Statement

Let f:MPf:M\to P and g:NPg:N\to P be between finite-dimensional without boundary. Using the , form the map

f×g:M×NP×P,(x,y)(f(x),g(y)).f\times g:M\times N\longrightarrow P\times P,\qquad (x,y)\longmapsto(f(x),g(y)).

Then ff and gg are if and only if f×gf\times g is transverse to the

ΔP={(z,z):zP}.\Delta_P=\{(z,z):z\in P\}.

Moreover, (f×g)1(ΔP)(f\times g)^{-1}(\Delta_P) is the set-theoretic fiber product M×PNM\times_PN.

Tangent-space calculation

At (z,z)ΔP(z,z)\in\Delta_P,

T(z,z)ΔP={(u,u):uTzP}.T_{(z,z)}\Delta_P=\{(u,u):u\in T_zP\}.

The quotient of TzPTzPT_zP\oplus T_zP by this diagonal subspace is identified with TzPT_zP by (u,v)uv(u,v)\mapsto u-v. Consequently, transversality of f×gf\times g to ΔP\Delta_P is equivalent to surjectivity of

(v,w)dfx(v)dgy(w),(v,w)\longmapsto df_x(v)-dg_y(w),

which is exactly the condition dfx(TxM)+dgy(TyN)=TzPdf_x(T_xM)+dg_y(T_yN)=T_zP. Compare Guillemin and Pollack, Chapter 2.

Fiber-product consequence

When the equivalent conditions hold, the makes M×PNM\times_PN an embedded submanifold of M×NM\times N. Its is

{(v,w):dfx(v)=dgy(w)},\{(v,w):df_x(v)=dg_y(w)\},

and its dimension is dimM+dimNdimP\dim M+\dim N-\dim P. This is the construction.

Scope

The criterion is a reformulation, not an additional hypothesis. It is especially useful because coincidence equations become a single inverse-image problem. Versions for manifolds with boundary or corners require a notion of transversality compatible with the relevant strata.

References
  1. Victor Guillemin and Alan Pollack, Differential Topology, AMS Chelsea Publishing, 2010 reprint. DOI record. Relevant: Chapter 2, diagonal formulation of transversality.
  2. Morris W. Hirsch, Differential Topology, Springer, 1976. DOI record. Relevant: Chapter 3, transverse maps and inverse images.