } equipped with W() is injective, continuous, and surjective by (ii). Since both spaces are complete (cf. (i)), the closed graph theorem then shows that the inverse mapping 1 is also continuous or equivalently, that the inequality of Korns type without boundary conditions holds. In order to establish a Korns inequality with boundary conditions, we have to identify classes of boundary conditions to be imposed on the elds = (i ) H 1 ( ) H 1 ( ) H 2 ( ) in order that we can get rid of the norms 0, and 3 1, in the right-hand side of the above inequality, i.e., situations where the semi-norm = (i )
,
( )
2 0,
+
,
( )
2 0,
1/2
becomes a norm, which should be in addition equivalent to the product norm.
Sect. 4.3]
175
To this end, the rst step consists in establishing in Theorem 4.3-3 an innitesimal rigid displacement lemma, which provides in particular one instance of boundary conditions implying that this semi-norm becomes a norm. The proof of this lemma relies on the preliminary observation, quite worthwhile per se, that a vector eld i ai on a surface may be canonically extended to a three-dimensional vector eld vi g i in such a way that all the components eij (v ) of the associated three-dimensional linearized change of metric tensor have remarkable expressions in terms of the components ( ) and ( ) of the linearized change of metric and curvature tensors of the surface vector eld. Theorem 4.3-2. Let be a domain in R2 and let C 3 (; E3 ) be an injective immersion. By Theorem 4.1-1, there exists > 0 such that the mapping dened by (y, x3 ) := (y ) + x3 a3 (y ) for all (y, x3 ) , where := ], [, is a C 2 -dieomorphism from onto () and thus the three vectors g i := i are linearly independent at all points of . With any vector eld i ai with covariant components in H 1 ( ) and 3 in H 2 ( ), let there be associated the vector eld vi g i dened on by
vi (y, x3 )g i (y, x3 ) = i (y )ai (y ) x3 ( 3 + b )(y )a (y ) i for all (y, x3 ) , where the vectors g i are dened by g i g j = j .
Then the covariant components vi of the vector eld vi g i are in H 1 () and the covariant components eij (v ) L2 () of the associated linearized change of metric tensor (Section 3.6) are given by e (v ) = ( ) x3 ( ) + ei3 (v ) = 0. x2 3 b ( ) + b ( ) 2b b ( ) , 2
Proof. As in the above expressions of the functions e (v ), the dependence on x3 is explicit, but the dependence with respect to y is omitted, throughout the proof. The explicit expressions of the functions ( ) and ( ) in terms of the functions i (Theorems 4.2-1 and 4.2-2) are used in this proof. (i) Given functions , X H 1 ( ) and 3 H 2 ( ), let the vector eld vi g i be dened on by vi g i = i ai + x3 X a .
176
[Ch. 4
Then the functions vi are in H 1 () and the covariant components ei j (v ) of the linearized change of metric tensor associated with the eld vi g i are given by 1 (| + | ) b 3 e (v ) = 2 x3 + X| + X | b (| b 3 ) b ( | b 3 ) 2 x2 + 3 b X| b X | , 2 1 e3 (v ) = (X + 3 + b ), 2 e33 (v ) = 0,
where | = and X| = X X designate the covariant i i derivatives of the elds i a and Xi a with X3 = 0 (Section 2.6). Since a3 = b a
by the second formula of Weingarten (Theorem 2.6-1), the vectors of the covariant basis associated with the mapping = + x3 a3 are given by g = a x3 b a and g 3 = a3 . The assumed regularities of the functions i and X imply that vi = (vj g j ) g i = (j aj + x3 X a ) g i H 1 () since g i C1 (). The announced expressions for the functions eij (v ) are obtained by simple computations, based on the relations vi j = {j (vk g k )} g i (Theorem 1.4-1) and eij (v ) = 1 2 (vi j + vj i ). (ii) When X = ( 3 + b ), the functions eij (v ) in (i) take the expressions announced in the statement of the theorem. 1 We rst note that X H 1 ( ) (since b C ( )) and that e3 (v ) = 0 when X = ( 3 + b ). It thus remains to nd the explicit forms of the functions e (v ) in this case. Replacing the functions X by their expressions and using the symmetry relations b | = b | (Theorem 4.2-2), we nd that 1 X| + X | b (| b 3 ) b ( | b 3 ) 2 = 3| b | b | b | + b b 3 , i.e., the factor of x3 in e (v ) is equal to ( ). Finally,
b X| b X | = b 3| + b| + b | + b 3| + b | + b | = b ( ) b | + b b 3 + b ( ) b | + b b 3 = b ( ) + b ( ) 2b b ( ),
i.e., the factor of
x2 3 2
in e (v ) is that announced in the theorem.
Sect. 4.3]
177
As shown by Destuynder [1985, Theorem 3.1] (see also Ciarlet & S. Mardare [2001]), the mapping F : (i ) H 1 ( ) H 1 ( ) H 2 ( ) (vi ) H1 () dened in Theorem 4.3-2 is in fact an isomorphism from the Hilbert space H 1 ( ) H 1 ( ) H 2 ( ) onto the Hilbert space VKL () = {v H1 (); ei3 (v ) = 0 in }. This identication of the image Im F as the space VKL () has an interest per se in linearized shell theory. This result shows that, inside an elastic shell, the Kirchho-Love displacement elds, i.e. those displacement elds vi g i that satisfy the relations ei3 (v ) = 0 in , are of the form
1 2 vi g i = i ai x3 ( 3 + b )a with H ( ) and 3 H ( ),
and vice versa. This identication thus constitutes an extension of the wellknown identication of Kirchho-Love displacement elds inside an elastic plate (cf. Ciarlet & Destuynder [1979] and also Theorem 1.4-4 of Ciarlet [1997]). We next establish an innitesimal rigid displacement lemma on a surface. The adjective innitesimal reminds that only the linearized parts ( ) and ( ) of the full change of metric and curvature tensors 1 2 (a ( ) a ) and (b ( ) b ) are required to vanish in . Thanks to Theorem 4.3-2, this lemma becomes a simple consequence of the three-dimensional innitesimal rigid displacement lemma in curvilinear coordinates (Theorem 3.8-2), to which it should be protably compared. This lemma is due to Bernadou & Ciarlet [1976, Theorems 5.1-1 and 5.2-1], who gave a more direct, but less transparent, proof (see also Bernadou [1994, Part 1, Lemma 5.1.4]). Part (a) in the next theorem is an innitesimal rigid displacement lemma on a surface, without boundary conditions, while part (b) is an innitesimal rigid displacement lemma on a surface, with boundary conditions. Theorem 4.3-3. Let there be given a domain in R2 and an injective immersion C 3 ( ; E3 ). (a) Let = (i ) H 1 ( ) H 1 ( ) H 2 ( ) be such that ( ) = ( ) = 0 in . Then there exist two vectors a, b R3 such that i (y )ai (y ) = a + b (y ) for all y . (b) Let 0 be a d -measurable subset of = that satises length 0 > 0 and let a vector eld = (i ) H 1 ( ) H 1 ( ) H 2 ( ) be such that ( ) = ( ) = 0 in and i = 3 = 0 on 0 . Then = 0 in .
178
[Ch. 4
Proof. Let the set = ], [ and the vector eld v = (vi ) H1 () be dened as in Theorem 4.3-2. By this theorem, ( ) = ( ) = 0 in implies that eij (v ) = 0 in .
Therefore, by Theorem 3.8-2 (a), there exist two vectors a, b R3 such that vi (y, x3 )g i (y, x3 ) = a + b {(y ) + x3 a3 (y )} for all (y, x3 ) . Hence i (y )ai (y ) = vi (y, x3 )g i (y, x3 )|x3 =0 = a + b (y ) for all y , and part (a) is established. Let 0 be such that length 0 > 0. If i = 3 = 0 on 0 , the functions ( 3 + b ) vanish on 0 , since 3 = 3 = 0 on 0 implies 3 = 0 on 0 . Theorem 4.3-2 then shows that vi = (vj g j ) g i = (j aj + x3 X a ) g i = 0 on 0 := 0 [, ] . Since area 0 > 0, Theorem 3.8-2 (b) implies that v = 0 in , hence that = 0 on . Remark. If a eld = (i ) H 1 ( ) H 1 ( ) H 2 ( ) satises ( ) = ( ) = 0 in , its three components i are automatically in C 2 ( ) since i = (j aj ) ai and the elds ai are of class C 2 on . Remarkably, the eld i ai = a + b inherits in this case even more regularity, as it is of class C 3 on . An innitesimal rigid displacement i ai of the surface S = ( ) is dened as one whose associated vector eld = (i ) H 1 ( ) H 1 ( ) H 2 ( ) satises ( ) = ( ) = 0 in . The vector elds associated with such an innitesimal rigid displacement thus span the vector space Rig( ) := { H 1 ( ) H 1 ( ) H 2 ( ); ( ) = ( ) = 0 in }, which, because of Theorem 4.3-3, is also given by Rig( ) = { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i ai = a + b for some a, b R3 }. This relation shows in particular that the innitesimal rigid displacements of the surface S span a vector space of dimension six. Furthermore, Ciarlet & C. Mardare [2004a] have shown that this vector space is precisely the tangent space at the origin to the manifold (also of dimension six) formed by the rigid displacements of the surface S = ( ), i.e., those whose associated deformed surface ( + i ai )( ) is obtained by means of a rigid deformation of the surface ( ) (see Section 2.9). In other words, this result shows that an innitesimal rigid
Sect. 4.3]
179
displacement of S is indeed the linearized part of a genuine rigid displacement of S , thereby fully justifying the use of the adjective innitesimal. We are now in a position to prove the announced Korns inequality on a surface, with boundary conditions. This inequality plays a fundamental r ole in the analysis of linearly elastic shells, in particular for establishing the existence and uniqueness of the solution to the linear Koiter equations (see Theorem 4.4-2). This inequality was rst proved by Bernadou & Ciarlet [1976]. It was later given other proofs by Ciarlet & Miara [1992] and Bernadou, Ciarlet & Miara [1994]; then by Akian [2003] and Ciarlet & S. Mardare [2001], who showed that it can be directly derived from the three-dimensional Korn inequality in curvilinear coordinates, with boundary conditions (Theorem 3.8-3), for ad hoc choices of set , mapping , and vector elds v (this idea goes back to Destuynder [1985]); then by Blouza & Le Dret [1999], who showed that it still holds under a less stringent smoothness assumption on the mapping . We follow here the proof of Bernadou, Ciarlet & Miara [1994]. Theorem 4.3-4. Let be a domain in R2 , let C 3 ( ; E3 ) be an injective immersion, let 0 be a d -measurable subset of = that satises length 0 > 0, and let the space V( ) be dened as: V( ) := { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on 0 }. Given = (i ) H 1 ( ) H 1 ( ) H 2 ( ), let ( ) := 1 ( a + a 2 L2 ( ),
2 ( ) := ( ) a3 L ( )
denote the covariant components of the linearized change of metric and linearized change of curvature tensors associated with the displacement eld := i ai of the surface S = ( ). Then there exists a constant c = c(, 0 , ) such that
2 1,
+ 3
2 2,
1/2
c
,
( )
2 0,
+
,
( )
2 0,
1/2
for all = (i ) V( ). Proof. Let

