Bourbaki

§ 13. Point distributions

13.1. Symmetric tensors and Banach spaces

13.1.1. Let E be a module over a commutative ring A. If n is an integer ≥0\geq 0, denote by TSn(E)TS^n(E) the module of symmetric tensors of rank n of E (A, III, p. 71 and A, IV, § 5, no 2); the direct sum of the TSn(E)TS^n(E) is denoted by TS(E)TS(E). If r is an integer, one of the symbols ∞,ω\infty,\omega, put

TS(r)(E)=⨁n≤rTSn(E)andTS(r)+(E)=⨁1≤n≤rTSn(E); TS^{(r)}(E)=\bigoplus_{n\leq r}TS^n(E)\quad\text{and}\quad TS^{(r)+}(E)=\bigoplus_{1\leq n\leq r}TS^n(E);

consequently, if r≥0r\geq 0, TSr(E)TS^{r}(E) is the direct sum of TS(0)(E)=ATS^{(0)}(E)=A and of TS(r)+(E)TS^{(r)+}(E). One has TS(1)+(E)=ETS^{(1)+}(E)=E and TS(∞)(E)=TS(ω)(E)=TS(E)TS^{(\infty)}(E)=TS^{(\omega)}(E)=TS(E).

If nn is an integer ≥0\geq 0, and if x∈Ex\in E, denote by γn(x)\gamma_n(x) the element x⊗⋯⊗xx\otimes\cdots\otimes x of TSn(E)TS^n(E); for n=0n=0, γ0(x)=1\gamma_0(x)=1.

13.1.2. (« Polynomial mappings »). Suppose that A is integral infinite and that E is a free a-module. Let F be an a-module. A mapping f:E→Ff:E\rightarrow F is said to be homogeneous polynomial of degree n if it satisfies the following equivalent conditions (cf. A, IV, § 5, no 9):

a) There exists a linear mapping f~:TSn(E)→F\tilde f:TS^n(E)\rightarrow F such that f(x)=f~(γn(x))f(x)=\tilde f(\gamma_n(x)) for all x∈Ex\in E.

b) There exists a multilinear mapping u:En→Fu:E^n\rightarrow F such that f(x)=u(x,…,x)f(x)=u(x,\ldots,x) for all x∈Ex\in E.

Suppose that this is the case. The mapping f~\tilde f satisfying a) is then unique; if t∈TSn(E)t\in TS^n(E), denote by ⟨f,t⟩\langle f,t\rangle the element f~(t)\tilde f(t). If u:En→Fu:E^n\rightarrow F satisfies b), the corresponding linear mapping of ⊗nE\otimes^nE into F coincides in the submodule TSn(E)TS^n(E). When n!n! is invertible in A, one may choose uu symmetric, and this in a unique way.

13.1.3. Let E be a Banach space over K and F a separated polynomial space over K. Let U be an open neighbourhood of 0 in E, f:U→Ff:U\rightarrow F a mapping of class CrC^r (r∈NKr\in\mathbf{N}_K), and tt an element of TS(r)(E)TS^{(r)}(E), cf. 13.1.1 (applied to the ring A=KA=K). Let kk be an integer ≤r\leq r such that t∈TS(k)(E)t\in TS^{(k)}(E). One can decompose tt and ff uniquely into

t=t0+⋯+tk,with ti∈TSi(E) t=t_0+\cdots+t_k,\qquad \text{with }t_i\in TS^i(E)

and

f=f0+⋯+fk+h, f=f_0+\cdots+f_k+h,

where fif_i is a homogeneous continuous polynomial of degree ii (0≤i≤k0 \leq i \leq k) and where hh has contact of order ≥k\geq k with 0 at the point 0. Put then

⟨f,t⟩=∑i=0k⟨fi,ti⟩; \langle f, t \rangle = \sum_{i=0}^{k} \langle f_i, t_i \rangle;

this is an element of F which does not depend on the choice of kk.

13.1.4. Let E and E’ be two Banach spaces, U an open neighbourhood of 0 in E and φ:U→E′\varphi : U \to E' a mapping of class CrC^r such that φ(0)=0\varphi(0) = 0. Let t∈TS(r)(E)t \in TS^{(r)}(E). There exists one and only one element t′∈TS(r)(E′)t' \in TS^{(r)}(E') such that, for every separated polynormed space F, every open neighbourhood U’ of 0 in E’, and every mapping f:U′→Ff : U' \to F of class CrC^r, one has

(*)

⟨f,t′⟩=⟨f∘φ,t⟩. \langle f, t' \rangle = \langle f \circ \varphi, t \rangle.

This element t′t' is denoted by φ∗(t)\varphi_*(t). The mapping

φ∗:TS(r)(E)→TS(r)(E′) \varphi_* : TS^{(r)}(E) \to TS^{(r)}(E')

thus defined is linear.

13.2. Point distributions

In this No., X denotes a manifold of class CrC^r, and xx a point of X.

13.2.1. Let A\mathscr{A} be the set of pairs (c,t)(c, t), where c=(U,φ,E)c = (U, \varphi, E) is a chart of X centred at xx, and where t∈TS(r)(E)t \in TS^{(r)}(E). Two elements ((U,φ,E),t)((U, \varphi, E), t) and ((U′,φ′,E′),t′)((U', \varphi', E'), t') of A\mathscr{A} are said to be equivalent if t′=γ∗(t)t' = \gamma_*(t), where γ=φ′∘φ−1\gamma = \varphi' \circ \varphi^{-1}. In this way one obtains an equivalence relation on A\mathscr{A}; an equivalence class for this relation is called a point distribution at xx on X.

Let Tx(r)(X)T_x^{(r)}(X) be the set of point distributions at xx on X. If cc is a chart of X centred at xx, the mapping

θc:TS(r)(E)→Tx(r)(X) \theta_c : TS^{(r)}(E) \to T_x^{(r)}(X)

which associates with tt the class of (c,t)(c, t), is a bijection. In all that follows, Tx(r)(X)T_x^{(r)}(X) is endowed with the structure of a K-vector space obtained by transporting that of TS(r)(E)TS^{(r)}(E) by means of θc\theta_c; this structure does not depend on the choice of cc. Analogously, if k≤rk \leq r, Tx(k)(X)T_x^{(k)}(X) and Tx(k)+(X)T_x^{(k)+}(X) denote the images of TS(k)(E)TS^{(k)}(E) and TS(k)+(E)TS^{(k)+}(E) under θc\theta_c; they do not depend on the choice of cc. An element of Tx(r)(X)T_x^{(r)}(X) is said to be of order ≤k\leq k if it belongs to Tx(k)(X)T_x^{(k)}(X). One has

Tx(k)(X)=Tx(0)(X)⊕Tx(k)+(X). T_x^{(k)}(X) = T_x^{(0)}(X) \oplus T_x^{(k)+}(X).

