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

Lemma 4.2.1 Let 1 < p < ∞, s1, s2 ∈ (0, 1), 0 < T2, 0 ≤ T1 ≤ T2, and Ω be an arbitrary domain in R n . Then |mf| Y T 2 ≤ |m| Y T 1 ∩L∞ |f| Y T 1 ∩L∞ + 2|m| Y T 1 , T 2 ∩L∞ |f| Y T 2 ∩L∞ (4.26) for all m, f ∈ Y T2 ∩ L∞([0, T2] × Ω). We turn now to products where one factor might be unbounded. Such a constellation arises for example when s1 < 1/p. We confine ourselves to the case Ω = R n . By means of extension and restriction, the subsequent multiplication property can be transferred to domains with sufficiently smooth boundary. Suppose f ∈ Y T2 and m ∈ C r1 ([0, T2]; C(R n )) ∩ C([0, T2]; C r2 (R n )) =: M T2 for some si < ri < 1, i = 1, 2. Letting J be a subinterval of [0, T2], we then estimate [mf] Lp(J;B s 2 pp(R n )) where � � I(m, f) ≤ ( J R n � = ( � � J Rn Rn ≤ |m| s L∞(J×Ω)[f] Lp(J;B 2 pp(Rn )) � B1(y) =: I1(m, f) + I2(m, f). |m(t, x)f(t, x) − m(t, y)f(t, y)| p |m(t, x) − m(t, y)| p |f(t, y)| p |x − y| n+s2p |x − y| n+s2p + I(m, f), dx dy dt) 1 � p + ( � J Rn � dx dy dt) 1 p · · ·) 1 p R n \B1(y) We put [m] C(J; Cr2 (Rn )) = supt∈J, x, y∈Rn |m(t, x) − m(t, y)| |x − y| −r2 . By hypothesis 1 − s2/r2 > 0. If η ∈ [0, 1 − s2/r2), then (1 − η)r2 − s2 > 0, and we obtain I1(m, f) ≤ (2|m| L∞(J×Rn )) η [m] 1−η C(J; Cr2 (Rn )) ( � � � |f(t, y)| p dx dy dt 1 ) p |x − y| n−((1−η)r2−s2)p Further, Hence [mf] Y T 2 2 J Rn B1(y) ≤ C(p, r2, s2, n)|m| η L∞(J×Rn ) [m]1−η C(J; Cr2 (Rn )) |f| Lp(J×Rn ). � � I2(m, f) ≤ 2|m| L∞(J×Rn )( J Rn � Rn \B1(y) ≤ C(p, s2, n)|m| L∞(J×Rn )|f| Lp(J×Rn ). |f(t, y)| p dx dy dt |x − y| n+s2p ≤ [mf] Lp([0,T1];B s2 pp(Rn )) + [mf] Lp([T1,T2];B s2 pp(Rn )) ≤ C(|m| L∞([0,T1]×Rn ) + [m] C([0,T1]; Cr2 (Rn )) )(|f| Lp([0,T1]×Rn ) + [f] T Y 1 ) 2 +|m| L∞([T1,T2]×Rn � ) [f] T Y 2 + C(1 + [m] C([T1,T2]; C 2 r2 (Rn )) )|f| Lp([0,T2]×Rn ) Let now [m] Cr1 (J; C(Rn )) = supt, τ∈J, x∈Rn |m(t, x) − m(τ, x)| |t − τ| −r1 . To estimate , we take (4.23) with Ω = Rn as starting point, apply Fubini’s theorem, and [mf] T Y 2 1 use the same estimation techniques as for [mf] T Y 2 2 [mf] Y T 2 1 to the result ≤ |m| L∞([0,T1]×Rn )[f] T Y 1 + C1[m] Cr1 ([0,T1]; C(R 1 n )) |f| Lp([0,T1]×Rn ) +2|m| L∞([T1,T2]×Rn )([f] T Y 2 + C2[m] Cr1 ([T1,T2]; C(R 1 n )) |f| Lp([0,T2]×Rn )). 66 ) 1 p � .

