Suppose we have a separable Hilbert space (thus with a countable basis) and that we to represent an operator in matrix form, i.e:

$$A: H \rightarrow H \\ \; \; \; \; \; \;x \;\rightarrow \sum_{j \in \mathbb{N}}\left(\sum_{k \in \mathbb{N}} a(j,k)\cdot<x,e_k> \right)e_j$$

Given that the series of complex numbers $\sum_{k \in \mathbb{N}} a(j,k)\cdot<x,e_k>$ converges and that $\sum_{j \in \mathbb{N}}\left(\sum_{k \in \mathbb{N}} a(j,k)\cdot<x,e_k> \right)e_j$ converges in $H$ how do I prove without further assumptions that $A$ is a bounded operator?

I can prove that for each $j \in \mathbb{N}$ the sucession $(a(j,k))_k \in l^2 (\mathbb{N},\mathbb{C})$ and that for each $k \in \mathbb{N}$ the sucession $(a(j,k))_j \in l^2 (\mathbb{N},\mathbb{C})$.

However I always get to a point where I require that $\sum_{j \in \mathbb{N}}\sum_{k \in \mathbb{N}} \left |a(j,k) \right|^2 < \infty$ which is a condition that I can show to be necessary and sufficient for $A$ to be a Hilbert-Schmidt operator. Any help would be welcome.

  • $\begingroup$ Yes, I meant a bounded operator. I edited it, thanks. $\endgroup$ – grizzlyjoker Apr 11 '13 at 19:52

As pointed out by Elias, the Uniform Boundedness Principle (UBP) is the key. But the assumption does not imply that $A$ be Hilbert-Schmidt, as it is satified by every bounded linear operator $A$ on $H$. Indeed, denoting by $P_n$ the finite rank projection onto $\mbox{span}(e_0,\ldots,e_n)$, we have that $P_n$ tends to the identity in the strong operator topology (i.e. pointwise). So for every $n$, $P_nAP_mx$ tends to $P_nAx$ as $m$ tends to $+\infty$, which proves that the coordinates of $Ax$ in the basis are given by the converging series $\sum_{k=0}^{+\infty} A_{jk}\langle x,e_k\rangle$ where $A_{jk}=\langle Ae_k,e_j\rangle$ are the matrix coordinates of $A$ in the basis. And then $P_nAx$ tends to $Ax$ as $n$ tends to $+\infty$, which proves the second part, namely that $\sum_{j=0}^{+\infty}\left(\sum_{k=0}^{+\infty}A_{jk}\langle x,e_k\rangle\right)e_j$ converges to $Ax$ in $H$.

Conversely, which is what you are asking about, assume that $(A_{jk})$ is a doubly infinite sequence such that $\sum_{j=0}^{+\infty}\left(\sum_{k=0}^{+\infty}A_{jk}\langle x,e_k\rangle\right)e_j$ has converging coordinates, and converges in $H$ for every $x\in H$. We will show altogether that this defines a bounded linear operator on $H$. For every $n,m$, consider the bounded linear operator $A^{(n,m)}$ defined by $A^{(n,m)}x:=\sum_{j=0}^n\left(\sum_{k=0}^mA_{jk}\langle x,e_k\rangle\right)e_j$. By the first part of the assumption, we see that $A^{(n,m)}$ converges to $A^{(n)}$ pointwise as $m$ tends to $+\infty$, where $A^{(n)}x:=\sum_{j=0}^n\left(\sum_{k=0}^{+\infty}A_{jk}\langle x,e_k\rangle\right)e_j$. By the UBP, it follows that $A^{(n)}$ is a bounded linear operator on $H$. Finally, note that the second part of the assumption yields that $A^{(n)}$ converges pointwise to $A$ given by $Ax:=\sum_{j=0}^{+\infty}\left(\sum_{k=0}^{+\infty}A_{jk}\langle x,e_k\rangle\right)e_j$ as $n$ tends to $+\infty$. By the UBP again, we can conclude that $A$ is a bounded linear operator on $H$.


Hint Use the Uniform boundedness principle or better yet just do $x=\sum_{v=1}^{\infty}\sum_{u=1}^{\infty}a(u,v)e_u$ in $\sum_{j \in \mathbb{N}}\left(\sum_{k \in \mathbb{N}} a(j,k)\cdot<x,e_k> \right)e_j<\infty$ for proving $\sum_{j \in \mathbb{N}}\sum_{k \in \mathbb{N}} \left |a(j,k) \right|^2 < \infty$.

  • 2
    $\begingroup$ The series $\sum_{u=1}^{+\infty}e_u$ does not define a vector of $H$. $\endgroup$ – Julien Apr 11 '13 at 20:25
  • $\begingroup$ Hmm, we didn't learn the uniform boundedness principle so far. About the second aproach: how to we know that this particular x converges? If it does that would seem to prove that every bounded operator is Hilbert-Schmidt operator $\endgroup$ – grizzlyjoker Apr 11 '13 at 20:28
  • $\begingroup$ Hmm, i got it by another means: using the Jensen inequality and using some results about $l^2$ sequences. Thanks for the help! $\endgroup$ – grizzlyjoker Apr 12 '13 at 9:03
  • $\begingroup$ @julien I corrected my little mistake in my answer. $\endgroup$ – MathOverview Apr 12 '13 at 11:29
  • $\begingroup$ Hmm nvm, it doesn't work: i can prove that $\sum_{k \in \mathbb{N}} \left |a(j,k) \right |$ converges but i also need (in order for my argument to work) that $\sum_{j \in \mathbb{N}} \left |a(j,k) \right |$ which I cannot prove $\endgroup$ – grizzlyjoker Apr 12 '13 at 21:38

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.