There exists one and only one element εx\varepsilon_x of Tx(0)(X)T_x^{(0)}(X) such that θc(1)=εx\theta_c(1) = \varepsilon_x for every chart cc of X centred at xx. If tt is a point distribution at xx, the constant term of tt is the element λ\lambda of K such that t−λεx∈Tx(r)+(X)t - \lambda \varepsilon_x \in T_x^{(r)+}(X); one says that tt is without constant term if its constant term is zero.

13.2.2. Let F be a separated polynormed space, and let ff be a function of class CrC^r with values in F, defined on a neighbourhood of xx. Let tt be a point-supported distribution at xx on X.

Consider a chart c=(U,φ,E)c = (U, \varphi, E) of XX centred at xx, and put fc=f∘φ−1f_c = f \circ \varphi^{-1} and tc=θc−1(t)t_c = \theta_c^{-1}(t); the element ⟨fc,tc⟩\langle f_c, t_c \rangle of FF defined in 13.1.3 does not depend on the choice of cc; we denote it by ⟨f,t⟩\langle f, t \rangle. One has εx(f)=f(x)\varepsilon_x(f) = f(x).¹ The constant term of tt is ⟨1,t⟩\langle 1, t \rangle.

If F′F' is a separated polynormed space and if u:F→F′u : F \to F' is continuous linear, one has ⟨u∘f,t⟩=u(⟨f,t⟩)\langle u \circ f, t \rangle = u(\langle f, t \rangle).

Suppose that FF is a Banach space, and that tt is of order ≤k\leq k, with kk finite. Let j∈Jxk(X,F)j \in J_x^k(X, F), cf. 12.1.2, and let f∈jf \in j. Then ⟨f,t⟩\langle f, t \rangle depends only on the jet jj; we denote it by ⟨j,t⟩\langle j, t \rangle. The mapping from Tx(k)(X)×Jxk(X,F)T_x^{(k)}(X) \times J_x^k(X, F) into FF thus defined is bilinear. When F=KF = K, we have put Jxk(X,F)=Pxk(X)J_x^k(X, F) = P_x^k(X), cf. 12.6.2, and we obtain a bilinear form on Tx(k)(X)×Pxk(X)T_x^{(k)}(X) \times P_x^k(X).

13.2.3. Let φ:X→Y\varphi : X \to Y be a morphism of manifolds of class CrC^r and let y=φ(x)y = \varphi(x). Let c=(U,ψ,E)c = (U, \psi, E) be a chart of XX centred at xx and let c′=(U′,ψ′,E′)c' = (U', \psi', E') be a chart of YY centred at yy; let φ~\tilde{\varphi} be the expression of φ\varphi in these charts (5.3.2), and let φ~∗\tilde{\varphi}_* be the corresponding mapping from TS(r)(E)TS^{(r)}(E) into TS(r)(E′)TS^{(r)}(E'), cf. 13.1.4. There exists one and only one linear mapping, denoted by Tx(r)(φ)T_x^{(r)}(\varphi) or φ∗\varphi_*, from Tx(r)(X)T_x^{(r)}(X) into Ty(r)(Y)T_y^{(r)}(Y), making the diagram commutative

TS(r)(E)→φ~∗TS(r)(E′)↓θc↓θcTx(r)(X)→φ∗Ty(r)(Y). \begin{array}{ccc} TS^{(r)}(E) & \xrightarrow{\tilde{\varphi}_*} & TS^{(r)}(E') \\ \downarrow_{\theta_c} & & \downarrow_{\theta_c} \\ T_x^{(r)}(X) & \xrightarrow{\varphi_*} & T_y^{(r)}(Y). \end{array}

This mapping does not depend on the choice of cc and c′c'. If t∈Tx(r)(X)t \in T_x^{(r)}(X), φ∗(t)\varphi_*(t) is called the image of tt by φ\varphi. We have φ∗(εx)=εy\varphi_*(\varepsilon_x) = \varepsilon_y. If k≤rk \leq r, we have

φ∗(Tx(k)(X))⊂Ty(k)(Y)andφ∗(Tx(k)+(X))⊂Ty(k)+(Y). \varphi_*(T_x^{(k)}(X)) \subset T_y^{(k)}(Y) \quad \text{and} \quad \varphi_*(T_x^{(k)+}(X)) \subset T_y^{(k)+}(Y).

If φ′:X→Y\varphi' : X \to Y is a morphism of class CrC^r mapping xx onto yy, we have φ∗=φ∗′\varphi_* = \varphi'_* if and only if jxr(φ)=jxr(φ′)j_x^r(\varphi) = j_x^r(\varphi'), cf. 12.1.

If φ\varphi is an immersion (resp. a submersion) at xx, φ∗\varphi_* is injective (resp. surjective); the converse is true if dim⁡yY<+∞\dim_y Y < +\infty.

If φ′:Y→Z\varphi' : Y \to Z is a morphism of CrC^r manifolds, and if t∈Tx(r)(X)t \in T_x^{(r)}(X), we have (φ′∘φ)∗(t)=φ∗′(φ∗(t))(\varphi' \circ \varphi)_*(t) = \varphi'_*(\varphi_*(t)).

Let FF be a separated polynormed space, and let ff be a function of class CrC^r with values in FF, defined in a neighbourhood of yy. We then have

(1)(1)
⟨f,φ∗(t)⟩=⟨f∘φ,t⟩for all t∈Tx(r)(X). \langle f, \varphi_*(t) \rangle = \langle f \circ \varphi, t \rangle \quad \text{for all } t \in T_x^{(r)}(X).
For given tt, the relations (1) (for variable ff and FF) characterize φ∗(t)\varphi_*(t); when dim⁡yY<+∞\dim_y Y < +\infty, we may restrict ourselves to F=KF = K.

