I want to prove Friedrichs' inequality: if $y \in C^1[0, l]$ and $y(0)=y(l)=0$, then $$\int_0^l y(x)^2 dx \leq C \int_0^l y'(x)^2dx.$$

I need to prove it with most optimal constant $C$. My approach is as follows:

  1. I prove the inequality with some non-optimal constant $C = C_0$
  2. I rewrite the inequality as $$\frac{1}{C} \leq \frac{\int_0^l y'(x)^2dx}{\int_0^l y(x)^2dx},$$ notice that WLOG $\int_0^l y(x)^2 dx = 1$ since $RHS[\alpha y] = RHS[y]$ for $\alpha \in \mathbb R$. So I get a conditional extrema problem $$J[y] = \int_0^l y'(x)^2dx \to min, \;\;\;\; G[y] = \int_0^l y(x)^2 dx = 1.$$

My questions are:

  1. If I find the unique extremal $\hat y$ of this variational problem and prove that it is a local minimum, can I conclude that $\hat y$ is actually a global minimum?

I'd thought so since the functional $J[y]$ has a lower-bound $1/C_0$, and the only way it can have a unique extremal is that the extremal is a global minimum (not a saddle or a maximum because of a known lower-bound).

  1. Can I conclude the same without proving that $\hat y$ is a minimum of any kind (just some extremal) by the same argument?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.