17.01.2013 Views

PDF (1016 kB)

PDF (1016 kB)

PDF (1016 kB)

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

commutes. By the elliptic estimates of Lemma 5.6, suitable domain and codomain for the operator<br />

on the lower horizontal arrow are the standard W 1,p and Lp spaces, for 1 < p < ∞. Moreover, if<br />

u ∈ Map(À,�n ) then<br />

�Ru� p<br />

Lp (À+ ) = 1<br />

�À1<br />

4 |z| |u(z)|p dsdt, (87)<br />

�D(Ru)� p<br />

Lp (À+ ) = 2 p−2<br />

�À|Du(z)| p |z| p/2−1 dsdt, (88)<br />

�T u� p<br />

Lp (À+ ) = 2 p−2<br />

�À|u(z)| p |z| p/2−1 dsdt. (89)<br />

Note also that by the generalized Poincaré inequality, the W 1,p norm onÀ+ ∩�r, where�r denotes<br />

the open disk of radius r, is equivalent to the norm<br />

and the ˜ W 1,p norm of Ru is<br />

�v� p<br />

˜W 1,p := �Dv�p<br />

(À+ ∩�r) Lp (À+ ∩�r) +<br />

�À+<br />

|v(ζ)|<br />

∩�r<br />

p |ζ| p dσdτ, (90)<br />

�Ru� p<br />

˜W 1,p �À∩�r<br />

1<br />

=<br />

(À+ ∩�r) 4 2<br />

|u(z)| p |z| p/2−1 dsdt + 2 p−2<br />

�À∩�r2 |Du(z)| p |z| p/2−1 dsdt. (91)<br />

So when dealing with bounded domains, both the transformations R and T involve the appearance<br />

of the weight |z| p/2−1 in the L p norms. Note also that when p = 2, this weight is just 1, reflecting<br />

the fact that the L 2 norm of the differential is a conformal invariant.<br />

By the commutativity of diagram (86) and by the identities (88), (89), Lemma 5.6 applied to<br />

Ru implies the following:<br />

5.8. Lemma. Let V and W be partially orthogonal linear subspaces ofÊn . For every p ∈]1, +∞[,<br />

there exists a constant c = c(p, n) such that<br />

�À|∇u(z)| p |z| p/2−1 ds dt ≤ c p<br />

�À|∂u(z)| p |z| p/2−1 ds dt<br />

for every compactly supported map u : Cl(À) →�n such that ζ ↦→ u(ζ 2 ) is smooth on Cl(À+ ), and<br />

u(s) ∈ N ∗ V ∀s ∈] − ∞, 0], u(s) ∈ N ∗ W ∀s ∈ [0, +∞[.<br />

5.3 Strips with jumping conormal boundary conditions<br />

Let us consider the following data: two integers k, k ′ ≥ 0, k + 1 linear subspaces V0, . . .,Vk of<br />

Ên such that Vj−1 and Vj are partially orthogonal, for every j = 1, . . .,k, k ′ + 1 linear subspaces<br />

′ , . . . , V k ofÊn ′ such that V j−1 and V ′<br />

j are partially orthogonal, for every j = 1, . . .,k′ , and real<br />

numbers<br />

V ′<br />

0<br />

−∞ = s0 < s1 < · · · < sk < sk+1 = +∞, −∞ = s ′ 0 < s′ 1 < · · · < s′ k ′ < −s′ k ′ +1<br />

Denote by V the (k + 1)-uple (V0, . . . , Vk), by V ′ the (k ′ + 1)-uple (V ′ ′<br />

0 , . . . , V ), and set<br />

Let Σ be the closed strip<br />

S := {s1, . . .,sk, s ′ 1 + i, . . .,s′ k ′ + i}.<br />

Σ := {z ∈�|0≤Im z ≤ 1}.<br />

k<br />

= +∞.<br />

The space C ∞ S (Σ,�n ) is the space of maps u : Σ →�n which are smooth on Σ \S , and such that<br />

the maps ζ ↦→ u(sj + ζ 2 ) and ζ ↦→ u(s ′ j + i − ζ2 ) are smooth in a neighborhood of 0 in the closed<br />

upper-right quadrant<br />

Cl(À+ ) = {ζ ∈�|Re ζ ≥ 0, Im ζ ≥ 0}.<br />

59

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

Saved successfully!

Ooh no, something went wrong!