Jet space or jet bundle

Note. In some texts the notation Jn(Rp,Rq) is used to mean what here we denote by Jn(Rp×Rq) being Rp×RqRp the trivial bundle.

First approach

(@saunders1989geometry)
Let π:EM be a smooth vector bundle, with dim(Ex)=q. We can interpret it as M being a manifold of independent variables and sections of π as functions on M. The set of sections forms a C(M)-module.

Two sections s,s~ are said to be k-equivalent at a point xM if their graphs are tangent to each other with order k at s(x)=s~(x)E. By "their graphs are tangent" we mean that if we express the sections as functions f and f~ in a local trivialization, the partial derivatives of f and f~ at x coincide up to order k. This fact does not depend on the chosen trivialization.

The equivalence class of s, denoted by [s]xk is called the k-jet of s at x, and the collection of all the classes, Jk(E)x, is the jet space at x. Finally, we can construct a bundle

Jk(E)=xMJk(E)x.

In this bundle we have special coordinates called the derivative coordinates. Let (U,u) be adapted coordinates on E, with u=(xi,uα). The derivative coordinates induced by (U,u) are (U~,u~) where

The dimension of Jk(E) is

p+q(p+kk).

If PJk(E)x, with P=[s]xk, we define π0(P)=s(x)E. See example 2.24 in @olver86.

The sections of Jk(E) of the form

x[s]xk,

being s:ME a section, are called holonomic sections or k-graphs (see @bryant2013exterior page 23 Theorem 3.2.), or kth-order jet prolongation of s, and will be denoted by jks.

The jet space or jet bundle Jk(E) has a natural distribution called the Cartan distribution.

Infinite jet bundle

The infinite jet space J(E) is defined as the inverse limit of the inverse system

Jk+1(π)πk+1,kJk(π)J1(π)π1,0J0(π)=EπM

In this case, Cartan distribution is involutive (see wikipedia) and its dimension agrees with dimM. We get a decomposition of TJ(E) in a vertical bundle (the subspace of TpJ(E) tangent to the fibre) and an horizontal one (the subspace belonging to the Cartan distribution). This decomposition induces a similar decomposition in the cotangent space TJ(E), and also in the ring of differential forms Ω(J(E)) giving rise to the variational bicomplex.

Second approach: Sketch of the formalization a la Olver for the extended jet bundle

Locally, any pair of submanifolds of a manifold M of the same dimensions is given by the graphs of smooth functions f and f~ (with independent and dependent variables between them of M). These functions are said to be k-equivalent at xM if there is a coordinate chart such that the partial derivatives at x coincide up to order k:

Jfα(x0)=Jf~α(x0),α=1,,q,0#Jk

This idea is independent of the coordinate chart, and therefore we can define an equivalence relation between submanifolds being tangent up to order k at x. The equivalence class [N]xk is called a jet. The union of all jets at x is the jet space at x, Jk(M)x and the jet bundle is

Jk(M)=xMJk(M)x

See @olver86 page 218 for more details. There he doesn't treat the case of sections of bundles but the construction let define the extended jet bundle (something like a projective version).

Third approach: Sketch of the formalization of Wikipedia

fj)(x)=gj)(x)

for j=0k. The equivalence class of f is denoted by [f]xk and is called the k-jet of f.

(φα)Exk(φβ)

This is like saying that curves α and β have k-order contact. The equivalence class of α is denoted by [α]xk and is called the k-jet of α. We will denote by

J0k(R,M)x$$thesetofall$k$jetsat$x$.As$x$variesover$M$theset$J0k(R,M)x$producesafibrebundleknownasthe$k$thordertangentbundle:$TkM$.If$f,g:MN$aremapsovertwomanifolds,wesaythat$fExkg$iffforevery$γ:RM$

(f\circ \gamma) E^k_x (g\circ \gamma)

Inthepreviouscase,ifwetake$N$asafiberbundleover$M$,wecanconsider$f$and$g$tobesections.Thesetofall$k$jetsat$x$isdenotedby$Jxk(M,N)$,andifweconsider$x$varyingat$M$wecanconstructafiberbundleover$M$denoted$Jk(N)$.