4
$\begingroup$

(First asked in MSE)

The Brauer algebra $B_n(x)$ is an algebra of matchings whose product is described here.

Given $A$ and $B$ two elements of $B_n(x)$, and given an integer $m$, there are in general many solutions $C\in B_n(x)$ to the equation $AC=x^m B$, where the left hand side is the Brauer product of $A$ and $C$.

I would like to know of sources that have a discussion of this problem. Can the set of such solutions be described?

$\endgroup$
7
  • $\begingroup$ Are you interested where $A$, $B$, and $C$ are basis elements corresponding to diagrams? Or when they are any elements of $B_n(x)$? $\endgroup$ Oct 13, 2019 at 3:15
  • $\begingroup$ @JohnMachacek Any elements $\endgroup$
    – thedude
    Oct 13, 2019 at 13:10
  • $\begingroup$ I thought a bit about it. Seems it can be worked out for diagrams, but I don’t know how to say anything about arbitrary elements. Unfortunately I don’t know any sources on this question. $\endgroup$ Oct 16, 2019 at 16:14
  • $\begingroup$ Every element is a linear combination of diagrams. I think I can work out some answer for equations in just diagrams. But for linear combinations I don’t know what to say. $\endgroup$ Oct 16, 2019 at 17:03
  • $\begingroup$ I tired writing down the details. It turns out I cannot classify all solutions nicely. I can construct a family of solutions, but I don’t have a way to get them all. $\endgroup$ Oct 17, 2019 at 1:17

1 Answer 1

1
+50
$\begingroup$

This answer does not give a full answer to the question. I started working out what I could; so, I'll post it in case it is of use to anyone.

Let $[n] = \{1,2,\dots, n\}$ and $[n'] = \{1',2', \dots, n'\}$. We will use the convention that $i'' = i$. A diagram $D$ is a perfect matching of the the complete graph on the vertex set $[n] \cup [n']$. We will think of any $D$ as $(S_D,T_D,f_D,M_D,M'_D)$. Here $S_D \subseteq [n]$, $T_D \subseteq [n']$, and $f:S_D \to T_D$ is a bijection while $M_D$ and $M'_D$ are perfect matchings on $[n] \setminus S_D$ and $[n'] \setminus T_D$ respectively. So, $D$ is the prefect matching $\{\{i,f_D(i)\} : i \in S_A\} \cup M_A \cup M'_A$ which partitions $D$ is edges which "cross" and edges which live entirely in $[n]$ or $[n']$.

Let us take two diagrams $A$ and $B$. We will assume that $M_A \subseteq M_B$. Otherwise there is no solution since $M_A \subseteq M_{AC}$ for any $C$. Here by abuse of notation we let $AC$ denote the both the product in the Brauer algebra and the diagram which is the product in the Brauer monoid. Now let us try to find $C$ such that $AC = x^m B$.

First, for all $\{i,j\} \in M_B \setminus M_A$ put $\{f_A(i)', f_A(j)'\}$ in $M_C$. This ensures that $\{i,j\} \in M_{AC}$.

Since $M_A \subseteq M_B$ we have $S_B \subseteq S_A$. For each $i \in S_B$ we set $f_C(f_A(i)') = f_B(i)$ while adding $f_A(i)'$ and $f_B(i)$ to $T_A$ and $T_C$ respectively. This will ensure that $f_{AC} = f_B$.

Now let $r = |M_A| = |M'_A|$ and $s = |M_B| = |M'_B|$. We have that $s \geq r$. Also, we must have $r \geq m$ to be able to obtain a solution. At this point in our construction of $C$ any unmatched vertices in $[n]$ will be glued to vertices in $M'_A$ when performing the Brauer product. Also all unmatched vertices in $[n']$ are matched in $M'_B$.

Choose $m$ edges from $M'_A$ and for each $\{i',j'\}$ chosen we put $\{i,j\}$ in $M_C$. This will give us the factor of $x^m$.

Next inject the remaining $r-m$ edges in $M'_A$ into $M'_B$. If $\{i',j'\}$ maps to $\{k',l'\}$ then set either $f_C(i) = k'$ and $f_C(j) = l'$ or else $f_C(i) = l'$ and $f_C(j) = k'$. Finally add the edges of $M'_B$ not covered by this injection directly into $M'_C$. This all results with $M'_B = M'_C$.

This construction gives $2^{r-m} \binom{r}{m} (s)_{r-m}$ solutions $C$ where $AC = x^mB$. Here $(s)_{r-m} = s(s-1) \cdots (s-r-m-1)$ is the falling factorial. In the case $r=m$ we get a single solution which should be the unique solution in that case. Though in general this doesn't cover all solutions.

$\endgroup$

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.