Let $(X, \| \cdot \|)$ be a normed vector space with algebraic basis $e_1, e_2, ...$ (dim $X = \infty$ and for all $x \in X, x = \sum_{i=1}^{\infty} a_i e_i$ for unique $a_i$.). Then $X$ is incomplete.

I have written a proof of this theorem, but it is quite long and I have a feeling there should be an easier way of doing it. Please let me know if my proof is correct, and if possible, how to write a better proof.

My proof:

Proof by contradiction: assume $X$ is complete, then $X$ is a Banach space. Then, as finite dimensional subspaces of a normed vector space, $C_n = \text{span} \{e_1, ..., e_n\}$ is closed for all $n \in \mathbb{N}$. Furthermore, since $X = \cup_n C_n$, by Baire's theorem there is some $n \in \mathbb{N}$ such that int$(C_n) \neq \emptyset$. Hence, let $x = \sum_{i=1}^n x_i e_i \in$ int$(C_n)$.

Note $C_{n+1}$ is itself a finite dimensional normed vector space, and as such all norms on $C_{n+1}$ are equivalent and $C_{n+1}$ is complete with respect to every norm.

Hence, define the norm $\| \sum_{i=1}^{n+1}a_ie_i\|_{\infty} = \sup_{i =1, ..., n}|a_i|$ on $C_{n+1}$ (it is easily verified that this is a norm). Consider the sequence $y_i = \sum_{i=1}^nx_ie_i + \frac{1}{i}e_{n+1} \in C_{n+1}$.

Clearly, $\lim_i\|y_i - x\|_{\infty} = \lim_i \frac{1}{i}=0$, which means that $y_i \to x$ in $(C_n,\| \cdot \|_{\infty})$. Since $\| \cdot \|_{\infty}$ is equivalent to $\| \cdot \|$ on $C_{n+1}$, we therefore have $y_i \to x$ in $\|\cdot\|$. But, $x \in \text{int}(C_n)$ so $y_i \to x$ implies that there is some $M \in \mathbb{N}$ such that for all $m \geq M, y_m \in C_n$, a contradiction since $y_m = \sum_{i=1}^n x_i e_i + \frac{1}{m} e_{n+1} \notin C_n$.

  • $\begingroup$ The statement in your title is wrong. Countable dimensional normed vector spaces are incomplete. $\endgroup$ Jun 19 '18 at 3:41
  • $\begingroup$ In your first sentence, since you are talking about an algebraic basis, why do you write $x = \sum_{i=1}^\infty a_i e_i$? Shouldn't the sum be finite? $\endgroup$ Sep 27 at 11:50

Now that you know BCT and the fact that all finite dimensional normed vector spaces are Euclidean, the proof is very straight-forward. As you argued, there is some $n\in\mathbb N$ such that $\operatorname{int}C_n\ne\emptyset$. In particular, $\operatorname{int}C_n$ is an non-empty open set in the subspace topology on $C_{n+1}$. However, $C_{n+1}$ is linearly homeomorphic to $\mathbb R^{n+1}$, whose $n$-dimensional subspace contains no non-empty open sets.

  • $\begingroup$ Thanks! Does that mean my proof is correct? I don't already have a theorem saying $C_{n+1}$ is linearly isomorphic to $\mathbb{R}^{n+1}$, or that $\mathbb{R}^{n+1}$ contains now non-empty open sets (although I can see that these things are true) $\endgroup$
    – user2139
    Jun 20 '18 at 13:51
  • 1
    $\begingroup$ @user2139 Yes, your proof is correct :-) $\endgroup$ Jun 21 '18 at 3:14

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.