MathLabs
TheoremProved

The Regular Value (Preimage) Theorem

Statement

Let f:Mm→Nnf: M^m \to N^n be smooth and q∈Nq \in N a regular value (meaning dfpdf_p is surjective for every p∈f−1(q)p \in f^{-1}(q)). Then f−1(q)f^{-1}(q) is a smooth submanifold of MM of dimension m−nm-n.

Why is it true?

This theorem is the primary tool for manufacturing manifolds: instead of exhibiting an atlas by hand, we describe a shape as the zero set of a map and just check a linear-algebra condition (surjectivity of the differential) at each solution point.

Proof sketch

Fix p∈f−1(q)p \in f^{-1}(q). Since dfp:TpM→TqNdf_p: T_pM \to T_qN is surjective and dim⁡TpM=m≥n=dim⁡TqN\dim T_pM = m \ge n = \dim T_qN, its kernel K=ker⁡dfpK = \ker df_p has dimension m−nm-n. Choose a linear complement WW so TpM=K⊕WT_pM = K \oplus W with dim⁡W=n\dim W = n; then dfp∣W:W→TqNdf_p|_W : W \to T_qN is an isomorphism.

Work in local coordinates centered at pp and qq (via charts), so ff becomes a smooth map Rm→Rn\mathbb{R}^m \to \mathbb{R}^n with f(0)=0f(0)=0 and df0df_0 surjective. Reorder coordinates (x,y)∈Rm−n×Rn(x,y) \in \mathbb{R}^{m-n}\times\mathbb{R}^n so that ∂f/∂y\partial f/\partial y at 00 is the invertible n×nn\times n block (possible since df0df_0 has rank nn).

Define Φ(x,y)=(x,f(x,y))\Phi(x,y) = (x, f(x,y)). Then dΦ0=(I0∂f/∂x∂f/∂y)d\Phi_0 = \begin{pmatrix} I & 0 \\ \partial f/\partial x & \partial f/\partial y \end{pmatrix} has det⁡dΦ0=det⁡(∂f/∂y)≠0\det d\Phi_0 = \det(\partial f/\partial y) \ne 0, so Φ\Phi is a local diffeomorphism by the Inverse Function Theorem. In the new coordinates (x,y′)=Φ(x,y)(x,y') = \Phi(x,y), the equation f=qf=q (i.e. f=0f=0) becomes exactly y′=0y'=0.

So near pp, f−1(q)f^{-1}(q) is the set {y′=0}\{y'=0\}, which in these coordinates is literally an (m−n)(m-n)-dimensional coordinate slice — a smooth chart for f−1(q)f^{-1}(q). Since pp was arbitrary, every point of f−1(q)f^{-1}(q) has such a chart, and the transition maps between these charts are restrictions of the (smooth) transition maps of MM, hence smooth. Therefore f−1(q)f^{-1}(q) is a smooth (m−n)(m-n)-dimensional submanifold.

Topics that use this theorem

Step-by-step proofs

No step-by-step proof yet for this theorem.

References

  1. John M. Lee (2012). Introduction to Smooth Manifolds
  2. Victor Guillemin, Alan Pollack (1974). Differential Topology
  3. F. Bullo, R. M. Murray (1999). Riemannian Manifolds in Robot Motion Planning and Control