Orthonormal sequence $L^2[0,1]$ (Brezis 5.30)

59 Views Asked by At

I'm trying to solve the following exercise, Let $\left(e_n\right)_{n \geq 1}$ be an orthonormal sequence in $H=L^2(0,1)$. Let $p(t)$ be a given function in $H$. Deduce that

$$ (1) \quad\sum_{n=1}^{\infty}\left|\int_0^t p(s) e_n(s) d s\right|^2 \leq \int_0^t|p(s)|^2 d s . $$ $$ (2) \quad \sum_{n=1}^{\infty} \int_0^1\left|\int_0^t p(s) e_n(s) d s\right|^2 d t \leq \int_0^1|p(t)|^2(1-t) d t. $$

I'm having trouble trying to deduce (2), I imagine that I have to integrate (1) and use monotone convergence to get left side, but how do i get that $(1-t)$ factor on right side.

Also, It seems like (2) with equality implies (1) with equality how does this happen?.

Any help would be appreciated, thank you in advance.

1

There are 1 best solutions below

0
On BEST ANSWER

The Fourier series for $\chi_{[0,t]}(s)p(s)$ is $$ \sum_{n=1}^{\infty}\int_0^1\chi_{[0,t]}(s')p(s')e_n(s')ds'\, e_n(s) \sim \chi_{[0,t]}(s)p(s) \\ \sum_{n=1}^{\infty}\int_0^tp(s')e_n(s')ds'e_n(s)\sim \chi_{[0,t]}(s)p(s). $$ The corresponding Parseval identity for this function of $s$ is: $$ \sum_{n=1}^{\infty}\left|\int_0^tp(s')e_n(s')ds'\right|^2 =\int_0^t|p(s')|^2ds' $$ The second inequality that you want is obtained by integrating the above in $t$ and by using the monotone convergence theorem: $$ \sum_{n=1}^{k}\left|\int_0^tp(s')e_n(s')ds'\right|^2 \le \int_0^t|p(s')|^2ds' \\ \sum_{n=1}^{k}\int_0^1\left|\int_0^tp(s')e_n(s')ds'\right|^2dt \le \int_0^1\int_0^t|p(s')|^2ds'dt $$ Now it's a matter of interchanging the order of integration on the right in order to obtained the desired inequality: $$ \int_0^1\int_0^t|p(s')|^2ds'dt = \int_0^1\int_0^1|p(s')|^2\chi_{[0,t]}(s')ds'\, dt \\ = \int_0^1|p(s')|^2\int_0^1\chi_{[0,t]}(s')dt\, ds' \\ = \int_t^1|p(s')|^2 (1-s')ds' $$