# Suppose that for all $g\in L^2(\mathbb{R}^d)$ we have $\int f_{n}g \to \int fg$. Prove convergence in $L^2$ problem.

How to solve the following problem:

Let $f_n, f \in L^2(\mathbb{R}^d)$ for all $n \geq 1$ be such that $\|f_n\|_2 \to \|f\|_2$ as $n \to \infty$. Suppose, moreover, that $$\int f_{n}g \to \int fg$$ for all $g \in L^2(\mathbb{R}^d)$. Then $f_n$ converges to $f$ in $L^2$-norm.

-
Try using the parallelogram law. – t.b. Aug 13 '12 at 11:06

You want to show $$\| f - f_n \|_2 \to 0$$

Equivalently, you can show $$\int (f-f_n)^2 d \mu = \| f - f_n \|_2^2 \to 0$$

We have $$\| f - f_n \|_2^2 = \int (f-f_n)^2 d \mu = \int f^2 d \mu - 2 \int f f_n d \mu + \int f_n^2 d \mu$$

Since we have $\|f_n\|_2 \to \|f\|_2$, we have $\int f_n^2 d \mu \to \int f^2 d \mu$ and since we have $\int f_n g d \mu \to \int fg d \mu$ we have $\int f f_n d \mu \to \int f^2 d \mu$ so that

$$\int f^2 d \mu - 2 \int f f_n d \mu + \int f_n^2 d \mu \to \int f^2 d \mu - 2 \int f^2 d \mu + \int f^2 d \mu = 0$$

that is, $$\| f - f_n \|_2^2 \to 0$$

and hence of course also

$$\| f - f_n \|_2 \to 0$$

-

Since $L^2$ is a Hilbert space, you can use the parallelogram identity, as suggested in a comment. More generally, you can also use a property of any uniformly convex Banach space, which is listed as the second property here. A very nice proof appears in Brezis' book on functional analysis.

-