Series Convergence Tests

The nth-term, p-series, and ratio tests.

The idea

A convergence test is a checkable condition on the terms of a series that carries a guaranteed verdict. The series $\sum_{n=1}^{\infty} a_{n}$ means the limit of its partial sums $a_{1} + a_{2} + \cdots + a_{N}$ as $N$ grows; whether that limit exists is rarely visible from the terms directly, so we apply a test. The first test comes first because it costs the least.

Theorem (The nth-term test).

If $a_{n}$ does not tend to $0$, then $\sum a_{n}$ diverges.

The test can only reject. Terms that do shrink to $0$ prove nothing, and the harmonic series $\sum \tfrac{1}{n}$ is the standard warning: its terms tend to $0$ while its partial sums grow past every bound.

Theorem (The p-series test).

The series $\sum_{n=1}^{\infty} \dfrac{1}{n^{p}}$ converges when $p > 1$ and diverges when $p \le 1$.

The harmonic series is the borderline case $p = 1$. Plotted side by side, the partial sums of $\sum \tfrac{1}{n^{2}}$ flatten toward their limit while the harmonic sums climb past every bound.

The last test suits terms built from powers and factorials, where each term is roughly a fixed multiple of the one before.

Theorem (The ratio test).

Let $L = \lim_{n \to \infty} \left| \dfrac{a_{n+1}}{a_{n}} \right|$. If $L < 1$ the series $\sum a_{n}$ converges; if $L > 1$ it diverges; if $L = 1$ the test gives no verdict.

When the ratio test is silent, another test must decide.

Ways to work on it

Builds on

Leads to

Not sure where to start? Take the ten-question placement test.