25.7. MAXIMAL MONOTONE OPERATORS 913
≤ tφ (y1)+(1− t)φ (y2)− (tφ 1 (x)+(1− t)φ (x))
−(tx∗ (y1− x)+(1− t)x∗ (y2− x))
≤ ta+(1− t)a = a
so C1 is indeed convex. The case of C2 is similar.C1 is nonempty because it contains (x,φ 1 (x)−φ 1 (x)− x∗ (x− x)) since
φ 1 (x)−φ 1 (x)− x∗ (x− x)≤ φ 1 (x)−φ 1 (x)− x∗ (x− x)
C2 is also nonempty because it contains (x,φ 2 (x)−φ 2 (x)) since
φ 2 (x)−φ 2 (x)≤ φ 2 (x)−φ 2 (x)
In addition to this,
(x,φ 1 (x)− x∗ (x− x)−φ 1 (x)+1) ∈ int(C1)
due to the assumed continuity of φ 1 at x and so int(C1) ̸= /0. If (y,a) ∈ int(C1) then
φ 1 (y)− x∗ (y− x)−φ 1 (x)≤ a− ε
whenever ε is small enough. Therefore, if (y,a) is also in C2, the assumption that x∗ ∈∂ (φ 1 +φ 2)(x) implies
a− ε ≥ φ 1 (y)− x∗ (y− x)−φ 1 (x)≥ φ 2 (x)−φ 2 (y)≥ a,
a contradiction. Therefore int(C1) ∩C2 = /0 and so by Theorem 18.2.14, there exists(w∗,β ) ∈ X ′×R with
(w∗,β ) ̸= (0,0) , (25.7.73)
andw∗ (y)+βa≥ w∗ (y1)+βa1, (25.7.74)
whenever (y,a) ∈C1 and (y1,a1) ∈C2.Claim: β > 0.Proof of claim: If β < 0 let
a = φ 1 (x)− x∗ (x− x)−φ 1 (x)+1,
a1 = φ 2 (x)−φ 2 (x) , and y = y1 = x.
Then from 25.7.74
β (φ 1 (x)− x∗ (x− x)−φ 1 (x)+1)≥ β (φ 2 (x)−φ 2 (x)) .
Dividing by β yields
φ 1 (x)− x∗ (x− x)−φ 1 (x)+1≤ φ 2 (x)−φ 2 (x)