I'm trying to prove that the norm of the multilinear symmetric operator $A$ is $\frac{1}{m!}$ where $A$ is defined as:

$$ A(x_1,\dots, x_m) = \frac{1}{m!} \sum_{\sigma \in S_m} \xi_1(x_{\sigma(1)} ) \dots \xi_m(x_{\sigma(m)} )$$

$\xi_i$ are the coordinate functionals ($\xi_i(e_j)=\delta_{ij}$) and $x_i \in l^1$. We consider the following norm on $(l^1)^m$, $\| (x_1, \dots, x_m) \| = max_j \|x_j\|$

Clearly, considering $x_i = e_i ~ \forall i$ I get the inequality $\| A \| \ge \frac{1}{m!}$ but I haven't been able to get the other inequality so far.

How should I proceed?

  • $\begingroup$ Just to be sure I understand: is $A:l^1\to \mathbb R$ (hence, the notation $A(x_1,\cdots,x_m)$ stands for $A((x_i\,:\,i\in\mathbb N))$ )? Or is $A:(l^1)^m\to \mathbb R$ ? $\endgroup$ – user228113 Dec 15 '15 at 13:51
  • 2
    $\begingroup$ @G.Sassatelli The second one, $A: (l^1)^m \to \mathbb{R}$. $\endgroup$ – A. A. Dec 15 '15 at 13:52
  • 1
    $\begingroup$ With what norm, then? There are several norms inducing the product topology, and $\|A\|$ depends on which one you pick. $\endgroup$ – user228113 Dec 15 '15 at 13:52
  • $\begingroup$ @G.Sassatelli Sorry I forgot to specify it. I've just added to the post. $\endgroup$ – A. A. Dec 15 '15 at 14:39

Aside: It's not a linear operator (for $m > 1$), but a multilinear (and symmetric) operator.

By homogeneity, we can assume that $\lVert x_i\rVert \leqslant 1$ for all $i$. Then split the sum,

\begin{align} \sum_{\sigma \in S_m} \xi_1(x_{\sigma(1)})\cdot \dotsc\cdot \xi_m(x_{\sigma(m)}) &= \sum_{k = 1}^m \Biggl(\sum_{\substack{\sigma \in S_m \\ \sigma^{-1}(m) = k}} \xi_1(x_{\sigma(1)})\cdot \dotsc\cdot \xi_m(x_{\sigma(m)})\Biggr)\\ &= \sum_{k = 1}^m \xi_k(x_m)\cdot\Biggl(\sum_{\substack{\sigma \in S_m \\ \sigma^{-1}(m) = k}}\prod_{\substack{i = 1 \\ i \neq k}}^m \xi_i(x_{\sigma(i)})\Biggr) \end{align}

to set up an induction proof.

The base case $m = 1$ is immediately verified, and then by the induction hypothesis, the sums in parentheses are all bounded in absolute value by $1$, which yields the induction step. (There's a small explanation needed why the sums in the parentheses are instances of the $m-1$ case.)

  • $\begingroup$ I've edited the post to specify the multilinearity of the operator. I'll write down your answer and fill the little details by myself to understand it. Thanks. $\endgroup$ – A. A. Dec 15 '15 at 15:21

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.