2418 CHAPTER 70. MEASURABILITY WITHOUT UNIQUENESS

It follows from equivalence of norms in finite dimensional spaces, the assumed esti-mates on M, A, and P and standard manipulations depending on compact embeddings thatthere exists an estimate suitable to apply Theorem 70.3.3 to obtain the existence of a solu-tion such that (t,ω)→ x(t,ω) is measurable into Rk which implies that (t,ω)→ vk (t,ω)is product measurable into V and H. This yields the measurable Galerkin approximation.

Also, the estimates and compact embedding results for Sobolev spaces imply an in-equality of the form

|vk (t)|2H +∫ T

0∥vk∥2

V ds+∥uk (t)∥2V ≤C (70.5.42)

where in fact C does not depend on ε,ω or k. Everything would work if C depended on ω

but because of our simplifying assumptions, we can get a single C as above.Next we need to estimate the time derivative in V ′. The integral equation implies that

for all w ∈Vk,⟨v′k (t) ,w

⟩V ′,V +

⟨Mvk +Auk +Puk + γ∗T F

((ukn−g(ω))+

µ(∣∣vkT − U̇T

∣∣)ψ ′ε(vkT − U̇T

) ,w⟩

V ′,V

= ⟨f,w⟩ (70.5.43)

where the dependence on t and ω is suppressed in most terms. In terms of inner productsin V this reduces to(

R−1v′k (t) ,w)

V +

(R−1

(Mvk +Auk +Puk + γ∗T F

((un−g(ω))+

µ(∣∣vkT − U̇T

∣∣)ψ ′ε(vkT − U̇T

) ),w)

V

=(R−1f,w

)V

In terms of Pk the orthogonal projection in V onto Vk, this takes the form(R−1v′k (t) ,Pkw

)V +(

R−1(

Mvk +Auk +Puk + γ∗T F((un−g(ω))+

µ(∣∣vkT − U̇T

∣∣)ψ ′ε(vkT − U̇T

) ),Pkw

)V

=(R−1f,Pkw

)V

for all w ∈V . Now v′k (t) ∈Vk and so the first term can be simplified and we can write(R−1v′k (t) ,w

)V +(

R−1(

Mvk +Auk +Puk + γ∗T F((un−g(ω))+

µ(∣∣vkT − U̇T

∣∣)ψ ′ε(vkT − U̇T

) ),Pkw

)V

=(R−1f,Pkw

)V

for all w ∈V . Then it follows that for all w ∈V,(R−1v′k (t) ,w

)V +

2418 CHAPTER 70. MEASURABILITY WITHOUT UNIQUENESSIt follows from equivalence of norms in finite dimensional spaces, the assumed esti-mates on M, A, and P and standard manipulations depending on compact embeddings thatthere exists an estimate suitable to apply Theorem 70.3.3 to obtain the existence of a solu-tion such that (t,@) + x(t,@) is measurable into R* which implies that (t,@) — v; (t, @)is product measurable into V and H. This yields the measurable Galerkin approximation.Also, the estimates and compact embedding results for Sobolev spaces imply an in-equality of the formTImei [ livellnds+ lime (lp SC (70.542)where in fact C does not depend on €, or k. Everything would work if C depended on @but because of our simplifying assumptions, we can get a single C as above.Next we need to estimate the time derivative in Y’. The integral equation implies thatfor allw Ee VY,Mvy + Aug + Puy + pF ((uin —8(@)))°O10 en tweet) yy= (f,w) (70.5.43)where the dependence on ¢ and @ is suppressed in most terms. In terms of inner productsin V this reduces to-lu _ Mvy,+Au,+ Pu +77F (un — g(@)) .(R Nyc(t)aw)y +(R '( k 1 (|ver —Ur|) v. iin 6 4) ))= (R'f,w) VvIn terms of P; the orthogonal projection in V onto V;, this takes the form(Ro'v; (t),Pew)y +(x ( My; + Aug + Puy + YF ((un — 8 (@)),.)- ) P )r / ° KWLt (|ver —Ur]) We (ver — Ur) V= (R"'f, ew),for all w € V. Now vy; (t) € Vz and so the first term can be simplified and we can write(R7!v;, (t) W)yt+a1 ( Mv +Auy + Pug + ¥7F ((un—g(@)),)-R : , : » PewLt (|ver —Ur]) We (ver — Ur) Vv= (R'f, Pew),for all w € V. Then it follows that for all w € V,(R-!v;, (t) W)y +