I need to demonstrate the previous.

What i said is... ...symmetrical endomorphism operator with all the eigenvalues positive means that he represents an "expansion" of the space $W$. this means that if i write every vector $v$ as:

$$v=\lambda_1 e_1 + \lambda_2 e_2 +...+\lambda_n e_n$$

with $e_1,...e_n$ indipendent vectors from the autospaces of $f$. Then $f(v)$ will be like:


But sign of $(e_i,e_i)$ will be $(+,+)$ or $(-,-)$

(because i said that $f$ is an "expansion" of the space $W$!) and $\Lambda_i$ will be all positive numbers!

So the euclidean scalar product of $\Omega(f(v),v)$ will be always $>0$ (if $v$ is different form $0$).

Im wrong? If yes, why?


Since $f$ is symmetric, there is an orthonormal basis $\{e_1,...,e_n\}$ of $W$, consisting of eigenvectors of $f$, such that $f$ has the representation (spectral theorem !):

$f(v)= \sum_{k=1}^n \lambda_k <v,e_k>e_k,$


(1) $f(e_k)=\lambda e_k$.

It follows:

(2) $<f(v),v>=\sum_{k=1}^n \lambda_k |<v,e_k>|^2$.

Can you take it from here ?

  • $\begingroup$ well, yes, $\lambda_k$ is always positive and is multiplied by a module squared wich is different from 0 because our $\Omega$ is euclidean. But is my "demonstration" right or it can be wrong in something? (for sure things like "expansion" needs to be formalized) $\endgroup$ Oct 9 '17 at 13:58

If $f$ is symmetric with positive eigenvalues then is an orthonormal basis $(e_i)_i$ such that each $e_i$ is an eigenvalue of $f$ and every vector $v$ has a unique the decompostion of the form $$v= \sum_i \langle e_i,v\rangle e_i~~\Longrightarrow f(v)=\sum_i \langle e_i,v\rangle f(e_i) = \sum_i \lambda_i\langle e_i,v\rangle e_i$$

Since $f(e_i) =\lambda_ie_i$


$$\langle f(v),v\rangle=\Big\langle\sum_i \lambda_i\langle e_i,v\rangle e_i,v\Big\rangle= \sum_i \lambda_i\langle e_i,v\rangle \langle e_i,v\rangle>0 $$

Conversely, if $$\langle f(v),v\rangle>0~~v\neq 0$$

Then , $$0< \langle f(e_i),e_i\rangle= \lambda_i\langle e_i,e_i\rangle =\lambda_i$$

that is $$\lambda_i>0$$


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.