MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $\Omega=(a,b)$ a finite interval, $g\in \mathcal{H}^k(\Omega)$ some integer $k$, with $g(a)=0$ and let $\epsilon>0$. Is there an $\alpha\geq 1+k$ such that:

$ \left\|g\right\|_{L_2(a,a+\epsilon)}\leq C\epsilon^{\alpha}\left\|g\right\|_{\mathcal{H}^k(\Omega)} $ where $C$ and $\alpha$ do not depend on $g$ or $\epsilon>0$?

I am thinking of $\epsilon>0$ small so the bound only has to hold for sufficiently small $\epsilon$.

If the above is not possible are there any extra conditions I can put on $g$ at the end point $a$ or extra smoothness of $g$ in $\Omega$ I can impose?

Poincare's inequality for $k=1$: I get $ \left\|g\right\|_{L_2(a,a+\epsilon)}\leq (1+C)\left\|g\right\|_{\mathcal{H}^1(a,a+\epsilon)}$ How do I continue from here?

share|cite|improve this question
Assuming that by $\|g\|_{H^k(\Omega)}$ you mean $\sum_{0\le j\le k}\|g^{(j)}\|_{L^2(\Omega)}$ then this is trivially true for $\alpha=0$. I think you forgot some extra condition. – Florian Nov 25 '10 at 9:50
Yes. I forgot the important condition on $\alpha$: $\alpha>1+k$!! – alext87 Nov 25 '10 at 9:58
This requirement is too strong: it's not even true for $k=0$. But $\alpha=k$ seems reasonable to me. – Florian Nov 25 '10 at 10:20
Thanks, sure. I was hoping to get $\alpha=1+k$ with the $1$ coming from the restricted domain $(a,a+\epsilon)$. Does $\alpha=k$ work? – alext87 Nov 25 '10 at 10:25
It's true at least for $\alpha=k=1$ (Poincaré's inequality). But in order to iterate the inequality, you'd need $g^{(j)}(a)=0$ for $j<k$. – Florian Nov 25 '10 at 10:32

You need to make the stronger assumption $g(a)=g'(a)=...=g^{(k-1)}(a)=0$. Then your statement is true with $\alpha=k$. You can see this by using the Cauchy-Schwarz inequality in $g^{(k-1)}(x)=\int_a^x g^{(k)}(y)\,dy$ to obtain $|g^{(k-1)}(x)|\le C(x-a)^{1/2}$, and then integrating repeatedly to get $|g(x)|\le C(x-a)^{k-1/2}$.

This is essentially optimal, since the function $(x-a)^{k-1/2}/\ln(x-a)$ satisfies all the hypotheses. In particular, you cannot get $\alpha=k+1$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.