Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to solve this problem:

Let $H$ a Hilbert space, $A:H\times H\rightarrow\mathbb{R}$ a bilinear form, bounded and $H$-elliptic, and $F\in H^{\prime}$ ($H^{\prime}$ = dual space). Besides, let $\{H_n\}_{n\in\mathbb{N}}$ a sequence of subspaces of finite dimension of $H$, and for each $n\in\mathbb{N}$ consider a bounded bilinear form $A_n:H_n\times H_n\rightarrow\mathbb{R}$ such that the sequence $\{A_n\}_{n\in\mathbb{N}}$ is uniform elliptic. This is, there is $\tilde{\alpha} > 0$, independent of $n$ m such that $A_n(v_n,v_n) \geq \tilde{\alpha}\|v_n\|_H^2$, $\forall\ v_n\in H_n$, $\forall\ n\in\mathbb{N}$.

  1. Show that there's unique $u\in H$ and $u_n\in H_n$ such that $$A(u,v)\ =\ F(v),\quad \forall\ v\in H$$ and $$A_n(u_n,v_n)\ =\ F(v_n),\quad \forall\ v_n\in H_n.$$
  2. Prove that there's $C>0$, independent of $n\in\mathbb{N}$, such that $$\|u - u_n\|_H\ \leq\ C\inf_{v_n\in H_n}\left\{\|u - v_n\|_H + \sup_{\mbox{$w_n\in H_n \atop w_n\neq 0$}}\frac{\left|A(v_n,w_n) - A_n(v_n, w_n)\right|}{\|w_n\|_H}\right\}.$$

I think I solved it 1., by direct applications of Lax-Milgram Lemma, but I really don't know how to solve 2. I tried to use the subordinate forms: $$A(w,v)\ =\ \langle\mathbb{A}(w),v\rangle, \quad \forall (w,v)\in H\times H,$$ $$A_n(w_n,v_n)\ =\ \langle\mathbb{A}_n(w_n),v_n\rangle, \quad \forall (w_n,v_n)\in H_n\times H_n,$$ and the best approximation theory, but I couldn't do it.

Please help me.

Thanks in advance.

share|cite|improve this question
Is there something like $\cup_n H_n$ dense in $H$? Also, is there any relation between $A$ and $A_n$? – TZakrevskiy Aug 7 '13 at 17:36
No, I wrote the problem exactly. – FASCH Aug 7 '13 at 17:42
The only relation, if I am correct, is $\mathbb{A}(u) = \mathcal{R}(F)$ and, $\langle\mathbb{A}_n(u_n), v_n\rangle = \langle\mathcal{R}(F), v_n\rangle$, for all $v_n\in H_n$. – FASCH Aug 7 '13 at 17:45

For all $v_n \in H_n$ we have \begin{align*} A(u-u_n, u-u_n) &= A(u-u_n, u - v_n) + A(u, v_n - u_n)-A(u_n, v_n - u_n) \\ &=A(u-u_n, u - v_n) + A_n(u_n, v_n - u_n)-A(u_n, v_n - u_n)\end{align*} This implies the estimate \begin{align*} \alpha \, \|u - u_n\|^2 &\le \|A\| \, \|u-u_n\| \, \|u -v_n\|\\ &\qquad + \|\mathbb A_nu_n-\mathbb A u_n\| \, \|u_n - v_n\|\\ &\le \|A\| \, \|u-u_n\| \, \|u -v_n\|\\ &\qquad + \|\mathbb A_nu_n-\mathbb A u_n\| \, \|u - u_n\|\\ &\qquad + \|\mathbb A_nu_n-\mathbb A u_n\| \, \|u - v_n\|\\ \end{align*} Here, $\alpha$ is the coercivity constant of $A$. Now, we use Young's inequality on the second term of the right-hand side and obtain \begin{align*} \frac\alpha2 \, \|u - u_n\|^2 &\le \|A\| \, \|u-u_n\| \, \|u -v_n\|\\ &\qquad + \|\mathbb A_nu_n-\mathbb A u_n\| \, \|u - v_n\|\\ &\qquad \frac1{2\,\alpha}\,\|\mathbb A_nu_n-\mathbb A u_n\|^2. \end{align*} Now, take Young's inequality on the first two terms on the right-hand side to obtain $$c \, \|u - u_n\|^2 \le \|u - v_n\|^2 + \|\mathbb A_nu_n-\mathbb A u_n\|^2$$ with some $c > 0$ (depending only on $\alpha$ and $\|A\|$). By taking the $\inf$ over all $v_n \in H_n$, using the equivalency of the $2$- and the $1$-norm in $\mathbb R^2$ and taking the square root, we get the claim. Note that this does not require the uniform coercivity of $A_n$.

share|cite|improve this answer
How, taking the inf over all $v_n\in H_n$, you can change $\|\mathbb A_nu_n-\mathbb A u_n\|^2$ for $\|\mathbb A_nv_n-\mathbb A v_n\|^2$ ? Because, in other hand, I don't see how get the claim. – FASCH Aug 7 '13 at 20:35
Sorry, my fault. Anyway, you might want to look for the first lemma of Strang, this is just your claim. – gerw Aug 8 '13 at 8:34
Thanks for the reference. ( – FASCH Aug 8 '13 at 12:37

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.