H 1 ( ) H 1 ( ) H 2 ( )
:=
2 1,
+ 3
2 2,
1/2
If the announced inequality is false, there exists a sequence ( k ) k=1 of vector elds k V( ) such that k
k H 1 ( ) H 1 ( ) H 2 ( )
= 1 for all k, ( k )
2 0, 1/2
lim
( k )
,
2 0,
+
,
= 0.
180
[Ch. 4
1 1 2 Since the sequence ( k ) k=1 is bounded in H ( ) H ( ) H ( ), a sub2 2 1 sequence ( ) converges in L ( ) L ( ) H ( ) by the Rellich-Kondra sov =1 theorem. Furthermore, each sequence ( ( )) and ( ( )) also con =1 =1 verges in L2 ( ) (to 0, but this information is not used at this stage) since
lim
( )
,
2 0,
+
,
( )
2 0,
1/2
= 0.
The subsequence ( ) =1 is thus a Cauchy sequence with respect to the norm
2 0,
+ 3
2 1,
+
,
( )
2 0,
+
,
| ( )|2 0,
1/2
hence with respect to the norm H 1 ()H 1 ()H 2 () by Korns inequality without boundary conditions (Theorem 4.3-1). The space V( ) being complete as a closed subspace of the space H 1 ( ) 1 H ( ) H 2 ( ), there exists V( ) such that in H 1 ( ) H 1 ( ) H 2 ( ), and the limit satises ( ) ( )
0, 0,
= lim ( )

0, 0,
= 0, = 0.
= lim ( )
Hence = 0 by Theorem 4.3-3. But this last relation contradicts the relations H 1 ()H 1 ()H 2 () = 1 for all 1, and the proof is complete. If the mapping is of the form (y1 , y2 ) = (y1 , y2 , 0) for all (y1 , y2 ) , the inequality of Theorem 4.3-4 reduces to two distinct inequalities (obtained by letting rst = 0, then 3 = 0): 3
2,
c
,
2 0,
1/2
for all 3 H 2 ( ) satisfying 3 = 3 = 0 on 0 , and

2 1, 1/2
1 ( + ) 2
2 0,
1/2
for all H1 ( ) satisfying = 0 on 0 . The rst inequality is a well-known property of Sobolev spaces. The second inequality is the two-dimensional Korn inequality in Cartesian coordinates. Both play a central r ole in the existence theory for linear two-dimensional plate equations (see, e.g., Ciarlet [1997, Theorems 1.5-1 and 1.5-2]). As shown by Blouza & Le Dret [1999], Le Dret [2004], and Anicic, Le Dret & Raoult [2005], the regularity assumptions made on the mapping and on
Sect. 4.3]
181
the eld = (i ) in both the innitesimal rigid displacement lemma and the Korn inequality on a surface of Theorems 4.3-3 and 4.3-4 can be substantially weakened. This improvement relies on the observation that the fully explicit expressions of the covariant components of the linearized change of metric and change of curvature tensors that have been used in the proofs of these theorems, viz., ( ) = and
( ) = 3 3 b b 3 + b ( ) + b ( ) + ( b + b b ) ,
1 ( + ) b 3 2
can be advantageously replaced by expressions such that ( ) = and ( ) = ( ) a3 , or ( ) = a ( a3 )a a3 , in terms of the eld := i ai . Note in passing that this last expression no longer involves Christoel symbols. The interest of such expressions is that they still dene bona de distributions under signicantly weaker smoothness assumptions than those of Theorem 4.3-4, viz., C 3 ( ; E3 ) and = (i ) H 1 ( ) H 1 ( ) H 2 ( ). For instance, it is easily veried that ( ) L2 ( ) and ( ) H 1 ( ) if W 2, ( ; E3 ) and H1 ( ); or that ( ) L2 ( ) and ( ) L2 ( ) if W 2, ( ; E3 ) and H1 () and a3 L2 ( ). This approach clearly widens the class of shells that can be modeled by Koiters linear equations, since discontinuities in the second derivatives of the mapping are allowed, provided these derivatives stay in L ( ). For instance, it aords the consideration of a shell whose middle surface is composed of a portion of a plane and a portion of a circular cylinder meeting along a segment and having a common tangent plane along this segment. We continue our study of Korns inequalities on a surface by showing that the Korn inequality without boundary conditions (Theorem 4.3-1) is equivalent to yet another Korns inequality on a surface, over the quotient space 1 ( a + a ) 2
182
[Ch. 4
H 1 ( ) H 1 ( ) H 2 ( )/ Rig( ). As we shall see, this inequality is the key to the existence theory for the pure traction problem for a shell modeled by the linear Koiter equations (cf. Theorem 4.4-3). We recall that the vector space Rig( ) = { H 1 ( ) H 1 ( ) H 2 ( ); ( ) = ( ) = 0 in } denotes the space of vector elds = (i ) H 1 ( ) H 1 ( ) H 2 ( ) whose associated displacement elds i ai constitute the innitesimal rigid displacements of the surface S and that the space Rig( ) is of dimension six (Theorem 4.3-3). designates the equivalence class of an In the next theorem, the notation element H 1 ( ) H 1 ( ) H 2 ( ) in the quotient space H 1 ( ) H 1 ( ) H 2 ( )/ Rig( ). In other words, := { H 1 ( ) H 1 ( ) H 2 ( ); ( ) Rig( )}. Theorem 4.3-5. Let be a domain in R2 and let C 3 ( ; E3 ) be an injective immersion. Dene the quotient space ( ) := (H 1 ( ) H 1 ( ) H 2 ( ))/ Rig( ), V which is a Hilbert space, equipped with the quotient norm
( ) V () V
dened by
:=
Rig( )
inf
H 1 ( ) H 1 ( ) H 2 ( )
V( ). for all
Then there exists a constant c =c (, ) such that

( ) V
c
,
) (
2 0,
+
,
) (
2 0,
1/2
( ). V for all
( ) is equivalent Moreover, this Korn inequality over the quotient space V to the Korn inequality without boundary condition of Theorem 4.3-1. Proof. To begin with, we observe that, thanks to the denition of the space ) L2 ( ) and ( ) L2 ( ) are unambiguously Rig( ), the functions ( ) = ( ) and ( ) = ( ) for any . In this dened, viz., as ( proof, we let V( ) := H 1 ( ) H 1 ( ) H 2 ( ) and for the sake of notational conciseness. (i) We rst show that the Korn inequality without boundary conditions (Theorem 4.3-1) implies the announced Korn inequality over the quotient space ( ). V
V ( )
2 1,
+ 3
2 2,
1/2
Sect. 4.3]
183
By the Hahn-Banach theorem, there exist six continuous linear forms on the space V( ), 1 6, with the following property: A vector eld Rig( ) is equal to 0 if and only if ( ) = 0, 1 6. It thus suces to show that there exists a constant c such that
V ( )
c
,
( )
2 0,
+
,
( )
2 0,
1/2
+
=1
| ( )|
for all V( ). For, given any V( ), let ( ) Rig( ) be dened by the relations ( + ( )) = 0, 1 6. The above inequality then implies that, ( ), V for all
( ) V
+ ( )
V ( )
c
,
( )
2 0,
+
,
( )
2 0,
1/2
Assume that there does not exist such a constant c . Then there exist k k V( ), k 1, such that V() = 1 for all k 1, ( )
, k 2 0,
+
,
( )
2 0,
1/2
+
=1
| ( k )|
0.
By Rellich theorem, there thus exists a subsequence ( ) =1 that converges in the space L2 ( ) L2 ( ) H 1 ( ) on the one hand; on the other hand, 2 each subsequence ( ( )) =1 and ( ( )) =1 converges in the space L ( ). Therefore, the subsequence ( ) =1 is a Cauchy sequence with respect to the norm = (i ) V( )
2 0, +
2 1, + ,
( )
2 0, + ,
( )
2 0,
1/2
hence also with respect to the norm V() by Korns inequality without boundary conditions. Consequently, there exists V( ) such that V() 0. But
then = 0, since ( ) = 0 and ( ) = ( ) = 0 in , in contradiction with the relations V() = 1 for all 1. (ii) We next show that, conversely, the Korn inequality over the quotient ( ) implies the Korn inequality without boundary condition of Thespace V orem 4.3-1. Assume that this Korn inequality does not hold. Then there exist k = k ) V( ), k 1, such that (i k
k 2 0, k + 3 2 1, V ( )
= 1 for all k 1,
2 0,
+
,
( k )
+
,
( k )
2 0,
1/2
0.
184
[Ch. 4
Let k Rig( ) denote for each k 1 the projection of k on Rig( ) with respect to the inner-product of the space V( ). This projection thus satises k k k
V ( ) V ( )
Rig( )
inf
k +
V ( )
V ( )
k =
V ( ) .
( ) , V
= k k
+ k
The space Rig( ) being nite-dimensional, the inequalities k V() 1 for all k 1 imply the existence of a subsequence ( ) =1 that converges in the space V( ) to an element = (i ) Rig( ). Besides, Korns inequality in the ( ) obtained in part (i) implies that quotient space V since ( )
, 2 0, V ( )
( ) V
0,
1/2
+
,
( )
2 0,
0.
Consequently, Hence { = 0 since {

