TheoremBase

Sketch (Brezis, Theorem 1.6): separate 0 from the open convex set A − B by the point-separation lemma, take α the supremum of F over A, and use openness of A for the strict inequality.

Proof

We separate the origin from the open convex set of differences A−BA-B and then move the level to the supremum of FF on AA.

Let

C={a−b:a∈A, b∈B}.C=\{a-b:a\in A,\ b\in B\}.

It is nonempty, as AA and BB are. It is convex: for a,a′∈Aa,a'\in A, b,b′∈Bb,b'\in B and 0≤t≤10\le t\le1,

t(a−b)+(1−t)(a′−b′)=(ta+(1−t)a′)−(tb+(1−t)b′)t(a-b)+(1-t)(a'-b')=\bigl(ta+(1-t)a'\bigr)-\bigl(tb+(1-t)b'\bigr)

by Elementary Identities in a Vector Space (ix), Elementary Identities in a Vector Space (vi), Vector Space over a Field (i) and Vector Space over a Field (ii), and the two convex combinations on the right lie in AA and in BB by Convex Subset of a Real Vector Space (i). It is open (Open Subset of a Metric Space (i)). Here we use that (u+w)−(v+w)=u−v(u+w)-(v+w)=u-v and (u+w)−w=u(u+w)-w=u for u,v,w∈Xu,v,w\in X (The Zero Vector, Negatives and Differences in a Vector Space (iii), Elementary Identities in a Vector Space (vi), Elementary Identities in a Vector Space (vii), The Zero Vector, Negatives and Differences in a Vector Space (i), Vector Space over a Field (i), Vector Space over a Field (ii)). Let a∈Aa\in A, b∈Bb\in B and ρ>0\rho>0 with BdX(a,ρ)⊆AB_{d_{X}}(a,\rho)\subseteq A, and let z∈BdX(a−b,ρ)z\in B_{d_{X}}(a-b,\rho). By The Norm Metric of a Real Normed Space (i), Elementary Identities in a Vector Space (viii) and Open Ball in a Metric Space (ii),

dX(a,z+b)=∥((a−b)+b)−(z+b)∥=∥(a−b)−z∥=dX(a−b,z)<ρ,d_{X}(a,z+b)=\bigl\lVert\bigl((a-b)+b\bigr)-(z+b)\bigr\rVert=\lVert(a-b)-z\rVert=d_{X}(a-b,z)<\rho ,

so z+b∈Az+b\in A, and therefore z=(z+b)−b∈Cz=(z+b)-b\in C. Hence BdX(a−b,ρ)⊆CB_{d_{X}}(a-b,\rho)\subseteq C. And 0X∉C0_{X}\notin C, since a−b=0Xa-b=0_{X} would give a=b+(a−b)=ba=b+(a-b)=b (Elementary Identities in a Vector Space (viii)), contradicting A∩B=∅A\cap B=\emptyset.

By Separating a Point from an Open Convex Set in a Separable Real Normed Space (i), applied to CC and the point 0X∉C0_{X}\notin C, there is F∈X∗F\in X^{*} with F(a−b)<F(0X)=0F(a-b)<F(0_{X})=0 for all a∈Aa\in A and b∈Bb\in B; by Linear Functional on a Linear Subspace (i) and Linear Functional on a Linear Subspace (ii) this says

F(a)<F(b)for all a∈A, b∈B.F(a)<F(b)\qquad\text{for all }a\in A,\ b\in B.

As C≠∅C\neq\emptyset, FF takes a negative value, so it is not identically zero.

Fix b0∈Bb_{0}\in B. The set {F(a):a∈A}\{F(a):a\in A\} is nonempty and bounded above by F(b0)F(b_{0}), so its supremum

α=sup⁡a∈AF(a)\alpha=\sup_{a\in A}F(a)

exists by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (vii). Every F(b)F(b) with b∈Bb\in B is an upper bound of this set, so α≤F(b)\alpha\le F(b) by Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order (iv).

It remains to see that F(a)<αF(a)<\alpha for a∈Aa\in A; here the openness of AA is used. Fix z0∈Xz_{0}\in X with F(z0)≠0F(z_{0})\neq0; replacing z0z_{0} by −z0-z_{0} if necessary (Linear Functional on a Linear Subspace (ii)), F(z0)>0F(z_{0})>0, and z0≠0Xz_{0}\neq0_{X} because F(0X)=0F(0_{X})=0. Let a∈Aa\in A, and fix ρ>0\rho>0 with BdX(a,ρ)⊆AB_{d_{X}}(a,\rho)\subseteq A (Open Subset of a Metric Space (i)). With s=ρ/(2∥z0∥)s=\rho/(2\lVert z_{0}\rVert) we have ∥sz0∥=ρ/2<ρ\lVert s z_{0}\rVert=\rho/2<\rho, so a+sz0∈Aa+sz_{0}\in A, and

F(a)<F(a)+sF(z0)=F(a+sz0)≤α.F(a)<F(a)+sF(z_{0})=F(a+sz_{0})\le\alpha .

Hence F(a)<α≤F(b)F(a)<\alpha\le F(b) for all a∈Aa\in A and b∈Bb\in B, which proves (i).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…