8
$\begingroup$

Let $L$ be a (non-restricted) Lie algebra over a field of prime characteristic $p,$ $UL$ be its universal enveloping algebra and $a_1,\dots, a_p \in L$ (the number of elements is equal to the characteristic). I can prove that the element $\sum_{\sigma \in S_p} a_{\sigma(1)}\cdot{}\dots{}\cdot a_{\sigma(p)}$ of $UL$ lies in $L.$ But I want to have an evident formula for this element on the language of commutators.

Question: Is it true that $$\sum_{\sigma \in S_p} a_{\sigma(1)}\cdot{}\dots{}\cdot a_{\sigma(p)} = \sum_{\sigma\in S_p\ :\ \sigma(1)=1} [a_{1},a_{\sigma(2)},\dots,a_{\sigma(p)}]$$ ?

Here $S_p$ denotes the symmetric group and $[x_1,\dots,x_n]=[[x_1,\dots,x_{n-1}],x_n].$ I can prove this formula for $p=2,3.$


Answer: Yes, it is true. This formula can be found here: http://msp.org/agt/2006/6-5/agt-v6-n5-p08-p.pdf on page 2252, line 6 (without prove). Jie Wu explained me a prove.

$\endgroup$
3
  • 1
    $\begingroup$ Also works for 5 and 7. $\endgroup$
    – F. C.
    May 28, 2015 at 7:43
  • $\begingroup$ It seems it is enough to prove that the right-hand element is invariant under the action of $S_p$. $\endgroup$ May 28, 2015 at 8:26
  • $\begingroup$ Yes, it is true. Jie Wu explained me this. $\endgroup$ May 28, 2015 at 17:34

1 Answer 1

5
$\begingroup$

If you expand the commutator by the possible $2^{n-1}$ sign choices you get $$[a_1,a_2,\dots,a_n]=\sum_{r=1}^{n-1}(-1)^r\sum_{k_1> k_2>\cdots> k_r>1}a_{k_1}a_{k_2}\cdots a_{k_r}W(k_1,k_2,\dots,k_r)$$ where the expression $W(k_1,k_2,\dots,k_r)$ stands for the word $a_1a_2\cdots a_n$ with the letters $a_{k_i}$ removed. Now this tells us that when you expand the sum $$\sum_{\sigma\in S_p, \sigma(1)=1}[a_{\sigma(1)},a_{\sigma(2)},\dots,a_{\sigma(p)}],$$ a word where $a_1$ is in the $r$-th position appears with coefficient $(-1)^r\binom{p-1}{r}$. Wison's theorem tells you that $(-1)^r\binom{p-1}{r}\equiv 1\pmod{p}$ and your equality follows.

$\endgroup$
2
  • $\begingroup$ Yes, it is very nice! Little correction: "..where a_1 is in the r+1-th position..". $\endgroup$ May 29, 2015 at 13:37
  • $\begingroup$ Superbe solution ! (I tried yesterday - unsuccessfully - with the descent classes) $\endgroup$ May 30, 2015 at 5:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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