Is there an easy way to see the following:

Suppose Q is an integral quadratic form in $n$ variables that is positive definite, that is $Q(x) \geq 1$ for all $0 \neq x \in \mathbb{Z}^n$. Then the number of solutions to the equation $Q(x)=m$, for some fixed $n \in \mathbb{N}$, is finite.


Suppose that $Q$ is represented by an $n \times n$ matrix that I will by abuse of notation refer to as $Q$, and that $Q$ as a positive definite matrix satisfies a "coercivity" condition:

$$ x^T Q x \ge c x^T x $$

for some constant $c \gt 0$.

Then the existence of at most finitely many solutions to $Q(x) = x^T Q x = m$ follows from the existence of at most finitely many $x$ such that $c x^T x \le m$, counting on the integer components of $x$.

Added: The basic definition of an integer n-ary quadratic form is a homogeneous of degree 2 polynomial $Q(x_1,\ldots,x_n) = \sum_{i \le j} a_{ij}x_i x_j$. As alluded to this can be expressed in matrix form $x^T Q x$ where $x$ is a column vector $(x_1,\ldots,x_n)$, whereby the off-diagonal entries become $\frac{1}{2} a_{ij}$ for $i \neq j$ to obtain real symmetric matrix $Q$ (indeed $Q$ is rational).

Now real symmetric matrices have a complete set of eigenvalues and eigenvectors, and the task is to show that quadratic form $Q(x) \geq 1$ for integer vectors implies matrix $Q$ is positive definite in the sense that all eigenvalues are bounded below by positive constant $c \gt 0$, from which the coercivity property is immediate.

We really just need, since the number of eigenvalues of $Q$ is finite, to eliminate the possibility of negative and/or zero eigenvalues. If there were negative eigenvalues, then taking $x$ to be an appropriate rational approximation to a corresponding eigenvector would produce a negative value for $Q(x)$, and by multiplying through by any denominator, a corresponding integer $x$ that also gives a negative value for the quadratic form.

The case of zero eigenvalues is even simpler to dispose of. If matrix $Q$ were to have a nontrivial nullspace, then $Qx = 0$ would have rational nontrivial solutions (since $Q$ is rational), and again some integer multiple of such a solution $x$ would give $Q(x) = 0$ over the integer vectors, contradicting the asserted property of $Q(x) \ge 1$.

| cite | improve this answer | |
  • $\begingroup$ Could you explain this coercivity a little bit, I don't see how $Q(x) \geq 1$ implies that your equation holds. $\endgroup$ – Boris Datsik Jun 14 '13 at 13:40
  • $\begingroup$ Sure, I'll add more details. $\endgroup$ – hardmath Jun 14 '13 at 15:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.