Choosing a Convergence Test
Most series cannot be summed exactly, but we can still decide whether they converge. Each test is a tool for a particular shape of series, and choosing the right one is the real skill.
Why tests?
For geometric and telescoping series we can find the exact sum. For most series, like \(\sum \frac{1}{n^2}\) or \(\sum \frac{n}{2^n}\), finding the exact sum is hard or impossible, but knowing whether it converges is still essential, for example to know whether an approximation method is valid. Convergence tests answer that yes-or-no question.
The integral test
If \(f\) is positive, continuous, and decreasing for \(x \ge 1\), with \(a_n = f(n)\), then \(\sum a_n\) and \(\int_1^{\infty} f(x)\,dx\) either both converge or both diverge. The terms of the series are like rectangles under the curve. This connects series to the improper integrals you already know.
\sum a_n \sim \int_1^{\infty} f(x)\,dxp-series
Applying the integral test to \(\sum \frac{1}{n^p}\) gives the most useful benchmark of all: a p-series converges if \(p > 1\) and diverges if \(p \le 1\). So \(\sum \frac{1}{n^2}\) converges (its sum is \(\frac{\pi^2}{6}\), a famous result of Euler), while \(\sum \frac{1}{\sqrt{n}}\) diverges.
\sum \frac{1}{n^p} \text{ converges} \iff p > 1The comparison test
For positive series, compare with something known. If \(0 \le a_n \le b_n\) and \(\sum b_n\) converges, so does \(\sum a_n\): smaller than finite is finite. If \(a_n \ge b_n \ge 0\) and \(\sum b_n\) diverges, so does \(\sum a_n\). For example, \(\frac{1}{n^2 + 5} < \frac{1}{n^2}\), so \(\sum \frac{1}{n^2 + 5}\) converges.
The limit comparison test
Inequalities can be awkward to arrange. Instead, if \(a_n, b_n > 0\) and \(\lim \frac{a_n}{b_n} = c\) with \(0 < c < \infty\), then both series behave the same way. Choose \(b_n\) by keeping only the dominant terms: \(\frac{3n + 1}{n^3 - 2}\) behaves like \(\frac{3n}{n^3} = \frac{3}{n^2}\), so compare with \(\frac{1}{n^2}\): it converges.
\lim \frac{a_n}{b_n} = c,\ 0 < c < \inftyAlternating series
A series whose signs alternate, like \(1 - \frac{1}{2} + \frac{1}{3} - \frac{1}{4} + \cdots\), converges if the sizes of the terms decrease to \(0\). This is the alternating series test. The alternating harmonic series converges (to \(\ln 2\)) even though the harmonic series diverges: the cancellation between positive and negative terms makes the difference. The error after stopping is at most the size of the first omitted term.
Absolute and conditional convergence
If \(\sum |a_n|\) converges, the series converges absolutely, and it converges regardless of signs. If \(\sum a_n\) converges but \(\sum |a_n|\) does not, it converges conditionally, like the alternating harmonic series. Absolutely convergent series behave well: you can rearrange their terms freely. Conditionally convergent ones can be rearranged to give any sum, a surprising result of Riemann.
The ratio test
Compute \(L = \lim \left|\frac{a_{n+1}}{a_n}\right|\). If \(L < 1\), the series converges absolutely; if \(L > 1\), it diverges; if \(L = 1\), the test gives no information. It is ideal for factorials and exponentials. For \(\sum \frac{2^n}{n!}\): \(\frac{a_{n+1}}{a_n} = \frac{2}{n+1} \to 0\), so it converges.
L = \lim_{n\to\infty}\left|\frac{a_{n+1}}{a_n}\right|The root test
Compute \(L = \lim \sqrt[n]{|a_n|}\), with the same conclusions as the ratio test. It works best when \(a_n\) is something raised to the \(n\)th power. For \(\sum \left(\frac{n}{2n + 1}\right)^n\), the \(n\)th root is \(\frac{n}{2n + 1} \to \frac{1}{2} < 1\), so the series converges.
A strategy for choosing a test
1. Do the terms go to \(0\)? If not, it diverges. 2. Is it geometric or a p-series? Use what you know. 3. Does it look like a known series after dropping small terms? Use comparison or limit comparison. 4. Does it alternate? Use the alternating series test. 5. Factorials or exponentials? Use the ratio test. 6. \(n\)th powers? Use the root test. 7. Easy to integrate? Use the integral test.
Common mistakes
Concluding convergence from the divergence test. Applying the ratio test to p-series, where it always gives \(L = 1\). Using comparison with the inequality in the wrong direction. Forgetting that the alternating series test requires the terms' sizes to decrease to \(0\). Confusing conditional with absolute convergence.
Try it yourself
Does \(\sum \frac{n^2}{3^n}\) converge? Answer: ratio test, \(\frac{(n+1)^2}{3n^2} \to \frac{1}{3} < 1\), so yes. Does \(\sum \frac{1}{2n + 7}\) converge? Answer: limit comparison with \(\frac{1}{n}\) gives \(\frac{1}{2}\), so it diverges like the harmonic series.
Key takeaways
p-series converge exactly when \(p > 1\). Compare positive series with known ones. Alternating series converge when term sizes decrease to \(0\). Ratio and root tests handle factorials, exponentials, and \(n\)th powers, but fail when \(L = 1\). Start every problem with the divergence test.
Flashcard
Series and integral of a positive decreasing f behave alike
Check your understanding
Which test is most natural for the series sum of n!/10^n?
