Benjamin McKay Introduction to Exterior Differential Systems February 13, 2022 ThisworkislicensedunderaCreativeCommonsAttribution-ShareAlike4.0UnportedLicense. iii Preface To the reader who wants to dip a toe in the water: read chapter 1. The reader who continues on to chapters 2 and 4 will pick up the rest of the tools. Subsequentchaptersprovethetheorems. Weassumethatthereaderisfamiliar with elementary differential geometry on manifolds and with differential forms. These lectures explain how to apply the Cartan–Kähler theorem to problems in differential geometry. Given some differential equations, we want to decide if they are locally solvable. The Cartan–Kähler theorem gives a linear algebra test: if the equations pass the test, they are locally solvable. We give the necessary background on partial differential equations in appendices A, B, and the (not so necessary) background on moving frames in appendices D, F, G. The reader should be aware of [4], which we will follow closely, and also the canonical reference work [3] and the canonical textbook [19]. I wrote these notes for lectures at the Banach Center in Warsaw, at the University of Paris Sud, and at the University of Rome Tor Vergata. I thank Jan Gutt, Gianni Manno, Giovanni Moreno, Nefton Pali, and Filippo Bracci for invitations to give those lectures, and Francesco Russo for the opportunity to write these notes at the University of Catania. v Contents 1 Exterior differential systems 1 2 Tableaux 13 3 Example: almost complex structures 25 4 Prolongation 33 5 Cartan’s test 39 6 The characteristic variety 47 7 Proof of the Cartan–Kähler theorem 53 8 Cauchy characteristics 59 9 Example: conformal maps of surfaces 69 10 Example: Weingarten surfaces 73 A The Cauchy–Kovalevskaya theorem 77 B Symbol and characteristic variety 87 C Analytic convergence 97 D The moving frame 101 E Curvature of surfaces 115 F The Gauss–Bonnet theorem 125 G Geodesics on surfaces 133 Hints 149 Bibliography 191 Index 195 vi Chapter 1 Exterior differential systems We define exterior differential systems, and state the Cartan–Kähler theorem. Background material from differential geometry By the expression analytic, we mean real analytic. We assume that all of our manifolds, maps, vector fields and differential forms are analytic. (In a few exercises, we describe some such thing as smooth, i.e. C∞, not necessarily analytic.) A submanifold is an immersion of manifolds, defined up to diffeomor- phism of the domain. Denote the dimension of a manifold M by dimM. The codimension of a submanifold S of M is dimM −dimS. A hypersurface in a manifold M is a submanifold of codimension one. A hyperplane in a vector space V is a linear subspace W ⊂V so that dim(V/W)=1. Differential equations encoded in differential forms To express a differential equation 0=f(cid:0)x,u,∂u(cid:1), add a variable p to represent ∂x the derivative ∂u. Let ϑ ..= du − pdx, ω ..= dx. Let M be the manifold ∂x M ..={(x,u,p)|0=f(x,u,p)} (assuming it is a manifold). Any submanifold of M on which 0=ϑ and 06=ω is locally the graph of a solution. It is easy to generalize this to any number of variables and any number of equations of any order. Exterior differential systems An integral manifold of a collection of differential forms is a submanifold on which the forms vanish. An exterior differential system on a manifold M is an ideal I ⊂ Ω∗ of differential forms, closed under exterior derivative, splitting into a direct sum I =I1⊕···⊕IdimM of forms of various degrees: Ik ..=I∩Ωk. Any collection of differential forms has the same integral manifolds as the exterior differential system it generates. 1 2 Exterior differential systems Some trivial examples: the exterior differential system generated by a. 0, b. Ω∗, c. the pullbacks of all forms via a submersion, d. dx1∧dy1+dx2∧dy2 in R4, e. dy−zdx on R3. 1.1 What are the integral manifolds of our trivial examples? By definition I0 = 0, i.e. there are no nonzero functions in I. If we wish some functions to vanish, we can replace our manifold M by the zero locus of those functions (which might not be a manifold, a technical detail we will ignore). Integral elements An integral element of an exterior differential system I is a linear subspace of a tangent space, on which all forms in I vanish. Every tangent space of any integral manifold is an integral element, but some integral elements of some exterior differential systems don’t arise as tangent spaces of integral manifolds. A 1-dimensional integral element is an integral line. A 2-dimensional integral element is an integral plane. 1.2 What are the integral elements of our trivial examples? 