Bourbaki

§ 11. Stokes Formula

In this paragraph, we suppose K=R\mathbf{K}=\mathbf{R}.

11.1. Pieces

11.1.1. Let E be a Banach space. A subset S of E is called a closed half-space if there exists a continuous linear form h≠0h\neq 0 and a real number kk such that S={x∣h(x)≤k}S=\{x|h(x)\leq k\} (cf. EVT, II, § 2, No. 6); the boundary of S is then the closed hyperplane {x∣h(x)=k}\{x|h(x)=k\}; it is also called the boundary of S, and is denoted by ∂S\partial S.

11.1.2. Let X be a variety of class CrC^r and A a subset of X. We say that A is a piece of X if, for every a∈Aa\in A, there exists a chart c=(U,φ,E)c=(U,\varphi,E) of X at a such that φ(A∩U)\varphi(A\cap U) is an open set of a closed half-space of E. We put ∂A=A−A˚\partial A=A-\mathring{A} (set of non-interior points of A). This is a subvariety of A, which is called the boundary of A.[^1] If a∈∂Aa\in\partial A, there exists a chart c=(U,φ,E)c=(U,\varphi,E) of X centred at a and a closed hyperplane H of E such that φ(A∩U)\varphi(A\cap U) is an open neighbourhood of 0 in one of the closed half-spaces ScS_c of E defined by H (EVT, II, § 2, No. 6), and such that φ(∂A∩U)=H∩φ(A∩U)\varphi(\partial A\cap U)=H\cap\varphi(A\cap U). Such a chart is said to be adapted to A at a. The closed half-space ScS_c is then the unique closed half-space of E of which φ(A∩U)\varphi(A\cap U) is an open set.

For a closed subset A of X to be a piece of X, it is necessary and sufficient that A˚\mathring{A} be dense in A and that A−A˚A-\mathring{A} be a subvariety of X of codimension 1 at each of its points.

11.1.3. Examples

a) In Rn\mathbf{R}^n, a closed ball of radius >0>0 is a piece; its boundary is the corresponding sphere.

b) If A is a piece of the variety X and B is a closed subset of ∂A\partial A, then A−BA-B is a piece of X.

c) Let φ:X→X′\varphi:X\rightarrow X' be a morphism of varieties of class CrC^r, and let A′A' be a piece of X′X'. If φ\varphi is transverse to ∂A′\partial A' (5.11.6), φ−1(A′)\varphi^{-1}(A') is a piece of X whose boundary is φ−1(∂A′)\varphi^{-1}(\partial A').

In particular, let h:X→Rh : X \to \mathbf{R} be a function of class CrC^r, and let a∈Ra \in \mathbf{R}; suppose that there is no x∈Xx \in X such that h(x)=ah(x) = a and dxh=0d_x h = 0. The set {x∣h(x)≤a}\{ x | h(x) \leq a \} is a closed piece of XX with boundary h−1(a)h^{-1}(a).

d) If r=∞r = \infty and if XX is locally compact, every compact subset of XX admits a fundamental system of neighbourhoods consisting of compact pieces.

11.1.4. Let XX be a CrC^r-class manifold and AA a piece of XX. Let a∈∂Aa \in \partial A and v∈Ta(X)v \in T_a(X). Let c=(U,φ,E)c = (U, \varphi, E) be a chart of XX adapted to AA at aa (11.1.2) and let h=θc−1(v)h = \theta_c^{-1}(v) be the element of EE corresponding to vv (5.5.1). Let ScS_c be the closed half-space of EE of which φ(A∩U)\varphi(A \cap U) is an open set (11.1.2). We say that vv is an inward vector (resp. strictly inward, resp. outward, resp. strictly outward) for AA at aa if hh belongs to ScS_c (resp. S˙c,−Sc,−S˙c\dot{S}_c, -S_c, -\dot{S}_c), a condition which is independent of the choice of the adapted chart cc. We denote by Ta+(A)T_a^+(A) (resp. Ta−(A)T_a^-(A)) the set of outward (resp. inward) vectors of Ta(X)T_a(X) for AA. These are closed half-spaces of Ta(X)T_a(X), whose boundary contains 0. We have

Ta(∂A)=Ta+(A)∩Ta−(A). T_a(\partial A) = T_a^+(A) \cap T_a^-(A).

