Quasilinear parabolic problems with nonlinear boundary conditions

Quasilinear parabolic problems with nonlinear boundary conditions Quasilinear parabolic problems with nonlinear boundary conditions

busch.felix24
from busch.felix24 More from this publisher
02.12.2012 Views

the natural regularity space of φ in (4.33), is equivalent to | · |Yl = | · | Lp([0,Tl]×R n ) + [ · ] Y T l 1 see (4.21) and (4.22) for the definition of [ · ] Y T l 1 , [ · ] T Y l . 2 + [ · ] T Y l , 2 Taking now for (g, φ) in (4.33) the right-hand sides of (4.5) and (4.31), application of the solution operator L ij H, i+1 yields where and (I − S ij H )ϕjv =L ij � ϕjf + k ∗ Cj(·, x, Dx)v H, i+1 ϕjh + γyCH, j(·, x ′ )v S ij Hw = Lij H, i+1 � j k ∗ (A# (Ti, xj, Dx) − A j # (·, x, Dx))w γy(B j # (Ti, x ′ j , Dx) − B j # (·, x′ , Dx))w � =: h ij H (f, h, v), (4.34) CH, j(t, x) = B#(t, x ′ , Dx)ϕj − ϕjb0, (4.35) γy denoting the trace operator at y = 0. One can then show an analogue to Claim 2 (cf. the proof of Theorem 4.1.1) asserting in particular existence of a small η0 > 0 which is such that whenever δ = maxi |Ti+1 − Ti|, η ≤ η0, the equation (I − S ij H )w = hij H (f, h, v) admits a unique solution w =: (I − S ij H )|−1 Zi+1(ϕjV −1 H, i (f, h))hij H (f, h, v) in Zi+1(ϕjV −1 H, i (f, h)) for all (f, h) ∈ ΞH, i+1, v ∈ Zi+1(V −1 H, i (f, h)), i = 0, . . . , M −1, and (f, h), i ≥ 1, refers to the solution of (4.30) on the time-interval [0, Ti], which is already known in the (i + 1)th time step. Further, Z1(ϕjV −1 H, 0 (f, h)) := {ϕjw : w ∈ Z1, ∂m t w|t=0 = ∂m t f|t=0, if α > m + 1/p, m = 0, 1}. The proof of these properties is similar to that of Claim 2 above, which is why we only consider the part that involves estimates on the boundary. Let w ∈ Zi+1(0). If p > n + 1 + 2/α, that is Y ↩→ BUC([0, T ] × Rn ), Lemma 4.2.1 yields j = 0, . . . , Nb. Here V −1 H, i |γy(B j # (Ti, x ′ j, Dx) − B j # (t, x′ , Dx))w|Yi+1 = |(b(Ti, x ′ j) − b j (·, ·)) · ∇x ′γyw|Yi+1 ≤ C|b(Ti, x ′ j) − b j (·, ·)| (Y T i , T i+1 ∩L∞) n| |∇x ′γyw| (Yi+1∩L∞) n ≤ C1|b(Ti, x ′ j) − b j (·, ·)| T (Y i , Ti+1 ∩L∞) n|w|Zi+1 =: κ1|w|Zi+1 , where the constant C1 does not depend on i, j. Similarly, if p ≤ n + 1 + 2/α, Lemma 4.2.2 shows that |γy(B j # (Ti, x ′ j, Dx) − B j # (t, x′ , Dx))w|Yi+1 ≤ ≤ C|(b(Ti, x ′ j)−b j (·, ·))| L∞([Ti,Ti+1]×R n ) n(1+[b(Ti, x ′ j)−b j (·, ·)] (M T i , T i+1 ) n)|w|Zi+1 ≤ C2η|w|Zi+1 =: κ2|w|Zi+1 , again with a constant not depending on i, j. In view of (4.24) and (4.25), it is clear that κ1, κ2 tend to zero if η and δ do so. This, together with Theorem 4.2.2 and an estimate 70 � ,

