Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

I asked this question here: http://math.stackexchange.com/questions/416885/checking-initial-condition-of-pde-is-satisfied-in-galerkin-method

But I did not receive the solution so I post it here.

The following is an argument I read, it is part of a proof of checking that the solution given by the Galerkin method satisfies the initial conditions.

From our weak formulation $$\langle u', v \rangle_{V',V} + a(u, v) = \langle f, v \rangle_{V',V}$$ which holds for all $v \in L^2(0,T;V)$, we can pick $v \in C^1(0,T;V)$ with $v(T) = 0$ and integrate this equation in time and use integration by parts to obtain $$\int_0^T -\langle u, v' \rangle_{V',V} + \int_0^T a(u,v) = \int_0^T\langle f, v \rangle_{V'V} + (u(0),v(0))_H.\tag{1}$$

Now we must use our Galerkin approximation $$\int_0^T\langle u_m', v_n \rangle_{V',V} + \int_0^Ta(u_m,v_n) = \int_0^T\langle f, v_n \rangle_{V',V}\tag{a}$$ and do some integration by parts and pass to the limit to obtain $$\int_0^T -\langle u, v' \rangle_{V',V} + \int_0^T a(u,v) = \int_0^T\langle f, v \rangle_{V'V} + (u_0,v(0))_H\tag{2}$$ where $v \in C^1(0,T;H)$ with $v(T) = 0$. This works because we know $u_m(0) \to u_0$, the initial condition.

Comparing (1) and (2), we can conclude $u(0) = u_0$.

Reasoning to obtain equation (2)

We have the convergences $u_m \rightharpoonup u$ in $L^2(0,T;V)$ and $u'_m \rightharpoonup u'$ in $L^2(0,T;V')$, where the $u_m \in V_m$. Pick $v_n \in V_n$ such that $v_n \to v$ in $L^2(0,T;V).$ Note that $v_n = \sum_{j=1}^n a_{j,n}w_j$ where $w_j$ is an o.n basis of $H$ and o.g basis of $V$.

We obtain from (a) the equation $$-\int_0^T\langle u_m, v_n' \rangle_{V',V} + \int_0^Ta(u_m, v_n) = \int_0^T\langle f, v_n \rangle + (u_m(0),v_n(0))_H$$ by IBP (and recall that $v \in C^1(0,T;V) \subset L^2(0,T;V)$ with $v(T) = 0$.)

My problem is, how to obtain (2) from here? Specificially, I don't understand why it is true that $$\int_0^T \langle u_m, v'_n \rangle \to \int_0^T \langle u, v' \rangle_{V'V}.$$

I am fine with the other terms in the equation. Why is true that $v'_n \to v$ in $L^2(0,T;V)$?

OK, so $v'$ exists and $v_n'$ probably exists assuming we can differentiate the coefficients. But why does it convergence to $v'$? No explanation is given anywhere.

share|improve this question
add comment

Your Answer

 
discard

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

Browse other questions tagged or ask your own question.