Differential (pushforward) of a smooth map

The linear map on tangent spaces induced by a smooth map, satisfying the chain rule.
Differential (pushforward) of a smooth map

Let M,NM,N be and let F:MNF:M\to N be a .

Definition (differential at a point). For each pMp\in M, the differential (or pushforward) of FF at pp is the linear map

dFp:TpMTF(p)N dF_p: T_pM \to T_{F(p)}N

between defined as follows (using the derivation model of tangent vectors): if vTpMv\in T_pM is a derivation at pp, then dFp(v)TF(p)NdF_p(v)\in T_{F(p)}N is the derivation given by

(dFp(v))(g):=v(gF) (dF_p(v))(g) := v(g\circ F)

for every smooth function gg defined near F(p)F(p).

Equivalently, if vv is represented by the velocity of a smooth curve γ:(ϵ,ϵ)M\gamma:(-\epsilon,\epsilon)\to M with γ(0)=p\gamma(0)=p, then dFp(v)dF_p(v) is represented by the velocity of FγF\circ\gamma at 00.

Bundle map form. The assignments pdFpp\mapsto dF_p assemble into a smooth map between total spaces of tangent bundles,

dF:TMTN, dF: TM \to TN,

covering FF (meaning πNdF=FπM\pi_N\circ dF = F\circ \pi_M, where πM:TMM\pi_M:TM\to M and πN:TNN\pi_N:TN\to N are the bundle projections). This viewpoint uses the functorially.

Chain rule. If G:NPG:N\to P is another smooth map, then for every pMp\in M,

d(GF)p=dGF(p)dFp. d(G\circ F)_p = dG_{F(p)}\circ dF_p.

The differential detects local rank properties: FF is a iff dFpdF_p is injective for all pp, and a iff dFpdF_p is surjective for all pp. In particular, the notion of is expressed in terms of surjectivity of dFpdF_p along a .

Examples

  1. A map R2R2\mathbb{R}^2\to\mathbb{R}^2. Let F(x,y)=(x2,y2)F(x,y)=(x^2,y^2). Then in standard coordinates,

    dF(x,y)(u,v)=(2xu,  2yv). dF_{(x,y)}(u,v) = (2x\,u,\;2y\,v).

    At (1,1)(1,1) the map has rank 22, while at (0,1)(0,1) it has rank 11, and at (0,0)(0,0) it has rank 00.

  2. Projection is a submersion. Let π:R2R\pi:\mathbb{R}^2\to\mathbb{R} be π(x,y)=x\pi(x,y)=x. Then

    dπ(x,y)(u,v)=u, d\pi_{(x,y)}(u,v)=u,

    which is surjective for every (x,y)(x,y). Thus π\pi is a , and its fibers are vertical lines.

  3. Inclusion is an immersion. Let i:S1R2i:S^1\hookrightarrow \mathbb{R}^2 be the standard inclusion. For each pS1p\in S^1, the map dip:TpS1TpR2di_p:T_pS^1\to T_p\mathbb{R}^2 is injective, so ii is a (in fact a ).