Convergent in $L^1(0,1)$ but not in $L^2(0,1)$ help understanding a paper from arxiv

Question

http://arxiv.org/pdf/math/0205003v1

In around equation (1.1) the author says

"By necessity all authors have been led in one way or another to the natural approximation

$$F(n) := \sum_{a=1}^n \mu(a) \rho_a$$

which tends to $-\chi$ both a.e. and in L1 norm when restricted to (0,1) but which has been shown to diverge in H."

My question is:

How to determine $F$ diverges in $L^2(0,1)$? because $$\int_{0}^{1} |1 + F|^2 dx < \infty$$ where $F \to -1$?

I am new to functional analysis, so little confused on this.

Thanks,

Roupam