Reason: First published version. Substitutes the increment kappa*h into the definition of total differentiability at x+tau_0*h, takes square roots by monotonicity of squares on nonnegative reals, and divides by the increment to obtain the one-variable difference quotient estimate. The degenerate case h=0 is treated separately.
We use repeatedly that squares are strictly monotone on nonnegative reals: if 0≤α, 0≤β and α<β, then 0<β by claim 2, so βα<ββ by claim 10, while αα≤βα; claim 2 gives α2<β2. Consequently α≤β holds whenever α2≤β2, since β<α would give β2<α2. Write ∥z∥ for the Euclidean distance from z∈Rn to the origin, the nonnegative real with ∥z∥2=∑i=1nzi2.
Case 1: every hi is 0. Then τh=0 for every τ, so F is the constant function with value f(x), and L=0. For κ=0 with τ0+κ∈(p,q) the difference quotient of F at τ0 is 0, so the quantity ∣0−L∣=∣0∣=0 is less than any ε with 0<ε; taking δ=1, positive by claim 6, shows F is differentiable at τ0 with F′(τ0)=0=L.
Case 2: some hi0=0. Then hi02 is positive by claim 5 applied to ∣hi0∣, using claim 4 of Properties of the Absolute Value in an Ordered Field, and every other hi2 is nonnegative, so ∥h∥2 is positive by claim 3 and hence 0<∥h∥. So ∥h∥−1 exists and is positive by claim 7.
Let ε∈R with 0<ε. Put ε1=ε⋅2−1⋅∥h∥−1, positive by claims 8, 7 and 5. Since f is differentiable at a, applied with m=1 and with ε1 in place of ε, there is δ1 with 0<δ1 such that every k=(k1,…,kn)∈Rn with 0<∑iki2<δ12 and a+k∈U satisfies
(f(a+k)−f(a)−i=1∑n∂if(a)ki)2≤ε12i=1∑nki2.
Put δ=δ1∥h∥−1, positive by claim 5. Let κ∈R satisfy 0<∣κ∣<δ and τ0+κ∈(p,q), and set k=κh, so ki=κhi for each i and a+k=x+(τ0+κ)h, which lies in U by hypothesis. Then
i=1∑nki2=κ2∥h∥2=(∣κ∣∥h∥)2,
using claim 4 of Properties of the Absolute Value in an Ordered Field. This is positive, since ∣κ∣ and ∥h∥ are positive and claim 5 applies; and ∣κ∣∥h∥<δ∥h∥=δ1 by claim 10, so ∑iki2<δ12 by the monotonicity of squares. Hence the displayed estimate applies to this k.
Since f(a+k)=F(τ0+κ), f(a)=F(τ0) and ∑i∂if(a)ki=κL, the estimate reads
(F(τ0+κ)−F(τ0)−κL)2≤ε12(∣κ∣∥h∥)2=(ε1∣κ∣∥h∥)2,
and both F(τ0+κ)−F(τ0)−κL and ε1∣κ∣∥h∥ are nonnegative with F(τ0+κ)−F(τ0)−κL2 equal to the left side, by claim 4 of Properties of the Absolute Value in an Ordered Field. By the monotonicity of squares,