Simpler proof of the Lax-Milgram Theorem if A is symmetric

383 Views Asked by At

In "Numerical Approximation of Partial Differential Equations" the authors Alfio Quarteroni and Alberto Valli state in remark 5.1.1, that the Riesz Representation Theorem suffices to prove the following Theorem, if $A$ is additionally assumed to by symmetric.

Theorem. (Lax-Milgram) Let $(V, (\cdot, \cdot), \| \cdot \|)$ be a (real) Hilbert space and $A \colon V \to V^*$ a linear, strongly positive, bounded operator, that is there exist $\mu, \beta > 0$ such that \begin{equation*} \langle A u, u \rangle \ge \mu \| u \|^2 \quad \text{and} \quad \langle A u, v \rangle \le \beta \| u \| \| v \| \quad \forall u,v \in V, \end{equation*} where $\langle f, v \rangle_{V^* \times V} =: f(v)$ is the dual pairing of $V^*$ and $V$, for $v \in V$ and $f \in V^*$. Then $A$ is bijective.

Here's is my attempt at a proof only using Riesz. As $A$ is symmetric, bounded and strongly positive, $(u, v)_A := \langle A u, v\rangle$ is a inner product on $V$. As $(V, (\cdot, \cdot)_A)$ is a Hilbert space (due to the boundedness and the strong positivity, the induced norms are equivalent to each other), by the Riesz Representation Theorem, there exists a isometric isomorphism $\iota \colon V^* \to V$ such that $\langle f, v \rangle = ( \iota(f), v )_A = \langle A \iota(f), v \rangle$ for all $f \in V^*$ and all $v \in V$. This implies $\langle f - A \iota(f), v \rangle = 0$ for all $f \in V^*$ and all $v \in V$ and hence $f = A\iota f$ for all $f \in V^*$ and hence $A \iota =$id, so $A$ is invertible with $A^{-1} = \iota$.

Is my proof correct?