The Ratio Test

A geometric series multiplies each term by the same ratio to get the next. Many other series multiply by a ratio that settles down as nn grows. The ratio test treats such a series as nearly geometric.

The test

Take positive terms with L<1L < 1, and pick a number rr with L<r<1L < r < 1. Eventually every ratio is below rr, so from some aNa_N on each term is at most rr times the one before. Then aN+k≤aNrka_{N+k} \le a_N r^k, and the series converges by comparison with a geometric series. For terms of mixed sign, the same argument shows that ∑∣an∣\sum |a_n| converges, and a series whose sizes have a finite sum converges too. The page on absolute and conditional convergence proves that. When L>1L > 1, the sizes of the terms eventually grow, so the terms can't approach 0 and the series diverges by the nth term test.
The test works best on factorials and on powers like 5n5^n, where the ratio of consecutive terms simplifies. These facts do most of the work.
(n+1)!n!=n+1(2n+2)!(2n)!=(2n+2)(2n+1)(3n+3)!(3n)!=(3n+3)(3n+2)(3n+1)\begin{aligned} \frac{(n + 1)!}{n!} &= n + 1 \\[4pt] \frac{(2n + 2)!}{(2n)!} &= (2n + 2)(2n + 1) \\[4pt] \frac{(3n + 3)!}{(3n)!} &= (3n + 3)(3n + 2)(3n + 1) \end{aligned}

When the limit is 1

Every p-series has L=1L = 1, whether it converges or diverges, so L=1L = 1 can't decide anything. The same happens for most series built only from powers of nn, and for those a comparison test is the right tool.