analogous to (4.17), shows that S ij −1 H : Zi+1(ϕjVi (f, h)) → Zi+1 is contractive if both η and δ are sufficiently small. Assuming this, we thus obtain, aside from (4.32), ϕjv = (I − S ij H )|−1 Zi+1(ϕjV −1 H, i (f, h))hij H (f, h, v), j = 0, . . . , Nb. Multiplying these equations by ψj and summing over all j then results in Nb � v = GH(v) := j=0 + ψj(I − S ij H )|−1 Zi+1(ϕjV −1 H, i (f, h))hij H Nb+N � j=Nb+1 (f, h, v) ψj(I − S ij )| −1 Zi+1(ϕjV −1 H, i (f, h))hij (f, v), (4.36) a fixed point equation for v ∈ Zi+1(v i ). Since GH leaves this space invariant, the contraction principle is applicable, provided that GH is a strict contraction. To verify that this can be arranged by selecting δ sufficiently small, we let v, ¯v ∈ Zi+1(v i ) and estimate with the aid of Theorem 4.1.1 and 4.2.2 |GH(v)−GH(¯v)|Zi+1 = � Nb � � � j=0 Nb+N � + j=Nb Nb+N � ≤ C( j=0 ψj(I − S ij H )|−1 Zi+1(0) Lij H, i+1 ψj(I − S ij )| −1 Zi+1(0) Lij � � k ∗ Cj(·, x, Dx)(v − ¯v) γyCH, j(·, x)(v − ¯v) i+1k ∗ Cj(·, � � x, Dx)(v − ¯v) � Zi+1 |Cj(·, x, Dx)(v − ¯v)|Xi+1 + Nb � |γyCH, j(t, x)(v − ¯v)|Yi+1 ), with C > 0 not depending on δ. By extension to J × Rn+1 , estimate (4.14), and restriction to J × R n+1 + , we obtain for the first sum Nb+N � j=0 |Cj(·, x, Dx)(v − ¯v)|Xi+1 ≤ C1(ε + Cε|k| L1(0,δ))|v − ¯v|Zi+1 , (4.37) where ε > 0 can be chosen arbitrary small and C1, Cε > 0 do not depend on δ. Turning to the second sum, we introduce the space which is normed by 0Z 1/2 i+1 = 0H α 2 |w| 0Z 1/2 i+1 j=0 p (J; Lp(R n+1 + )) ∩ Lp(J; H 1 p(R n+1 = |Bk w|Xi+1 α/2 + |∇xw| n+1 X , i+1 + )), where k α/2(t) = t α/2−1 , t > 0, and Bk α/2 = (k α/2∗) −1 in Xi+1. Suppose that u ∈ Zi+1(0). By causality, B 2 k α/2 u| [0,Ti] = 0, and so we have |u| 0Z 1/2 = |(kα/2χ [0,Ti+1−Ti]) ∗ B i+1 2 k u|Xi+1 α/2 + |∇xu| n+1 Xi+1 ≤ |kα/2| L1(0,Ti+1−Ti)|B 2 k u|Xi+1 α/2 + ε|∇2xu| + Cε|u|Xi+1 ≤ C0|k α/2| L1(0,δ)|Bku|Xi+1 + ε|∇2 xu| + Cε|k| L1(0,δ)|Bku|Xi+1 ≤ (C0|k α/2| L1(0,δ) + ε + Cε|k| L1(0,δ))|u|Zi+1 ∀u ∈ Zi+1(0), (4.38) 71

