the rank in $M_n(\mathbb{Z}_2)$ of an adjacency matrix

286 Views Asked by At

This is a replay of a problem recently posed by @Algebra-Math-Arash.

I was the only one to post an answer, and although my posted proof had a few gaps (helpfully identified by @Andreas Blass), it was, I think, basically correct.

However the OP showed no work, so the question was closed for lack of context, and eventually deleted.

Shown below is a revised (hopefully correct) version of my prior proof.

What I'm looking for in this thread are two things:

  1. Is my proof correct?$\\[8pt]$
  2. Is there an alternative, perhaps simpler approach?

OK, so here's the problem, followed by my proposed proof . . .

Problem:

Show that if $A$ is the adjacency matrix of a simple graph $G$ of order $n$, then the rank of $A$ in $M_n(\mathbb{Z}_2)$ is even.

Proof:

Proceed by induction on $n$.

If $n=1$, then $A=0$, so the rank of $A\;$is zero, hence is even.

Next, let $n > 1$, and assume the claimed result is true for all simple graphs of order $n-1$.

Let $r\;$be the rank of $A\;$in $M_n(\mathbb{Z}_2)$.

Suppose $r\;$is odd.

Our goal is to derive a contradiction.

Claim $r < n$.$\;$Consider two cases . . .

Case $(1)$:$\;n\;$is even.

Then since $r\;$is odd, we must have $r < n$.

Case $(2)$:$\;n\;$is odd.

Temporarily regard $A\;$as an element of $M_n(\mathbb{Z})$, and let $S\;$be the element of $M_n(\mathbb{Z})$ whose entries are the same as $A\;$except negated below the main diagonal. Then $S\;$is skew-symmetric, hence, since $n\;$is odd, it follows that in $M_n(\mathbb{Z})$, we have $\det(S)=0$.

By considering the sum of signed products of generalized diagonals, it follows that $\det(A) \equiv \det(S)\;(\text{mod}\; 2)$, so $\det(A)$ is even.

Returning to $M_n(\mathbb{Z}_2)$, we have $\det(A) = 0$, hence $r < n$.

Thus, in both cases, we have $r < n$.

Since $r < n$, there is some vertex $v$ of $G$ such that the row of $A$ corresponding to $v$, row $i$ say, is dependent on the other rows of $A$. Then, since $A$ is symmetric, column $i$ of $A$ is dependent on the other columns of $A$.

Deleting row $i$ from $A$ yields an $(n{-}1){\times}n$ matrix, $B_1$ say.

Since row $i$ in $A$ is dependent on the other rows of $A$, it follows that $\text{rank}(B_1)=\text{rank}(A)$.

The dependency relation for column $i$ in $A$ still holds for column $i$ in $B_1$, hence if you delete column $i$ from $B_1$, you get an $(n{-}1){\times}(n{-}1)$ matrix, $B_2$ say, such that $\text{rank}(B_2) = \text{rank}(B_1)$.

Thus, $\text{rank}(B_2) = r$.

But $B_2$ is the adjacency matrix for $G-v$, hence by the inductive hypothesis, the rank of $B_2\;$in $M_{n-1}(\mathbb{Z}_2)$ is even, contrary to the assumption that $r$ is odd.

Thus, the induction is complete.

This completes the proof.