· 8,216 chars · 16 deps · depth 11 Reason: Proof of existence, localisation and compactness for maximisers of linearly perturbed continuous functions on a closed ball, via the extreme value theorem, the strictly smaller maximum on the closed annulus, and Bolzano-Weierstrass.
Existence comes from the extreme value theorem on the closed ball; localisation from comparing the value at a maximiser with the value at the centre against the strictly smaller maximum of the function on the closed annulus; compactness from Bolzano-Weierstrass applied to the perturbing vectors together with the sequential characterisation of closed sets.
Step 1 (the perturbed functions are continuous). Let p∈Rn and S⊆Bˉ. We check that the restriction of ψp to S satisfies the continuity hypothesis of Extreme Value Theorem on a Compact Subset of a Metric Space. Let x∈S and let ε∈R with 0<ε. For y∈Bˉ,
By (H1) there is a positive real δ1 such that every z∈Bˉ with ∥z−x∥<δ1 satisfies ∣φ(z)−φ(x)∣<ε/2. Let δ be the smaller of δ1 and ε/(2∥p∥+2), a positive real number. If y∈S satisfies dE(x,y)=∥y−x∥<δ, then
ψp(y)−ψp(x)<2ε+∥p∥2∥p∥+2ε≤2ε+2ε=ε,
since 2∥p∥≤2∥p∥+2 gives ∥p∥/(2∥p∥+2)≤21.
Proof of claim 1. Let p∈Rn. By step 1 with S=Bˉ, the function ψp satisfies the hypotheses of Extreme Value Theorem on a Compact Subset of a Metric Space on the nonempty compact set Bˉ, so there is xmax∈Bˉ with ψp(y)≤ψp(xmax) for every y∈Bˉ; that is, xmax∈M(p) and M(p)=∅. Now let δ be a positive real number. The origin0Rn satisfies ∥0Rn∥=0≤δ by claim 3 of Elementary Properties of the Euclidean Norm on Rn, so any element of M(0Rn) belongs to Kδ, and Kδ=∅.
Proof of claim 2. Let ρ∈R with 0<ρ.
Suppose first that r<ρ. Every x∈Bˉ satisfies dE(x^,x)≤r<ρ, so Bˉ⊆BdE(x^,ρ); since Kδ⊆Bˉ for every positive real δ, the choice δρ=1 works.
Suppose now that ρ≤r, and put
A={x∈Bˉ:ρ≤∥x−x^∥}.
The set A is nonempty: let e1∈Rn have first component 1 and all other components 0, so that ∥e1∥2=1 and hence ∥e1∥=1 by claim 1 of Elementary Properties of the Euclidean Norm on Rn; then z=x^+ρe1 satisfies ∥z−x^∥=∥ρe1∥=ρ by claim 5 there, so dE(x^,z)=ρ≤r and z∈A.
Let δ be real with 0<δ≤δρ and let x∈Kδ, say x∈M(p) with ∥p∥≤δ. Since x^∈Bˉ we have φ(x^)+p⋅x^≤φ(x)+p⋅x, hence, using Cauchy-Schwarz and ∥x−x^∥=dE(x^,x)≤r,
Let (x(q))q∈N be a sequence with x(q)∈Kδ for every q, converging to a point x∈Rn, and for each q choose p(q)∈Rn with ∥p(q)∥≤δ and x(q)∈M(p(q)). As Bˉ is closed, x∈Bˉ.
Fix y∈Bˉ and let ε∈R with 0<ε. By (H1) there is a positive real δ2 such that every z∈Bˉ with ∥z−x∥<δ2 satisfies ∣φ(z)−φ(x)∣<ε/4. By convergence, choose t∈N so large that
As ε was an arbitrary positive real number, φ(y)+p⋅y≤φ(x)+p⋅x: were the reverse strict inequality to hold, taking ε to be half of the positive difference would contradict the display. Since y∈Bˉ was arbitrary, x∈M(p) with ∥p∥≤δ, so x∈Kδ.