V ( )
0.
2 0, +
1/2 3 2 1, }
0 a fortiori, which shows that
2 0,
1/2 3 2 1, } V ( )
0 on the other hand. We have thus
reached the conclusion that
0, a contradiction.
The various Korn inequalities established or mentioned so far, which apply to general surfaces (i.e., without any restrictions bearing on their geometry save some regularity assumptions), all involve both the linearized change of metric, and the linearized change of curvature, tensors. It is remarkable that, for specic geometries and boundary conditions, a Korn inequality can be established that only involves the linearized change of metric tensors. More specically, Ciarlet & Lods [1996a] and Ciarlet & SanchezPalencia [1996] have established the following Korn inequality on an elliptic surface: Let be a domain in R2 and let C 2,1 ( ; E3 ) be an injective immersion with the property that the surface S = ( ) is elliptic, in the sense that all its points are elliptic (this means that the Gaussian curvature is > 0 everywhere on S ; cf. Section 2.5). Then there exists a constant cM = cM (, ) > 0 such that
1 1 for all = (i ) H0 ( ) H0 ( ) L2 ( ). 2 1,
+ 3
2 0,
1/2
cM
,
( )
2 0,
1/2
Remarks. (1) The norm 3 2, appearing in the left-hand side of the Korn inequality on a general surface (Theorem 4.3-4) is now replaced by the norm
Sect. 4.4]
Existence theorems for the linear Koiter shell equations
185
3 0, . This replacement reects that it is enough that = (i ) H 1 ( ) H 1 ( ) L2 ( ) in order that ( ) L2 ( ). As a result, no boundary condition can be imposed on 3 . (2) The Korn inequality on an elliptic surface was rst established by Destuynder [1985, Theorem 6.1 and 6.5], under the additional assumptions that the surface S can be covered by a single system of lines of curvature (Section 2.5) and that the C 0 ( )-norms of the corresponding Christoel symbols are small enough. Only compact surfaces dened by a single injective immersion C 3 ( ) have been considered so far. By contrast, a compact surface S without boundary (such as an ellipsoid or a torus) is dened by means of a nite number I 2 of injective immersions i C 3 ( i ), 1 i I , where the sets i are domains in R2 , in such a way that S = iI i (i ). As shown by S. Mardare [2003a], the Korn inequality without boundary conditions (Theorem 4.3-1) and the Korn inequality on the quotient space H 1 ( ) H 1 ( ) H 2 ( )/ Rig( ) (Theorem 4.3-5) can be both extended to such surfaces without boundary. Likewise, Slicaru [1998] has shown that the above Korn inequality on an elliptic surface can be extended to elliptic surfaces without boundary.
4.4
EXISTENCE AND UNIQUENESS THEOREMS FOR THE LINEAR KOITER SHELL EQUATIONS; COVARIANT DERIVATIVES OF A TENSOR FIELD DEFINED ON A SURFACE
V( ) := { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on 0 },
Let the space V( ) be dened by
where 0 is a d -measurable subset of := that satises length 0 > 0. Our primary objective consists in showing that the bilinear form B : V( ) V( ) R dened by B ( , ) :=
a ( ) ( ) +
3 a ( ) ( ) a dy 3
for all ( , ) V( ) V( ) is V( )-elliptic. As a preliminary, we establish the uniform positive-deniteness of the elasticity tensor of the shell, given here by means of its contravariant components e constants are the same as in a . Note that the assumptions on the Lam three-dimensional elasticity (Theorem 3.9-1). Theorem 4.4-1. Let be a domain in R2 , let C 3 ( ; E3 ) be an injective immersion, let a denote the contravariant components of the metric tensor of the surface (), let the contravariant components of the two-dimensional elasticity tensor of the shell be given by a = 4 a a + 2(a a + a a ), + 2
186
[Ch. 4
and assume that 3 + 2 > 0 and > 0. Then there exists a constant ce = ce (, , , ) > 0 such that |t |2 ce a (y )t t
,
for all y and all symmetric matrices (t ). Proof. This proof is similar to that of Theorem 3.9-1 and for this reason, is only sketched. Given any y and any symmetric matrix (t ), let A(y ) = (a (y )) and T = (t ), let K(y ) S2 be the unique square root of A(y ), and let B(y ) := K(y )TK(y ) S2 . Then 1 a (y )t t = tr B(y ) 2
2
+ 2 tr B(y )T B(y ) with :=
2 . + 2
By the inequality established in part (i) of the proof of Theorem 3.9-1 (with d = 2 in this case), there thus exists a constant (, ) > 0 such that 1 a (y )t t tr B(y )T B(y ) 2 if + > 0 and > 0, or equivalently, if 3 + 2 > 0 and > 0. The proof is then concluded as the proof of Theorem 3.9-1. Combined with Korns inequality with boundary conditions (Theorem 4.3-4), the positive deniteness of the elasticity tensor leads to the existence of a weak solution, i.e., a solution to the variational equations of the linear Koiter shell equations. Theorem 4.4-2. Let be a domain in R2 , let 0 be a subset of = with length 0 > 0, and let C 3 ( ; E3 ) be an injective immersion. Finally, let there be given constants and that satisfy 3 + 2 > 0 and > 0, and functions p Lr ( ) for some r > 1 and p3 L1 ( ). Then there is one and only one solution = (i ) to the variational problem: V( ) = { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on 0 }, a ( ) ( ) +
3 a ( ) ( ) a dy 3 pi i a dy for all = (i ) V( ), =
Sect. 4.4] where
187
4 a a + 2(a a + a a ), + 2 1 ( ) = ( a + a ) and ( ) = ( ) a3 , 2 a = where := i ai . The eld V( ) is also the unique solution to the minimization problem: j ( ) = where j ( ) := 1 2 a ( ) ( ) +
V ( )
inf
j ( ),
3 a ( ) ( ) a dy 3
pi i a dy.
Proof. As a closed subspace of H 1 ( ) H 1 ( ) H 2 ( ), the space V( ) is a Hilbert space. The assumptions made on the mapping ensure in particular that the vector elds ai and ai belong to C 2 ( ; R3 ) and that the functions a , , and a are continuous on the compact set . Hence the bilinear form dened by the left-hand side of the variational equations is continuous over the space H 1 ( ) H 1 ( ) H 2 ( ). The continuous embeddings of the space H 1 ( ) into the space Ls ( ) for any s 1 and of the space H 2 ( ) into the space C 1 ( ) show that the linear form dened by the right-hand side is continuous over the same space. Since the symmetric matrix (a (y )) is positive-denite for all y , there exists a0 such that a(y ) a0 > 0 for all y . Finally, the Korn inequality with boundary conditions (Theorem 4.3-4) and the uniform positive deniteness of the elasticity tensor of the shell (Theorem 4.4-1) together imply that min ,
3 1 2 c c a0 3 e
2 1,
+ 3
2 2,
a ( ) ( ) +
3 a ( ) ( ) a dy 3
for all = (i ) V( ). Hence the bilinear form B is V( )-elliptic. The Lax-Milgram lemma then shows that the variational equations have one and only one solution. Since the bilinear form is symmetric, this solution is also the unique solution of the minimization problem stated in the theorem. The above existence and uniqueness result applies to linearized pure displacement and displacement-traction problems, i.e., those that correspond to length 0 > 0.
188
[Ch. 4
We next consider the linearized pure traction problem, i.e., corresponding to the case where the set 0 is empty. In this case, we seek a vector eld = (i ) H 1 ( ) H 1 ( ) H 2 ( ) that satises a ( ) ( ) +
pi i a dy for all = (i ) H 1 ( ) H 1 ( ) H 2 ( ).
3 a ( ) ( ) a dy 3
Clearly, such variational equations can have a solution only if their righthand side vanishes for any vector eld = (i ) Rig( ), where Rig( ) := H 1 ( ) H 1 ( ) H 2 ( ); ( ) = ( ) = 0 in , since replacing by ( + ) for any Rig( ) does not aect their left-hand side. We now show that this necessary condition is in fact also sucient for the existence of solutions, because in this case of the Korn inequality over the quotient space H 1 ( ) H 1 ( ) H 2 ( )/ Rig( ) (Theorem 4.3-5). Evidently, the existence of solutions can then hold only up to the addition of vector elds Rig( ). This means that the solution is naturally sought in this quotient space. Theorem 4.4-3. Let be a domain in R2 and let C 3 ( ; E3 ) be an injective immersion. Let there be given constants and that satisfy 3 + 2 > 0 and > 0 and functions p Lr ( ) for some r > 1 and p3 L1 ( ). Dene the quotient space ( ) := H 1 ( ) H 1 ( ) H 2 ( )/ Rig( ), V and assume that the functions pi satisfy pi i a d = 0 for all Rig( ).
Finally, let the functions a be dened as in Theorem 4.4-1. V ( ) to the variational equaThen there is one and only one solution tions 3 ) ( ) ( ) + a ( ) a ( a dy 3 ( ). = (i ) V = pi i a dy for all
V ( ) is also the unique solution to the minimizaThe equivalence class tion problem ) = inf j ( ), j (
( ) V
Sect. 4.4] where
189
) := j (
1 2
) ( )+ a (
3 ) ( ) a dy a ( 3
pi i a dy.
Proof. The proof is analogous to that of Theorem 4.4-2, the Korn inequality of Theorem 4.3-4 being now replaced by that of Theorem 4.3-5. An immediate consequence of Theorems 4.4-2 and 4.4-3 is the existence and uniqueness (up to innitesimal rigid displacements when the set 0 is empty) of a displacement eld i ai of the middle surface S , whose components H 1 ( ) and 3 H 2 ( ) are thus obtained by nding the solution = (i ) to the variational equations of either theorem. Since the vector elds ai formed by the covariant bases belong to the space C 2 ( ; R3 ) by assumption, the vector elds a and 3 a3 belong respectively to the spaces H1 ( ) and H2 ( ). We next derive the boundary value problem that is, at least formally, equivalent to the variational equations of the Theorems 4.4-2 or 4.4-3, the latter corresponding to the case where the set 0 is empty. In what follows, 1 := 0 , ( ) is the unit outer normal vector along , 1 := 2 , 2 := 1 , and := denotes the tangential derivative of in the direction of the vector ( ). Theorem 4.4-4. Let be a domain in R2 and let C 3 (; E3 ) be an injective immersion. Assume that the boundary of and the functions pi are smooth enough. If the solution = (i ) to the variational equations found in either Theorem 4.4-2 or Theorem 4.4-3 is smooth enough, then is also a solution to the following boundary value problem:
b n = p3 in , m | b b m (n + b m )| b (m | ) = p in , i = 3 = 0 on 0 ,
m = 0 on 1 , (m | ) + (m ) = 0 on 1 ,
(n + 2b m ) = 0 on 1 ,
3 a ( ), 3 and, for an arbitrary tensor eld with smooth enough covariant components t : R, n := a ( ) and m :=
t | := t + + , t t t | := (t | ) + | ). (t
where
190
[Ch. 4
Proof. For simplicity, we give the proof only in the case where 0 = , i.e., when the space V( ) of Theorem 4.4-2 reduces to
1 1 2 V( ) = H0 ( ) H0 ( ) H0 ( ).
The extension to the case where length 1 > 0 is straightforward. In what follows, we assume that the solution is smooth enough in the sense that n H 1 ( ) and m H 2 ( ). (i) We rst establish the relations a = a . Let A denote the matrix of order three with a1 , a2 , a3 as its column vectors, so that a = det A (see part (i) of the proof of Theorem 4.2-2). Consequently, a = det( a1 , a2 , a3 ) + det(a1 , a2 , a3 ) + det(a1 , a2 , a3 ) 2 3 a = (1 1 + 2 + 3 ) det(a1 , a2 , a3 ) = since a = a + b a3 (Theorem 2.6-1). (ii) Using the Green formula in Sobolev spaces (see, e.g., Ne cas [1967]) and assuming that the functions n = n are in H 1 ( ), we rst transform the rst integral appearing in the left-hand side of the variational equations. This 1 1 ( ) H0 ( ) L2 ( ), hence a fortiori for all = gives, for all = (i ) H0 1 1 2 (i ) H0 ( ) H0 ( ) H0 ( ), a ( ) ( ) a dy =

