Let $A$ be a real $2\times2$ symmetric matrix that is indefinite (it has a positive and a negative eigenvalues). Let $Q = \langle Av, v\rangle$ be the associated quadratic form. If $Q(v) = 0$, is it true that $$|Av|^2 = -\det(A)|v|^2?$$


Denote the eigenvalues of $A$ by $\lambda_1 < 0 <\lambda_2$ with corresponding eigenvectors $v_i$ with norm one. These are orthogonal: $v_1^Tv_2=0$, thus $\{v_1,v_2\}$ is an orthonormal basis. Then $v= v_1 (v^Tv_1) + v_2(v^Tv_2)$ for all $v$, and $Av=\lambda_1v_1 (v^Tv_1) + \lambda_2v_2(v^Tv_2)$. $Q(v)=0$ implies $0=v^TAv=\lambda_1 (v^Tv_1)^2 + \lambda_2(v^Tv_2)^2$. Then $$\begin{split} \|Av\|^2 &= \lambda_1^2 (v^Tv_1)^2 + \lambda_2^2 (v^Tv_2)^2\\ &= -\lambda_1\lambda_2(v^Tv_2)^2 - \lambda_2\lambda_1(v^Tv_1)^2\\ &= -\lambda_1\lambda_2 \|v\|^2 \\ &=-\det(A)\|v\|^2. \end{split}$$

The original question was without squaring the norms...

No, this is not true. If the claim would be true for $A$, it would be true for $2A$. But $\|2Av\|=2\|Av\|$ and $\det(2A)=4\det(A)$.

  • $\begingroup$ You are indeed correct, I forgot to square the norm! $\endgroup$ – Henrique Augusto Souza Jul 8 at 11:31

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.