11.2. Stokes formula for pieces

In this No., let XX be a separated, pure CrC^r-class manifold (r≥2r \geq 2) of finite dimension nn. Let AA be a piece of XX and ii the canonical injection of ∂A\partial A into XX. Let EE be a Banach space.

11.2. Stokes formula for pieces

11.2.1. Let x∈∂Ax \in \partial A and let ξ\xi be an orientation of Tx(∂A)T_x(\partial A); we denote by i~x(ξ)\tilde{i}_x(\xi) the orientation of Tx(X)T_x(X) containing the elements v∧uv \wedge u, where vv is a vector strictly outward for AA at xx (11.1.4) and where uu is a nonzero element of ∧n−1Tx(∂A)\wedge^{n-1} T_x(\partial A) belonging to the orientation ξ\xi. The mapping i~x\tilde{i}_x is a bijection of Or(Tx(∂A))\mathrm{Or}(T_x(\partial A)) onto Or(Tx(X))\mathrm{Or}(T_x(X)). The mappings i~x\tilde{i}_x for x∈∂Ax \in \partial A, define a morphism i~:∂A~→X~\tilde{i} : \tilde{\partial A} \to \tilde{X} which is an orientation of ii (10.2.5). If ξ\xi is an orientation of XX, the orientation of ∂A\partial A associated with ξ\xi by i~\tilde{i} (10.2.5) is said to be defined by ξ\xi.

Example

When X=RnX = \mathbf{R}^n, ξ\xi is the usual orientation of Rn\mathbf{R}^n and AA is a closed ball of radius >0> 0, the orientation of the sphere ∂A\partial A defined by ξ\xi is the canonical orientation (10.2.8, b)).

11.2.2. Let ω\omega be a twisted differential form of degree pp on XX with values in EE (10.4.1); the inverse image i∗(ω)i^*(\omega) of ω\omega by the oriented morphism i:∂A→Xi : \partial A \to X is denoted by ω∣∂A\omega|_{\partial A} and is called the form induced by ω\omega on ∂A\partial A (cf. 10.4.3).

11.2.3. Let us make one of the following two hypotheses:

(i) ω\omega is a twisted differential form of degree n−1n-1 on XX, with values in EE; (ii) XX is oriented, ∂A\partial A is provided with the corresponding orientation (11.2.1) and ω\omega is a differential form of degree n−1n-1 on XX with values in EE.

Suppose in addition that ω\omega is of class C1C^1 and that the intersection of AA and the support of ω\omega is compact. The exterior differential dωd\omega of ω\omega is continuous (8.3.5 and 10.3.4).

The characteristic function of AA is essentially integrable for the vector measure defined by dωd\omega on XX and the differential form ω∣∂A\omega|_{\partial A} of degree n−1n - 1 on ∂A\partial A is integrable (10.4.3 and 10.4.4). One has

∫Adω=∫∂Aω(« Stokes formula »). \int_A d\omega = \int_{\partial A} \omega \quad \text{(« Stokes formula »)}.
11.2.4. Let α\alpha be a twisted differential form of degree nn on XX with values in EE, with compact support, and of class C1C^1. In order that there exist a twisted differential form ω\omega of degree n−1n - 1 on XX, with values in EE, with compact support and of class C1C^1, such that α=dω\alpha = d\omega, it is necessary that ∫Xα=0\int_X \alpha = 0; if XX is connected, this condition is sufficient; if in addition α\alpha is of class CkC^k (k≤r−1,k≠ωk \leq r - 1, k \neq \omega), one can choose ω\omega of class CkC^k.

