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.

Given φ, φ ′ ∈ (0, π] we denote by H ∞ (Σφ × Σφ<br />

′) the Banach algebra of all bounded<br />

φ′<br />

holomorphic scalar-valued functions on Σφ × Σφ ′ equipped <strong>with</strong> the norm |f|φ, ∞<br />

sup{|f(λ, λ ′ )| : |arg λ| < φ, |arg λ ′ | < φ ′ }, and we put H0(Σφ × Σφ ′) = {f ∈ H∞ (Σφ ×<br />

Σφ ′) : ∃(f1, f2) ∈ H0(Σφ)×H0(Σφ ′), f1 and f2 non-vanishing, and f(f1f2) −1 ∈ H ∞ (Σφ×<br />

Σφ ′)}. Let A, B ∈ S(X) <strong>with</strong> spectral angles φA and φB, respectively, commute in the<br />

resolvent sense. For φ ∈ (φA, π), φ ′ ∈ (φB, π), and f ∈ H0(Σφ × Σφ<br />

f(A, B) = − 1<br />

4π 2<br />

�<br />

Γψ×Γ ψ ′<br />

′) one defines<br />

:=<br />

f(λ, λ ′ )(λ − A) −1 (λ − B) −1 dλ dλ ′ , (2.12)<br />

where (ψ, ψ ′ ) ∈ (φA, φ) × (φB, φ ′ ). This integral converges in B(X) and does not depend<br />

on the choice of ψ and ψ ′ . Via ΦA, B(f) = f(A, B), it defines a joint functional calculus<br />

ΦA, B : H0(Σφ × Σφ ′) → B(X). In analogy to Definition 2.2.1, we say that (A, B) admits<br />

a bounded joint H∞-calculus (symbolized by (A, B) ∈ H∞ (X)) if there exist φ ∈ (φA, π),<br />

φ ′ ∈ (φB, π), and a constant Kφ, φ ′ < ∞ such that<br />

φ′<br />

|f(A, B)| ≤ Kφ, φ ′|f|φ, ∞ for all f ∈ H0(Σφ × Σφ ′). (2.13)<br />

If (A, B) ∈ H∞ (X), then the functional calculus for (A, B) on H0(Σφ × Σφ ′) extends<br />

uniquely to H∞ (Σφ × Σφ ′).<br />

An interesting question is the following: what are the Banach spaces X for which<br />

(A, B) admits a bounded joint H∞-calculus as soon as A and B, each, admit a bounded<br />

H∞-calculus? In [1] Albrecht was able to prove that this is the case if X = Lp(Ω, Σ, µ),<br />

1 < p < ∞, where (Ω, Σ, µ) is a σ-finite measure space, see also [2, Section 5]. Lancien et<br />

al. [51] extended this result to a class of Banach spaces which enjoy a certain geometric<br />

property. They also give an example for a Banach space not possessing the joint calculus<br />

property. We would further like to mention a result by Kalton and Weis [47] which<br />

asserts, <strong>with</strong>out additional assumption on X, that if A ∈ H∞ (X) and B ∈ RH∞ (X)<br />

<strong>with</strong> commuting resolvents, then (A, B) admits a bounded joint H∞-calculus. We consider now an important example.<br />

Example 2.4.1 Let 1 < p < ∞, X = Lp(R+ × Rn ), and denote the independent variables<br />

by t (∈ R+) resp. x (∈ Rn ). Take B = ∂t <strong>with</strong> domain D(B) = 0H 1 p(R+; Lp(Rn )),<br />

and define A as the natural extension of −∆x in Lp(Rn ) <strong>with</strong> D(−∆x) = H2 p(Rn ) to X,<br />

i.e. D(A) = Lp(R+; H2 p(Rn )) and Af = −∆xf, f ∈ D(A). Then A and B commute in<br />

the resolvent sense, and A, B ∈ H ∞ (X) <strong>with</strong> H ∞ -angles φ ∞ A = 0 resp. φ∞ B<br />

= π/2. Since<br />

X has the joint calculus property, we have (A, B) ∈ H ∞ (X). More precisely, for each<br />

η ∈ (0, π/2), there exists Cη > 0 such that for all f ∈ H ∞ (Ση × Σ π<br />

2 +η), f(A, B) ∈ B(X)<br />

η, π/2+η<br />

and |f(A, B)| B(X) ≤ Cη|f| ∞ .<br />

Regarding functions in X as elements in Lp(R+; Lp(Rn )), the resolvent of B admits<br />

the kernel representation<br />

(λ − B) −1 w(t) = −<br />

� t<br />

0<br />

e λ(t−s) w(s) ds, t ∈ R+,<br />

for all w ∈ X. Thus, for f ∈ H0(Ση × Σ π<br />

2 +η), the operators f(A, B) admit a kernel<br />

representation as well, namely<br />

� �<br />

t �<br />

−1<br />

f(A, B)w(t) =<br />

e<br />

0 2πi Γψ ′<br />

λ′ �<br />

(t−s) ′ ′<br />

f(A, λ )dλ w(s) ds, t ∈ R+. (2.14)<br />

20

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

Saved successfully!

Ooh no, something went wrong!