n ( ) a dy
= =
1 ( + ) an b 3 dy 2 an dy

an dy
an b 3 dy
an dy
an dy
an b 3 dy
dy a n + + n n
an b 3 dy
n | + b n 3 dy.
Sect. 4.4]
191
(iii) We then likewise transform the second integral appearing in the lefthand side of the variational equations, viz., 1 3 =
a ( ) ( ) a dy =

m ( ) a dy
a m 3 dy +
a m (2b 3 ) dy a m (2b + b | b b 3 ) dy,
1 1 2 for all = (i ) H0 ( ) H0 ( ) H0 ( ). Using the symmetry m = m , the relation a = a (cf. part (i)), and the same Green formula as in part (ii), we obtain
m ( ) a dy =

a( m + + m m ) 3 dy
+2
am b dy
am (2b + b | b b 3 ) dy.
The same Green formula further shows that
a( m + + m m ) 3 dy a(m | ) 3 dy = a(m | )3 dy,

= = 2
( a m | )3 dy
a m b dy = 2
a (b ) + dy. m b m
Consequently, m ( ) a dy =

dy a 2(b m )| + (b | )m 3 dy. a m | b b m
Using in this relation the easily veried formula

(b + b m )| = (b | )m (m | )
192
[Ch. 4
and the symmetry relations b | = b | (Theorem 4.2-2), we nally obtain
m ( ) a dy =

a (b m )| + b (m | ) dy a b m | 3 dy. b m
(iv) By parts (ii) and (iii), the variational equations 1 a ( ) ( ) + a ( ) ( ) pi i a dy = 0 3
imply that a (n + b dy m )| + b (m | ) + p a b n + b m | + p3 3 dy = 0 b m
1 1 2 for all (i ) H0 ( ) H0 ( ) H0 ( ). The announced partial dierential equations are thus satised in .
The functions n = a ( ) are the contravariant components of the linearized stress resultant tensor eld inside the shell, and the functions m = 3 a ( ) 3
are the contravariant components of the linearized stress couple, or linearized bending moment, tensor eld inside the shell. The functions
t | = t + + , t t t | = (t | ) + | ), (t
which have naturally appeared in the course of the proof of Theorem 4.4-4, constitute examples of rst-order, and second order, covariant derivatives of a tensor eld dened on a surface, here by means of its contravariant components t : R. Finally, we state a regularity result that provides an instance where the weak solution, viz., the solution of the variational equations, is also a classical solution, viz., a solution of the associated boundary value problem. The proof of this result, which is due to Alexandrescu [1994], is long and delicate and for this reason is only briey sketched here (as expected, it follows the same pattern as in the three-dimensional case, considered in Theorem 3.9-4).
Sect. 4.5]
A brief review of linear shell theories
193
Theorem 4.4-5. Let be a domain in R2 with boundary and let : E3 be an injective immersion. Assume that, for some integer m 0 and some real number q > 1, is of class C m+4 , C m+4 ( ; E3 ), p W m+1,q ( ), and p3 W m,q ( ). Finally, assume that 0 = . Then the weak solution found in Theorem 4.4-2 satises = (i ) W m+3,q ( ) W m+3,q ( ) W m+4,q ( ). Sketch of the proof. To begin with, assume that the boundary is of class C 4 and the mapping belongs to the space C 4 ( ; E3 ). One rst veries that the linear system of partial dierential equations found in Theorem 4.4-4 (which is of the third order with respect to the unknowns and of the fourth order with respect to the unknown 3 ) is uniformly elliptic and satises the supplementing condition on L and the complementing boundary conditions, in the sense of Agmon, Douglis & Nirenberg [1964]. One then veries that the same system is also strongly elliptic in the sense of Ne cas [1967, p. 185]. A regularity result of Ne cas [1967, Lemma 3.2, p. 260] then shows that the weak solution = (i ) found in Theorem 4.4-2, which belongs 1 1 2 to the space H0 ( ) H0 ( ) H0 ( ) since 0 = by assumption, satises = (i ) H 3 ( ) H 3 ( ) H 4 ( ) if p H 1 ( ) and p3 L2 ( ). A result of Geymonat [1965, Theorem 3.5] about the index of the associated linear operator then implies that
= (i ) W 3,q ( ) W 3,q ( ) W 4,q ( )
if p W 1,q ( ) and p3 Lq ( ) for some q > 1. Assume nally that, for some integer m 1 and some real number q > 1, is of class C m+4 and C m+4 ( ; E3 ). Then a regularity result of Agmon, Douglis & Nirenberg [1964] implies that = (i ) W m+3,q ( ) W m+3,q ( ) W m+4,q ( ). if p W m+1,q ( ) and p3 W m,q ( ).
4.5
A BRIEF REVIEW OF LINEAR SHELL THEORIES
In order to put the linear Koiter shell equations in their proper perspective, we briey review the genesis of those two-dimensional linear shell theories that can be found, and rigorously justied, as the outcome of an asymptotic analysis of the equations of three-dimensional linearized elasticity as 0.
194
[Ch. 4
The asymptotic analysis of elastic shells has been a subject of considerable attention during the past decades. After the landmark attempt of Goldenveizer [1963], a major step for linearly elastic shells was achieved by Destuynder [1980] in his Doctoral Dissertation, where a convergence theorem for membrane shells was almost proved. Another major step was achieved by SanchezPalencia [1990], who clearly delineated the kinds of geometries of the middle surface and boundary conditions that yield either two-dimensional membrane, or two-dimensional exural, equations when the method of formal asymptotic expansions is applied to the variational equations of three-dimensional linearized elasticity (see also Caillerie & Sanchez-Palencia [1995] and Miara & SanchezPalencia [1996]). Then Ciarlet & Lods [1996a,b] and Ciarlet, Lods & Miara [1996] carried out an asymptotic analysis of linearly elastic shells that covers all possible cases : Under three distinct sets of assumptions on the geometry of the middle surface, on the boundary conditions, and on the order of magnitude of the applied forces, they established convergence theorems in H 1 , in L2 , or in ad hoc completion spaces, that justify either the linear two-dimensional equations of a membrane shell, or those of a generalized membrane shell, or those of a exural shell. More specically, consider a family of linearly elastic shells of thickness 2 that satisfy the following assumptions : All the shells have the same middle surface S = ( ) E3 , where is a domain in R2 with boundary , and C 3 ( ; E3 ). Their reference congurations are thus of the form ( ), > 0, where := ], [ , and the mapping is dened by
(y, x 3 ) := (y ) + x3 a3 (y ) for all (y, x3 ).
All the shells in the family are made with the same homogeneous isotropic elastic material and that their reference congurations are natural states. Their elastic material is thus characterized by two Lam e constants and satisfying 3 + 2 > 0 and > 0. The shells are subjected to body forces and that the corresponding applied body force density is O(p ) with respect to , for some ad hoc power p (which will be specied later). This means that, for each > 0, the contravariant components f i, L2 ( ) of the body force density are of the form f i, (y, x3 ) = p f i (y, x3 ) for all (y, x3 ) := ]1, 1[ , and the functions f i L2 () are independent of (surface forces acting on the upper and lower faces of the shell could be as well taken into account but will not be considered here, for simplicity of exposition). Let then the functions pi, L2 ( ) be dened for each > 0 by pi, :=

