MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given an integer $d$, let $\alpha_d(N)$ denote the number of symmetric integral positive definite matrices of size $d\times d$ with coefficients in $\lbrace -N,-N+1,\dots,N-1,N\rbrace$.

Asymptotically, the number $\alpha_d(N)$ is given by $\alpha_d(N)\sim c_dN^{d+1\choose 2}$ for some strictly positive real number $c_d$ (with $c_d\leq 2^{d+1\choose 2}$).

What is known on the constants $c_2,c_3,\dots$?

Added: $c_d$ is in fact equal to the proportion (with respect to the obvious Lebesgue measure) of $d\times d$ matrices with coefficients in $[-1,1]$ which are positive definite. It is also equal to $2^{d+1\choose 2}$ times the probability that a random matrix of size $d\times d$ (with respect to the uniform probability) with all coefficients in $[-1,1]$ is positive definite.

share|cite|improve this question
Roland: nice question, as usual. $c_2=2/9$? – Did May 4 '11 at 14:43
$c2\geq 1/4$ by considering $\left(\begin{array}{cc}a&b\\b&c\end{array}\right)$ with $a,c\in\lbrace N+1,\dots,2N\rbrace$ and $b\in \lbrace −N,…,N\rbrace$. – Roland Bacher May 4 '11 at 14:48
In fact, $c_2=2\left(\int_0^1 t^{3/2}dt\right)^2=\frac{8}{25}$. – Roland Bacher May 4 '11 at 15:13
I thought $|a|$ and $|c|$ were a priori at most $N$? And that the conditions for positive definiteness were $a\ge0$, $c\ge0$ and $b^2\le ac$. – Did May 4 '11 at 15:49
For a symmetric 2 x 2 matrix with entries uniformly distributed in $[-1,1]$ the probability of being positive definite is $\frac{1}{8} \int_0^1 da \int_0^1 dc \int_{-\sqrt{ac}}^{\sqrt{ac}} db = \frac{1}{9}$. – Robert Israel May 4 '11 at 17:48

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.