Sketch: under the isometry (x, y) to x + iy, multiplication by a + ib is the linear map with matrix [[a, -b], [b, a]], so the complex difference-quotient condition and the Euclidean derivative condition for F translate into each other.
So the error of the linear approximation of F at p0 by Aa,b is exactly the error of the complex-linear approximation of f at z0 by a+ib, measured at corresponding points.
Step 3: complex differentiability gives the derivative matrix. This proves (i) from left to right. Suppose f is complex differentiable at z0 with f′(z0)=a+ib. Since U is open and z0∈U, first fix r>0 with BdC(z0,r)⊆U (Open Subset of a Metric Space (i)); r does not depend on what follows. Now let ε>0 be given. Take δ0>0 as in The Complex Derivative at a Point (i) for this ε, so that ∣Q(w)−(a+ib)∣<ε whenever w∈U and 0<∣w−z0∣<δ0, and put δ=min(r,δ0).
Let h∈R2 with ∥h∥<δ, and put w=z0+Φ(h)=Φ(p0+h). Then ∣w−z0∣=∥h∥<r, so w∈U, and hence p0+h∈V. If h=0, the error in Step 2 is 0. If h=0, then 0<∣w−z0∣<δ0, and writing f(w)−f(z0)−(a+ib)(w−z0)=(w−z0)(Q(w)−(a+ib)), Step 2 gives
Step 4: the derivative matrix gives complex differentiability. This proves (i) from right to left. Suppose F is differentiable at p0 with derivative matrix Aa,b, and let ε>0 be given. Take δ>0 as in Differentiability at a Point for Functions Between Euclidean Spaces (i) for ε/2 in place of ε. Let w∈U with 0<∣w−z0∣<δ, and put h=Φ−1(w−z0). Then ∥h∥=∣w−z0∣<δ and Φ(p0+h)=z0+(w−z0)=w∈U, so p0+h∈V, and Step 2 gives
Step 5: the Cauchy–Riemann equations and the derivative. This proves (ii) and (iii). Recall that u=F1, v=F2, and that ∂/∂x and ∂/∂y are the partial derivatives ∂1 and ∂2 (The Complex Plane as ℝ²: Standing Notation (v)).
Suppose first that f is complex differentiable at z0, and write f′(z0)=a+ib with a,b real. By (i), F is differentiable at p0 with derivative matrix Aa,b, and by The Derivative Is the Jacobian Matrix, and Is Unique (ii) the partial derivatives of u and v at p0 exist and are the entries of Aa,b: