The zero divisor conjecture and Mealy automata

The statement in the first item is equivalent to saying that for every row i the map j\mapsto j' defined on \{1,2,\ldots, m\} by the relation c_{ij}=c_{i'j'} is a permutation. This exactly means that the automaton \mathsf{A}_C is invertible. Another reformulation of invertability: for every pair (i,j') there exists a unique pair (i',j) such that c_{ij}=c_{i'j'}. It follows that for every (i',j') there exists a unique (i,j) such that c_{i'j'}=c_{ij}, and for every (i',j) there exists a unique (i,j') with c_{i'j'}=c_{ij}. This means that \mathsf{A}_C is bireversible.

The Helix graph of a bireversible automaton consists of cycles. If \mathsf{A} corresponds to a paring matrix C, then the pairing c_{ij}=c_{i'j'} implies a cycle in the Helix graph \mathcal{H}_\mathsf{A}:

The absence of loops and arrows s\xrightarrow{x|x} t is equivalent to absence of repeated values in rows and columns of C. Conversely, if all cycles have length two, then we can construct a matrix C=(c_{sx}) with rows labeled by the states and columns labeled by the letters such that

The matrix C is a pairing matrix, because (t,y)\neq (s,x) and (t,y) is uniquely determined by (s,x).

The arrow s\xrightarrow{xx\ldots x|yy\ldots y}s in \mathsf{A}_C^{*} means that we have the following sequence of arrows in \mathsf{A}_C:

These arrows correspond to the following relations in \Gamma_C:

Then

Each row and column of C do not contain two equal coefficients, since otherwise a_i=a_j or b_i=b_j for i\neq j. We want to show that by permuting the columns/rows and relabeling equal coefficients, the matrix C can be brought to the following form:

where (c_{2,n+1}, c_{2,n+2}, \ldots, c_{2,2n}) is a permutation of 1,2,\ldots,n, (c_{3,1}, c_{3,2}, \ldots, c_{3,n}) is a permutation of n+1, n+2, \ldots, 2n, and (c_{3,n+1}, c_{3,n+2}, \ldots, c_{3,2n}) is a permutation of 2n+1, 2n+2, \ldots, 3n. This gives certain “bipartite” structure on C.

First, half of the numbers 1,2,\ldots,2n should be in the second row and the other half in the third one, since otherwise there would be two equal elements in one row. By permuting the rows if necessary, we can assume that 1,2,\ldots,n are in the second row.

Second, the numbers 1,2,\ldots,n in the second row cannot belong to columns 1, 2,…, n. Indeed, if c_{2,i}\in\{1,2,\ldots,n\} for 1\leq i\leq n, then the ith column and 3rd row do not contain common coefficient except for their intersection; this contradicts invertibility. Therefore, (c_{2,n+1}, c_{2,n+2}, \ldots, c_{2,2n}) is a permutation of 1,2,\ldots,n. By the same reason, (c_{3,1}, c_{3,2}, \ldots, c_{3,n}) is a permutation of n+1, n+2, \ldots, 2n and (c_{3,n+1}, c_{3,n+2}, \ldots, c_{3,2n}) is a permutation of 2n+1, 2n+2, \ldots, 3n.

Let us denote x_i=b_i and y_i=b_{n+i} for i=1,\ldots,n in order to reflect the bipartite structure of C. Then we get the following relations in \Gamma_C:

where (i_1,i_2,\ldots,i_n), (j_1,j_2,\ldots,j_n), and (k_1,k_2,\ldots,k_n) are permutations of 1,2,\ldots,n. Notice that if we write the corresponding relations for each t_1^{-1}, t_2^{-1}, t_3^{-1}, we will see y_1x_{l_1}^{-1}, y_2x_{l_2}^{-1},\ldots, y_nx_{l_n}^{-1}, where (l_1,l_2,\ldots,l_n) is a permutation of 1,2,\ldots,n.

We want to show that t_1t_2^{-1} has finite order. Let us construct a directed (n,n) bipartite graph: the left part consists of vertices x_1y_{i_1}^{-1}, …, x_ny_{i_n}^{-1}, which represent t_1, the right part consists of vertices y_1x_{l_1}^{-1}, y_2x_{l_2}^{-1},\ldots, y_nx_{l_n}^{-1}, which represent t_2^{-1}. We put a directed edge from x_sy_{i_s}^{-1} to y_{i_s}x_{l_{is}}^{-1} for s=1,\ldots,n, and a directed edge from y_{s}x_{l_s}^{-1} to x_{l_s}y^{-1}_{i_{l_s}} for s=1,\ldots,n. Now there is a directed edge passing from any vertex. Hence, there is a directed cycle of some length 1\leq 2m_{12}\leq 2n. By multiplying consequently the elements along this cycle, we get (t_1t_2^{-1})^{m_{12}}=e.

The same observation holds for t_2t_3^{-1} and t_3t_1^{-1}. Hence

for some 1\leq m_{ij}\leq n. If t_i\neq t_j for some i\neq j, we get a nontrivial element t_it_j^{-1} of finite order. If t_1=t_2=t_3, then

and t_1=a_1^{-1}a_2\neq e by assumption. The theorem is proved.