Let $V$ be a closed subspace of $\mathcal H$ and if $y \in \mathcal H$ but $y \notin V$ prove that $W = \text{Span}(V \cup\{y\})$ is a closed subspace.

Here's my attempted proof.


Note first that theorem 4.11 in Big Rudin says that $\mathcal{H} = V \oplus V^\perp$. Therefore, since $y \in \mathcal{H}$ but $y \notin V$, $y = z+z^\perp$ with $z \in V$ and $z^\perp \in V^\perp$. Now, since span is the set of all finite linear combinations, $$\text{Span}(V \cup\{y\}) = \text{Span}(V \cup\{z+z^\perp\}) = \text{Span}(V \cup\{z^\perp\})$$ because $z \in V$ already. Thus, it suffices to show the rhs is closed. Now, take a sequence in this subspace, $\{x_n\}_{n \in \mathbb{N}} \to x \in \mathcal{H}$. We need to show $x \in \text{Span}(V \cup\{z^\perp\})$. Let $\epsilon >0$; since $x_n \to x$, there exists $n$ such that

$$\|x_n - x\|^2 < \epsilon.$$ We know there exists linear maps (since $V$ is closed) $P: \mathcal{H} \to V$ and $P: \mathcal{H} \to V^\perp$ giving unique orthogonal decomopsitions. We have $P(x_n) \in V$, so that $\beta_nz^\perp=Q(x_n)$, and let $x = P(x)+Q(x)$.

Then, using the fact that $P(x_n) - P(x) \perp Q(x_n)-Q(x)$, \begin{eqnarray*} \|x_n-x\|^2 &=& \|(P(x_n)+\beta_nz^\perp)-(P(x)+Q(x))\|^2\\ &=& \|P(x_n) - P(x)\|^2 + \|\beta_nz^\perp - Q(x)\|^2 \\ &<& \epsilon \end{eqnarray*}

and the above says precisely that $P(x) = \lim \limits_{n \to \infty} P(x_n)$ and $Q(x) = \lim \limits_{n \to \infty} Q(x_n) = \beta z^\perp$ in norm for some $\beta \in \mathbb{C}$. \ Thus, since $V$ is closed, $P(x) \in V$, so thus $x \in \text{Span}(V \cup\{z^\perp\})$ since it is of the form $P(x) + \beta z^\perp$.

Are there any issues with this proof? I've EDITED this post substantially


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.