11.2.5. Suppose that XX is the real variety underlying C\mathbf{C}, that EE is a complex Banach space and that AA is a compact piece of XX. Equip XX with the orientation defined by its complex structure (10.2.7). Let ff be a continuous mapping from AA into EE whose restriction to the interior A˚\mathring{A} of AA is holomorphic. Let dzdz denote the differential of the injection z:∂A→Cz : \partial A \to \mathbf{C}. The form f.dzf.dz, product of ff and dzdz, is a differential form of degree 1 on ∂A\partial A, with values in EE, of class C0C^0, and one has:

∫∂Af.dz=0(« Cauchy formula »). \int_{\partial A} f.dz = 0 \quad \text{(« Cauchy formula »)}.
When ff extends to a holomorphic mapping, still denoted ff, of an open neighbourhood UU of AA, with values in EE, the differential form f.dzf.dz (where zz denotes this time the canonical injection of UU into C\mathbf{C}) is of class C∞C^\infty on UU and its exterior differential is zero.

11.2.6 (Derivative of an integral). Suppose XX is oriented. Let YY be a variety of class CrC^r, and let α\alpha be a differential form of degree nn on YY, with values in EE, of class C1C^1, and with compact support. Let II be an open subset of R\mathbf{R} containing 0 and let g:I×X→Yg : I \times X \to Y be a morphism of class CrC^r; for t∈It \in I, denote by gtg_t the mapping x↦g(t,x)x \mapsto g(t, x) from XX into YY. Denote by ψ\psi the morphism from XX into T(Y)T(Y) defined by ψ(x)=T(0,x)(g)(1,0)\psi(x) = T_{(0, x)}(g)(1, 0); it is a lifting of g0g_0 (8.6.5). Suppose that the restriction of pr1:I×A→I\mathrm{pr}_1 : I \times A \to I to the intersection of I×AI \times A and the support of g∗(α)g^*(\alpha) is proper (TG, I, § 10). For every t∈It \in I, the intersection of AA and the support of gt∗(α)g_t^*(\alpha) is then compact; the mapping t↦∫Agt∗(α)t \mapsto \int_A g_t^*(\alpha) is of class C1C^1 on II and its derivative at the origin is given by the formula

ddt(∫Agt∗(α))t=0=∫Aθψ⋅α=∫Ai(ψ) dα+∫∂Ai(ψ) α, \frac{d}{dt} \left( \int_A g_t^*(\alpha) \right)_{t=0} = \int_A \theta_\psi \cdot \alpha = \int_A i(\psi) \, d\alpha + \int_{\partial A} i(\psi) \, \alpha,
cf. No. 8.6.

11.3. Stokes formula for locally polyhedral sets[^2]

In this No., XX denotes a real purely finite-dimensional variety of class CrC^r (r≥2r \geq 2); we suppose XX separated.

11.3.1. Let AA be a subset of a real finite-dimensional vector space. We say that AA is polyhedral if it is a finite union of finite intersections of closed half-spaces. A subset AA of XX is said to be locally polyhedral if, for every x∈Xx\in X, there exists a chart c=(U,φ,E)c=(U,\varphi,E) of XX at xx and a polyhedral subset AcA_c of EE such that φ(A∩U)=φ(U)∩Ac\varphi(A\cap U)=\varphi(U)\cap A_c. A piece of XX is a locally polyhedral subset.

11.3.2. Let AA be a closed subset of XX, and let Fr⁡(A)=A−A˚\operatorname{Fr}(A)=A-\mathring{A} be its frontier. A point x∈Fr⁡(A)x\in\operatorname{Fr}(A) is said to be regular if there exists an open neighbourhood UU of xx such that A∩UA\cap U is a piece of the manifold UU (in which case xx belongs to the boundary of A∩UA\cap U). We denote by ∂A\partial A the set of regular points of Fr⁡(A)\operatorname{Fr}(A) and call it the regular boundary (or simply the boundary) of AA. The set A′=A˚∪∂AA'=\mathring{A}\cup\partial A is a piece of XX, with boundary ∂A\partial A.