13.2.4. Suppose that XX is an open submanifold of a Banach space EE. Denote by φx\varphi_x the mapping y↦y−xy \mapsto y - x of XX into EE; the chart c=(X,φx,E)c = (X, \varphi_x, E) is centred

¹ When K=RK = \mathbf{R} or C\mathbf{C} and XX is locally compact, the preceding formula leads to identifying the point distribution εx\varepsilon_x with the Dirac measure εx\varepsilon_x defined in JNT, III, § 1, No. 3. More generally, every measure with finite support in XX is identified with a distribution with finite support in the sense of No. 13.6.

at xx. We then identify Tx(r)(X)T_x^{(r)}(X) with TS(r)(E)TS^{(r)}(E) by means of θc−1\theta_c^{-1}; in particular, we have Tx(∞)(X)=TS(E)T_x^{(\infty)}(X) = TS(E).

Let FF be a Banach space, let u:E→Fu : E \to F be a continuous linear mapping, and let x∈E,y∈Fx \in E, y \in F be such that u(x)=yu(x) = y. Identify as above Tx(∞)(E)T_x^{(\infty)}(E) with TS(E)TS(E) and Ty(∞)(F)T_y^{(\infty)}(F) with TS(F)TS(F). The mapping

u∗:TS(E)→TS(F)(cf. 13.2.3) u_* : TS(E) \to TS(F) \quad (\text{cf. } 13.2.3)
coincides with the mapping TS(u)TS(u) induced by the canonical extension of uu to the tensor algebra.

13.2.5. If k≤rk \leq r, denote by T(k)(X)T^{(k)}(X) (resp. T(k)+(X)T^{(k)+}(X)) the sum set of the Ta(k)(X)T_a^{(k)}(X) (resp. of the Ta(k)+(X)T_a^{(k)+}(X)) for a∈Xa \in X. A mapping t:X→T(k)(X)t : X \to T^{(k)}(X) such that t(a)∈Ta(k)(X)t(a) \in T_a^{(k)}(X) for all a∈Xa \in X is called a field of point distributions of order ≤k\leq k.

Suppose kk finite and XX locally of finite dimension. For every a∈Xa \in X, the bilinear form (j,t)↦⟨j,t⟩(j, t) \mapsto \langle j, t \rangle (cf. 13.2.2) defines an isomorphism iai_a of Ta(k)(X)T_a^{(k)}(X) onto the dual Pak(X)∗P_a^k(X)^* of Pak(X)P_a^k(X). The iai_a define a bijection i:T(k)(X)→Pk(X)∗i : T^{(k)}(X) \to P^k(X)^*, where Pk(X)∗P^k(X)^* denotes the dual of the vector bundle Pk(X)P^k(X); by transport of structure by means of i−1i^{-1}, one endows T(k)(X)T^{(k)}(X) with a structure of vector bundle with base XX and of class Cr−kC^{r-k}. A field of point distributions of order ≤k\leq k is said to be of class CsC^s, with s≤r−ks \leq r - k, if it is a section of class CsC^s of the vector bundle T(k)(X)T^{(k)}(X).

Let φ:X→Y\varphi : X \to Y be a morphism of manifolds of class CrC^r and suppose that YY is, like XX, locally of finite dimension. Then φ∗:T(k)(X)→T(k)(Y)\varphi_* : T^{(k)}(X) \to T^{(k)}(Y) is a φ\varphi-morphism of vector bundles of class Cr−kC^{r-k}.

13.2.6. Let kk be an integer such that 0≤k≤r0 \leq k \leq r. If UU is an open set of a Banach space EE of finite dimension, the vector bundle T(k)(U)T^{(k)}(U) is identified, thanks to 13.2.4, with the trivial bundle of fibre TS(k)(E)TS^{(k)}(E).

Take in particular E=KnE = K^n, with n≥0n \geq 0, and let (e1,…,en)(e_1, \ldots, e_n) be the canonical basis of EE. If α=(α1,…,αn)\alpha = (\alpha_1, \ldots, \alpha_n) is an element of Nn\mathbf{N}^n, denote by Δα\Delta^\alpha the element γα1(e1)…γαn(en)\gamma_{\alpha_1}(e_1) \ldots \gamma_{\alpha_n}(e_n) of TS(E)TS(E) (the product used being the symmetric product of symmetric tensors, cf. A, IV, § 5, No. 3).(1) The Δα\Delta^\alpha, for ∣α∣≤k|\alpha| \leq k, form a basis of TS(k)(E)TS^{(k)}(E); if a∈Kna \in K^n, denote by Δaα\Delta_a^\alpha the corresponding elements of Ta(k)(E)T_a^{(k)}(E). Let FF be a separated polynormed space, and ff a function of class CrC^r, with values in FF, defined in a neighbourhood of aa; if α∈Nn\alpha \in \mathbf{N}^n, the element ⟨f,Δaα⟩\langle f, \Delta_a^\alpha \rangle coincides with the element (Δαf)(a)(\Delta^\alpha f)(a) defined in 2.5.3, 3.2.1 and 4.2.1.

Suppose XX is locally of finite dimension, and let ξ=(ξ1,…,ξn)\xi = (\xi^1, \ldots, \xi^n) be a system of coordinates of XX in an open set UU. If a∈Ua \in U and if α\alpha is a multi-index such that ∣α∣≤k|\alpha| \leq k, denote by (Δξα)a(\Delta_\xi^\alpha)_a the point distribution at aa whose image by ξ\xi is the point distribution Δξ(a)α\Delta_{\xi(a)}^\alpha on KnK^n; the fields of distributions

Δξα:a↦(Δξα)a(∣α∣≤k) \Delta_\xi^\alpha : a \mapsto (\Delta_\xi^\alpha)_a \quad (|\alpha| \leq k)
form a frame on UU of the vector bundle T(k)(X)T^{(k)}(X). If f:U→Ff : U \to F is of class CrC^r, put ⟨f,(Δξα)a⟩=Δξαf(a)\langle f, (\Delta_\xi^\alpha)_a \rangle = \Delta_\xi^\alpha f(a) and denote by Δξαf\Delta_\xi^\alpha f the function a↦Δξαf(a)a \mapsto \Delta_\xi^\alpha f(a); this is a function of class Cr−kC^{r-k} in UU.

