Let $V \subset H$ be separable Hilbert spaces with continuous embedding and suppose $\{v_n\}$ be a (non-orthogonal) basis for $V$. If we let $V_n = \text{span}(v_1, ..., v_n)$ and given $h \in H$ we define an operator $P$ by $$(P_n(h)-h, v)_H = 0$$ for all $v \in V_n$, then what space does $P_n(h)$ lie in?

I hoped it was in $V_n$. But I don't know how to prove that.


Fix $n$. Let $\{ e_{1},e_{2},\cdots,e_{n} \}$ be an equivalent orthonormal basis obtained from $\{ v_{1},v_{2},\cdots,v_{n}\}$ by Graham-Schmidt using the inner product on $H$. Each $e_{j} \in V_{n}$ for $1 \le j \le n$ because $e_{j}$ is a linear combination of $\{ v_{1},v_{2},\cdots,v_{j}\}$; therefore, $$ P_{n}h = \sum_{j=1}^{n}(h,e_{j})_{H}e_{j} \in V_{n},\;\;\; h \in H. $$ This is true for any fixed $n =1,2,3,\cdots$.

  • $\begingroup$ Why does $P_nh$ as defined in the OP equal the right hand side in your answer? Your formula defines the the projection onto the first $n$ basis functions of $h$ of course but I don't see why it's the same as the OP's projection. $\endgroup$ – riem Sep 29 '14 at 18:01
  • $\begingroup$ @riem : Verify that $(P_{n}h-h,e_{j})_{H}=0$ for $1 \le j \le n$. It then follows that $(P_{n}h-h,v)_{H}=0$ for all $v \in V_{n}$. $\endgroup$ – DisintegratingByParts Sep 29 '14 at 18:29

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.