2

What are the most general conditions for which the commutation relations $[A,BC] = 0$ and $[AB,C] = 0$ equivalent? Alternatively, can it at least be shown to hold if $A$, $B$, and $C$ are all involutant? If $A$, $B$, and $C$ are involutant permutation matrices? I am particularly interested in cases for which none of the commutators $[A,B]$, $[B,C]$, and $[C,A]$ are zero.

user773458
  • 1,353
  • 1
    Your question needs context: what are $A,B,C$? (matrices, it seems, near the end of your post. Only permutation matrices or general (but square?) matrices?) What does "involutant" mean? And last but not least: what were your attempts to solve your problem(s!)? It also needs more focus. – Anne Bauval Feb 23 '24 at 18:46
  • Interesting. I am looking for the answer for any $A$, $B$, and $C$ that are mathematical objects for which commutators are defined since the objects in which I'm interested could be viewed as members of either a group or a ring, (Those are the only structures for which I know commutators to be defined.) Matrices are one of the possible representations, but not the only one. Those matrices are square, unitary, involutory, orthogonal, and nonnegative. However, I don't know which of those properties are important to the proof. – xphileprof Feb 24 '24 at 20:57
  • I have an exhaustive collection of a certain type of object and of a certain size that can be represented as the product of matrices $A$, $B$, and $C$. Excluding those for which $[A,B]=0$, $[B,C]=0$, or $[C,A]=0$, the vast majority satisfy neither $[A,BC]=0$ nor $[AB,C]=0$. The remainder satisfy both. I am trying to understand why in order to gain insight into this type of object. – xphileprof Feb 24 '24 at 21:17
  • 2
    I tried first to prove the result using only the fact that the matrices are involutory, but got nowhere. Next, I thought this must be a solved problem, so I searched for answers on the web. I then realized that the matrices also represent permutations, so I researched the composition and commutation properties of permutations but did not find those properties in a form that seems to lend itself to this problem. – xphileprof Feb 24 '24 at 21:17
  • 4
    Well, I'm surprised and disappointed. I thought that math.stackexchange was a place to visit for expert guidance on mathematical questions. Instead, it seems to be a place where newcomers can be told that they don't know the rules and be publicly shamed by having their mathematically interesting questions anonymously downvoted. Pity. – xphileprof Feb 27 '24 at 19:52
  • 2
    Late response, but I just read your post on the Meta page and I voted to reopen this question. This question doesn't seem like a typical homework question. It looks pretty high-level enough for people to give a little slack to a newcomer. Such askers of high-level questions already know the basics of taking the initiative and doing research on their own, unlike some undergraduates who just want others to do their work for them without a single effort. Sorry that your first experience on this site wasn't pleasant. – Accelerator Feb 29 '24 at 03:43

1 Answers1

1

For involutant matricies in $M_n$ lets take

$$[A,BC] = ABC - BCA = 0$$

and

$$[AB,C] = ABC - CBA = 0$$

Which holds iff

$$BCA = CBA = ABC$$

If $A,B,C$ are all involutant, this is equivalent to $BC = CB = ABCA$.We also have that $ABCA = BC \iff ABC = BCA$. This can be rewritten as:

$$[A,BC] = 0, [B,C]=0$$

Take for example the pauli matricies. They are involutary but no two pauli matricies commute, therefore it is not the case that for involutary matricies we necessarily have that $[A,BC] = [AB,C]=0$.

What about for involutant matricies coming from a group? We can say more.

Let's take our matricies to lie in the representation of a group $\langle a,b,c \rangle \subseteq G \to M_n$ as involutions $a,b,c \mapsto A,B,C$.

Assume for now that $\langle a \rangle, \langle b \rangle, \langle c \rangle$ are normal. We have that $\langle b \rangle \cap \langle c \rangle = \{1\}$, so $\langle b , c \rangle$ is a commutative subgroup of $G$, since both elements are of order two. Likewise $\langle a, b,c \rangle$ is commutative.

What if $\langle a \rangle$ is not normal? For the following let's just assume that $ \langle b,c \rangle$ is commutative and that $ \langle a \rangle$ is not normal.

We know that $\langle bc \rangle$ is of order two. By orbit-stabilizer, $|G:C_G(bc)|=|Cl(bc)|$, where Cl is the conjugacy class of $bc$. If $a \in C_G(bc)$ then the above commutator identity is satisfied. If $a \notin C_G(bc)$ then $a$ is in $gbcg^{-1}C_g(bc)$ for some $g$, in which case, we use the fact that $gbg^{-1}$ is also involutive, and so $a,gbg^{-1},gcg^{-1}$ satisfy your commutator relation.

When $\langle b,c \rangle$ is commutative we have two cases:

(1) $\langle a \rangle$ is normal in $G$, so your commutator identity is satisfied with $[A,B] = [B,C]=0$.

(2) $a, gbg^{-1}, gcg^{-1}$ satisfies your identity for some $g \in G$.

Now you can use the fact that permutation matricies are representations of $S_n \ .$ You can also take $G = \langle a,b,c \rangle$ so that its easier to find those elements $g$.

  • 1
    Thank you! This is a lot (for me) to digest, but that is to be expected. :-) I'm sure I'll have questions as I dig into it, but I really appreciate your response! – xphileprof Feb 29 '24 at 17:22
  • 1
    Where do your matricies come from? Some context of the problem youre working on may help! – Opisthokont Feb 29 '24 at 23:36