(1) If m=∣α∣m = |\alpha| and if Σ\Sigma is the set of mappings σ\sigma of {1,…,m}\{1, \ldots, m\} into {1,…,n}\{1, \ldots, n\} such that Card σ−1(i)=αi\mathrm{Card}\, \sigma^{-1}(i) = \alpha_i, we have

Δα=γα1(e1)…γαn(en)=∑σ∈Σeσ(1)⊗⋯⊗eσ(n),cf. A, IV, § 5, n° 4. \Delta^\alpha = \gamma_{\alpha_1}(e_1) \ldots \gamma_{\alpha_n}(e_n) = \sum_{\sigma \in \Sigma} e_{\sigma(1)} \otimes \cdots \otimes e_{\sigma(n)}, \quad \text{cf. A, IV, § 5, n° 4.}

13.3. Point distributions and tangent spaces

In this No., XX denotes a variety of class CrC^r.

13.3.1. Let x∈Xx \in X and let kk be an integer ≤r\leq r. Let c=(U,φ,E)c=(U,\varphi,E) be a chart of XX centred at xx; the isomorphism θc:TS(X)(E)→Tx(X)\theta_c:T^S(X)(E)\to T_x(X) defined in 13.2.1 maps TS(k)(E)T^S{}^{(k)}(E) onto Tx(k)(X)T_x^{(k)}(X) and TS(k−1)(E)T^S{}^{(k-1)}(E) onto Tx(k−1)(X)T_x^{(k-1)}(X); by restriction and passing to the quotient, it induces an isomorphism

θc,k:TS(k)(E)=TS(k)(E)/TS(k−1)(E)⟶Tx(k)(X)/Tx(k−1)(X). \theta_{c,k}:T^S{}^{(k)}(E)=T^S{}^{(k)}(E)/T^S{}^{(k-1)}(E)\longrightarrow T_x^{(k)}(X)/T_x^{(k-1)}(X).

On the other hand, cc defines an isomorphism, already denoted by θc\theta_c, of EE onto the tangent space Tx(X)T_x(X), cf. 5.5.1. Let iki_k denote the composite

Tx(k)(X)/Tx(k−1)(X)→θc,k−1TS(k)(E)→TS(θc)TS(k)(Tx(X)). T_x^{(k)}(X)/T_x^{(k-1)}(X)\xrightarrow{\theta_{c,k}^{-1}}T^S{}^{(k)}(E)\xrightarrow{T^S(\theta_c)}T^S{}^{(k)}(T_x(X)).

The isomorphism iki_k is independent of the choice of cc; for k=1k=1, it is used to identify Tx(1)(X)=Tx(1)(X)/Tx(0)(X)T_x^{(1)}(X)=T_x^{(1)}(X)/T_x^{(0)}(X) with the tangent space Tx(X)T_x(X). If t∈Tx(k)(X)t\in T_x^{(k)}(X), we allow ourselves to denote by ik(t)i_k(t) the image by iki_k of the class of tt modulo Tx(k−1)(X)T_x^{(k-1)}(X).

Put

gr⁡kTx(r)(X)=Tx(k)(X)/Tx(k−1)(X) \operatorname{gr}_k T_x^{(r)}(X)=T_x^{(k)}(X)/T_x^{(k-1)}(X)

and

gr⁡Tx(r)(X)=⨁0≤k≤rgr⁡kTx(r)(X). \operatorname{gr}T_x^{(r)}(X)=\bigoplus_{0\leq k\leq r}\operatorname{gr}_k T_x^{(r)}(X).

We say that gr⁡Tx(r)(X)\operatorname{gr}T_x^{(r)}(X) is the graded associated with the increasing filtration (Tx(k)(X))k≤r(T_x^{(k)}(X))_{k\leq r} of Tx(r)(X)T_x^{(r)}(X). The iki_k define an isomorphism of graded vector spaces

i:gr⁡Tx(r)(X)⟶TS(r)(Tx(X)). i:\operatorname{gr}T_x^{(r)}(X)\longrightarrow T^S{}^{(r)}(T_x(X)).

If r=∞r=\infty or ω\omega, ii is an isomorphism of gr⁡Tx(r)(X)\operatorname{gr}T_x^{(r)}(X) onto TS(Tx(X))T^S(T_x(X)). When one wishes to specify xx (or XX, or both), one writes ixi_x (or iXi_X, or ix,Xi_{x,X}) in place of ii.

13.3.2. In addition to the above hypotheses, suppose that XX is locally finite-dimensional. The isomorphisms iki_k corresponding to the different points of XX define an isomorphism of vector bundles

ik:T(k)(X)/T(k−1)(X)⟶TS(k)(T(X)) i_k:T^{(k)}(X)/T^{(k-1)}(X)\longrightarrow TS^{(k)}(T(X))

where TS(k)(T(X))TS^{(k)}(T(X)) denotes the vector bundle of class Cr−1C^{r-1} obtained from T(X)T(X) by the finite-dimensional vector functor TS(k)TS^{(k)} (cf. 7.6.5). For k≥1k\geq 1, iki_k is an isomorphism of class Cr−kC^{r-k}; for k=0k=0, it is the identity mapping of the trivial bundle KXK_X.

13.3.3. Example. — The hypotheses being as above, let ξ=(ξ1,…,ξn)\xi=(\xi^1,\ldots,\xi^n) be a system of coordinates of XX at xx. Let (∂i)1≤i≤n(\partial_i)_{1\leq i\leq n} be the basis of Tx(X)T_x(X) defined by ξ\xi (cf. 5.5.8) and let (Δxα)∣α∣≤k(\Delta^\alpha_x)_{|\alpha|\leq k} be that of Tx(k)(X)T_x^{(k)}(X) defined in 13.2.6. We have:

ik(Δxα)=0if ∣α∣<k i_k(\Delta^\alpha_x)=0 \qquad \text{if } |\alpha|<k
ik(Δxα)=γα1(∂1,x)⋯γαn(∂n,x)if ∣α∣=k. i_k(\Delta^\alpha_x)=\gamma_{\alpha_1}(\partial_1,x)\cdots\gamma_{\alpha_n}(\partial_n,x) \qquad \text{if } |\alpha|=k.