We remark that the constant C2 stems from finding an upper bound for the integral � T2 T1 |t − τ|−β dt, τ ∈ [0, T2], where β < 1 is a fixed number. Thus if T1 and T2 vary within the set [0, T ], C2 can be chosen to be independent of those numbers. Set and |m| M T 1 = |m| L∞([0,T1]×R n ) + [m] C r 1 ([0,T1]; C(R n )) + [m] C([0,T1]; C r 2 (R n )) [m] M T 1 , T 2 = [m] C r 1 ([T1,T2]; C(R n )) + [m] C([T1,T2]; C r 2 (R n )) . Then our observations can be summarized as follows. Lemma 4.2.2 Suppose that 1 < p < ∞, 0 < T , 0 ≤ T1 ≤ T2 ≤ T , and 0 < si < ri < 1 for i = 1, 2. Let Y T2 s1 = Bpp([0; T2]; Lp(Rn )) ∩ Lp([0, T2]; Bs2 pp(Rn )), and M T2 r1 = C ([0, T2]; C(Rn )) ∩ C([0, T2]; Cr2 (Rn )). Then there exists a constant C > 0 not depending on T1 and T2, such that � |mf| Y T2 ≤ C (4.27) for all m ∈ M T2 and f ∈ Y T2 . 4.2.3 Variable coefficients � |m| M T 1 |f| Y T 1 + |m| L∞([T1,T2]×R n )(1 + [m] M T 1 , T 2 )|f| Y T 2 The goal of this subparagraph is to extend the results proven in Subsection 4.2.1 to variable coefficients. Besides we no longer stick to differential operators consisting only of the principal part but consider general operators of second (Volterra equation) respectively first order (boundary condition). To start with, recall the notation R n+1 + = {(x′ , y) ∈ Rn+1 : x ′ ∈ Rn , y > 0}. Define the operators A(t, x, Dx) and B(t, x ′ , Dx) by and A(t, x, Dx) = −a(t, x) : ∇ 2 x + a1(t, x) · ∇x + a0(t, x), t ∈ J, x ∈ R n+1 + , B(t, x ′ , Dx) = −∂y + b(t, x ′ ) · ∇x ′ + b0(t, x ′ ), t ∈ J, x ′ ∈ R n , (4.28) respectively. We are concerned with the two separate problems � v + k ∗ A(·, x, Dx)v = f, t ∈ J, x ∈ R n+1 + , v = g, t ∈ J, x ′ ∈ R n , y = 0 � v + k ∗ A(·, x, Dx)v = f, t ∈ J, x ∈ R n+1 + , B(t, x ′ , Dx)v = h, t ∈ J, x ′ ∈ R n , y = 0 and seek, as in Subsection 4.2.1, unique solutions in the regularity class Z = H α p (J; Lp(R n+1 + )) ∩ Lp(J; H 2 p(R n+1 + )). Concerning (4.29), we have the following result. , (4.29) , (4.30) Theorem 4.2.3 Let 1 < p < ∞, J = [0, T ], n ∈ N, and k ∈ K 1 (α, θ), where θ < π, and α ∈ (0, 2) \ � 1 p , 2 2p−1 1 3 , 1 + p , 1 + 2p−1 � . Suppose a ∈ Cul(J × R n+1 + , Sym{n + 1}), a1 ∈ L∞(J × R n+1 + , Rn+1 ), a0 ∈ L∞(J × R n+1 + ), and assume further that there exists c0 > 0 such that a(t, x)ξ · ξ ≥ c0|ξ| 2 , t ∈ J, x ∈ Rn+1 , ξ ∈ Rn+1 . Then (4.29) has a unique solution in the space Z if and only if the data f and g are subject to the conditions (i)-(vi) stated in Theorem 4.2.1. 67

Lemma 4.2.1 Let 1 < p < ∞, s1, s2 ∈ (0, 1), 0 < T2, 0 ≤ T1 ≤ T2, and Ω be an<br />

arbitrary domain in R n . Then<br />

|mf| Y T 2 ≤ |m| Y T 1 ∩L∞ |f| Y T 1 ∩L∞ + 2|m| Y T 1 , T 2 ∩L∞ |f| Y T 2 ∩L∞ (4.26)<br />

for all m, f ∈ Y T2 ∩ L∞([0, T2] × Ω).<br />

We turn now to products where one factor might be unbounded. Such a constellation<br />

arises for example when s1 < 1/p. We confine ourselves to the case Ω = R n . By means<br />

of extension and restriction, the subsequent multiplication property can be transferred<br />

to domains <strong>with</strong> sufficiently smooth <strong>boundary</strong>.<br />

