1
$\begingroup$

I need a really good explication with details of this proof of Hilbert - Schmidt theorem :

Let $(H,\left< , \right>)$ be a complex Hilbert space and let $A:H\rightarrow H$ be a bounded, compact, self-adjoint operator and $(\lambda_n)_n$ a sequence of non-zero real eigenvalues where each eigenvalue of A is repeated in the sequence according to its multiplicity, then there exists an orthonormal set $(v_n)_n$ of corresponding eigenfunctions, i.e. $Av_n=\lambda_nv_n.$ Then A can be written as $$Ax=\sum_n \lambda_n\left<x,v_n\right>v_n,\;\;\;\;\forall\;x\in H. $$

Proof: It's enough to show that $Im A\subset \overline{Sp}(v_n)_n$ using the fact that the closed linear span is dense in H. Because $H=\overline{Sp}(v_n)_n\oplus(\overline{Sp}(v_n)_n)^{\bot}$ and how both subspaces are invariant, it's only remains to show that $$B=A|_{(\overline{Sp}(v_n)_n)^{\bot}}=0.$$ Assume that $B\neq 0$, then there exists a non-zero number $\lambda$ and an eigenvector $v\in(\overline{Sp}(v_n)_n)^{\bot}$ so that $Bv=\lambda v$ and $||v||=1.$ Then $Av=\lambda v$ so $\lambda=\lambda_{n_0},$ for a certain $n_0.$ Therefore $v\in Ker(\lambda_{n_0}I-A)\cap(\overline{Sp}(v_n)_n)^{\bot}=\{0\}$, contradiction.

$\endgroup$
  • $\begingroup$ Is there a specific line at which you get confused? $\endgroup$ – Mauve Apr 20 '18 at 19:15
  • $\begingroup$ I would say for now the first 4 rows of the proof. $\endgroup$ – Andrew Apr 20 '18 at 19:23
0
$\begingroup$

Because $\{v_n\}$ is an orthonormal basis of the closure of its span, if the image of $A$ is contained in this closure then for any $x$ $$ Ax=A\left(\sum_n\langle x,v_n\rangle\,v_n\right)=\sum_n \langle x,v_n\rangle\,Av_n =\sum_n\lambda_n\langle x,v_n\rangle\,v_n $$ (we are using that $\langle Ax,y\rangle=0$ if $y\in\overline{\operatorname{span}}\{v_n:\ n\}^\perp$).

Now, because $A$ is selfadjoint and ${\operatorname{span}}\{v_n:\ n\}$ is invariant for $A$, one can easily show that $\overline{\operatorname{span}}\{v_n:\ n\}^\perp$ is also invariant for $A$.

$\endgroup$

Your Answer

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

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