[^1]

13.3.4. Suppose that K is of characteristic zero or that X is locally of finite dimension. Let k be an integer such that 0≤k≤r0 \leq k \leq r and let x∈Xx \in X; let F be a Banach space. By 12.6.8, we have an exact sequence

(i)

0→Pk(Tx(X);F)→iPxk(FX)→σPxk−1(FX)→0, 0 \to P_k(T_x(X); F) \xrightarrow{i} P_x^k(F_X) \xrightarrow{\sigma} P_x^{k-1}(F_X) \to 0,
with σ=rk,k−1\sigma = r^{k, k-1}. On the other hand, by 13.3.1, we have an exact sequence

(ii)

0←TSk(Tx(X))←iTx(k)(X)←sTx(k−1)(X)←0, 0 \leftarrow TS^k(T_x(X)) \xleftarrow{i} T_x^{(k)}(X) \xleftarrow{s} T_x^{(k-1)}(X) \leftarrow 0,
where s is the inclusion of Tx(k−1)(X)T_x^{(k-1)}(X) in Tx(k)(X)T_x^{(k)}(X) and i=iki = i_k. These exact sequences are “coupled in F”. More precisely, write them in the form:

(i)

0→A→iB→σC→0 0 \to A \xrightarrow{i} B \xrightarrow{\sigma} C \to 0

(ii)

0←A′←iB′←sC′←0. 0 \leftarrow A' \xleftarrow{i} B' \xleftarrow{s} C' \leftarrow 0.

If a∈Aa \in A and a′∈A′a' \in A', the element ⟨a,a′⟩\langle a, a' \rangle of F is defined by 13.1.2; if b∈Bb \in B and b′∈B′b' \in B', (resp. if c∈Cc \in C and c′∈C′c' \in C'), the element ⟨b,b′⟩\langle b, b' \rangle (resp. ⟨c,c′⟩\langle c, c' \rangle) of F is defined by 13.2.2; we then have:

⟨ia,b′⟩=⟨a,ib′⟩and⟨σb,c′⟩=⟨b,sc′⟩ifa∈A, b∈B, b′∈B′, c′∈C′. \langle ia, b' \rangle = \langle a, ib' \rangle \quad \text{and} \quad \langle \sigma b, c' \rangle = \langle b, sc' \rangle \quad \text{if} \quad a \in A,\ b \in B,\ b' \in B',\ c' \in C'.

13.3.5. Let Y be a variety of class CrC^r, let φ:X→Y\varphi : X \to Y be a morphism of class CrC^r and let x∈X, y∈Yx \in X,\ y \in Y be such that y=φ(x)y = \varphi(x). The mapping φ∗:Tx(r)(X)→Ty(r)(Y)\varphi_* : T_x^{(r)}(X) \to T_y^{(r)}(Y) defined in 13.2.3 is compatible with the filtrations of these spaces; it defines by passing to the associated graded spaces a linear mapping

gr⁡(φ∗):gr⁡ Tx(r)(X)→gr⁡ Ty(r)(Y). \operatorname{gr}(\varphi_*) : \operatorname{gr}\ T_x^{(r)}(X) \to \operatorname{gr}\ T_y^{(r)}(Y).
On the other hand, let TS(r)(Tx(φ))TS^{(r)}(T_x(\varphi)) be the mapping of TS(r)(Tx(X))TS^{(r)}(T_x(X)) into TS(r)(Ty(Y))TS^{(r)}(T_y(Y)) induced by the canonical extension TS(Tx(φ))TS(T_x(\varphi)) of Tx(φ)T_x(\varphi). The diagram

gr⁡ Tx(r)(X)→gr⁡(φ∗)gr⁡ Ty(r)(Y)↓iX,x↓iY,yTS(r)(Tx(X))→TS(r)(Tx(φ))TS(r)(Ty(Y)) \begin{array}{ccc} \operatorname{gr}\ T_x^{(r)}(X) & \xrightarrow{\operatorname{gr}(\varphi_*)} & \operatorname{gr}\ T_y^{(r)}(Y) \\ \downarrow^{i_{X,x}} & & \downarrow^{i_{Y,y}} \\ TS^{(r)}(T_x(X)) & \xrightarrow{TS^{(r)}(T_x(\varphi))} & TS^{(r)}(T_y(Y)) \end{array}

is commutative.

13.3.6. Suppose that K is of characteristic zero. Using the isomorphisms φM:Sk(M)→TSk(M)\varphi_M : S^k(M) \to TS^k(M) defined in A, IV, § 5, No. 8, we can replace, in all that precedes, the TSk(Tx(X))TS^k(T_x(X)) by the k-th symmetric powers Sk(Tx(X))S^k(T_x(X)) of the tangent space Tx(X)T_x(X).

13.4. Tensor product of point distributions

13.4.1. Let X1X_1 and X2X_2 be two manifolds of class CrC^r, let x1∈X1, x2∈X2x_1 \in X_1,\ x_2 \in X_2 and t1∈Tx1(k1)(X1), t2∈Tx2(k2)(X2)t_1 \in T_{x_1}^{(k_1)}(X_1),\ t_2 \in T_{x_2}^{(k_2)}(X_2) with k1+k2≤rk_1 + k_2 \leq r. Put X=X1×X2, x=(x1,x2)X = X_1 \times X_2,\ x = (x_1, x_2) and k = k_1 + k_2. For i = 1, 2, let c_i = (U_i, \varphi_i, E_i) be a chart of X_i centred at x_i and denote by t~i\tilde{t}_i the element of TS(E_i) such that θc(t~i)=ti\theta_c(\tilde{t}_i) = t_i, cf. 13.2.1. Let σ\sigma be the canonical isomorphism of TS(E_1) ⊗\otimes TS(E_2) onto TS(E_1 ×\times E_2), cf. A, IV, § 5, no. 5. The element σ(t~1⊗t~2)\sigma(\tilde{t}_1 \otimes \tilde{t}_2) is the symmetric product of t~1\tilde{t}_1 and t~2\tilde{t}_2 (identified with elements of TS(E_1 ×\times E_2) by means of the canonical injections TS(E_i) →\to TS(E_1 ×\times E_2)); it belongs to TS^{(k)}(E_1 ×\times E_2). Put c=c1×c2c = c_1 \times c_2; this is a chart of X centred at x. The image by θc\theta_c of σ(t~1⊗t~2)\sigma(\tilde{t}_1 \otimes \tilde{t}_2) is an element of T_x^{(k)}(X), which does not depend on the choice of the charts cic_i. It is called the direct product, or the symmetric tensor product (or simply the tensor product) of t1t_1 and t2t_2 and is denoted by t1×t2t_1 \times t_2 or t1⊗t2t_1 \otimes t_2.