11.3.3. \ Example.* — Let H\mathscr{H} be a locally finite set of hyperplanes of a finite-dimensional real affine space EE, and let CC be a chamber of EE with respect to H\mathscr{H} (LIE, V, § 1, No. 3). The closure C‾\overline{C} of CC is a locally polyhedral subset of the manifold EE; the regular boundary of C‾\overline{C} is the union of the faces of CC (loc. cit., No. 4). In particular, if CC is an open simplex (loc. cit., No. 6) with vertices a0,…,ana_0,\ldots,a_n, the boundary of C‾\overline{C} is the union of the C(i)C_{(i)} (0≤i≤n0\leq i\leq n), where C(i)C_{(i)} is the open simplex with vertices a0,…,ai−1,ai+1,…,ana_0,\ldots,a_{i-1},a_{i+1},\ldots,a_n in the affine space generated by these vertices.*\*

11.3.4. Let AA be a locally polyhedral subset of XX, and let ω\omega be a twisted differential form of degree n−1n-1 on XX with values in a Banach space EE; suppose that ω\omega is of class C1C^1 and that the intersection of its support with AA is compact. Then the characteristic function of AA (resp. A′A', A^\hat{A}) is essentially integrable for the vector measure defined by dωd\omega on XX and one has

∫Adω=∫A′dω=∫A^dω. \int_A d\omega=\int_{A'}d\omega=\int_{\hat{A}}d\omega.

The differential form ω∣∂A\omega|\partial A (for the canonical orientation of the canonical injection of ∂A\partial A, considered as the boundary of the piece A′A', into XX) is integrable in ∂A\partial A and one has

∫Adω=∫∂Aω(« Stokes formula »). \int_A d\omega=\int_{\partial A}\omega \qquad\text{(« Stokes formula »).}

When XX is provided with an orientation, and ∂A\partial A is provided with the corresponding orientation (11.2.1), this formula is still valid if ω\omega is an ordinary differential form of degree n−1n-1 with values in EE, of class C1C^1 and such that A∩Supp⁡ωA\cap\operatorname{Supp}\omega is compact.

11.4. Relative Stokes formula (integration over the fibres)

In this number, let XX and SS be two (real) manifolds of class CrC^r and let π:X→S\pi:X\to S be a submersion. Let nn be an integer and suppose that, for every s∈Ss\in S, the fibre Xs=π−1(s)X_s=\pi^{-1}(s) of π\pi at ss (which is a submanifold of XX (5.10.5)) is purely of dimension nn.

4—B.

11.4.1. Let E and H be two vector spaces and let F be a finite-dimensional vector subspace of dimension n of H. Let p be an integer ⩾0\geqslant 0, let u be an alternating (n+p)(n + p)-linear mapping of Hn+pH^{n+p} into E and let t1,…,tpt_1, \ldots, t_p be elements of H/F; let t1′,…,tp′t'_1, \ldots, t'_p be representatives of t1,…,tpt_1, \ldots, t_p in H. The mapping

(x1,…,xn)↦u(t1′,…,tp′,x1,…,xn) (x_1, \ldots, x_n) \mapsto u(t'_1, \ldots, t'_p, x_1, \ldots, x_n)

is an alternating nn-linear mapping of FnF^n into E, which depends only on u and the tit_i; it is denoted by u⊥(t1,…,tp)u \perp (t_1, \ldots, t_p) (cf. A, III, p. 158 for the particular case E=KE = K). The mapping

(t1,…,tp)↦u⊥(t1,…,tp) (t_1, \ldots, t_p) \mapsto u \perp (t_1, \ldots, t_p)

is alternating pp-linear.

11.4.2 (« π\pi-torsion »). Consider the vector bundle T(X/S) on X (8.1.3); it is of finite rank n at every point of X. Put R~π=R~T(X/S)\tilde{R}_\pi = \tilde{R}_{T(X/S)} (10.3.1) and X~π=OrT(X/S)\tilde{X}_\pi = \mathrm{Or}_{T(X/S)} (10.2.2). Let s∈Ss \in S; in view of the natural identification of T(X/S)∣XsT(X/S)|_{X_s} with T(Xs)T(X_s) (8.1.3), the fibre at s of the mapping π~:X~π→S\tilde{\pi}: \tilde{X}_\pi \to S composed of π\pi and the canonical mapping X~π→X\tilde{X}_\pi \to X, is X~s\tilde{X}_s (10.2.4). A differential form π\pi-twisted on X is called a T(X/S)T(X/S)-twisted differential form (10.3.3).

