2572 CHAPTER 76. IMPLICIT STOCHASTIC EQUATIONS

where Ĉ is a constant used in the Burkholder Davis Gundy inequality. This is a restrictionon the size of K. Thus we only give a solution if K is small enough. Later, this will beremoved in the most interesting case. This will give a local solution valid for a fixed T > 0and then the global solution can be obtained by applying this result on the succession ofintervals [0,T ] , [T,2T ] , [3T,4T ] , and so forth.

From Theorem 76.4.9, if w ∈W , there exists a unique solution u to

Bu(t,ω)−Bu0 (ω)+∫ t

0A(s,u(s,ω) ,ω)ds =

∫ t

0f ds+B

∫ t

0σ (w)dW.

holding in the sense described there. Let ui result from wi. Then from the implicit Itoformula and the above monotonicity estimate,

12⟨B(u1−u2) ,u1−u2⟩(t)+ r

∫ t

0∥u1−u2∥2

W ds

−λ

∫ t

0⟨B(u1−u2) ,u1−u2⟩ds

−∫ t

0⟨Bσ (u1)−Bσ (u2) ,σ (u1)−σ (u2)⟩L2

ds≤M∗ (t)

where the right side is of the form sups∈[0,t] |M (s)| where M (t) is a local martingale havingquadratic variation dominated by

C∫ t

0∥σ (w1)−σ (w2)∥2 ⟨B(u1−u2) ,u1−u2⟩ds (76.5.56)

Therefore, since M∗ is increasing in t, it follows from the Lipschitz condition on σ that

12⟨B(u1−u2) ,u1−u2⟩(t)+ r

∫ t

0∥u1−u2∥2

W ds

−λ

∫ t

0⟨B(u1−u2) ,u1−u2⟩ds−∥B∥K2

∫ t

0∥u1−u2∥2

W ds≤M∗ (t)

Thus, from the assumption about r,

sups∈[0,t]

⟨B(u1−u2) ,u1−u2⟩(s)+4∫ t

0∥u1−u2∥2

W ds

≤ λ

∫ t

0⟨B(u1−u2) ,u1−u2⟩ds+2M∗ (t)

Then applying Gronwall’s inequality,

sups∈[0,t]

⟨B(u1−u2) ,u1−u2⟩(s)+4∫ t

0∥u1−u2∥2

W ds≤ 2eλT M∗ (t)

Now take expectations and use the Burkholder Davis Gundy inequality. The expectation ofthe right side is then dominated by

2ĈeλT∫

(∫ t

0∥σ (w1)−σ (w2)∥2

L2⟨B(u1−u2) ,u1−u2⟩ds

)1/2

dP

2572 CHAPTER 76. IMPLICIT STOCHASTIC EQUATIONSwhere C is a constant used in the Burkholder Davis Gundy inequality. This is a restrictionon the size of K. Thus we only give a solution if K is small enough. Later, this will beremoved in the most interesting case. This will give a local solution valid for a fixed T > 0and then the global solution can be obtained by applying this result on the succession ofintervals [0,7], [7,27], [37,47] , and so forth.From Theorem 76.4.9, if w € WY, there exists a unique solution u toBu( 1,00) ~Buo(@) + [A(s,u(s,00),0)ds= | fas+B |’ o(wyaw.holding in the sense described there. Let u; result from w;. Then from the implicit Itoformula and the above monotonicity estimate,1 t5 (B (ui — ua) a1 09) (1) +r lla —ug||2, dst-2 | (B (uy — uz) ,u) — U2) dsJO-[ (Bo (u) — Bo (ux) ,o (1) — 0 (u2)) ds < M* (1)where the right side is of the form sup, jo, |M (s)| where M (1) is a local martingale havingquadratic variation dominated bytc| Io (w1) — 0 (wr) ||? (B (uy — up) 5 —up) ds (6.5.56)Therefore, since M* is increasing in f, it follows from the Lipschitz condition on o that12t 4 t 2A [ (B(u; — uz) uj — up) ds —||BI| K [ lla — u2|[2, ds <M? (1)t(B (uy — uz) , uy — ur) +r \lu1 — ually dsThus, from the assumption about r,tsup (B(u; — uo) ,u1 —u2)(s) +4 / lla — up|, asse[0,z] 0t<A i (B (ui — uo) ,uy —ur)ds-++2M* (t)JoThen applying Gronwall’s inequality,tsup (B(uy —uz) U4 — up) (s) +4 [ || ve ~up|\q ds < 2047 M* (t)s€[0,t]Now take expectations and use the Burkholder Davis Gundy inequality. The expectation ofthe right side is then dominated byacer | (/ I]o (w1) — 6 (wa) [2 (B (ur — 00) a - tn) ” iP