We have εx1⊗εx2=εx\varepsilon_{x_1} \otimes \varepsilon_{x_2} = \varepsilon_x. The constant term of t1⊗t2t_1 \otimes t_2 is the product of the constant terms of t1t_1 and t2t_2.

The isomorphism (y1,y2)↦(y2,y1)(y_1, y_2) \mapsto (y_2, y_1) of X1×X2X_1 \times X_2 onto X2×X1X_2 \times X_1 transforms t1⊗t2t_1 \otimes t_2 into t2⊗t1t_2 \otimes t_1.

13.4.2 (« Associativity and functoriality »). Let XiX_i (i=1,2,3i = 1, 2, 3) be manifolds of class CrC^r, and let xi∈Xix_i \in X_i, ti∈Txi(ki)(Xi)t_i \in T^{(k_i)}_{x_i}(X_i) with k1+k2+k3≤rk_1 + k_2 + k_3 \leq r. We have

(t1⊗t2)⊗t3=t1⊗(t2⊗t3); (t_1 \otimes t_2) \otimes t_3 = t_1 \otimes (t_2 \otimes t_3);
we denote this point distribution by t1⊗t2⊗t3t_1 \otimes t_2 \otimes t_3. Finite tensor products of any number of factors are defined analogously.

Let φ1:X1→Y1\varphi_1 : X_1 \to Y_1 and φ2:X2→Y2\varphi_2 : X_2 \to Y_2 be morphisms of manifolds of class CrC^r, and let t1∈T(k1)(X1)t_1 \in T^{(k_1)}(X_1), t2∈T(k2)(X2)t_2 \in T^{(k_2)}(X_2) with k1+k2≤rk_1 + k_2 \leq r. We have

(φ1×φ2)∗(t1⊗t2)=φ1∗(t1)⊗φ2∗(t2). (\varphi_1 \times \varphi_2)_*(t_1 \otimes t_2) = \varphi_{1*}(t_1) \otimes \varphi_{2*}(t_2).

13.4.3. The notation being that of 13.4.1, let F1,F2F_1, F_2 and F be separated polynormed spaces, and let (u1,u2)↦u1.u2(u_1, u_2) \mapsto u_1 . u_2 be a continuous bilinear mapping of F1×F2F_1 \times F_2 into F. Let

fi:Xi→Fi(i=1,2) f_i : X_i \to F_i \quad (i = 1, 2)
be a mapping of class CrC^r, and define f1⊗f2:X→Ff_1 \otimes f_2 : X \to F by
(f1⊗f2)(y1,y2)=f1(y1).f2(y2). (f_1 \otimes f_2)(y_1, y_2) = f_1(y_1) . f_2(y_2).
The mapping f1⊗f2f_1 \otimes f_2 is of class CrC^r, and we have
⟨f1⊗f2,t1⊗t2⟩=⟨f1,t1⟩.⟨f2,t2⟩. \langle f_1 \otimes f_2, t_1 \otimes t_2 \rangle = \langle f_1, t_1 \rangle . \langle f_2, t_2 \rangle.

13.4.4. The notation being that of 13.4.1, let ff be a mapping of class CrC^r of X into a separated polynormed space F. If y1∈X1y_1 \in X_1, denote by fy1f_{y_1} the mapping y2↦f(y1,y2)y_2 \mapsto f(y_1, y_2) of X2X_2 into F and put g(y1)=⟨fy1,t2⟩g(y_1) = \langle f_{y_1}, t_2 \rangle. The function g:X1→Fg : X_1 \to F thus defined is of class Cr−k1C^{r - k_1}, and we have

⟨f,t1⊗t2⟩=⟨g,t1⟩, \langle f, t_1 \otimes t_2 \rangle = \langle g, t_1 \rangle,
in other words
⟨f,t1⊗t2⟩=⟨y1↦⟨y2↦f(y1,y2),t2⟩,t1⟩. \langle f, t_1 \otimes t_2 \rangle = \langle y_1 \mapsto \langle y_2 \mapsto f(y_1, y_2), t_2 \rangle, t_1 \rangle.
Analogously
⟨f,t1⊗t2⟩=⟨y2↦⟨y1↦f(y1,y2),t1⟩,t2⟩. \langle f, t_1 \otimes t_2 \rangle = \langle y_2 \mapsto \langle y_1 \mapsto f(y_1, y_2), t_1 \rangle, t_2 \rangle.

13.4.5. The notation being that of 13.4.1, let

αk1,k2:Tx1(k1)(X1)⊗Tx2(k2)(X2)→Tx(k)(X) \alpha_{k_1, k_2} : T^{(k_1)}_{x_1}(X_1) \otimes T^{(k_2)}_{x_2}(X_2) \to T^{(k)}_x(X)

be the linear mapping defined by the bilinear mapping (t1,t2)↦t1⊗t2(t_1, t_2) \mapsto t_1 \otimes t_2; this mapping is injective; it is bijective if k1k_1 and k2k_2 are infinite.

If n≤rn \leq r, denote by Tx(n)(X1,X2)T^{(n)}_x(X_1, X_2) the vector subspace of Tx1(n)(X1)⊗Tx2(n)(X2)T^{(n)}_{x_1}(X_1) \otimes T^{(n)}_{x_2}(X_2) generated by the subspaces Tx1(k1)(X1)⊗Tx2(k2)(X2)T^{(k_1)}_{x_1}(X_1) \otimes T^{(k_2)}_{x_2}(X_2) for k1+k2≤nk_1 + k_2 \leq n. There exists one and only one linear mapping

αn:Tx(n)(X1,X2)→Tx(n)(X) \alpha_n : T^{(n)}_x(X_1, X_2) \to T^{(n)}_x(X)

extending the mappings αk1,k2\alpha_{k_1, k_2} defined above; it is an isomorphism. In what follows, we identify Tx(n)(X)T^{(n)}_x(X), by means of αn−1\alpha_n^{-1}, with the vector subspace Tx(n)(X1,X2)T^{(n)}_x(X_1, X_2) of Tx1(n)(X1)⊗Tx2(n)(X2)T^{(n)}_{x_1}(X_1) \otimes T^{(n)}_{x_2}(X_2).