Suppose f ∈ Y T2 and m ∈ C r1 ([0, T2]; C(R n )) ∩ C([0, T2]; C r2 (R n )) =: M T2 for some<br />

si < ri < 1, i = 1, 2. Letting J be a subinterval of [0, T2], we then estimate<br />

[mf] Lp(J;B s 2<br />

pp(R n ))<br />

where<br />

� �<br />

I(m, f) ≤ (<br />

J<br />

R n<br />

�<br />

= (<br />

�<br />

�<br />

J Rn Rn ≤ |m| s<br />

L∞(J×Ω)[f]<br />

Lp(J;B 2<br />

pp(Rn ))<br />

�<br />

B1(y)<br />

=: I1(m, f) + I2(m, f).<br />

|m(t, x)f(t, x) − m(t, y)f(t, y)| p<br />

|m(t, x) − m(t, y)| p |f(t, y)| p<br />

|x − y| n+s2p<br />

|x − y| n+s2p<br />

+ I(m, f),<br />

dx dy dt) 1<br />

�<br />

p + (<br />

�<br />

J Rn �<br />

dx dy dt) 1<br />

p<br />

· · ·) 1<br />

p<br />

R n \B1(y)<br />

We put [m] C(J; Cr2 (Rn )) = supt∈J, x, y∈Rn |m(t, x) − m(t, y)| |x − y| −r2 . By hypothesis<br />

1 − s2/r2 > 0. If η ∈ [0, 1 − s2/r2), then (1 − η)r2 − s2 > 0, and we obtain<br />

I1(m, f) ≤ (2|m| L∞(J×Rn )) η [m] 1−η<br />

C(J; Cr2 (Rn )) (<br />

� � �<br />

|f(t, y)| p dx dy dt 1<br />

) p<br />

|x − y| n−((1−η)r2−s2)p<br />

Further,<br />

Hence<br />

[mf] Y T 2<br />

2<br />

J Rn B1(y)<br />

≤ C(p, r2, s2, n)|m| η<br />

L∞(J×Rn ) [m]1−η<br />

C(J; Cr2 (Rn )) |f| Lp(J×Rn ).<br />

� �<br />

I2(m, f) ≤ 2|m| L∞(J×Rn )(<br />

J Rn �<br />

Rn \B1(y)<br />

≤ C(p, s2, n)|m| L∞(J×Rn )|f| Lp(J×Rn ).<br />

|f(t, y)| p dx dy dt<br />

|x − y| n+s2p<br />

≤ [mf] Lp([0,T1];B s2 pp(Rn )) + [mf] Lp([T1,T2];B s2 pp(Rn ))<br />

≤ C(|m| L∞([0,T1]×Rn ) + [m] C([0,T1]; Cr2 (Rn )) )(|f| Lp([0,T1]×Rn ) + [f] T<br />

Y 1 )<br />

2<br />

+|m| L∞([T1,T2]×Rn �<br />

) [f] T<br />

Y 2 + C(1 + [m] C([T1,T2]; C<br />

2<br />

r2 (Rn )) )|f| Lp([0,T2]×Rn )<br />

Let now [m] Cr1 (J; C(Rn )) = supt, τ∈J, x∈Rn |m(t, x) − m(τ, x)| |t − τ| −r1 . To estimate<br />

, we take (4.23) <strong>with</strong> Ω = Rn as starting point, apply Fubini’s theorem, and<br />

[mf] T<br />

Y 2<br />

1<br />

use the same estimation techniques as for [mf] T<br />

Y 2<br />

2<br />

[mf] Y T 2<br />

1<br />

to the result<br />

≤ |m| L∞([0,T1]×Rn )[f] T<br />

Y 1 + C1[m] Cr1 ([0,T1]; C(R<br />

1<br />

n )) |f| Lp([0,T1]×Rn )<br />

+2|m| L∞([T1,T2]×Rn )([f] T<br />

Y 2 + C2[m] Cr1 ([T1,T2]; C(R<br />

1<br />

n )) |f| Lp([0,T2]×Rn )).<br />

66<br />

) 1<br />

p<br />

�<br />

.

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

Saved successfully!

Ooh no, something went wrong!