Given a positive sequence $(a_n)$ which satisfies $\displaystyle \sum_{n=1}^{\infty}\frac{1}{a_n}$ converges. Let $S_n=a_1+a_2+\cdots+a_n$. Prove that $\displaystyle \sum_{n=1}^{\infty} \frac{n^2a_n}{S_n^2}$ also converges.
This is the first part of the solution:
Translate the hypothesis into function language: if $f:\mathbb{R}^+\rightarrow \mathbb{R}^+$ is an increasing function and $\displaystyle \int_{0}^{\infty}\frac{dx}{f(x)}$ converges, then $\displaystyle \int_{0}^{\infty} \frac{x^2f(x)}{F^2(x)}dx$ also converges, where $F(x)$ is the anti-derivative of $f(x)$.
I don't understand this. What is the function $f(x)$? And why $S_n$ can be also understood as $F(x)$?
Thanks a lot.
If $f:\mathbb R^+\to\mathbb R^+$ is given, then \begin{align*} \sum_{n=1}^\infty\inf_{x\in(n-1,n)}f(x)\leq\sum_{n=1}^\infty\int_{n-1}^nf(x)dx=\int_0^\infty f(x)dx=\sum_{n=1}^\infty\int_{n-1}^nf(x)dx\leq\sum_{n=1}^\infty\sup_{x\in(n-1,n)}f(x), \end{align*} so the integral converges if the right sum converges, and if the integral converges, then so does the left sum. If we define $f(x):=a_n$ for $x\in(n-1,n),n\in\mathbb N$, then we have equality in the above inequalities. Moreover, $F(x):=\int_0^xf(t)dt$ is piecewise linear and $F'(x)=f(x)$ for $x\in(n-1,n),n\in\mathbb N$, and \begin{align*} F(n)=\int_0^{n}f(t)dt=\sum_{k=1}^{n}\int_{k-1}^kf(t)dt=\sum_{k=1}^{n}a_k=S_n. \end{align*} Now, if one proves that \begin{align*} \int_0^\infty\frac{x^2f(x)}{F(x)^2}dx \end{align*} converges, then the above considerations show that the series \begin{align*} \sum_{n=1}^\infty\inf_{x\in(n-1,n)}\frac{x^2f(x)}{F(x)^2} \end{align*} converges. Due to \begin{align*} \sum_{n=1}^\infty\frac{n^2a_n}{S_n^2}\leq\sum_{n=1}^\infty\inf_{x\in(n-1,n)}\frac{x^2f(x)}{F(x)^2}, \end{align*} the left series also converges.