I want to prove this:

Given a vector space $V$ on $\mathbb{R}$ with a positive definite inner product $\left \langle .,. \right \rangle$. Show that there exist a natural number $p$ and a linear map $T:V\rightarrow \mathbb{R}^{p}$ such that $\left \langle u,v \right \rangle=T(u).T(v)$ for all $u,v\in V$ where $.$ is the standard dot product.

My try: Let $V$ be $\mathbb{R}^{n}$. If we pick any $p \geq n$ and choose $T(u)=proj_{\mathbb{R}^{n}}(u)$ where $proj_{\mathbb{R}^{n}}(u)$ is the linear map that projects vectors from $\mathbb{R}^{p}$ to $\mathbb{R}^{n}$ then we have the inner product of elements $T(V)$ reduced to the one desired of $\mathbb{R}^{n}$. Is my proof legitimate? What about if I had been asked for the case $p < n$?

Thanks in advance.

  • $\begingroup$ Hint: For $p=dim(V)$ the statement is equivalent to the fact that there is an orthormal basis for the inner product space $(V,\langle .,. \rangle)$. $\endgroup$ – Andreas Cap Jun 19 '16 at 8:38

Let $\;U:=\{u_1,...,u_p\}\;$ be an orthonormal basis of $\;V\;$ ,and let $\;T:V\to\Bbb R^p\;$ be the map $\;Tu_i:=e_i\;$ , with $\;\{e_i\}\;$ the usual, standard (and orthonormal with the usual inner product) basis of $\;\Bbb R^p\;$.

Thus, we can write with resepct to the basis $\;U\;$ above:

$$\begin{cases}u=\sum\limits_{k=1}^n \langle u,u_k\rangle u_k\\{}\\v=\sum\limits_{k=1}^n \langle v,u_k\rangle u_k \end{cases}\;\;\implies\;\;\langle u,v\rangle=\sum_{k,i=1}^n \langle u,u_k\rangle\langle v,u_i\rangle\langle u_k,u_i\rangle$$$${}$$

$$=\sum_{k=1}^n\langle u,u_k\rangle\langle v,u_k\rangle=Tu\cdot Tv\;,\;\;\text{since}\;\;$$

$$Tu=T\left(\sum_{k=1}^n\langle u,u_k\rangle u_k\right)=\sum_{k=1}^n\langle u,u_k\rangle Tu_k=\sum_{k=1}^n\langle u,u_k\rangle e_k$$

  • $\begingroup$ Thank you! But why is the basis $U$ consisting of $p$ vectors while the vectors $u$ and $v$ go from $1$ to $n$? $\endgroup$ – John11 Jun 19 '16 at 9:45
  • $\begingroup$ Well, I really didn't see that but it you say "if we pick..", then I presume we can pick as we want, so for me $\;p=n=\dim V\;$ , to avoid problems...:) $\endgroup$ – DonAntonio Jun 19 '16 at 9:49

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.