# How to prove a Banach normed vector space is NOT a Hilbert space?

We know that the Banach space $\big(\Bbb R^n,\|\cdot\|_2\big)$ is a Hilbert space with inner product $\langle x,y\rangle := \sum_{k=1}^n x_ky_k$. However, how to prove that $\big(\Bbb R^n,\|\cdot\|_3\big)$ is NOT a Hilbert space where $\|v\|_3 := \left(\sum_{k=1}^n|v_k|^3\right)^{1/3}$?

More generally is there a characterization of Hilbert spaces other than the existence of an inner product? And supposing $(V,\|\cdot\|_V)$ satisfies this characterization, how can we build an associated inner product?

• Show that the Parallelogram law doesn't hold. Jan 6 '15 at 21:56
• @DavidMitra Nice and the polarization formula (just under in the wikipedia article) shows how to build the inner product. Jan 6 '15 at 21:58

## 2 Answers

It's possible to prove that norm comes from inner product if only if Parallelogram law holds, that means:

$$2\|x\|^2+2\|y\|^2=\|x+y\|^2+\|x-y\|^2$$

For all $x,y \in \mathbb{R}^n$.

For example, let $x=(1,0,0,0,\ldots)$, $y=(0,1,0,0,\ldots)$ , then:

$$\|x\|^2=\|y\|^2=1$$

$$\|x+y\|^2=(2)^{\frac{2}{3}}$$

$$\|x-y\|^2=(2)^{\frac{2}{3}}$$

So Parallelogram law doesn't hold.

• I don't understand your example... Jan 6 '15 at 22:09
• More formally $x_{1}=1$ and $x_{n}=0$ for $n>1$, $y_2=1$ and $y_n =0$ for $n \neq 2$.
– agha
Jan 6 '15 at 22:11
• the norm of $\|x+y\|$ should be $2^{1/3}$ I think Jan 6 '15 at 22:14
• Yes, it's true, so $\|x+y\|^2=((2)^{\frac{1}{3}})^2$.
– agha
Jan 6 '15 at 22:15

The completeness properties associated with Banach spaces and Hilbert spaces are not very relevant: a norm on a real vector space is called euclidean if it is induced by an inner product (so a Banach space is euclidean iff it is a Hilbert space). The $p$-norms on $\mathbb{R}^n$ for $p$ other than $2$ can be seen not to be euclidean in lots of ways, e.g., using the parallelogram law as in agha's answer.

To answer your "more generally" questions: there has been a huge cottage industry producing characterisations of euclidean normed spaces. The book by Dan Amir referred to in my minor contribution to this subject gives a huge list. If you know that a normed space is euclidean, then the inner product is given by the formula:

$$\langle v, w \rangle = \frac{1}{2}(\|v + w\|^2 - \|v\|^2 - \|w\|^2))$$

• Thank you for the reference, it seems to be a great starting point for other reference in the subject. Jan 7 '15 at 7:44