8

Let $p$ be an odd prime, and $\mathbb{Z}_p$ be the finite field of size $p$, and let $\mathbb{Z}_p^{2 \times 2}$ to be the set of all $2 \times 2$ matrices with entries from $\mathbb{Z}_p$.

Denote by $[k] = \{ 1, 2, \ldots, k \}$ (for any positive integer $k$). For functions $f, g : [n] \to [n]$, denote $f \sim g$ if there exists $k \in [n]$ such that $f(i) = g((i + k) \pmod{n} + 1)$ for all $i \in [n]$. Another equivalent definition is that there exists $k$ where $f \circ \sigma^k = g$ where $\sigma$ is the permutation $(1 \, 2\, 3 \dots n)$.

Suppose $f, g$ are functions where $f \sim g$ doesn't hold. Show that for any sufficiently large prime $p$, there exists $A_1, A_2, \dots, A_n \in \mathbb{Z}_p^{2 \times 2}$ such that

$$\operatorname{Tr}(A_{f(1)} A_{f(2)} \dots A_{f(n)}) \ne \operatorname{Tr}(A_{g(1)} A_{g(2)} \dots A_{g(n)}).$$

I have been thinking about this problem for a while, and I found a solution here for the case when $f$ and $g$ are permutations. The general case seems to be tricky.

What I tried is that we can define $A_i = \begin{pmatrix}x_{i, 1, 1} & x_{i, 1, 2} \\ x_{i, 2, 1} & x_{i, 2, 2} \end{pmatrix}$ in terms of indeterminites, and let $P(\mathbf{x}) = \operatorname{Tr}(A_{f(1)} A_{f(2)} \dots A_{f(n)}) - \operatorname{Tr}(A_{g(1)} A_{g(2)} \dots A_{g(n)})$ and show that it is not identically zero.

One easy observation is that if the multisets $\{ f(1), f(2), \dots, f(n) \} \ne \{ g(1), g(2), \dots, g(n) \}$, then you can consider an element $k$ that appears in, WLOG, the first more than the second. Then, you can argue that $\deg_{x_{k, 1, 1}}\operatorname{Tr}(A_{f(1)} A_{f(2)} \dots A_{f(n)}) > \deg_{x_{k, 1, 1}}\operatorname{Tr}(A_{g(1)} A_{g(2)} \dots A_{g(n)})$.

I can't seem to figure out how to show what happens when the multisets are equal. Any help is appreciated.

Update: It seems that from here, the theorem turned out to be incorrect. The counterexample is an exhibit of functions $f = (1, 1, 2, 1, 2, 2), g = (1, 1, 2, 2, 1, 2)$.

I would love to see a way to characterize good functions $f$ and $g$ (find necessary and sufficient conditions for $f$ and $g$) that make this theorem true. One sufficient condition is that $f$ and $g$ are permutations/bijections as pointed out before.

  • I do think that it is true if you allow $m\times m$ matrices, for $m$ large enough – Enguerrand Moulinier Apr 19 '25 at 09:09
  • If you're talking about the block-quoted conjecture then the post already mentioned a counterexample. – Anas A. Ibrahim Apr 19 '25 at 10:11
  • Its is fals for $m=2$, but I do think it becomes true if you allow large enough matrixes (not large nough primes). I have no clue of a proof, but I think it would make sens when looking at what is saying in termes of polynomial equations – Enguerrand Moulinier Apr 19 '25 at 10:16
  • I think the same counterexample listed in the post already works for any $m$ – Anas A. Ibrahim Apr 19 '25 at 10:24
  • You mean the counter example with $f=(1,1,2,1,2,2),g=(1,1,2,2,1,2)$ ? It works only for $m=2$. We don't have $tr((AB-BA)^3)=0$ in general when $m\neq 2$. Indeed matrices of the form $AB-BA$ are exactly traceless matrices. For general $m$, if $tr(N)=0$, we don't have $tr(N^3)=0$. This is true when $m=2$ because of cayley hamilton – Enguerrand Moulinier Apr 19 '25 at 10:37
  • Sure, I'd love a proof for $m > 2$ and necessary and sufficient conditions for $m = 2$. Seems like an interesting question in general. I do feel it has something to do with Invariant Theory and the research done by Procesi and Razmyslov. – Anas A. Ibrahim Apr 21 '25 at 01:11

0 Answers0