f i, dx 3.
Sect. 4.5]
195
Then, for each > 0, the associated equations of linearized three-dimensional elasticity in curvilinear coordinates, viz., (y1 , y2 , x 3 ) , have one and only 1 one solution (Theorem 3.9-2) u = (u i ) H ( ), where the functions ui are the covariant components of the displacement eld of the reference conguration ( ). Finally, each shell is subjected to a boundary condition of place on the portion (0 [, ]) of its lateral face, where 0 is a xed portion of , with length 0 > 0. Incidentally, such particular instances of sets and mappings provide a fundamental motivation for studying the equations of linear elasticity in curvilinear coordinates, since they constitute a most natural point of departure of any asymptotic analysis of shells. Let ( ) = 1 ( a + a ) and ( ) = ( ) a3 , 2
where = i ai , denote as usual the covariant components of the linearized change of metric, and linearized change of curvature, tensors. In Ciarlet, Lods & Miara [1996] it is rst assumed that the space of linearized inextensional displacements (introduced by Sanchez-Palencia [1989a]) VF ( ) := { = (i ) H 1 ( ) H 1 ( ) H 2 ( ); i = 3 = 0 on ( ) = 0 in } contains non-zero functions. This assumption is in fact one in disguise about the geometry of the surface S and on the set 0 . For instance, it is satised if S is a portion of a cylinder and (0 ) is contained in one or two generatrices of S , or if S is contained in a plane, in which case the shells are plates. Under this assumption Ciarlet, Lods & Miara [1996] showed that, if the applied body force density is O(2 ) with respect to , then 1 2
1 u i dx3 i in H ( ) as 0,
where the limit vector eld := (i ) belongs to the space VF ( ) and satises the equations of a linearly elastic exural shell , viz., 3 3 a ( ) ( ) a dy =

pi, i a dy
for all = (i ) VF ( ). Observe in passing that the limit is indeed independent of , since both sides of these variational equations are of the same order (viz., 3 ), because of the assumptions made on the applied forces. Equivalently, the vector eld satises the following constrained minimization problem : ( ) = inf jF ( ), VF ( ) and jF
196 where
jF ( ) :=
[Ch. 4
1 2
3 a ( ) ( ) a dy 3
pi, i a dy
for all = (i ) VF ( ), where the functions a = 4 a a + 2(a a + a a ) ( + 2)
are precisely the familiar contravariant components of the shell elasticity tensor. If VF ( ) = {0}, the two-dimensional equations of a linearly elastic exural shell are therefore justied. If VF ( ) = {0}, the above convergence result still applies. However, the 1 1 only information it provides is that u dx 3 0 in H ( ) as 0. 2 i Hence a more rened asymptotic analysis is needed in this case. A rst instance of such a renement was given by Ciarlet & Lods [1996a], where it was assumed that 0 = and that the surface S is elliptic, in the sense that its Gaussian curvature is > 0 everywhere. As shown in Ciarlet & Lods [1996a] and Ciarlet & Sanchez-Palencia [1996], these two conditions, together with ad hoc regularity assumptions, indeed imply that VF ( ) = {0}. In this case, Ciarlet & Lods [1996b] showed that, if the applied body force density is O(1) with respect to , then 1 2
1 u dx3 in H ( ) and
1 2
2 u 3 dx3 3 in L ( ) as 0,
where the limit vector eld := (i ) belongs to the space

1 1 VM ( ) := H0 ( ) H0 ( ) L2 ( ),
and solves the equations of a linearly elastic membrane shell, viz., a ( ) ( ) a dy =

pi, i a dy
for all = (i ) VM ( ), where the functions a , ( ), a, and pi, have the same meanings as above. If 0 = and S is elliptic, the two-dimensional equations of a linearly elastic membrane shell are therefore justied. Observe that the limit is again independent of , since both sides of these variational equations are of the same order (viz., ), because of the assumptions made on the applied forces. Equivalently, the eld satises the following unconstrained minimization problem : ( ) = inf jM ( ), VM ( ) and jM
VM ( )
where
jM ( ) :=
1 2
a ( ) ( ) a dy

pi, i a dy.
Sect. 4.5]
197
Finally, Ciarlet & Lods [1996d] studied all the remaining cases where VF ( ) = {0}, e.g., when S is elliptic but length 0 < length , or when S is for instance a portion of a hyperboloid of revolution, etc. To give a avor of their results, consider the important special case where the semi-norm || : = (i ) | |M =
, M
( )
2 0,
1/2
becomes a norm over the space W( ) := { H1 ( ); = 0 on 0 }. In this case, Ciarlet & Lods [1996d] showed that, if the applied body forces are admissible in a specic sense (but a bit too technical to be described here), and if their density is again O(1) with respect to , then 1 2 where VM ( ) := completion of W( ) with respect to || . Furthermore, the limit eld VM ( ) solves limit variational equations of the form , BM ( , ) = LM ( ) for all VM ( ), where BM is the unique extension to VM ( ) of the bilinear form BM dened by 1 BM ( , ) := a ( ) ( ) a dy for all , W( ), 2 i.e., BM is the bilinear form found above for a linearly elastic membrane shell, , and LM : VM ( ) R is an ad hoc linear form, determined by the behavior as 0 of the admissible body forces. In the last remaining case, where VF ( ) = {0} but ||M is not a norm over the space W( ), a similar convergence result can be established, but only in ( ) with respect of ||M of the quotient space W( )/W0 ( ), the completion V M where W0 ( ) = { W( ); ( ) = 0 in }. Either one of the above variational problems corresponding to the remaining cases where VF = {0} constitute the equations of a linearly elastic generalized membrane shell, whose two-dimensional equations are therefore justied. The proofs of the above convergence results are long and technically dicult. Suce it to say here that they crucially hinge on the Korn inequality with boundary conditions (Theorem 4.3-4) and on the Korn inequality on an elliptic surface mentioned at the end of Section 4.3.
M
u dx 3 in VM ( ) as 0,
198
[Ch. 4
Combining these convergences with earlier results of Destuynder [1985] and Sanchez-Palencia [1989a,b, 1992] (see also Sanchez-Hubert & Sanchez-Palencia [1997]), Ciarlet & Lods [1996b,c] have also justied as follows the linear Koiter shell equations studied in Sections 4.2 to 4.4, again in all possible cases. Let K denote for each > 0 the unique solution (Theorem 4.4-2) to the linear Koiter shell equations, viz., the vector eld that satises
1 1 2 K V( ) = { = (i ) H ( ) H ( ) H ( ); i = 3 = 0 on 0 }, 3 a ( a ( a dy K ) ( ) + K ) ( ) 3 = pi, i a dy for all = (i ) V( ),
or equivalently, the unique solution to the minimization problem