11.4.3 (« S-orientation of an S-morphism »). Let X′X' be a manifold provided with a submersion π′:X′→S\pi': X' \to S whose fibres XsX_s are purely of finite dimension n′n', and let φ:X′→X\varphi: X' \to X be a morphism such that π′=π∘φ\pi' = \pi \circ \varphi. An S-orientation of φ\varphi is called a morphism φ~:X~π′→X~π\tilde{\varphi}: \tilde{X}'_\pi \to \tilde{X}_\pi, commuting with the action of the group {±1}\{ \pm 1 \} and such that the diagram

X~π′→φ~X~π∣∣X′→φX \begin{array}{ccc} \tilde{X}'_\pi & \xrightarrow{\tilde{\varphi}} & \tilde{X}_\pi \\ | & & | \\ X' & \xrightarrow{\varphi} & X \end{array}

is commutative. The data of an S-orientation of φ\varphi makes it possible, as in 10.4.2, to identify the bundles R~π′\tilde{R}_{\pi'} and φ∗(R~π)\varphi^*(\tilde{R}_\pi) and to define the inverse image of a π\pi-twisted differential form by the S-oriented morphism φ\varphi: it is a π′\pi'-twisted differential form on X′X'.

11.4.4. Let A be a piece of X such that the restriction of π\pi to ∂A\partial A is a submersion. For every s∈Ss \in S, the fibre Xs=π−1(s)X_s = \pi^{-1}(s) is a submanifold of X transverse to ∂A\partial A and A∩XsA \cap X_s is a piece of XsX_s, denoted by AsA_s, with boundary ∂A∩Xs\partial A \cap X_s.

Let i be the canonical injection of ∂A\partial A into X. There exists one and only one S-orientation i~\tilde{i} of i such that, for every s∈Ss \in S, the restriction of i~\tilde{i} to (∂As)∗(\partial A_s)^* (identified with the fibre at s of the submersion ∂A~π∣∂A→S\partial \tilde{A}_{\pi|\partial A} \to S) is the orientation defined in 11.2.1 of the canonical injection of ∂As\partial A_s into XsX_s. If ω\omega is a π\pi-twisted differential form on X, the inverse image of ω\omega by the morphism i thus oriented is denoted by ω∣∂A\omega|\partial A (cf. 11.2.2).

11.4.5 (« Interior product »). Let pp be an integer ⩾0\geqslant 0 and let ω\omega be an nn-twisted differential form of degree n+pn + p on XX, with values in a vector bundle EE with base XX. Let s∈Ss \in S and x∈Xsx \in X_s; we have ω(x)=ξ⊗u\omega(x) = \xi \otimes u, where ξ\xi is an orientation of Tx(Xs)=T(X/S)xT_x(X_s) = T(X/S)_x, identified with an element of (R~Xs)x=(R~π)x(\tilde{R}_{X_s})_x = (\tilde{R}_\pi)_x (10.3.2), and where uu is an alternating (n+p)(n + p)-linear mapping of Tx(X)T_x(X) into ExE_x. Let t1,…,tpt_1, \ldots, t_p be tangent vectors to SS at ss. Put

θ(x)=ξ⊗(u∧(t1,…,tp))∈(R~Xs)x⊗(Altn(T(X);E))x \theta(x) = \xi \otimes (u \wedge (t_1, \ldots, t_p)) \in (\tilde{R}_{X_s})_x \otimes (\mathrm{Alt}^n(T(X); E))_x

(cf. 11.4.1). This defines a twisted differential form θ:x↦θ(x)\theta : x \mapsto \theta(x) of degree nn on XsX_s, with values in E∣XsE|X_s. We denote it by ω∧(t1,…,tp)\omega \wedge (t_1, \ldots, t_p).

In particular, let MM be a vector bundle of class Cr−1C^{r-1} over SS and take E=π∗(M)E = \pi^*(M). The bundle E∣XsE|X_s is naturally identified with the trivial bundle over XsX_s defined by the Banach space MsM_s, and the form ω∧(t1,…,tp)\omega \wedge (t_1, \ldots, t_p) is identified with a twisted differential form on XsX_s, with values in MsM_s.

