0
$\begingroup$

Given an $n \times n$ matrix $(c_{ij})$ with entries in $\mathbb{R}$ and such that $c_{ij} \leq B$ for some $B > 0$, then it is obvious that the determinant, which we call $\Delta$, is at most $B^n$. However it is frequently the case that this bound is not reached.

My question is, can there be an substantial improvements over the estimate $|\Delta| \leq B^n$? It seems that the upper bound is not reachable. Is there an example of a matrix where the upper bound is close to being obtained?

Thank you for any insights.

$\endgroup$
2

1 Answer 1

8
$\begingroup$

Presumably you mean to say that $|c_{ij}|\le B$, since if you just say $c_{ij}<B$, then the $c_{ij}$ could be very negative.

But even then your bound $B^n$ is false. Here's a counterexample: $B=\begin{pmatrix} 2&1\\-1&2\\\end{pmatrix}$, then entries are bounded by $2$, but $\Delta=5>2^2$. The most trivial estimate is that $\Delta$ is a sum of $n!$ terms, each of which is at most $B^n$, so the trivial bound is $n!B^n$. One can, of course, do much better. Hadamard's inequality gives $|\Delta|\le n^{n/2}B^n$.

$\endgroup$
1
  • $\begingroup$ Furthermore, for many n which are multiples of 4, the bound is sharp. I don't know how far away from Omega((BBn)^(n/2)) reality is in the maximal case using only real coefficients, but that is the belief backed up by a lot of evidence. Gerhard "Ask Me About Binary Matrices" Paseman, 2012.04.18 $\endgroup$ Apr 18, 2012 at 23:17

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.