Let $H$ be a Hilbert space and $S\subseteq H$ a closed subspace. Moreover, let $\{s_{n}\}_{n=1}^{\infty}\subseteq S$ a complete and linear independent sequence in $S$, i.e.

  • $S=\overline{\text{Span}\big(\{s_{n}\}_{n=1}^{\infty}\big)}$ and
  • for $N\geq 1$ and $\lambda\in\mathbb{R}^{N}$, $\sum_{n=1}^{N}\lambda_{n}\,s_{n}=0$ implies $\lambda=0$.

Denote by $\mathcal{P}$ the orthogonal projection from $H$ onto $S$ and by $\mathcal{P}_{N}$ the orthogonal projection from $H$ onto $\text{Span}\big(\{s_{n}\}_{n=1}^{N}\big)$.

Is it true that, for all $x\in H$, $$\mathcal{P}_{N}(x) \quad\rightarrow\quad \mathcal{P}(x)$$ for $N\rightarrow\infty$ ?

This is used in a paper without proof or comment and I am wondering how to show this rigorously.

Any help or comment is highly appreciated! Thanks.


Apply Gram-Schmidt process on the sequence $\{s_n\}_{n=1}^\infty$ to obtain an orthonormal sequence $\{e_n\}_{n=1}^\infty$ in $S$ such that $\operatorname{span} \{e_1, \ldots, e_n\} = \operatorname{span} \{s_1, \ldots, s_n\}, \forall n \in \mathbb{N}$ and $\operatorname{span} \{e_n\}_{n=1}^\infty = \operatorname{span} \{s_n\}_{n=1}^\infty$.

Then clearly $\{e_n\}_{n=1}^\infty$ is an orthonormal basis for $S$ so

$$P_nx = \sum_{k=1}^n \langle x, e_k\rangle e_k \xrightarrow{n\to\infty} \sum_{k=1}^\infty \langle x, e_k\rangle e_k = Px$$

for all $x \in H$.

  • $\begingroup$ Clear and concise. Thanks! $\endgroup$ – Mark Sep 5 '18 at 10:08

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.