11.4.6. Retain the preceding notation (in particular E=π(∗M)E = \pi(*M)) and suppose in addition that XX is separated and ω\omega continuous. Let AA be a piece of XX such that the restriction of π\pi to ∂A\partial A is a submersion and the restriction of π\pi to the intersection of AA and the support of ω\omega is proper. For s∈Ss \in S and t1,…,tpt_1, \ldots, t_p in Ts(S)T_s(S), the form ω∧(t1,…,tp)\omega \wedge (t_1, \ldots, t_p) (11.4.5) is a continuous twisted differential form of degree nn on XsX_s, with values in the Banach space MsM_s, and its support meets AsA_s in a compact set. There exists one and only one differential form α\alpha of degree pp on SS, with values in the vector bundle MM, such that

α(s)(t1,…,tp)=∫Asω∧(t1,…,tp) \alpha(s)(t_1, \ldots, t_p) = \int_{A_s} \omega \wedge (t_1, \ldots, t_p)

for all s∈Ss \in S and t1,…,tpt_1, \ldots, t_p in Ts(S)T_s(S). The form α\alpha is denoted by ∫π∣Aω\int_{\pi|A} \omega (or simply ∫πω\int_\pi \omega when A=XA = X); it is said to be obtained by integrating ω\omega over AA along the fibres of π\pi. If ω\omega is of class CkC^k (0≤k≤r−1,k≤∞0 \leq k \leq r - 1, k \leq \infty), the same is true of ∫π∣Aω\int_{\pi|A} \omega.

When ω\omega is a form of degree <n< n, we agree that ∫π∣Aω=0\int_{\pi|A} \omega = 0.

11.4.7. Retain the preceding hypotheses and notation.

a) For every continuous scalar differential form β\beta on SS, we have

∫π∣A(π∗β)∧ω=β∧∫π∣Aω. \int_{\pi|A} (\pi^*\beta) \wedge \omega = \beta \wedge \int_{\pi|A} \omega.

b) Let η\eta be a continuous vector field on XX and ζ\zeta a continuous vector field on SS, such that η\eta is π\pi-related to ζ\zeta (8.2.6). We have

i(ζ)∫π∣Aω=∫π∣Ai(η)ω. i(\zeta) \int_{\pi|A} \omega = \int_{\pi|A} i(\eta)\omega.

c) Suppose in addition that η,ζ\eta, \zeta and ω\omega are of class C1C^1, that η(x)∈Tx(∂A)\eta(x) \in T_x(\partial A) for all x∈∂Ax \in \partial A and that MM is the trivial vector bundle defined by a Banach space EE. We have

θζ⋅∫π∣Aω=∫π∣Aθη⋅ω, \theta_\zeta \cdot \int_{\pi|A} \omega = \int_{\pi|A} \theta_\eta \cdot \omega,

(cf. 8.4.2 and 10.3.4).

d) If ω\omega is of degree n+pn + p and of class C1C^1, we have

d(∫π∣Aω)=∫π∣Adω+(−1)p∫π∣∂Aω∣∂A. d \left( \int_{\pi|A} \omega \right) = \int_{\pi|A} d\omega + (-1)^p \int_{\pi|\partial A} \omega|_{\partial A}.
e) Suppose that S is reduced to a point. The π\pi-twisted forms on X are then the ordinary twisted forms (10.4.1). If ω\omega is of degree nn, the form ∫π∣Aω\int_{\pi|A} \omega of degree 0 on S is the constant ∫Aω\int_A \omega (10.4.3).

11.4.8. Let π′:S→S′\pi': S \to S' be a submersion such that the fibres of π′\pi' are pure submanifolds of constant finite dimension n′n'. Put π′′=π′∘π\pi'' = \pi' \circ \pi; it is a submersion of X onto S' whose fibres are pure submanifolds of dimension m=n+n′m = n + n'.