13.4.6. Retain the notation of 13.4.5 and 13.3.1. By passing to the quotient, αk1,k2\alpha_{k_1, k_2} defines a linear mapping

εk1,k2:gr⁡k1Tx1(r)(X1)⊗gr⁡k2Tx2(r)(X2)→gr⁡kTx(r)(X). \varepsilon_{k_1, k_2} : \operatorname{gr}_{k_1} T^{(r)}_{x_1}(X_1) \otimes \operatorname{gr}_{k_2} T^{(r)}_{x_2}(X_2) \to \operatorname{gr}_k T^{(r)}_x(X).

The mappings εk1,k2\varepsilon_{k_1, k_2} (k1+k2≤rk_1 + k_2 \leq r) are the components of a graded linear mapping of degree zero

ε:⨁k1+k2≤r(gr⁡k1Tx1(r)(X1)⊗gr⁡k2Tx2(r)(X2))→⨁k≤rgr⁡kTx(r)(X) \varepsilon : \bigoplus_{k_1 + k_2 \leq r} (\operatorname{gr}_{k_1} T^{(r)}_{x_1}(X_1) \otimes \operatorname{gr}_{k_2} T^{(r)}_{x_2}(X_2)) \to \bigoplus_{k \leq r} \operatorname{gr}_k T^{(r)}_x(X)

which is an isomorphism. The diagram

⨁k1+k2≤r(gr⁡k1Tx1(r)(X1)⊗gr⁡k2Tx2(r)(X2))→εgr⁡Tx(r)(X)↓i12↓iX⨁k1+k2≤r(TSk1(Tx1(X1))⊗TSk2(Tx2(X2)))→σ⨁k≤rTSk(Tx1(X1)×Tx2(X2)) \begin{array}{ccc} \bigoplus_{k_1 + k_2 \leq r} (\operatorname{gr}_{k_1} T^{(r)}_{x_1}(X_1) \otimes \operatorname{gr}_{k_2} T^{(r)}_{x_2}(X_2)) & \xrightarrow{\varepsilon} & \operatorname{gr} T^{(r)}_x(X) \\ \downarrow i_{12} & & \downarrow i_X \\ \bigoplus_{k_1 + k_2 \leq r} (\mathrm{TS}^{k_1}(T_{x_1}(X_1)) \otimes \mathrm{TS}^{k_2}(T_{x_2}(X_2))) & \xrightarrow{\sigma} & \bigoplus_{k \leq r} \mathrm{TS}^k(T_{x_1}(X_1) \times T_{x_2}(X_2)) \end{array}

is commutative (in this diagram, i12i_{12} denotes the homomorphism induced by ix1⊗ix2i_{x_1} \otimes i_{x_2} and σ\sigma is the isomorphism defined in A, IV, § 5, No. 5). If rr is infinite, this diagram is simply written:

gr⁡Tx1(∞)(X1)⊗gr⁡Tx2(∞)(X2)→εgr⁡Tx(∞)(X)↓ix1⊗ix2↓iXTS(Tx1(X1))⊗TS(Tx2(X2))→σTS(Tx(X)). \begin{array}{ccc} \operatorname{gr} T^{(\infty)}_{x_1}(X_1) \otimes \operatorname{gr} T^{(\infty)}_{x_2}(X_2) & \xrightarrow{\varepsilon} & \operatorname{gr} T^{(\infty)}_x(X) \\ \downarrow i_{x_1} \otimes i_{x_2} & & \downarrow i_X \\ \mathrm{TS}(T_{x_1}(X_1)) \otimes \mathrm{TS}(T_{x_2}(X_2)) & \xrightarrow{\sigma} & \mathrm{TS}(T_x(X)). \end{array}

13.5. Coproducts

In this number, XX denotes a variety of class CrC^r and xx a point of XX.

13.5.1. Let k≤rk \leq r. If Δ\Delta is the diagonal mapping y↦(y,y)y \mapsto (y, y) of XX into X×XX \times X, then Δ∗\Delta_* maps Tx(k)(X)T^{(k)}_x(X) into T(x,x)(k)(X×X)T^{(k)}_{(x,x)}(X \times X), which is a vector subspace of

Tx(k)(X)⊗Tx(k)(X)T^{(k)}_x(X)\otimes T^{(k)}_x(X), cf. 13.4.5. One thus obtains a linear mapping, again denoted by Δ∗\Delta_* (or cc):

Tx(k)(X)⟶Tx(k)(X)⊗Tx(k)(X); T^{(k)}_x(X)\longrightarrow T^{(k)}_x(X)\otimes T^{(k)}_x(X);

it is called the coproduct attached to Tx(k)(X)T^{(k)}_x(X). Endowed with this coproduct, Tx(k)(X)T^{(k)}_x(X) is a coassociative and cocommutative cogebra (A, III, p. 144-145); its counit is the linear form which associates to a point distribution its constant term (13.2.1). If k′≤kk'\leq k, the inclusion of Tx(k′)(X)T^{(k')}_x(X) in Tx(k)(X)T^{(k)}_x(X) is a morphism of cogebras. If c=(U,φ,E‾)c=(U,\varphi,\overline{E}) is a chart of XX centred at xx, the isomorphism

θc:T(k)(E)⟶Tx(k)(X) \theta_c:T^{(k)}(E)\longrightarrow T^{(k)}_x(X)

is an isomorphism of cogebras, T(k)(E)T^{(k)}(E) being endowed with the coproduct induced by that of T(E)T(E), cf. A, IV, § 5, no 7. If φ:X⟶Y\varphi:X\longrightarrow Y is a morphism of manifolds of class CrC^r, the mapping of T(k)(X)T^{(k)}(X) into Tφ(x)(k)(Y)T^{(k)}_{\varphi(x)}(Y) induced by φ∗\varphi_* is a morphism of cogebras.