1.3 Supposethatthe1-formsinanexteriordifferentialsystemspanasubspace of constant rank in each cotangent space. Prove that there is an integral curve tangent to each integral line. 1.4 Give an example of an embedded integral manifold, whose every tangent space is a hyperplane in a unique integral element, but which is not a hyper- surface in an integral manifold. 1.5 Provethatak-dimensionallinearsubspaceofatangentspaceisanintegral element of an exterior differential system I just when all k-forms in I vanish on it. The polar equations of an integral element E are the linear functions w ∈T M 7→ϑ(w,e ,e ,...,e ) m 1 2 k where ϑ ∈ Ik+1 and e ,e ,...,e ∈ E, for any k. They vanish on a vector 1 2 k w just when the span of {w}∪E is an integral element. The set of polar equations of E is a linear subspace of T∗M. If an integral element E lies in m Integral elements as points of the Grassmann bundle 3 a larger one F, then all polar equations of E occur among those of F: larger integral elements have more (or at least the same) polar equations. 1.6 What are the polar equations of the integral elements of our trivial exam- ples? An integral flag of an exterior differential system is a sequence of nested integral elements 0=E ⊂E ⊂E ⊂···⊂E 0 1 2 p with dimensions 0,1,2,...,p. Danger: most authors require that a flag have subspaces of all dimensions; we don’t: we only require that the subspaces have all dimensions 0,1,2,...,p up to some dimension p. Successive subspaces have successivelylargerspace ofpolarequations. The characters s ,s ,...,s ofthe 0 1 p flag are the increments in rank of polar equations: E has polar equations of k rank s +s +···+s . 0 1 k ConsiderallflagsinsideagivenE . Polarequationsremainlinearlyindepen- p dent under small motions of a flag. So if a flag in E has maximal dimensional p polar equations, i.e. character sums s ,s +s ,...,s +···+s , then so do all 0 0 1 0 p nearby flags in E . The characters of the integral element E are those of any p p such flag. 1.7 What are the characters of the integral flags of our trivial examples? Integral elements as points of the Grassmann bundle The rank p Grassmann bundle of a manifold M is the set of all p-dimensional linear subspaces of tangent spaces of M. 1.8 Recall how charts are defined on the Grassmann bundle. Prove that the Grassmann bundle is a fiber bundle. The integral elements of an exterior differential system I form a subset of the Grassmann bundle. We would like to see if that subset is a submanifold, and predict its dimension. The Cartan–Kähler theorem An integral element, say of dimension p and with characters s ,...,s , pre- 0 p dicts the dimension dimM +s +2s +···+ps . It predicts correctly if the 1 2 p nearby integral elements form a submanifold of the Grassmann bundle of the predicted dimension. (We will see that they always sit in a submanifold of the predicteddimension.)Anintegralelementwhichcorrectlypredictsdimensionis involutive. An exterior differential system on a manifold M is involutive if each component of M contains an involutive maximal integral element. (We will see that involutivity of an exterior differential system implies that involutive integral elements occur at a dense open subset of points of M.) 4 Exterior differential systems Theorem 1.1 (Cartan–Kähler). Every involutive integral element of any ana- lytic exterior differential system is tangent to an analytic integral manifold. 1.9 Giveanexampleofanexteriordifferentialsystemwhosecharactersarenot constant. Example: Lagrangian submanifolds We employ the Cartan–Kähler theorem to prove the existence of Lagrangian submanifolds of complex Euclidean space. Let ϑ..=dx1∧dy1+dx2∧dy2+···+dxn∧dyn. Let I be the exterior differential system generated by ϑ on M ..=R2n. Writing spans of vectors in angle brackets, Flag Polar equations Characters E ={0} {0} s =0 0 0 E =h∂ i (cid:10)dy1(cid:11) s =1 1 x1 1 . . . . . . . . . E =h∂ ,∂ ,...,∂ i (cid:10)dy1,dy2,...,dyn(cid:11) s =1 n x1 x2 xn n The flag predicts dimM +s +2s +···+ns =2n+1+2+···+n. 1 2 n The nearby integral elements at a given point of M are parameterized by dy = adx, which we plug in to ϑ = 0 to see that a can be any symmetric matrix. So the space of integral elements has dimension n(n+1) n(n+1) dimM + =2n+ , 2 2 correctly predicted. The Cartan–Kähler theorem proves the existence of La- grangian submanifolds of complex Euclidean space, one (at least) through each point, tangent to each subspace dy =adx, at least for any symmetric matrix a close to 0. 1.10 On a complex manifold M, take a Kähler form ϑ and a holomorphic volume form Ψ, i.e. closed forms expressed in local complex coordinates as √ −1 ϑ= g dzµ∧dzν¯, 2 µν¯ Ψ =f(z)dz1∧dz2∧···∧dzn, withf(z)aholomorphicfunctionandg apositivedefiniteself-adjointcomplex µν¯ matrix offunctions. Prove the existence of special Lagrangian submanifolds, i.e. integral manifolds of both ϑ and the imaginary part of Ψ.