25.7. MAXIMAL MONOTONE OPERATORS 915
Corollary 25.7.52 Let φ : X → (−∞,∞] be convex, proper, and lower semicontinuous.Here X is a Banach space. Then ∂φ is maximal monotone.
Proof: Let ψ (x) = 12 ∥x∥
2. There exists x∗ and some number b such that φ (x) ≥b+ ⟨x∗,x⟩ . Therefore, ψ + φ is convex, lower semicontinuous, and bounded. It follows∂ (ψ +φ) is onto by Theorem 25.7.49. However, ψ is continuous everywhere, in particu-lar at every point of the domain of φ . Therefore, ∂ψ + ∂φ = ∂ (φ +ψ) and by Theorem25.7.48, this shows that F +∂φ is onto.
It seems to me that the above are the most important results about convex properlower semicontinuous functions. However, there are many other very interesting propertiesknown.
Proposition 25.7.53 Let φ : X → (−∞,∞] be convex proper and lower semicontinuous.Then D(∂φ) is dense in D(φ) and so D(∂φ) = D(φ).
Proof: Let xλ be the solution to 0 ∈ F (xλ − x)+λ∂φ (xλ ) . Here x ∈ D(φ). Say u∗λ∈
∂φ (xλ ) such that the inclusion becomes an equality. Then
0 =〈F (xλ − x)+λu∗
λ,xλ − x
〉= ∥xλ − x∥2−λ
〈u∗
λ,x− xλ
〉≥ ∥xλ − x∥2−λ (φ (x)−φ (xλ ))
Hence, letting z∗,b be such that φ (y)≥ b+ ⟨z∗,y− x⟩ ,
λ (φ (x)− [b+ ⟨z∗,xλ − x⟩])≥ λ (φ (x)−φ (xλ ))≥ ∥xλ − x∥2
λφ (x)−λb≥ ∥xλ − x∥2−λ ∥z∗∥∥xλ − x∥
≥ ∥xλ − x∥2−λ
(∥z∗∥2
2+∥xλ − x∥2
2
)Thus
λφ (x)−λb+λ∥z∗∥2
2≥(
1− λ
2
)∥xλ − x∥2
It follows that xλ → x. This shows that D(φ) ⊆ D(∂φ) and so D(φ) ⊆ D(∂φ) ⊆ D(φ).
There is a really amazing theorem, Moreau’s theorem. It is in [24], [13] and [116]. Itinvolves approximating a convex function with one which is differentiable, at least in thecase where you have a Hilbert space. In the general case considered in this chapter, thefunction is continuous.
Theorem 25.7.54 Let φ be a convex lower semicontinuous proper function defined on X.Define A≡ ∂φ ,Aλ = (∂φ)
λ
φ λ (x)≡miny∈X
(1
2λ∥x− y∥2 +φ (y)
)