# How does the Hahn-Banach theorem implies the existence of weak solution?

I came across the following question when I read chapter 17 of Hormander's book "Tha Analysis of Linear Partial Differential Operators", and the theorem is

Let $a_{jk}(x)$ be Lipschitz continuous in an open set $X\subset\mathbb{R}^n$, $a_{ij}=a_{ji}$, and assume that $(\Re a_{ij}(x))$ is positive definite. Then $$\sum_{ij} D_j(a_{jk}D_ku)=f$$ has a solution $u\in H_{(2)}^{loc}(X)$ for every $f\in L_{loc}^2(X)$

The auther then says if we can show that $$|(f,\phi)|\leq \|M \cdot\sum_{ij} D_j(\bar{a_{jk}}D_k\phi) \|_{L^2}, \quad \phi\in C_c^{\infty}(X)$$ for some positive continuous function $M$, then by Hahn-Banach theorem there exists some $g\in L^2$ $$(f,\phi)=\left(g,M\cdot\sum_{ij} D_j(\bar{a_{jk}}D_k\phi)\right)$$ which inplies that the weak solution is $u=Mg$. what confuses me is how the Hahn-Banach theorem is used here to show the existence of $g$.

-
This $\Sigma$ is the sum $\sum_{ij}$ \sum_{ij}? –  Ilya Jun 5 '13 at 16:02
@Ilya,yes, thanks –  sun Jun 5 '13 at 16:06
There is missing something in your equations. In the inequality and in the equality, you have to add the terms containing $\phi$. –  Tomás Jun 5 '13 at 23:20
@Tomás, ah,sorry about that... –  sun Jun 5 '13 at 23:45
Your notaation is different form that on the book. There he consider the product $D_j\overline{a}_{jk}D_k\phi$, instead of $D_j(\overline{a}_{jk} D_k\phi)$ –  Tomás Jun 6 '13 at 0:00

Recall that $C_c^\infty$ is a dense linear subset of $L^2$. So by the Hahn-Banach theorem there is a norm preserving extension of $(f,\phi)$ to $(L^2)^*$. Then since $(L^2)^*$ is a Hilbert space the Riesz representation theorem tells us this extension has the form $\phi^{**}\longmapsto (g,\phi^{**})$ for some $g \in (L^2)^{**}$. Identifying $(L^2)^{**}=L^2$ then tells us $(f,\phi) = (g,\phi)$ for all $\phi \in L^2$.