Reduction to the normalised case.
Let A=Jfβ(a). Since detJfβ(a)ξ =0, the matrix A is invertible. The map f~β(x)=Aβ1(f(x+a)βf(a)) is C1 by the chain rule and linearity of Aβ1, and satisfies f~β(0)=0 and Jf~ββ(0)=Aβ1A=Inβ. Since f=f(a)+Aβf~β(β
βa) and composition with the invertible affine map yβ¦f(a)+Ay preserves the local-C1-diffeomorphism property, it suffices to prove the theorem under the additional assumptions a=0, f(0)=0, Jfβ(0)=Inβ.
Step 1: Finding the contractive radius.
Define g:UβRn by g(x)=xβf(x). Then g is C1 and, since Jfβ(0)=Inβ, we have Jgβ(0)=InββInβ=0. Because the entries of Jgβ are continuous and vanish at 0, there exists r>0 with B(0,r)βU such that
(j,k=1βnβ(βxkββgjββ(x))2)1/2β€21βforΒ allΒ xβB(0,r).
Step 2: Mean-value bound and injectivity.
By the chain rule and the Fundamental Theorem of Calculus applied to tβ¦gjβ(xβ²+t(xβxβ²)), for any x,xβ²βB(0,r) and each index j:
gjβ(x)βgjβ(xβ²)=β«01βk=1βnββxkββgjββ(xβ²+t(xβxβ²))(xkββxkβ²β)dt.
By the Cauchy--Schwarz inequality and the bound from Step 1,
β₯g(x)βg(xβ²)β₯2ββ€21ββ₯xβxβ²β₯2β.
Hence β₯f(x)βf(xβ²)β₯2ββ₯β₯xβxβ²β₯2βββ₯g(x)βg(xβ²)β₯2ββ₯21ββ₯xβxβ²β₯2β, so f is injective on B(0,r).
Step 3: Existence of a local inverse via CMT.
Set W={yβRn:β₯yβ₯2β<r/2}. For each yβW, define Tyβ:B(0,r)βRn by Tyβ(x)=g(x)+y.
Tyβ is a contraction: β₯Tyβ(x)βTyβ(xβ²)β₯2β=β₯g(x)βg(xβ²)β₯2ββ€21ββ₯xβxβ²β₯2β.
Tyβ maps B(0,r) into itself: since g(0)=0βf(0)=0, for any xβB(0,r),
β₯Tyβ(x)β₯2ββ€β₯g(x)βg(0)β₯2β+β₯yβ₯2ββ€21ββ₯xβ₯2β+2rββ€r.
The space (B(0,r),dEβ) is complete, since B(0,r) is a closed subset of the complete Euclidean space (Rn,dEβ).
By the Contraction Mapping Theorem, each Tyβ has a unique fixed point xβ(y)βB(0,r). The fixed point equation Tyβ(xβ(y))=xβ(y) gives f(xβ(y))=y. Injectivity from Step 2 makes xβ:WβB(0,r) the unique local inverse of f on W.
Step 4: The local inverse is C1.
Let V=fβ1(W)β©B(0,r). Since f is continuous and W, B(0,r) are open, V is open, and fβ£Vβ:VβW is a bijection with inverse xβ.
From the contraction estimate β₯xβ(y)βxβ(yβ²)β₯2ββ€2β₯yβyβ²β₯2β, the inverse xβ is Lipschitz, hence continuous. Since detJfβ is continuous and nonzero at 0, shrinking r if necessary ensures Jfβ(x) is invertible for all xβV. Differentiating the identity f(xβ(y))=y by the chain rule gives
Jfβ(xβ(y))Jxββ(y)=Inβ,soJxββ(y)=(Jfβ(xβ(y)))β1.
The right-hand side is continuous in y (composition of continuous xβ with smooth Jfβ and matrix inversion), so xβ is C1. Hence fβ£Vβ:VβW is a local C1 diffeomorphism at a=0.