K V( ) and j ( K ) = V ( )
inf
j ( )
where j ( ) = 1 2
a ( ) ( ) +
3 a ( ) ( ) a dy 3
pi i a dy.
Observe in passing that, for a linearly elastic shell, the stored energy function found in Koiters energy, viz., 3 a ( ) ( ) + a ( ) ( ) 2 6
is thus exactly the sum of the stored energy function of a linearly elastic membrane shell and of that of a linearly elastic exural shell (a similar, albeit less satisfactory, observation holds for a nonlinearly elastic shell, cf. Section 4.1). Then, for each category of linearly elastic shells (membrane, generalized 1 u dx membrane, or exural), the vector elds K and 3 , where u = (ui ) 2 denotes the solution of the three-dimensional problem, have exactly the same asymptotic behavior as 0, in precisely the same function spaces that were found in the asymptotic analysis of the three-dimensional solution. It is all the more remarkable that Koiters equations can be fully justied for all types of shells, since it is clear that Koiters equations cannot be recovered as the outcome of an asymptotic analysis of the three-dimensional equations, the two-dimensional equations of linearly elastic, membrane, generalized membrane, or exural, shells exhausting all such possible outcomes! So, even though Koiters linear model is not a limit model, it is in a sense the best two-dimensional one for linearly elastic shells !
Sect. 4.5]
199
One can thus only marvel at the insight that led W.T. Koiter to conceive the right equations, whose versatility is indeed remarkable, out of purely mechanical and geometrical intuitions! We refer to Ciarlet [2000a] for a detailed analysis of the asymptotic analysis of linearly elastic shells, for a detailed description and analysis of other linear shell models, such as those of Naghdi, Budiansky and Sanders, Novozilov, etc., and for an extensive list of references.
REFERENCES
Adams, R.A. [1975]: Sobolev Spaces, Academic Press, New York. Agmon, S.; Douglis, A.; Nirenberg, L. [1964]: Estimates near the boundary for solutions of elliptic partial dierential equations satisfying general boundary conditions. II, Comm. Pure Appl. Math. 17, 3592. Akian, J.L. [2003]: A simple proof of the ellipticity of Koiters model, Analysis and Applications 1, 116. Alexandrescu, O. [1994]: Th eor` eme dexistence pour le mod` ele bidimensionnel de coque non lin eaire de W.T. Koiter, C.R. Acad. Sci. Paris, S er. I, 319, 899902. Amrouche, C.; Girault, V. [1994]: Decomposition of vector spaces and application to the Stokes problem in arbitrary dimension, Czech. Math. J. 44, 109140. Anicic, S.; Le Dret, H.; Raoult, A. [2005]: The innitesimal rigid displacement lemma in Lipschitz coordinates and application to shells with minimal regularity, Math. Methods Appl. Sci. 27, 12831299. Antman, S.S. [1976]: Ordinary dierential equations of non-linear elasticity I: Foundations of the theories of non-linearly elastic rods and shells, Arch. Rational Mech. Anal. 61, 307351. Antman, S.S. [1995]: Nonlinear Problems of Elasticity, Springer-Verlag, Berlin (Second Edition: 2004). Ball, J.M. [1977]: Convexity conditions and existence theorems in nonlinear elasticity, Arch. Rational Mech. Anal. 63, 337403. Bamberger, Y. [1981]: M ecanique de lIng enieur, Volume II, Hermann, Paris. Berger, M. [2003]: A Panoramic View of Riemannian Geometry, Springer, Berlin. Berger M.; Gostiaux, B. [1987]: G eom emtrie Di erentielle: Vari et es, Courbes et Surfaces, Presses Universitaires de France, Paris. Bernadou, M. [1994]: M ethodes dEl ements Finis pour les Coques Minces, Masson, Paris (English translation: Finite Element Methods for Thin Shell Problems, John Wiley, New York, 1995). Bernadou, M.; Ciarlet, P.G. [1976]: Sur lellipticit e du mod` ele lin eaire de coques de W.T. Koiter, in Computing Methods in Applied Sciences and Engineering (R. Glowinski & J.L. Lions, Editors), pp. 89136, Lecture Notes in Economics and Mathematical Systems, 134, Springer-Verlag, Heidelberg. Bernadou, M.; Ciarlet, P.G.; Miara, B. [1994]: Existence theorems for twodimensional linear shell theories, J. Elasticity 34, 111138. Blouza, A.; Le Dret, H. [1999]: Existence and uniqueness for the linear Koiter model for shells with little regularity, Quart. Appl. Math. 57, 317337.
201
202
References
Bolley, P.; Camus, J. [1976]: R egularit e pour une classe de probl` emes aux limites elliptiques d eg en er es variationnels, C.R. Acad. Sci. Paris, S er. A, 282, 4547. Bonnet, O. [1848]: M emoire sur la th eorie g en erale des surfaces, Journal de lEcole Polytechnique 19, 1146. Boothby, W.M. [1975]: An Introduction to Dierentiable Manifolds and Riemannian Geometry, Academic Press, New York. Borchers, W.; Sohr, H. [1990]: On the equations rot v = g and div u = f with zero boundary conditions, Hokkaido Math. J. 19, 6787. Brezis, H. [1983]: Analyse Fonctionnelle, Th eorie et Applications, Masson, Paris. Caillerie, D.; Sanchez-Palencia, E. [1995]: Elastic thin shells: asymptotic theory in the anisotropic and heterogeneous cases, Math. Models Methods Appl. Sci. 5, 473496. Cartan, E. [1927]: Sur la possibilit e de plonger un espace riemannien donn e dans un espace euclidien, Annales de la Soci et e Polonaise de Math ematiques 6, 17. do Carmo, M.P. [1976]: Dierential Geometry of Curves and Surfaces, Prentice-Hall, Englewood Clis. do Carmo, M.P. [1994]: Dierential Forms and Applications, Universitext, Springer-Verlag, Berlin (English translation of: Formas Diferenciais e Apli c oes, Instituto da Matematica, Pura e Aplicada, Rio de Janeiro, 1971). Chen, W.; Jost, J. [2002]: A Riemannian version of Korns inequality, Calculus of Variations 14, 517530. Choquet-Bruhat, de Witt-Morette & Dillard-Bleick [1977]: Analysis, Manifolds and Physics, North-Holland, Amsterdam (Revised Edition: 1982). Ciarlet, P.G. [1982]: Introduction ` a lAnalyse Num erique Matricielle et ` a lOptimisation, Masson, Paris (English translation: Introduction to Numerical Linear Algebra and Optimisation, Cambridge University Press, Cambridge, 1989). Ciarlet, P.G. [1988]: Mathematical Elasticity, Volume I: Three-Dimensional Elasticity, North-Holland, Amsterdam. Ciarlet, P.G. [1993]: Mod` eles bi-dimensionnels de coques: Analyse asymptotique et th eor` emes dexistence, in Boundary Value Problems for Partial Dierential Equations and Applications (J.L. Lions & C. Baiocchi, Editors), pp. 6180, Masson, Paris. Ciarlet, P.G. [1997]: Mathematical Elasticity, Volume II: Theory of Plates, NorthHolland, Amsterdam. Ciarlet, P.G. [2000a]: Mathematical Elasticity, Volume III: Theory of Shells, NorthHolland, Amsterdam. Ciarlet, P.G. [2000b]: Un mod` ele bi-dimensionnel non lin eaire de coque analogue ` a celui de W.T. Koiter, C.R. Acad. Sci. Paris, S er. I, 331, 405410. Ciarlet, P.G. [2003]: The continuity of a surface as a function of its two fundamental forms, J. Math. Pures Appl. 82, 253274. Ciarlet, P.G.; Destuynder, P. [1979]: A justication of the two-dimensional plate model, J. M ecanique 18, 315344. Ciarlet, P.G.; Larsonneur, F. [2001]: On the recovery of a surface with prescribed rst and second fundamental forms, J. Math. Pures Appl. 81, 167185. Ciarlet, P.G.; Laurent, F. [2003]: Continuity of a deformation as a function of its Cauchy-Green tensor, Arch. Rational Mech. Anal. 167, 255269.
References
203
Ciarlet, P.G.; Lods, V. [1996a]: On the ellipticity of linear membrane shell equations, J. Math. Pures Appl. 75, 107124. Ciarlet, P.G.; Lods, V. [1996b]: Asymptotic analysis of linearly elastic shells. I. Justication of membrane shell equations, Arch. Rational Mech. Anal. 136, 119 161. Ciarlet, P.G.; Lods, V. [1996c]: Asymptotic analysis of linearly elastic shells. III. Justication of Koiters shell equations, Arch. Rational Mech. Anal. 136, 191 200. Ciarlet, P.G.; Lods, V. [1996d]: Asymptotic analysis of linearly elastic shells: Generalized membrane shells, J. Elasticity 43, 147188. Ciarlet, P.G.; Lods, V.; Miara, B. [1996]: Asymptotic analysis of linearly elastic shells. II. Justication of exural shell equations, Arch. Rational Mech. Anal. 136, 163190. Ciarlet, P.G.; Mardare, C. [2003]: On rigid and innitesimal rigid displacements in three-dimensional elasticity, Math. Models Methods Appl. Sci. 13, 15891598. Ciarlet, P.G.; Mardare, C. [2004a]: On rigid and innitesimal rigid displacements in shell theory, J. Math. Pures Appl. 83, 115. Ciarlet, P.G.; Mardare, C. [2004b]: Recovery of a manifold with boundary and its continuity as a function of its metric tensor, J. Math. Pures Appl. 83, 811843. Ciarlet, P.G.; Mardare, C. [2004c]: Continuity of a deformation in H 1 as a function of its Cauchy-Green tensor in L1 , J. Nonlinear Sci. 14, 415427. Ciarlet, P.G.; Mardare, C. [2005]: Recovery of a surface with boundary and its continuity as a function of its two fundamental forms, Analysis and Applications, 3, 99117. Ciarlet, P.G.; Mardare, S. [2001]: On Korns inequalities in curvilinear coordinates, Math. Models Methods Appl. Sci. 11, 13791391. Ciarlet, P.G.; Miara, B. [1992]: Justication of the two-dimensional equations of a linearly elastic shallow shell, Comm. Pure Appl. Math. 45, 327360. as, J. [1987]: Injectivity and self-contact in nonlinear elasticity, Ciarlet, P.G.; Nec Arch. Rational Mech. Anal. 19, 171188. Ciarlet, P.G.; Roquefort, A. [2001]: Justication of a two-dimensional nonlinear shell model of Koiters type, Chinese Ann. Math. 22B, 129244. Ciarlet, P.G.; Sanchez-Palencia, E. [1996]: An existence and uniqueness theorem for the two-dimensional linear membrane shell equations, J. Math. Pures Appl. 75, 5167. Dacorogna [1989]: Direct Methods in the Calculus of Variations, Springer, Berlin. Dautray, R.; Lions, J.L. [1984]: Analyse Math ematique et Calcul Num erique pour les Sciences et les Techniques, Tome 1, Masson, Paris. Destuynder, P. [1980]: Sur une Justication des Mod` eles de Plaques et de Coques par les M ethodes Asymptotiques, Doctoral Dissertation, Universit e Pierre et Marie Curie, Paris. Destuynder, P. [1985]: A classication of thin shell theories, Acta Applicandae Mathematicae 4, 1563. Duvaut, G.; Lions, J.L. [1972]: Les In equations en M ecanique et en Physique, Dunod, Paris (English translation: Inequalities in Mechanics and Physics, SpringerVerlag, Berlin, 1976).
204
References
Flanders, H. [1989]: Dierential Forms with Applications to the Physical Sciences, Dover, New York. ller, S. [2003]: Derivation of Friesecke, G.; James, R.D.; Mora, M.G.; Mu nonlinear bending theory for shells from three dimensional nonlinear elasticity by Gamma-convergence, C.R. Acad. Sci. Paris, S er. I, 336, 697702. Friesecke, G.; James, R.D.; Muller, S. [2002]: A theorem on geometric rigidity and the derivation of nonlinear plate theory from three dimensional elasticity, Comm. Pure Appl. Math. 55, 14611506. Gallot, S.; Hulin, D.; Lafontaine, J. [2004]: Riemannian Geometry, Third Edition, Springer, Berlin. Gauss, C.F. [1828]: Disquisitiones generales circas supercies curvas, Commentationes societatis regiae scientiarum Gottingensis recentiores 6, G ottingen. Germain, P. [1972]: M ecanique des Milieux Continus, Tome I, Masson, Paris. Geymonat, G. [1965]: Sui problemi ai limiti per i sistemi lineari ellittici, Ann. Mat. Pura Appl. 69, 207284. Geymonat, G.; Gilardi, G. [1998]: Contre-exemples a ` lin egalit e de Korn et au lemme de Lions dans des domaines irr eguliers, in Equations aux D eriv ees Partielles et Applications. Articles D edi es a ` Jacques-Louis Lions, pp. 541548, GauthierVillars, Paris. Geymonat, G.; Suquet, P. [1986]: Functional spaces for Norton-Ho materials, Math. Methods Appl. Sci. 8, 206222. Goldenveizer, A.L. [1963]: Derivation of an approximate theory of shells by means of asymptotic integration of the equations of the theory of elasticity, Prikl. Mat. Mech. 27, 593608. Gurtin, M.E. [1981]: Topics in Finite Elasticity, CBMS-NSF Regional Conference Series in Applied Mathematics, SIAM, Philadelphia. Hartman, P.; Wintner, A. [1950]: On the embedding problem in dierential geometry, Amer. J. Math. 72, 553564. Jacobowitz, H. [1982]: The Gauss-Codazzi equations, Tensor, N.S. 39, 1522. e dans un Janet, M. [1926]: Sur la possibilit e de plonger un espace riemannien donn espace euclidien, Annales de la Soci et e Polonaise de Math ematiques 5, 3843. John, F. [1961]: Rotation and strain, Comm. Pure Appl Math. 14, 391413. John, F. [1965]: Estimates for the derivatives of the stresses in a thin shell and interior shell equations, Comm. Pure Appl. Math. 18, 235267. John, F. [1971]: Rened interior equations for thin elastic shells, Comm. Pure Appl. Math. 18, 235267. John, F. [1972]: Bounds for deformations in terms of average strains, in Inequalities III (O. Shisha, Editor), pp. 129144, Academic Press, New York. Klingenberg, W. [1973]: Eine Vorlesung u ber Dierentialgeometrie, SpringerVerlag, Berlin (English translation: A Course in Dierential Geometry, Springer-Verlag, Berlin, 1978). Kohn, R.V. [1982]: New integral estimates for deformations in terms of their nonlinear strains, Arch. Rational Mech. Anal. 78, 131172. Koiter, W.T. [1966]: On the nonlinear theory of thin elastic shells, Proc. Kon. Ned. Akad. Wetensch. B69, 154.
References
205