13.5.2. Let F1F_1, F2F_2 and FF be separated polynomial spaces, and (u1,u2)↦u1.u2(u_1,u_2)\mapsto u_1.u_2 a continuous bilinear mapping of F1×F2F_1\times F_2 into FF. Let t∈Tx(r)(X)t\in T^{(r)}_x(X) and let

c(t)=∑juj⊗vj(uj,vj in Tx(r)(X)) c(t)=\sum_j u_j\otimes v_j\quad(u_j,v_j\text{ in }T^{(r)}_x(X))

be its image by the coproduct. Let fi:X⟶Fif_i:X\longrightarrow F_i (i=1,2)(i=1,2) be mappings of class CrC^r. We have

⟨f1.f2,t⟩=∑j⟨f1,uj⟩⟨f2,vj⟩, \langle f_1.f_2,t\rangle=\sum_j\langle f_1,u_j\rangle\langle f_2,v_j\rangle,

which we write, by abuse of notation:

⟨f1.f2,t⟩=⟨f1⊗f2,c(t)⟩. \langle f_1.f_2,t\rangle=\langle f_1\otimes f_2,c(t)\rangle.

13.5.3. Let t∈Tx(k)(X)t\in T^{(k)}_x(X), where k≤rk\leq r.

a) For c(t)=t⊗1c(t)=t\otimes 1, it is necessary and sufficient that tt be equal to 00 or to εx\varepsilon_x.

b) For c(t)=1⊗t+εx⊗tc(t)=1\otimes t+\varepsilon_x\otimes t, it is necessary and sufficient that tt be a tangent vector, i.e. that t∈Tx(1)(X)t\in T^{(1)}_x(X).

13.6. Distributions with finite support

13.6.1. Let XX be a manifold of class CrC^r, and let k≤rk\leq r. We denote by T(k)(X)\mathscr{T}^{(k)}(X) the direct sum of the spaces Tx(k)(X)T^{(k)}_x(X), for x∈Xx\in X. An element of T(k)(X)\mathscr{T}^{(k)}(X) is called a distribution with finite support on XX, of order ≤k\leq k. If f:X⟶Ff:X\longrightarrow F is a function of class CrC^r with values in a separated polynomial space FF, and if

t=∑x∈Xtx, t=\sum_{x\in X}t_x,

where tx∈Tx(k)(X)t_x\in T^{(k)}_x(X) for all x∈Xx\in X is a distribution with finite support, we put

⟨f,t⟩=∑x∈X⟨f,tx⟩. \langle f,t\rangle=\sum_{x\in X}\langle f,t_x\rangle.

Analogously, we put

c(t)=∑x∈Xc(tx), c(t)=\sum_{x\in X}c(t_x),

which equips T(k)(X)\mathscr{T}^{(k)}(X) with a coassociative, cocommutative cogebra structure, possessing a counit t↦⟨1,t⟩t\mapsto\langle1,t\rangle. The inclusions

T(k′)(X)⟶T(k)(X), \mathscr{T}^{(k')}(X)\longrightarrow\mathscr{T}^{(k)}(X),

where k′≤kk'\leq k, are cogebra morphisms.

If φ:X→Y\varphi : X \to Y is a morphism of varieties of class CrC^r, the mappings Tx(k)(φ):Tx(k)(X)→Tφ(x)(k)(Y)T_x^{(k)}(\varphi) : T_x^{(k)}(X) \to T_{\varphi(x)}^{(k)}(Y) relative to the various points xx of XX define a linear mapping φ∗\varphi_* of T(k)(X)\mathscr{T}^{(k)}(X) into T(k)(Y)\mathscr{T}^{(k)}(Y), which is a cogebra morphism.

If X1X_1 and X2X_2 are two varieties of class CrC^r, the mappings αk−1\alpha_k^{-1} (cf. 13.4.5) define a linear mapping

T(k)(X1×X2)→T(k)(X1)⊗T(k)(X2) \mathscr{T}^{(k)}(X_1 \times X_2) \to \mathscr{T}^{(k)}(X_1) \otimes \mathscr{T}^{(k)}(X_2)
which is injective; it is a cogebra morphism. If k=∞k = \infty or ω\omega, it is an isomorphism of cogebras.

13.6.2. Let XX be a variety of class CrC^r, and let VV be a KK-vector space of finite dimension. An element of T(r)(X)⊗V\mathscr{T}^{(r)}(X) \otimes V is called a distribution with finite support on XX with values in VV. When K=RK = \mathbf{R}, V=CV = \mathbf{C}, such a distribution is also called a complex distribution with finite support on XX. The definitions and results of the preceding No.s extend immediately, by linearity, to distributions with values in VV.

13.7. Weakening of structure

Suppose K=RK = \mathbf{R}. Let r′∈NKr' \in \mathbf{N}_K such that r′≤rr' \leq r. Let XX be a variety of class CrC^r, and let X′X' be the variety of class Cr′C^{r'} obtained by weakening of structure (5.13.1). Let x∈Xx \in X, and let t∈Tx(k)(X)t \in T_x^{(k)}(X), with k≤r′k \leq r'. Choose a chart c=(U,φ,E)c = (U, \varphi, E) of XX centred at xx, and let t~\tilde{t} be the element of TS(k)(E)TS^{(k)}(E) such that θc(t~)=t\theta_c(\tilde{t}) = t. Since cc is a chart of X′X' centred at xx, the element t′=θc(t~)t' = \theta_c(\tilde{t}) of Tx(k)(X′)T_x^{(k)}(X') is defined; it is independent of the choice of cc. The mapping t↦t′t \mapsto t' is a bijection of Tx(k)(X)T_x^{(k)}(X) onto Tx(k)(X′)T_x^{(k)}(X') by which these two spaces are identified. The identifications thus obtained are compatible with the operations ⟨f,t⟩,φ∗(t),t1⊗t2,c(t),…\langle f, t \rangle, \varphi_*(t), t_1 \otimes t_2, c(t), \ldots of the preceding No.s.

[^1]: Here again, the product of the γαi(∂i,x)\gamma_{\alpha_i}(\partial_i,x) is a symmetric product in TS(Tx(X))TS(T_x(X)).

Machine translation of the English, by gpt-5-6, gpt-5-6-mini, against glossary version 34. Not checked by a person. That model is a cut down version of the one the rest of this language was written by.