Let x∈Xx \in X. The sequence

0⟶T(X/S)x→IdT(X/S′)x→Tx(π)T(S/S′)π(x)⟶0 0 \longrightarrow T(X/S)_x \xrightarrow{\mathrm{Id}} T(X/S')_x \xrightarrow{T_x(\pi)} T(S/S')_{\pi(x)} \longrightarrow 0
is exact. There exists one and only one isomorphism jj of the vector bundle R~π⊗π∗(R~π′)\tilde{R}_{\pi} \otimes \pi^*(\tilde{R}_{\pi'}) onto R~π′′\tilde{R}_{\pi''} such that, if ξ\xi (resp. η\eta) is an orientation of T(X/S)xT(X/S)_x (resp. of (π∗T(S/S′))x(\pi^*T(S/S'))_x) identified with T(S/S′)π(x)T(S/S')_{\pi(x)}, we have j(ξ⊗η)=ηξj(\xi \otimes \eta) = \eta \xi (the product being defined by the preceding exact sequence (10.2.1)).

Let M′M' be a vector bundle with base S′S'; put M=π′∗(M′)M = {\pi'}^*(M'), and E=π∗(M)=π′′∗(M′)E = \pi^*(M) = {\pi''}^*(M'). If ω\omega is a π′′\pi''-twisted differential form on X with values in E, the isomorphism jj allows it to be identified with a π\pi-twisted form with values in π∗(R~π′)⊗E\pi^*(\tilde{R}_{\pi'}) \otimes E, or equivalently with values in π∗(R~π′⊗M)\pi^*(\tilde{R}_{\pi'} \otimes M).

Soit de plus A une pièce de X telle que π∣∂A:∂A→S\pi|\partial A : \partial A \to S soit une submersion ; the same then holds for π′′∣∂A:∂A→S′\pi''|\partial A : \partial A \to S'. Suppose in addition that ω\omega is continuous and that the restriction of π′′\pi'' to A∩Supp⁡ωA \cap \operatorname{Supp} \omega is proper; the same then holds for the restriction of π\pi to A∩Supp⁡ωA \cap \operatorname{Supp} \omega. One can consider on the one hand the differential form ∫π′′∣Aω\int_{\pi''|A} \omega on S′S', which is a form with values in M′M', and on the other hand the differential form ∫π∣Aω\int_{\pi|A} \omega on S, which is a form with values in R~π′⊗M\tilde{R}_{\pi'} \otimes M, or equivalently a π′−\pi'-twisted form with values in M=π′∗(M′)M = {\pi'}^*(M'). Moreover, π(A)\pi(A) is an open submanifold of S and the restriction of π′\pi' to π(A)∩Supp⁡∫π∣Aω\pi(A) \cap \operatorname{Supp} \int_{\pi|A} \omega is proper. The differential form ∫π′∣π(A)∫π∣Aω\int_{\pi'|_{\pi(A)}} \int_{\pi|A} \omega is defined; it is a differential form on S′S', with values in M′M'. We have

∫π′∘π∣Aω=∫π′∣π(A)∫π∣Aω. \int_{\pi' \circ \pi|A} \omega = \int_{\pi'|_{\pi(A)}} \int_{\pi|A} \omega.
If A=XA = X, we have
∫π′∘πω=∫π′∫πω. \int_{\pi' \circ \pi} \omega = \int_{\pi'} \int_{\pi} \omega.

[^1]: This terminology comes from the fact that A is naturally endowed with a structure of a “manifold with boundary”, with boundary ∂A\partial A. For the definition of this category, as well as for that, more generally, of the category of “manifolds with corners”, the reader may refer to H. CARTAN, Séminaire 1961/62, Topologie Différentielle, exposés 1-2-3 (par A. DOUADY), Benjamin, New York, 1969. Let us point out that one can show that every “manifold with boundary” whose boundary is paracompact is isomorphic to a closed piece of a manifold. [^2]: The reader interested in more general cases may consult H. WHITNEY, Geometric Integration Theory, Chap. III, § 18 (Princeton Univ. Press, 1957).

Machine translation of the English, by gpt-5-6-mini, gpt-5-6, 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.