Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I ran into the following question about bounded operators on Hilbert spaces; I could really use some help.

It goes like this: Suppose that $\left\{T_{k}\right\}$ is a collection of bounded operators on a Hilbert space $\mathcal{H}$, with $||T_{k}|| \leq 1$ for all $k$. Suppose also that $T_{k}T_{j}^{*}=T_{k}^{*}T_{j}=0$ for all $k \neq j$. Let $S_{N}=\sum_{k=-N}^{N}{T_{k}}$. Then, we have that $S_{N}(f)$ converges as $N \rightarrow \infty$, for every $f \in \mathcal{H}$. Moreover, if $T(f)$ denotes the limit, then $||T|| \leq 1$.

What if the condition $T_{k}T_{j}^{*}=T_{k}^{*}T_{j}=0$ is replaced by $||T_{k}T_{j}^{*}|| \leq a_{k-j}^{2}$ and $||T_{k}^{*}T_{j}|| \leq a_{k-j}^{2}$ for positive constants $a_{n}$ with the property $\sum_{-\infty}^{\infty}{a_{n}}=A < \infty$? Is is true that the convergence preserves and $||T|| \leq A$?

share|cite|improve this question
Hint: let $P_j$ be the self-adjoint projection on $\text{Ran}(T_j)$. – Robert Israel Apr 16 '12 at 7:28
More details please? :) – Anna Apr 16 '12 at 7:59
More hints: $\text{Ran}(T_j)$ are orthogonal to each other, and $S_N(P_j f) = T_j P_j f$ if $|j| \le N$. – Robert Israel Apr 16 '12 at 18:23
Can anyone provide a proof with more details? I'm not sure I understand after all.. – Anna Apr 19 '12 at 2:33
Oops, that wasn't quite right. $P_j$ should be the projection on $\overline{\text{Ran}(T_j^*)} = (\text{Ker}(T_j))^\perp$. I'll write out an answer. – Robert Israel Apr 19 '12 at 5:09
up vote 2 down vote accepted

In the first version, let $P_j$ be the orthogonal projection of $\cal H$ on $\overline{\text{Ran}(T_j^*)} = (\text{Ker}(T_j))^\perp$. Note that these are orthogonal to each other since $T_i T_j^* = 0$ for $i \ne j$: we have $P_i P_j = T_j P_i = 0$, while $T_i P_i = T_i$. Thus for any $f \in \cal H$ and any $N$, we can write $f = g_N + \sum_{j=-N}^N P_j f$ where $g_N = (I - \sum_{j=-N}^N P_j) f$ and $\|f\|^2 = \|g_N\|^2 + \sum_{j=-N}^N \|P_j f\|^2$. In particular, $\sum_{j=-\infty}^\infty \|P_j f\|^2 \le \|f\|^2$.

Now $T_i f = T_i P_i f$ so $S_N f = \sum_{i=-N}^N T_i P_i f$ with

$$\eqalign{\|S_N f\|^2 &= \sum_{i=-N}^N \sum_{j=-N}^N \langle P_i f, T_i^* T_j P_j f \rangle = \sum_{i=-N}^N \langle P_i f, T_i^* T_i P_i f \rangle\cr & = \sum_{i=-N}^N \|T_i P_i f\|^2 \le \sum_{i=-N}^N \|P_i f\|^2 \le \|f\|^2\cr}$$

Similarly, for $M > N$, $\|S_M f - S_N f\|^2 \le \sum_{N < |i|\le M} \|P_i f\|^2$. For any $f$ and any $\epsilon > 0$, we can take $K$ large enough that $\sum_{|i|>K} \|P_i f\|^2 < \epsilon$, and so $\|S_M f - S_N f\|^2 < \epsilon$ for all $M, N > K$. Thus $\lim_N S_N f$ exists. Of course the limit is linear in $f$, so $T$ exists, and our bound on $\|S_N f\|^2$ shows that $\|T\| \le 1$.

I'm not sure what happens in the generalized version.

share|cite|improve this answer
Very nice. Thanks! – Anna Apr 23 '12 at 18:41

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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