Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. 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

I have found an exercice in a calculus book, which I have problems to solve.

$$\text{Let}\, R:\mathbb{R}^n-{0}\to \mathbb{R},\quad R(x)=\frac{x^tAx}{x^tx} = \frac{\langle x,Ax\rangle}{\langle x,x\rangle},$$ where $\langle \cdot,\cdot \rangle$ is the euclidean inner product, $A\in\mathbb{R}^{n\times n}$.

$1)$ $R$ has a minimum in $\mathbb{R}-{0}$.

$2)$ Every critical point of $R$ is an Eigenvector of $\frac{1}{2}(A^t+A)$ corresponding to an Eigenvalue of $A$. In particular, every symmetric real matrix has real Eigenvalues.

I have solved $1)$ (by seeing that it is sufficient to study $R|S^n$ and $S^n$ is compact, so $R$ has a minimum and maximum which it attains.) For $2)$ I can't find the desired result. I know that $\frac{1}{2}(A^t+A)$ is symmetric, and that $\langle x,Ax \rangle=\langle x,A^tx \rangle$.

share|cite|improve this question
Did you try to compute $$\left. \frac{d}{d\epsilon} \right|_{\epsilon =0} R(x+\epsilon v)$$ at a critical point $x$ of $R$? – Siminore Sep 16 '12 at 11:53
Ah yes this is the idea. Now I don't really know how to conclude from this. – Lucien Sep 16 '12 at 12:19
Of course the derivative that I wrote is zero, since $x$ is a critical point of $R$. Write down the derivative, and you will find a relation that must hold for every $v$. This will imply that $x$ is an eigenvalue. Anyway, this is an easy case of the Lagrange multiplier rule. – Siminore Sep 16 '12 at 12:46
Thank you for your hint, I have solved it. I will post the answer later. – Lucien Sep 16 '12 at 15:57
up vote 0 down vote accepted

Here is the solution using Siminore's hint.

Let $x\in\mathbb{R}$ be an critical point. Then $\left. \frac{d}{d\epsilon} \right|_{\epsilon =0} R(x+\epsilon v)=0$ $(*)$.

Computing $R(x+\epsilon v)$ explicitly, deriving with respect to $\epsilon$, and putting $\epsilon=0$ we find, using $(*)$ for all $v\in\mathbb{R}$ we find:

$\langle v, \frac{A+A^t}{2}\; x\rangle\langle x,x\rangle=\langle x, Ax\rangle\langle x,v\rangle$.

For all $v\bot\, x$ we find that $\langle v, \frac{A+A^t}{2}\; x\rangle=0$ which shows that $\frac{A+A^t}{2}\;x=\lambda x$ for some $\lambda\in \mathbb{R}$. Moreover $\lambda$ is an Eigenvalue of $A$: with the choice $x=v$ we get: $\lambda \langle x,x\rangle=\langle x,Ax\rangle \iff \langle x,(A-\lambda Id)x\rangle=0$, and so $Ax=\lambda x$ $(x\neq 0)$.

share|cite|improve this answer

Your Answer


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

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