0

This is a linear algebra problem regarding a sufficient condition for the determinant of an odd-order square matrix to be even. As said above, if all elements along the main diagonal are even, and the sum of any pair of symmetric elements is even, how can we prove that the determinant is even? I've learned the methods of calculating determinants and the properties of matrices, but I struggle to prove it. I tried induction along with Laplace expansion, but it doesn't seem to help. Or regarding the definition of determinant, I thought the key point might be a product is even as long as one factor is even, but I find it challeging as the statement only provides that the sum of symmetric entries is even. Can anyone help? Thanks.

1 Answers1

2

$\det(A)$ is a polynomial over the entries of $A$. Thus, if $A \equiv B \pmod 2$, then it follows that $\det(A) \equiv \det(B) \pmod 2$. So, let $B$ be the matrix obtained by replacing each entry $a_{i,j}$ with $0$ if $a_{i,j}$ is even, with $1$ if $a_{i,j}$ is odd and $i < j$, and with $-1$ if $a_{i,j}$ is odd and $i > j$.

Note that $B$ is an antisymmetric matrix with an odd size. It follows that $\det(B) = 0$. Thus, we have $$ \det(A) \equiv \det(B) = 0 \pmod 2. $$ So, $\det(A)$ is equal to $0$ modulo $2$, which is to say that $\det(A)$ is even, which is what we wanted.

Ben Grossmann
  • 234,171
  • 12
  • 184
  • 355
  • Thanks for rescuing my botched attempt! – TonyK Jan 08 '24 at 20:34
  • @TonyK It was the right general idea for sure – Ben Grossmann Jan 08 '24 at 20:38
  • Thanks! I just worked on a problem related to anti-symmetric matrix of odd order, but didn't link them together. – Alex_123 Jan 08 '24 at 20:44
  • @DanielV I am not merely saying that $B$ is an antisymmetric matrix modulo 2, I am saying that $B$ is an antisymmetric matrix with integer entries (rather than with entries in $\Bbb F_2$). Unlike $B$, $I_3$ is not an antisymmetric matrix with integer entries, though it would be antisymmetric over $\Bbb F_2$. And no, stating that antisymmetric matrices of odd sizes (with entries in a field of characteristic 0) have a 0 determinant is not "restating the question" in any reasonable sense – Ben Grossmann Jan 09 '24 at 03:44