MathOverflow is a question and answer site for professional mathematicians. 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

The standard polynomial in $r$ non-commuting indeterminates $x_1,\ldots,x_r$ is defined by $${\mathcal S}_r(x_1,\ldots,x_r):=\sum_{\sigma\in S_r}\epsilon(\sigma)x_{\sigma(1)}x_{\sigma(2)}\cdots x_{\sigma(r)}\,$$ where $S_r$ is the symmetric group in $r$ letters and $\epsilon$ is the signature. Each monomial is a word in the letters $x_j$, affected of a sign $\pm1$. Finally, let us normalize $$T_r:={1\over r!}{\mathcal S}_r.$$

When $r=2$, ${\mathcal S}_2(x_1,x_2)=x_1x_2-x_2x_1$ is the commutator. When using the Frobenius norm $\|M\|=\sqrt{Tr(M^*M)}$ , we know from [1] that $\|[A,B]\|\le\sqrt2\|A\|\cdot\|B\|$, and $\sqrt2$ is the least constant in this inequality, whatever the dimension $n\ge2$.

If $r\ge2$, what is the least constant $c(r,n)$ in the inequality $$\|T_r(A_1,\ldots,A_r)\|\le c(r;n)\prod_{j=1}^r\|A_j\|,\quad\forall A_1,\ldots,A_r\in M_n(\mathbb{C})?$$

Because $n\mapsto c(r,n)$ is non-decreasing, we set $\tau(r)=\lim_{n\rightarrow+\infty}=\sup_n c(r,n)$.

By Amitsur--Levitski Theorem (see [2] for a short proof, or Section 4.4 of the second edition of my book on Matrices), we have $c(r,n)=0$ when $r\ge2n$. From above, we have $c(2,n)=\sqrt2/2$. I can also prove the inequalities $$c(r+s,n)\le c(r,n)\cdot c(s,n),\qquad \tau(r+s)\le\tau(r)\tau(s).$$ In particular, the limit $$\rho:=\lim_{r\rightarrow+\infty}\tau(r)^{1/r}$$ exists and is also the infimum of the sequence. What is its value ?

Finally, is there a quantitative counterpart of Amitsur--Levitski, in the sense that $r\mapsto c(r,n)$ would drop to a small value at some threshold value, less than $2n$ ?

[1] Seak-Weng Vong, Xiao-Qing Jin, Proof of Böttcher and Wenzel’s conjecture. Oper. Matrices, 2 (2008), pp 435--442.

[2] S. Rosset. A new proof of the Amitsur-Levitski identity. Israel J. Math., 23 (1976), pp 187--188.

share|cite|improve this question
Is it really $1/n!$ and not $1/r!$ in the definition of $T_r$? – Darsh Ranjan Sep 14 '10 at 16:23
Of course, it is $1/r!$. Sorry for the typo. I shall correct it. – Denis Serre Sep 14 '10 at 19:15
Isn't $\tau(r)$ equal to the supremum of the $c(r,n)$ as $n$ tends to infinity? – Andreas Thom Oct 5 '10 at 20:22
Is $\tau(3)$ known? – Andreas Thom Oct 5 '10 at 20:23
@Denis: Any updates on this? Also, could you point me out to more literature on standard polynomials applied to matrices? Thanks! – Suvrit Feb 26 '12 at 23:44

P.G. Dixon has studied topologically nilpotent Banach algebras, those Banach algebras $A$ for which $$ \sup\{||x_1x_2... x_n||^{1/n}: x_i \in A,\; ||x_i|| < 1 \;(1 < i < n)\}\to 0 $$ as $n \to \infty$. Possibly his papers: "Topologically nilpotent Banach algebras and factorization", Proc. Roy. Soc. Edin. A, 119 (1991), 329-341 and (with V. Müller) , "A note on topologically nilpotent Banach algebras" Studia Math., 102 (1992), 269-275 would be of interest. The latter can be downloaded here

share|cite|improve this answer
Interesting, although I do not think that Dixon's paper has something to do with my question. Strange yet that Dixon did not cite (did not know ?) the paper A note on the joint spectral radius by G.-C. Rota and G. Strang (1960). – Denis Serre Oct 6 '10 at 13:13

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.