Koiter, W.T. [1970]: On the foundations of the linear theory of thin elastic shells, Proc. Kon. Ned. Akad. Wetensch. B73, 169195. Kuhnel, W. [2002]: Dierentialgeometrie, Fried. Vieweg & Sohn, Wiesbaden (English translation: Dierential Geometry: Curves-Surfaces-Manifolds, American Mathematical Society, Providence, 2002). Lebedev, L.P.; Cloud, M.J. [2003]: Tensor Analysis, World Scientic, Singapore. Le Dret, H. [2004]: Well-posedness for Koiter and Naghdi shells with a G1 midsurface, Analysis and Applications 2, 365388. Le Dret, H.; Raoult, A. [1996]: The membrane shell model in nonlinear elasticity: A variational asymptotic derivation, J. Nonlinear Sci. 6, 5984. Lions, J.L. [1961]: Equations Di erentielles Op erationnelles et Probl` emes aux Limites, Springer-Verlag, Berlin. Lions, J.L. [1969]: Quelques M ethodes de R esolution des Probl` emes aux Limites Non Lin eaires, Dunod, Paris. Lions, J.L. [1973]: Perturbations Singuli` eres dans les Probl` emes aux Limites et en Contr ole Optimal, Lecture Notes in Mathematics, Vol. 323, Springer-Verlag, Berlin. Lions, J.L.; Magenes, E. [1968]: Probl` emes aux Limites Non Homog` enes et Applications, Vol. 1, Dunod, Paris (English translation: Non-Homogeneous Boundary Value Problems and Applications, Vol. 1, Springer-Verlag, Berlin, 1972). Lods, V. & Miara, B. [1998]: Nonlinearly elastic shell models. II. The exural model, Arch. Rational Mech. Anal. 142, 355374. Magenes, E.; Stampacchia, G. [1958]: I problemi al contorno per le equazioni dierenziali di tipo ellittico, Ann. Scuola Norm. Sup. Pisa 12, 247358. Mardare, C. [2003]: On the recovery of a manifold with prescribed metric tensor, Analysis and Applications 1, 433453. Mardare, S. [2003a]: Inequality of Korns type on compact surfaces without boundary, Chinese Annals Math. 24B, 191204 Mardare, S. [2003b]: The fundamental theorem of surface theory for surfaces with little regularity, J. Elasticity 73, 251290. Mardare, S. [2004]: On isometric immersions of a Riemannian space with little regularity, Analysis and Applications 2, 193226. Mardare, S. [2005]: On Pfa systems with Lp coecients and their applications in dierential geometry, J. Math. Pures Appl., to appear. Marsden, J.E.; Hughes, T.J.R. [1983]: Mathematical Foundations of Elasticity, Prentice-Hall, Englewood Clis (Second Edition: 1999). Miara, B. [1998]: Nonlinearly elastic shell models. I. The membrane model, Arch. Rational Mech. Anal. 142, 331353. Miara, B.; Sanchez-Palencia, E. [1996]: Asymptotic analysis of linearly elastic shells, Asymptotic Anal. 12, 4154.. Morrey, Jr., C.B. [1952]: Quasi-convexity and the lower semicontinuity of multiple integrals, Pacic J. Math. 2, 2553. Naghdi, P.M. [1972]: The theory of shells and plates, in Handbuch der Physik, gge & C. Truesdell, Editors), pp. 425640, Springer-Verlag, Vol. VIa/2 (S. Flu Berlin. Nash, J. [1954]: C 1 isometric imbeddings, Annals of Mathematics, 60, 383396.
206
References
as, J. [1967]: Les M Nec ethodes Directes en Th eorie des Equations Elliptiques, Masson, Paris. Nirenberg, L. [1974]: Topics in Nonlinear Functional Analysis, Lecture Notes, Courant Institute, New York University (Second Edition: American Mathematical Society, Providence, 1994). Reshetnyak, Y.G. [1967]: Liouvilles theory on conformal mappings under minimal regularity assumptions, Sibirskii Math. J. 8, 6985. Reshetnyak, Y.G. [2003]: Mappings of domains in Rn and their metric tensors, Siberian Math. J. 44, 332345. Sanchez-Palencia, E. [1989a]: Statique et dynamique des coques minces. I. Cas de exion pure non inhib ee, C.R. Acad. Sci. Paris, S er. I, 309, 411417. Sanchez-Palencia, E. [1989b]: Statique et dynamique des coques minces. II. Cas de exion pure inhib ee Approximation membranaire, C.R. Acad. Sci. Paris, S er. I, 309, 531537. Sanchez-Palencia, E. [1990]: Passages ` a la limite de l elasticit e tri-dimensionnelle a la th ` eorie asymptotique des coques minces, C.R. Acad. Sci. Paris, S er. II, 311, 909916. Sanchez-Palencia, E. [1992]: Asymptotic and spectral properties of a class of singular-sti problems, J. Math. Pures Appl. 71, 379406. Sanchez-Hubert, J.; Sanchez-Palencia, E. [1997]: Propri et es Asymptotiques, Masson, Paris. Schwartz, L. [1992]: Hermann, Paris. Coques Elastiques Minces:
Schwartz, L. [1966]: Th eorie des Distributions, Hermann, Paris. Analyse II: Calcul Di erentiel et Equations Di erentielles,
Simmonds, J.G. [1994]: A Brief on Tensor Analysis, Second Edition, Springer-Verlag, Berlin. Slicaru, S. [1998]: Quelques R esultats dans la Th eorie des Coques Lin eairement Elastiques ` a Surface Moyenne Uniform ement Elliptique ou Compacte Sans Bord, Doctoral Dissertation, Universit e Pierre et Marie Curie, Paris. Spivak, M. [1999]: A Comprehensive Introduction to Dierential Geometry, Volumes I to V, Third Edition, Publish or Perish, Berkeley. Stein, E. [1970]: Singular Integrals and Dierentiability Properties of Functions, Princeton University Press. Stoker, J.J. [1969]: Dierential Geometry, John Wiley, New York. Szczarba, R.H. [1970]: On isometric immersions of Riemannian manifolds in Euclidean space, Boletim da Sociedade Brasileira de Matem atica 1, 3145. Szopos, M. [2005]: On the recovery and continuity of a submanifold with boundary, Analysis and Applications 3, 119143. Tartar, L. [1978]: Topics in Nonlinear Analysis, dOrsay No. 78.13, Universit e de Paris-Sud, Orsay. Publications Math ematiques
Tenenblat, K. [1971]: On isometric immersions of Riemannian manifolds, Boletim da Sociedade Brasileira de Matem atica 2, 2336. Truesdell, C.; Noll, W. [1965]: The non-linear eld theories of mechanics, in gge, Editor), pp. 1-602, Springer-Verlag, Handbuch der Physik, Vol. III/3 (S. Flu Berlin.
References
207
Valent, T. [1988]: Boundary Value Problems of Finite Elasticity, Springer Tracts in Natural Philosophy, Vol. 31, Springer-Verlag, Berlin. Wang, C.C.; Truesdell, C. [1973]: Introduction to Rational Elasticity, Noordho, Groningen. Whitney, H. [1934]: Analytic extensions of dierentiable functions dened in closed sets, Trans. Amer. Math. Soc. 36, 6389. Yosida, K. [1966]: Functional Analysis, Springer-Verlag, Berlin. Zeidler, E. [1986]: Nonlinear Functional Analysis and its Applications, Vol. I: FixedPoint Theorems, Springer-Verlag, Berlin.
INDEX
applied body force: 113 applied surface force: 113 area element: 16, 17, 67, 68 asymptotic analysis: of linearly elastic shells: 159, 163, 193, 194 of nonlinearly elastic shells: 159, 163 asymptotic line: 77 boundary condition of place: 113 boundary condition of traction: 114 boundary value problem for a linearly elastic shell: 189 boundary value problem of linearized elasticity: 132 boundary value problem of nonlinear elasticity: in Cartesian coordinates: 116 in curvilinear coordinates: 130 Cauchy-Green strain tensor: 44, 57 in Cartesian coordinates: 115 center of curvature: 70 change of curvature tensor: 161 linearized : 169 change of metric tensor: 115, 159 linearized : 116, 165 Christoffel symbols: of the rst kind: 25 of the second kind: 22, 25 Christoffel symbols on a surface: of the rst kind: 84 of the second kind: 81, 84 Codazzi-Mainardi equations: 82, 83, 87 compact surface without boundary: 185 209
210
Index
constitutive equation: in Cartesian coordinates: 116 in curvilinear coordinates: 130 continuity of an immersion: 44 continuity of a deformation: 55 continuity of a surface: 100 contravariant basis: 14, 15 of the tangent plane: 62, 66, 79 contravariant components of: applied body force density: 122 applied surface force density: 122 rst Piola-Kirchhoff stress tensor: 129 rst fundamental form: 66 linearized stress couple: 192 linearized stress resultant tensor: 192 metric tensor: 14 second Piola-Kirchhoff stress tensor: 126 three-dimensional elasticity tensor: 130 shell elasticity tensor: 162, 171 coordinate line on a surface: 65 coordinate line in a three-dimensional set: 13 coordinate: : 11,12 cylindrical spherical : 11, 12, 62, 63 : 62, 63 stereographic covariant basis: 12, 13, 15 of the tangent plane: 62, 65, 66, 79 covariant components of: change of curvature tensor: 161 change of metric tensor: 159 displacement eld: 121 displacement eld on a surface: 159, 161 rst fundamental form: 66 Green-St Venant strain tensor: 126 linearized change of curvature tensor: 169 linearized change of metric tensor: 165 linearized strain tensor: 133 metric tensor: 14, 24 metric tensor of a surface: 66 Riemann curvature tensor: 25 Riemann curvature tensor of a surface: 84 second fundamental form: 73 third fundamental form: 170 vector eld: 20, 79, 80
Index covariant derivative of the rst order: of a tensor eld: 129, 192 of a vector eld: 22, 23, 81 covariant derivative of the second order: of a tensor eld: 192 of a vector eld: 169 curvature: of a planar curve: 70, 71 center of : 70 Gaussian : 76 : 76 line of mean : 76 principal : 76 : 76 radius of curvilinear coordinates: for a shell: 158 in a three-dimensional open set: 11, 12, 20 on a surface: 62, 64 deformation: 44, 56, 113 deformation gradient: 44 deformed conguration: 44, 113 deformed surface: 159 developable surface: 77 displacement gradient: 115 displacement traction-problem: in Cartesian coordinates: 117 in curvilinear coordinates: 131 linearized in curvilinear coordinates: 134, 151, 187 displacement vector eld: 113 in Cartesian coordinates: 113 in curvilinear coordinates: domain in Rn : 16 elastic material: 116 elasticity tensor: in Cartesian coordinates: 119 in curvilinear coordinates: 130 shell : 162, 171, 185 elliptic point: 77, 78 elliptic surface: 184, 196 elliptic surface without boundary: 185
211
212 energy:
Index
in Cartesian coordinates: 118 in curvilinear coordinates: 132 for a linearly elastic shell: 171 Koiter Koiter for a nonlinearly elastic shell: 162 equations of equilibrium: in Cartesian coordinates: 114 in curvilinear coordinates: 129 linearized in curvilinear coordinates: 134 Euclidean space: 11 Euler characteristic: 77 existence of solutions: 147, 180, 186 rst fundamental form: 66 contravariant components of : 66 : 66 covariant components of exural shell: linearly elastic : 195 nonlinearly elastic : 163 fundamental theorem of Riemannian geometry: 27, 86 fundamental theorem of surface theory: 87 -convergence: 163 Gauss-Bonnet theorem: 77 Gauss equations: 82, 87 Gauss formula: 81 Gauss Theorema Egregium: 77, 85 Gaussian curvature: 76, 77, 78, 85 genus of a surface: 77 Greens formula: 114 Green-St Venant strain tensor: in Cartesian coordinates: 115 in curvilinear coordinates: 126 Hookes law in curvilinear coordinates: 133 hyperbolic point: 77, 78 hyperelastic material: 117, 131 immersion: 13, 15, 26, 65, 66 innitesimal rigid displacement: 140 on a surface: 178 innitesimal rigid displacement lemma: in curvilinear coordinates: 140 on a surface: 177
Index isometry: 39 proper : 97 Kirchhoff-Love assumption: 162 Kirchhoff-Love displacement eld: 177 Koiter energy: for a linearly elastic shell: 171 for a nonlinear elastic shell: 162 Koiter linear shell equations: 170 Koiter nonlinear shell equations: 159 Korn inequality: in Cartesian coordinates: 180 in curvilinear coordinates: 137, 141 on a surface: 172, 179, 181 on an elliptic surface: 184 nonlinear : 57 Lam e constants: 118 Lax-Milgram lemma: 148, 172 Lemma of J.L. Lions: 135, 137 length element: 17, 67, 68 line of curvature: 76 Liouville theorem: 40 Mazur-Ulam theorem: 40 mean curvature: 76 membrane shell: linearly elastic : 196 : 197 linearly elastic generalized nonlinearly elastic : 163 metric tensor: 14, 24, 44 contravariant components of the : 14 : 24 covariant components of the metric tensor on a surface: 66 : 66 contravariant components of the covariant components of the : 66 middle surface of a shell: 156 minimization problem: 147, 150, 187, 188, 195, 196, 198 mixed components of the second fundamental form: 76, 169 Nash theorem: 27 natural state: 118 nonlinear Korn inequality: 57 parabolic point: 77, 78
213
214
Index
Piola-Kirchhoff stress tensor (rst ): 114 in Cartesian coordinates: 114 in curvilinear coordinates: 129 Piola-Kirchhoff stress tensor (second ): 114 in Cartesian coordinates: 114 in curvilinear coordinates: 126 planar point: 76 Poincar e lemma: 34 principal curvature: 73, 76 principal radius of curvature: 76 principal direction: 76 principle of virtual work: in Cartesian coordinates: 114 in curvilinear coordinates: 126 in curvilinear coordinates: 134 linearized pure displacement problem: in Cartesian coordinates: 117 in curvilinear coordinates: 131 linearized in curvilinear coordinates: 134, 187 pure traction problem: in Cartesian coordinates: 117 in curvilinear coordinates: 131 linearized in curvilinear coordinates: 134, 149, 188 radius of curvature: 70 principal : 76 reference conguration: 113, 156, 157 regularity of solutions: 150, 151, 192 response function: in Cartesian coordinates: 116 in curvilinear coordinates: 130 Riemann curvature tensor: 25 of a surface: 84 Riemannian geometry: 26, 27 Riemannian metric: 26, 27 rigid transformation: 36, 96, 140 rigidity theorem: 27, 36 for surfaces: 96 second fundamental form: 69 covariant components of : 73 mixed components of : 76, 169
Index shell: curvilinear coordinates for a : 158 elastic : 156, 157, 159, 160 elasticity tensor: 162, 171, 185 middle surface of a : 156 : 156, 157 reference conguration of a thickness of a : 156 spherical coordinates: 62 stationary point of a functional: 118, 134, 162, 171 stereographic coordinates: 62 stored energy function: in Cartesian coordinates: 117 in curvilinear coordinates: 131 in Koiter energy: 162, 198 strain tensor: Cauchy-Green : 44, 57, 115 : 116, 133 linearized stress couple: 192 stress resultant: 192 stress tensor: rst Piola-Kirchhoff : 114 : 114 second Piola-Kirchhoff strongly elliptic system: 150, 193 surface: 61, 64 compact without boundary: 185 : 77 developable elliptic : 184, 196 without boundary: 185 elliptic genus of a : 77 middle of a shell: 156 : 172, 179, 182 Korn inequality on a Theorema Egregium of Gauss: 77, 85 thickness of a shell: 156 third fundamental form: 95, 170 umbilical point: 76 volume element: 16, 17 Weingarten formula: 81
215

