I am trying to analyze the sequence $g(n)$, defined as follows:
- $g(1)=0$
- $g(2)=1$
- For $n\geq 2$, $g(n+1)=g(n)\left(1+\displaystyle{\frac{n}{n-1}\ln\left(\frac{n+1}{n}\right)}\right)$.
A computational experiment suggests that, as $n\rightarrow\infty$, $\displaystyle{\frac{g(n)}{n}}$ approaches a constant of $\approx 0.758097$, but I don't know how to prove it. How would you do so?
Define $\displaystyle h(n) = 1-\frac1n+\ln\biggl( 1+\frac1n \biggr)$, so that \begin{align*} g(n+1) &= \prod_{k=2}^n \biggl( 1+\frac k{k-1}\ln\frac{k+1}k \biggr) \\ &= \prod_{k=2}^n \biggl( \frac k{k-1}h(k) \biggr) = \prod_{k=2}^n \frac k{k-1} \prod_{k=2}^n h(k) = n \prod_{k=2}^n h(k). \end{align*} But from the Maclaurin series for $\ln(1+x)$, we see that $h(n) \displaystyle = 1+O\biggl( \frac1{n^2} \biggr)$, and so this last product converges as $n\to\infty$. Indeed this equation implies $$ g(n) = n \prod_{k=2}^\infty h(k)+O(1) $$ where this infinite product is indeed approximately $0.75809696$, accurate to that many decimal places (rounded).