23.10.2014 Views

Regularity near the characteristic boundary for sub-laplacian operators

Regularity near the characteristic boundary for sub-laplacian operators

Regularity near the characteristic boundary for sub-laplacian operators

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

382 DIMITER VASSILEV<br />

(2) It is enough to prove that <strong>the</strong> L q () norms of u are uni<strong>for</strong>mly bounded by<br />

some sufficiently large but fixed L q 0<br />

norm of u, q 0 ≥ p ∗ , which is finite from (1).<br />

We shall do this by iteration. We use <strong>the</strong> function F(u) from part (1) in <strong>the</strong> weak<br />

<strong>for</strong>m (4-3) of our equation. The left-hand side is estimated from below as be<strong>for</strong>e,<br />

in (4-5). This time, though, we use Hölder’s inequality to estimate from above <strong>the</strong><br />

right-hand side:<br />

∫<br />

V |u| p−2 ∥<br />

(4-8)<br />

uF(u) d H ≤ ‖V ‖ t |u| p−1 F(u) ∥ t ′<br />

<br />

With <strong>the</strong> estimate from below, we come to<br />

Letting l → ∞, we obtain<br />

(4-9)<br />

≤ ‖V ‖ t<br />

∥ ∥ C(q)|G(u)| p∥ ∥<br />

t ′ ≤ C(q)‖V ‖ t ‖u‖ q qt ′ .<br />

S p ‖G(u)‖ p p ∗ ≤ C(q)‖V ‖ t ‖u‖ q qt ′ .<br />

∥ |u|<br />

q/p ∥ ∥ p p ∗ ≤ C(q)<br />

S p<br />

‖V ‖ t ‖u‖ q qt ′ .<br />

Set δ = p ∗ /(pt ′ ). The assumption t > Q/p implies δ > 1, since <strong>the</strong> latter is equivalent<br />

to t ′ < p ∗ /p = t ′ 0 , as t 0 = Q/p. With this notation we can rewrite (4-9) as<br />

( C(q)<br />

) 1/q‖V<br />

1/q<br />

(4-10) ‖u‖ δqt ′ ≤<br />

‖ t ‖u‖ qt ′.<br />

S p<br />

Recall that C(q) ≤ Cq p−1 . At this point, we define q 0 = p ∗ t ′ and q k = δ k q 0 . After<br />

a simple induction, we obtain<br />

k−1<br />

∏ ( p−1) t<br />

(4-11) ‖u‖ qk ≤ Cq ′ /q ∑ j<br />

j<br />

‖V ‖ t′ k−1<br />

j=0 1/q j<br />

t ‖u‖ q0 .<br />

j=0<br />

We observe that <strong>the</strong> right-hand side is finite,<br />

(4-12)<br />

∞∑<br />

j=0<br />

1<br />

= 1 ∑<br />

∞ 1<br />

q j<br />

q 0 δ j < ∞<br />

j=1<br />

and<br />

∞∑<br />

j=1<br />

log q j<br />

q j<br />

< ∞,<br />

because δ > 1. Letting j → ∞, we obtain<br />

‖u‖ ∞ ≤ C‖u‖ q0 .<br />

□<br />

Remark 4.2. When is a bounded open set, we trivially have V ∈ L Q/p ()<br />

whenever V ∈ L t () with t > Q/p. Also, in this case one can obtain a uni<strong>for</strong>m<br />

estimate of <strong>the</strong> L ∞ () norm of u by its L p∗ () norm that does not depend on <strong>the</strong><br />

distribution function of V , as we had in <strong>the</strong> preceding <strong>the</strong>orem. This can even be

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

Saved successfully!

Ooh no, something went wrong!