I have found that the $H^2(D)$ norm of a field with zero Cauchy data on $\partial D$ (i.e. in $H_0^2(D)$) is equivalent to the $L^2(D)$ norm of its Laplacian, where D is simply connected with smooth boundary in $\mathbb{R}^n$. How can i prove this using Poincare inequality;
-
1There's also a similar result for functions without zero boundary data, but it's harder to prove: see this question – Feb 12 '13 at 18:21
1 Answers
Note that we have the following Poincaré inequality for $H^2_0(D)$: $$\|u\|_{H^2_0(D)} \leq C \|D^2 u\|_{L^2(D)}^2.$$ This is obtained by chaining the Poincaré inequality for $u$ with the Poincaré inequality for $Du$. Therefore we may take $$\|u\|_\ast^2 = \|D^2 u\|_{L^2(D)}^2$$ as our norm on $H^2_0(D)$, as it is equivalent to the standard $H^2_0(D)$ norm.
We claim that $$\|\Delta u\|_{L^2(D)} = \|D^2 u\|_{L^2(D)} = \|u\|_\ast$$ for any $u \in H_0^2(D)$. To see this, first consider $u \in C_0^\infty(D)$. Then integration by parts and commutativity of partial derivatives for smooth functions implies $$\int_D u_{x_i x_i} u_{x_j x_j} ~dx = -\int_D u_{x_i} u_{x_j x_j x_i} ~dx = - \int_D u_{x_i} u_{x_j x_i x_j} ~dx = \int_D u_{x_i x_j} u_{x_i x_j} ~dx$$ for all $1 \leq i, j \leq n$. Summing over all $i$ and $j$ then gives $$\|\Delta u\|_{L^2(D)} = \|D^2 u\|_{L^2(D)}$$ for all $u \in C_0^\infty(D)$. Since $C_0^\infty(D)$ is dense in $H_0^2(D)$, passing to limits we find that $$\|\Delta u\|_{L^2(D)} = \|D^2 u\|_{L^2(D)} \text{ for all } u \in H_0^2(D).$$ This gives the desired equality of norms.
- 18,318
- 5
- 60
- 77
-
thanks a lot, the only thing i can't see is the poincare inequality for $H_0^2$. I thought that it was $∥u∥{H^2_0}(D)^2≤C\sum{|a|=2}∥D^au∥^2L_2(D)$..i must be missing something! – nikosp Feb 12 '13 at 18:48
-
1In my notation, $|D^2 u|{L^2(D)}$ means $$|D^2 u|{L^2(D)} = \left( \sum_{|\alpha| = 2} |D^\alpha u|{L^2(D)}^2 \right)^{1/2}.$$ This is the notation used throughout Evans's _Partial Differential Equations. – Henry T. Horton Feb 12 '13 at 18:54
-
-
Does this holds if we consider $H^2$ instead of $H^2_0$? – Danilo Gregorin Afonso Dec 06 '19 at 18:07
-
Thank you. I am wondering whether this generalizes to other $p \neq 2$. Do you know anything about that? – shuhalo Jun 03 '23 at 09:52