Note: Please do not give a solution; I would prefer guidance to help me complete the question myself. Thank you.

Let $\alpha_{0},\ldots,\alpha_{n}\in\mathbb{C}$ be given. Compute the spectrum of $T:\ell^{1}\rightarrow\ell^{1}$ with \begin{align} (Tx)_{i}=\sum_{j=0}^{n}\alpha_{j}x_{i+j}\text{ for all }i\geq 1. \end{align}

My Solution

I am only to find a bound on the spectrum. Let us look at the $\ell^{1}$ norm of the image of $T$. \begin{align} |Tx|_{1}&=\sum_{i=1}^{\infty}\bigg|\sum_{j=0}^{\infty}\alpha_{j}x_{i+j}\bigg|\\ &\leq \sum_{i=1}^{\infty}\sum_{j=0}^{\infty}|\alpha_{j}||x_{i+j}|\\ &= \sum_{i=1}^{\infty}(|\alpha_{0}||x_{i+0}|+|\alpha_{1}||x_{i+1}|+\ldots+|\alpha_{n}||x_{i+n}|)\\ &= |\alpha_{0}|\sum_{i=1}^{\infty}|x_{i}|+|\alpha_{1}|\sum_{i=1}^{\infty}|x_{i+1}|+\ldots+|\alpha_{n}|\sum_{i=1}^{\infty}|x_{i+n}|\\ &=|\alpha_{0}||x|_{1}+|\alpha_{1}||Lx|_{1}+|\alpha_{2}||L^{2}x|_{1}+\ldots+|\alpha_{n}||L^{n}x|_{1}\\ &\leq \sum_{j=0}^{n}|\alpha_{j}|, \end{align} where $L:\ell^{1}\rightarrow\ell^{1}$ is the left shift operator. By picking the sequence $(L^{n-1}x)_{n\geq 1}$ this shows that $\|T\|=\sum_{j=0}^{\infty}|\alpha_{j}|$. So $\text{spr}(T)\leq \sum_{j=0}^{\infty}|\alpha_{j}|$, where $\text{spr}(T)$ denotes the spectral radius of $T$.

Can someone give me some guidance on where to progress to from here? Please no answers, just hints.

Completion of the proof

Observe by expanding the series you can rewrite $Tx$ as the polynomial of some other operator. \begin{align} Tx &= (\sum_{j=0}^{n}\alpha_{j}x_{1+j},\sum_{j=0}^{n}\alpha_{j}x_{2+j},\sum_{j=0}^{n}\alpha_{j}x_{3+j},\ldots)\\ &= (\alpha_{0}x_{1}+\alpha_{1}x_{2}+\ldots+\alpha_{n}x_{n+1},\\ &\hspace{22.5mm} \alpha_{0}x_{2}+\alpha_{1}x_{3}+\ldots+\alpha_{n}x_{n+2},\\ &\hspace{37.5mm} \alpha_{0}x_{3}+\alpha_{1}x_{4}+\ldots+\alpha_{n}x_{n+3},\ldots)\\ &=(\alpha_{0}x_{1},\alpha_{0}x_{2},\ldots,\ldots) + (\alpha_{1}x_{2},\alpha_{1}x_{3},\ldots) + \cdots +(\alpha_{n}x_{n+1},\alpha_{n}x_{n+2},\ldots)\\ &=\alpha_{0}x + \alpha_{1}Lx + \alpha_{2}L^{2}x + \cdots +\alpha_{n}L^{n}x\\ &=(\alpha_{0}+\alpha_{1}L+\alpha_{2}L^{2}+\cdots+\alpha_{n}L^{n})x. \end{align}

Therefore the spectrum of $T$ is equal to the polynomial of the spectrum by the spectral mapping theorem. That is, \begin{align} \sigma(T)&=\sigma(\alpha_{0}+\alpha_{1}L+\alpha_{2}L^{2}+\cdots+\alpha_{n}L^{2})\\ &=\alpha_{0}\sigma(I)+\alpha_{1}\sigma(L)+\alpha_{2}\sigma(L^{2})+\cdots+\alpha_{n}\sigma(L^{n})\\ &= \alpha_{0}\sigma(I)+\alpha_{1}\sigma(L)+\alpha_{2}\sigma(L)^{2}+\cdots+\alpha_{n}\sigma(L)^{n}. \end{align}

Now, $\sigma(L)=\{\lambda\in\mathbb{C}||\lambda|\leq 1\}$, from a previous question I completed.Therefore, \begin{align} \sigma(T) = \{\alpha_{0}+\alpha_{1}\lambda +\alpha_{2}\lambda^{2}+\cdots+\alpha_{n}\lambda^{n}\in\mathbb{C}||\lambda|\leq 1\}. \end{align} We know $\text{spr}(T)\leq\sum_{j=0}^{n}|\alpha_{j}|$ and, \begin{align} \sup_{\lambda\in\sigma(T)}|\alpha_{0}+\alpha_{1}\lambda+\alpha_{2}\lambda^{2}+\cdots+\alpha_{n}\lambda^{n}|\leq|\alpha_{0}+\alpha_{1}+\alpha_{2}+\cdots+\alpha_{n}|\leq\sum_{j=0}^{n}|\alpha_{j}|, \end{align} hence we have found the entire spectrum.


I don't think there is an elementary way of doing this. The key observation is that $T=p(S)$, where $p$ is a polynomial and $S$ is the unilateral shift $$ Sx=(0,x_1,x_2,\ldots). $$ So the Spectral Mapping Theorem reduces the problem to finding the spectrum of $S$. This is still non-obvious, as far as I can tell, but one can consider the adjoint $S^*:\ell^\infty\to\ell^\infty$. The spectrum of $S^*$ is easy to calculate, as it is almost entirely made up of eigenvalues. Then using $\sigma(S)=\sigma(S^*)$ (another non-trivial fact for operators on Banach spaces as opposed to Hilbert spaces) we are done.

  • $\begingroup$ Yes. I have completed the problem now. It does require the spectral mapping theorem. $\endgroup$ – Zeta-Squared May 2 '19 at 8:06

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.