An Introduction To Differential Geometry With Applications To Elasticity - Ciarlet

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Introduction To Differential Geometry With Applications To Elasticity - Ciarlet

Uploaded by

Copyright:

Available Formats

Philippe G. Ciarlet City University of Hong Kong

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . linear Koiter shell eld dened on a . . . . . . . . . . . . . . . . . . . . . . . .

185 193 201 209

Chapter 1 THREE-DIMENSIONAL DIFFERENTIAL GEOMETRY

Three-dimensional dierential geometry

gij (x)g j (x) for i = 1, 2, 3.

Three-dimensional dierential geometry

Let be an open subset of R3 and let

Three-dimensional dierential geometry

Three-dimensional dierential geometry

VOLUMES, AREAS, AND LENGTHS IN CURVILINEAR COORDINATES

Volumes, areas, and lengths in curvilinear coordinates

Three-dimensional dierential geometry

In particular, the volume of D is given by vol D :=

ni (x)g ij (x)nj (x) d(x).

In particular, the area of is given by area :=

ni (x)g ij (x)nj (x) d(x).

df i df j (t) (t) dt. dt dt

Covariant derivatives of a vector eld

COVARIANT DERIVATIVES OF A VECTOR FIELD

Three-dimensional dierential geometry

vi (x)g i (x) g2 (x) v2 (x) v1 (x) g1 (x)

Covariant derivatives of a vector eld

Three-dimensional dierential geometry

Covariant derivatives of a vector eld

can be also dened by the relations j (vi g i ) = vi j g i vi

p p 0 (b) The Christoel symbols p ij := g i g j = ji C () satisfy the relations j i i g p = p ij g and j g q = jq g i .

To establish the other relations j g q = i jq g i , we note that

Three-dimensional dierential geometry

1 (j giq + i gjq q gij ) = i g j g q = jiq 2

EXISTENCE OF AN IMMERSION DEFINED ON AN OPEN SET IN R3 WITH A PRESCRIBED METRIC TENSOR

Three-dimensional dierential geometry

Sect. 1.6] Existence of an immersion with a prescribed metric tensor

Three-dimensional dierential geometry

where ijq := 1 (j giq + i gjq q gij ), 2 p ij := g pq ijq with (g pq ) := (gij )1 .

1 (j giq + i gjq q gij ), 2

dene the functions

which are equivalent to the relations

Three-dimensional dierential geometry

Gi j = p p , so that we also have ij t t j t =

dj Gi (t, ) = q (t, )q (t, ), 0 t 1, ij (G(t, )) dt t j (0, ) = 0.

Three-dimensional dierential geometry

j (1, ) = 0 for all 0 1, since G(1, ) = x1 for all 0 1.

possesses partial derivatives in

i F p (x) = p ij (x)F p (x).

Sect. 1.6] Existence of an immersion with a prescribed metric tensor

1 (j giq + i gjq q gij ), 2

Three-dimensional dierential geometry

Hence the matrix eld (gij ) C 2 (; S3 > ) satises

Sect. 1.6] Existence of an immersion with a prescribed metric tensor

Three-dimensional dierential geometry

UNIQUENESS UP TO ISOMETRIES OF IMMERSIONS WITH THE SAME METRIC TENSOR

Uniqueness of immersions with the same metric tensor

Three-dimensional dierential geometry

Uniqueness of immersions with the same metric tensor

Three-dimensional dierential geometry

Uniqueness of immersions with the same metric tensor

Three-dimensional dierential geometry

Uniqueness of immersions with the same metric tensor

Three-dimensional dierential geometry

CONTINUITY OF AN IMMERSION AS A FUNCTION OF ITS METRIC TENSOR

An immersion as a function of its metric tensor