77.6. VARIATIONAL INEQUALITIES 2637

Proof: This follows from Theorem 77.2.10. This is because there is an estimate of theright sort for the measurable functions un (·,ω) and u∗n (·,ω) and the above argument whichshows that a subsequence has a convergent subsequence which converges appropriately toa solution.

You can have K = K (ω) . There would be absolutely no change in the above the-orem. You just need to have the operator of penalization satisfy ω → P(u(ω) ,ω) =

F(

u(ω)−projK (ω) u(ω))

is measurable into V ′ provided ω → u(ω) is measurable intoV . What are the conditions on the set valued ω →K (ω) which will cause this to takeplace?

Lemma 77.6.2 Let ω →K (ω) be measurable into V . Then ω → projK (ω) u(ω) is alsomeasurable into V if ω → u(ω) is measurable.

Proof: It follows from standard results on measurable multi-functions [70] that thereis a countable collection {wn (ω)} , ω → wn (ω) being measurable and wn (ω) ∈K (ω)for each ω such that for each ω, K (ω) = ∪nwn (ω). Let

dn (ω)≡min{∥u(ω)−wk (ω)∥ ,k ≤ n}

Let u1 (ω)≡ w1 (ω) . Letu2 (ω) = w1 (ω)

on the set{ω : ∥u(ω)−w1 (ω)∥< {∥u(ω)−w2 (ω)∥}}

andu2 (ω)≡ w2 (ω) off the above set.

Thus ∥u2 (ω)−u(ω)∥= d2. Let

u3 (ω) = w1 (ω) on{

ω : ∥u(ω)−w1 (ω)∥<∥∥u(ω)−w j (ω)

∥∥ , j = 2,3

}≡ S1

u3 (ω) = w2 (ω) on S1∩{

ω : ∥u(ω)−w1 (ω)∥<∥∥u(ω)−w j (ω)

∥∥ , j = 3

}u3 (ω) = w3 (ω) on the remainder of Ω

Thus ∥u3 (ω)−u(ω)∥= d3. Continue this way, obtaining un (ω) such that

∥un (ω)−u(ω)∥= dn (ω)

and un (ω) ∈K (ω) with un measurable. Thus, in effect one picks the closest of all thewk (ω) for k ≤ n as the value of un (ω) and un is measurable and by density in K (ω) of{wn (ω)} for each ω,{un (ω)} must be a minimizing sequence for

λ (ω)≡ inf{∥u(ω)− z∥ : z ∈K (ω)}

Then it follows that un (ω)→ projK (ω) u(ω) weakly in V . Here is why: Suppose it failsto converge to projK (ω) u(ω). Since it is minimizing, it is a bounded sequence. Thus

77.6. VARIATIONAL INEQUALITIES 2637Proof: This follows from Theorem 77.2.10. This is because there is an estimate of theright sort for the measurable functions uw, (-,@) and u; (-,@) and the above argument whichshows that a subsequence has a convergent subsequence which converges appropriately toasolution. JYou can have .# = .# (@). There would be absolutely no change in the above the-orem. You just need to have the operator of penalization satisfy @ > P(u(@),@) =F (u (@) — proj y(@) 4 (0)) is measurable into VY’ provided @ — u(@) is measurable into¥. What are the conditions on the set valued @ > .% (@) which will cause this to takeplace?Lemma 77.6.2 Let © + (@) be measurable into V. Then @ + proj 7(q)u(@) is alsomeasurable into V if @ — u(@) is measurable.Proof: It follows from standard results on measurable multi-functions [70] that thereis a countable collection {w,(@)}, @ + w,(@) being measurable and w,(@) € % (@)for each @ such that for each @, # (@) = Unwn (@). Letdy (@) = min {]|u(@) — wx (@)|| ,& <n}Let u; (@) = w; (@). Letud (@) =w| (@)on the set{@ : ||u(@) —w) (@)|| < {||u(@) — we (@) || }}anduz (@) = w2(@) off the above set.Thus ||u2 (@) —u(@)|| = do. Letu3(@) = wy(Q) on { @ : ||u(@) —w1 (@)I| has,< ||u(@) —w;(@)||, j= 2,3u3(@) = wa (0) on Sif @ : ||u(@) — wi (@)||< |ju(@) —w;(@)|| 7 =3u3(@) = w3(@) on the remainder of QThus ||u3 (@) — u(@)|| = d3. Continue this way, obtaining u, (@) such that\|un (@) —u(@)|| = dn (@)and un (@) € 4 (@) with u, measurable. Thus, in effect one picks the closest of all thewx (@) for k <n as the value of u,(@) and uy, is measurable and by density in .~ (@) of{wn (@)} for each @, {u, (@)} must be a minimizing sequence forA (a) = inf {Ju (@) —z|| :2€ % (@)}Then it follows that up, (@) — proj y-() u(@) weakly in ¥. Here is why: Suppose it failsto converge to proj.y(~) u(@). Since it is minimizing, it is a bounded sequence. Thus