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.

Example 2.8.1 For J = [0, T ] and aα(t) = tα−1 /Γ(α), t > 0, α ∈ (0, 1), the operator<br />

B in Corollary 2.8.1 takes the form<br />

Bu(t) = d<br />

dt<br />

� t<br />

0<br />

a1−α(t − s)u(s) ds, t > 0, u ∈ 0H α p (J; X),<br />

thus coincides <strong>with</strong> (d/dt) α , the derivation operator of (fractional) order α. The function<br />

Bu is called the fractional derivative of u of order α.<br />

We point out that Corollary 2.8.1 will prove extremely useful in establishing maximal<br />

Lp-regularity results for <strong>parabolic</strong> Volterra equations. On the one hand it describes<br />

precisely the mapping properties of the convolution operators associated to a K-kernel,<br />

on the other hand it enables us to apply the Dore-Venni theorem to operator sums in<br />

Lp(J; X) which involve an inverse convolution operator B.<br />

We next show how Volterra operators can be used to introduce equivalent norms for<br />

the vector-valued Bessel-potential spaces Hα p (J; X). Suppose we are in the situation of<br />

Corollary 2.8.1, where we restrict ourselves to the case J = [0, T ]. Assume further that<br />

α ∈ (0, 2) \ {1/p, 1 + 1/p} and let µ > 0. For f ∈ Hα p (J; X), we put <strong>with</strong> some abuse of<br />

language<br />

|f| (a,µ)<br />

H α p (J;X) =<br />

⎧<br />

⎪⎨ |Bf| Lp(J;X)<br />

: α ∈ (0,<br />

⎪⎩<br />

1<br />

p )<br />

|B(f − f(0))| Lp(J;X) + µ|f(0)|X<br />

: α ∈ ( 1 1<br />

p , 1 + p )<br />

|B(f − f(0) − t ˙ f(0))| Lp(J;X) + µ|f(0)|X + µ| ˙ f(0)|X : α ∈ (1 + 1<br />

p , 2).<br />

(2.30)<br />

This is a well-defined expression in view of Sobolev’s embedding theorem. Observe that<br />

| · | (a,µ)<br />

H α p (J;X) enjoys the properties of a norm for the space Hα p (J; X). To verify that it<br />

is equivalent to the usual norm | · | H α p (J;X) we can employ Corollary 2.8.1 and Sobolev<br />

embeddings. Indeed, if α > 1 + 1/p and f ∈ H α p (J; X), we may estimate<br />

|f| Hα p (J;X) ≤ |f − f(0) − t ˙ f(0)| Hα p (J;X) + |f(0)| Hα p (J;X) + |t ˙ f(0)| Hα p (J;X)<br />

= |a ∗ B(f − f(0) − t ˙ f(0))| Hα p (J;X) + |{t ↦→ 1}| Hα p (J)|f(0)|X<br />

+|{t ↦→ t}| Hα p (J)| ˙ f(0)|X<br />

≤ c1 (|B(f − f(0) − t ˙ f(0))| Lp(J;X) + µ|f(0)|X + µ| ˙ f(0)|X)<br />

= c1 |f| (a,µ)<br />

H α p (J;X),<br />

<strong>with</strong> c1 not depending on f. Conversely, by<br />

and<br />

|f(0)|X + | ˙<br />

f(0)|X ≤ |f| C 1 (J;X) ≤ CSob|f| H α p (J;X)<br />

|B(f − f(0) − t ˙<br />

f(0))| Lp(J;X) ≤ c|f − f(0) − t ˙<br />

f(0)| H α p (J;X)<br />

≤ c(|f| H α p (J;X) + |f(0)| H α p (J;X) + |t ˙<br />

f(0)| H α p (J;X))<br />

we also get an inequality of the form<br />

≤ c(|f| H α p (J;X) + ˜c(|f(0)|X + | ˙<br />

f(0)|X)),<br />

|f| (a,µ)<br />

H α p (J;X) ≤ c2 |f| H α p (J;X),<br />

29<br />

(2.31)

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

Saved successfully!

Ooh no, something went wrong!