3
$\begingroup$

Let $u, \eta$ be smooth functions and $\eta$ compactly supported in $(0,1)$. Integrating by parts, we can easily prove $$-\int_0^1 u_{xxx}u_x \eta = \int_0^1 (u_{xx})^2\eta - \int_0^1 \frac{1}{2} (u_x)^2 \eta_{xx}$$

Can we obtain the same result using only the Fourier series
$$u(x) = \sum_{n=1}^\infty \sqrt 2 \cos (\pi n x) \, a_n, \qquad \eta(x) = \sum_{n=1}^\infty \sqrt 2 \cos (\pi n x) \, b_n $$ as a tool?

$\endgroup$
3
  • $\begingroup$ without partial integration, you can just evaluate all integrals and the result will follow, but it will be a tedious calculation; any motivation for avoiding P.I. ? $\endgroup$ Commented Jan 26, 2021 at 10:12
  • $\begingroup$ @CarloBeenakker I want to study a similar identity for a more general operator, where the Fourier expansion is all I have, but then I realized that I'm not even able to compute the integrals in the case of the Laplacian. How would you evaluate them? $\endgroup$
    – user139844
    Commented Jan 26, 2021 at 10:37
  • $\begingroup$ see answer box, quite a tedious calculation, but it works out. $\endgroup$ Commented Jan 26, 2021 at 13:20

1 Answer 1

2
$\begingroup$

term on left hand side: $$L=-\int_0^1 dx\, u_{xxx}u_x \eta =2^{3/2}\pi^4 \sum_{n,m,k=1}^\infty n^3m a_na_mb_k \int_0^1 dx\,\sin(n\pi x)\sin(m\pi x)\cos(k\pi x)=$$ $$=2^{3/2}\pi^4 \frac{1}{4}\sum_{n,m,k=1}^\infty n^3m a_na_mb_k \left(\delta_{m,n+k}+\delta_{n,m+k}-\delta_{k,n+m}\right).$$ two terms on right hand side $$R_1=\int_0^1 dx\, u_{xx}^2 \eta =2^{3/2}\pi^4 \sum_{n,m,k=1}^\infty n^2m^2 a_na_mb_k \int_0^1 dx\,\cos(n\pi x)\cos(m\pi x)\cos(k\pi x)=$$ $$=2^{3/2}\pi^4 \frac{1}{4}\sum_{n,m,k=1}^\infty n^2m^2 a_na_mb_k \left(\delta_{m,n+k}+\delta_{n,m+k}+\delta_{k,n+m}\right).$$ $$R_2=-\frac{1}{2}\int_0^1 dx\, u_{x}^2 \eta_{xx} =\frac{1}{2}2^{3/2}\pi^4 \sum_{n,m,k=1}^\infty nmk^2 a_na_mb_k \int_0^1 dx\,\sin(n\pi x)\sin(m\pi x)\cos(k\pi x)=$$ $$=\frac{1}{2}2^{3/2}\pi^4 \frac{1}{4}\sum_{n,m,k=1}^\infty nmk^2 a_na_mb_k \left(\delta_{m,n+k}+\delta_{n,m+k}-\delta_{k,n+m}\right).$$ left hand side minus right hand side $$L-R_1-R_2=2^{3/2}\pi^4 \frac{1}{4}\left(\frac{1}{2}\sum_{n=1}^\infty\sum_{m=n+1}^\infty (mn^3-nm^3)a_na_m b_{m-n}+\frac{1}{2}\sum_{m=1}^\infty\sum_{n=m+1}^\infty (mn^3-nm^3)a_{n}a_m b_{n-m}\right.$$ $$\left.+\frac{1}{2}\sum_{n,m=1}^\infty m n (m^2 -n^2)a_na_m b_{n+m}\right)=0.$$

$\endgroup$
12
  • $\begingroup$ Thank you very much. Please, let me know if you manage to solve the right-hand side $\endgroup$
    – user139844
    Commented Jan 26, 2021 at 11:05
  • $\begingroup$ Thank you very much! Could you add some details to the last line? Why is $L-R_1-R_2 \equiv 0$? $\endgroup$
    – user139844
    Commented Jan 26, 2021 at 13:19
  • $\begingroup$ Also in that line, why is $k$ gone? $\endgroup$
    – user139844
    Commented Jan 26, 2021 at 13:20
  • $\begingroup$ the double sum over $n,m$ vanishes by symmetry, exchanging $n$ and $m$; I have used the Kronecker delta to carry out the sum over $k$. $\endgroup$ Commented Jan 26, 2021 at 13:21
  • $\begingroup$ Thanks! Among the factors $(mn^3-nm^3), \quad (mn^3-nm^3), \quad m n (m^2 -n^2)$, where did the contribution of $k$ end up? $\endgroup$
    – user139844
    Commented Jan 26, 2021 at 13:29

You must log in to answer this question.