On a Bessel Inequality proof

354 Views Asked by At

I've been working on a couple of algebra identities' demonstrations, among which the Bessel inequality. Here is the problem set:

We consider a vector space $V$, a vector $u$ in $V$, a subspace $E$ of $V$ of dimension $n$ and a basis $B$ for $E$ such that $B=\{u_1, u_2, ..., u_n\}$.

The orthogonal projection of $u$ on $E$, named $v$, is given by

$v=\sum_{i=1}^n {\lt u,u_i \gt\over \lt u_i,u_i \gt}\cdot u_i$

We define the vector $w$ as $u-v$ (hence $w$ is orthogonal to $E$).

We want to prove that

$||u||^2\geq \sum_{i=1}^n ({\lt u,u_i \gt^2\over \lt u_i, u_i \gt})\cdot u_i$ (Bessel Inequality)

By the generalization of the Pythagorean Theorem, we have:

$ \begin{align} {\|u\|}^2 & = {\|w\|}^2 + {\|v\|}^2 \\ & = {\|w\|}^2 + {\|\sum_{i=1}^n {\lt u,u_i \gt\over \lt u_i,u_i \gt}\cdot u_i \|}^2 \\ \end{align}$

Hence,

$ {\|u\|}^2 \geq {\|\sum_{i=1}^n {\lt u,u_i \gt\over \lt u_i,u_i \gt}\cdot u_i \|}^2 $

But I'm stuck here. I have to show that

$ {\|\sum_{i=1}^n {\lt u,u_i \gt\over \lt u_i,u_i \gt}\cdot u_i \|}^2 = {\sum_{i=1}^n ({\lt u,u_i \gt^2\over \lt u_i, u_i \gt})\cdot u_i}$

And I don't know how. I have tried the inner product properties, but I just do not understand how the norm squared gives this sum. I hope you can help me on this.

By the way, as this is my first time using MathJax, sorry for any errors in my format style. Thank you.

1

There are 1 best solutions below

2
On BEST ANSWER

In order to see that $ {\|\sum_{i=1}^n {\lt u,u_i \gt\over \lt u_i,u_i \gt}\cdot u_i \|}^2 = {\sum_{i=1}^n ({\lt u,u_i \gt^2\over \lt u_i, u_i \gt})\cdot u_i}, $ it suffices to (inductively) use the generalization of the Pythagorean theorem. In particular, $$ \left\|\sum_{i=1}^n {\langle u,u_i \rangle\over \langle u_i,u_i \rangle}\cdot u_i \right\|^2 = \left\|\left(\sum_{i=1}^{n-1} {\langle u,u_i \rangle\over \langle u_i,u_i \rangle}\cdot u_i \right) + {\langle u,u_n \rangle\over \langle u_n,u_n \rangle}\right\|^2 \\= \left\|\sum_{i=1}^{n-1} {\langle u,u_i \rangle\over \langle u_i,u_i \rangle}\cdot u_i\right\|^2 + \left\|{\langle u,u_n \rangle\over \langle u_n,u_n \rangle} \cdot u_n\right\|^2 $$

From there, note that (by the homogeneity of norms) $$ \left\|{\langle u,u_n \rangle\over \langle u_n,u_n \rangle} \cdot u_n\right\|^2 = \left| {\langle u,u_n \rangle\over \langle u_n,u_n \rangle} \right|^2 \cdot \|u_n\|^2 = \frac{|\langle u,u_n \rangle|^2}{\langle u_n,u_n \rangle} \cdot \langle u_n,u_n \rangle = |\langle u,u_n\rangle|^2. $$