analogous to (4.17), shows that S ij<br />

−1<br />

H : Zi+1(ϕjVi (f, h)) → Zi+1 is contractive if both η<br />

and δ are sufficiently small.<br />

Assuming this, we thus obtain, aside from (4.32),<br />

ϕjv = (I − S ij<br />

H )|−1<br />

Zi+1(ϕjV −1<br />

H, i (f, h))hij H<br />

(f, h, v), j = 0, . . . , Nb.<br />

Multiplying these equations by ψj and summing over all j then results in<br />

Nb �<br />

v = GH(v) :=<br />

j=0<br />

+<br />

ψj(I − S ij<br />

H )|−1<br />

Zi+1(ϕjV −1<br />

H, i (f, h))hij H<br />

Nb+N �<br />

j=Nb+1<br />

(f, h, v)<br />

ψj(I − S ij )| −1<br />

Zi+1(ϕjV −1<br />

H, i (f, h))hij (f, v), (4.36)<br />

a fixed point equation for v ∈ Zi+1(v i ). Since GH leaves this space invariant, the contraction<br />

principle is applicable, provided that GH is a strict contraction.<br />

To verify that this can be arranged by selecting δ sufficiently small, we let v, ¯v ∈<br />

Zi+1(v i ) and estimate <strong>with</strong> the aid of Theorem 4.1.1 and 4.2.2<br />

|GH(v)−GH(¯v)|Zi+1 =<br />

� Nb<br />

�<br />

�<br />

�<br />

j=0<br />

Nb+N �<br />

+<br />

j=Nb<br />

Nb+N �<br />

≤ C(<br />

j=0<br />

ψj(I − S ij<br />

H )|−1<br />

Zi+1(0) Lij<br />

H, i+1<br />

ψj(I − S ij )| −1<br />

Zi+1(0) Lij<br />

� �<br />

k ∗ Cj(·, x, Dx)(v − ¯v)<br />

γyCH, j(·, x)(v − ¯v)<br />

i+1k ∗ Cj(·,<br />

�<br />

�<br />

x, Dx)(v − ¯v)<br />

� Zi+1<br />

|Cj(·, x, Dx)(v − ¯v)|Xi+1 +<br />

Nb �<br />

|γyCH, j(t, x)(v − ¯v)|Yi+1 ),<br />

<strong>with</strong> C > 0 not depending on δ. By extension to J × Rn+1 , estimate (4.14), and<br />

restriction to J × R n+1<br />

+ , we obtain for the first sum<br />

Nb+N �<br />

j=0<br />

|Cj(·, x, Dx)(v − ¯v)|Xi+1 ≤ C1(ε + Cε|k| L1(0,δ))|v − ¯v|Zi+1 , (4.37)<br />

where ε > 0 can be chosen arbitrary small and C1, Cε > 0 do not depend on δ.<br />

Turning to the second sum, we introduce the space<br />

which is normed by<br />

0Z 1/2<br />

i+1 = 0H α<br />

2<br />

|w| 0Z 1/2<br />

i+1<br />

j=0<br />

p (J; Lp(R n+1<br />

+ )) ∩ Lp(J; H 1 p(R n+1<br />

= |Bk w|Xi+1 α/2 + |∇xw| n+1<br />

X ,<br />

i+1<br />

+ )),<br />

where k α/2(t) = t α/2−1 , t > 0, and Bk α/2 = (k α/2∗) −1 in Xi+1. Suppose that u ∈ Zi+1(0).<br />

By causality, B 2 k α/2 u| [0,Ti] = 0, and so we have<br />

|u|<br />

0Z 1/2 = |(kα/2χ [0,Ti+1−Ti]) ∗ B<br />

i+1<br />

2 k u|Xi+1 α/2 + |∇xu| n+1<br />

Xi+1 ≤ |kα/2| L1(0,Ti+1−Ti)|B 2 k u|Xi+1 α/2 + ε|∇2xu| + Cε|u|Xi+1<br />

≤ C0|k α/2| L1(0,δ)|Bku|Xi+1 + ε|∇2 xu| + Cε|k| L1(0,δ)|Bku|Xi+1<br />

≤ (C0|k α/2| L1(0,δ) + ε + Cε|k| L1(0,δ))|u|Zi+1 ∀u ∈ Zi+1(0), (4.38)<br />

71

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!