02.12.2012 Views

Quasilinear parabolic problems with nonlinear boundary conditions

Quasilinear parabolic problems with nonlinear boundary conditions

Quasilinear parabolic problems with nonlinear boundary conditions

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

the data, and R(v) contains only terms of lower order, see equation (4.52) below. By<br />

means of the contraction principle, this fixed point equation can be solved first on a<br />

small interval [0, T1] (denote the (unique) solution by v 1 ), then on [0, T2], where T2 > T1<br />

and the unknown v equaling v 1 on [0, T1], and proceeding in this way, finally, after<br />

finitely many steps, it can be solved on the entire interval [0, T ]. Here it is essential that<br />

maxi |Ti+1 − Ti| is sufficiently small to ensure that, in each step of this procedure, the<br />

mapping G is a strict contraction.<br />

Let us start <strong>with</strong> the spatial localization. By boundedness of Γ, there exists r0 > 0<br />

such that Γ is entirely contained in the open ball Br0 (0). If Ω is unbounded we set<br />

U0 = {x ∈ Rn+1 : |x| > r0}, otherwise we may assume that Ω ⊂ Br0 (0) and put<br />

U0 = ∅. The other assumptions on Ω allow us to cover Br0 (0) by finitely many open sets<br />

Uj, j = 1, . . . , N, which are subject to the following <strong>conditions</strong>.<br />

(L1) Uj ∩ Γ = ∅ and Uj = Brj (xj) for all j = 1, . . . , N1;<br />

(L2) Uj ∩ ΓD �= ∅ for N1 + 1 ≤ j ≤ N2, Uj ∩ ΓN �= ∅ for N2 + 1 ≤ j ≤ N, and<br />

for each j in either index set, there exist xj ∈ Uj ∩ ΓD resp. Uj ∩ ΓN and ˜ ζj ∈ C 2 (R n )<br />

<strong>with</strong> compact support such that - using coordinates corresponding to xj (i.e. xj = 0<br />

and n(xj) = (0, . . . , 0, −1)) - Γ ∩ Uj = {x = (x ′ , y) ∈ Uj : y = ˜ ζj(x ′ )} as well as<br />

Ω ∩ Uj = {x = (x ′ , y) ∈ Uj : y > ˜ ζj(x ′ )}, and Uj = ˜ ϑ −1<br />

j (Brj (xj)), where ˜ ϑj is related to<br />

˜ζj as described above.<br />

(L3) Ui ∩ Uj = ∅ for all N1 + 1 ≤ i ≤ N2 and N2 + 1 ≤ j ≤ N.<br />

It is then not difficult to construct local operators Aj = Aj(t, x, Dx), j = 0, . . . , N,<br />

and Bj = Bj(t, x, Dx), j = N2 + 1, . . . , N, of second resp. first order which enjoy the<br />

subsequent properties.<br />

(LO1) Aj is defined on J ×R n+1 if 0 ≤ j ≤ N1, and on J ×Ωj otherwise; here the set<br />

Ωj is given in coordinates corresponding to xj by means of Ωj = {x = (x ′ , y) ∈ R n+1 :<br />

y ≥ ˜ ζj(x ′ )}; Bj is defined on J × Γj for all j = N2 + 1, . . . , N, where Γj = {x = (x ′ , y) ∈<br />

R n+1 : y = ˜ ζj(x ′ )};<br />

(LO2) the coefficients of Aj coincide <strong>with</strong> the corresponding coefficients of A(t, x, Dx)<br />

on Ω∩Uj, for all j = 0, . . . , N, and the coefficients of Bj coincide <strong>with</strong> those of B(t, x, Dx)<br />

on Γ ∩ Uj, for all j = N2 + 1, . . . , N;<br />

(LO3) Aj satisfies the assumptions of Theorem 4.1.1 for all j = 0, . . . , N1; A ˜ ϑj<br />

j =<br />

Θ −1<br />

j AjΘj defined on J × R n+1<br />

+<br />

fulfills the assumptions of Theorem 4.2.3 for all j =<br />

N1 + 1, . . . , N; finally, B ˜ ϑj<br />

j = Θ−1<br />

j BjΘj defined on J × Rn satisfies the assumptions of<br />

Theorem 4.2.4 for all j = N2 + 1, . . . , N.<br />

Here we use the fact that ellipticity and normality, as well as smoothness of the<br />

coefficients of A(t, x, Dx) resp. B(t, x, Dx) are preserved in Ω ∩ Uj resp. Γ ∩ Uj under<br />

the coordinate transformations ¯x = ˜ ϑj(x) for all j = N1 + 1, . . . , N. We refer to [29,<br />

Section 8.2], where appropriate extensions of the coefficients are constructed by means<br />

of reflection and cut-off techniques.<br />

We choose next a partition of unity {ϕj} N j=0 ⊂ C∞ (R n+1 ) such that � N<br />

j=0 ϕj(x) ≡ 1<br />

on Ω, 0 ≤ ϕj(x) ≤ 1, and supp ϕj ⊂ Uj; we fix also a family {ψj} N j=0 ⊂ C∞ (R n+1 ) that<br />

satisfies ψj ≡ 1 on an open set Vj containing supp ϕj, as well as supp ψj ⊂ Uj. As to<br />

localization w.r.t. time, we subdivide the interval [0, T ] according to 0 =: T0 < T1 <<br />

. . . < TM−1 < TM := T and put δ := maxi |Ti+1 − Ti|. Then, owing to (LO1) and (LO2),<br />

76

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

Saved successfully!

Ooh no, something went wrong!