Calculus › Section J · Infinite series › Topic J.6

The ratio and root tests

Factorials and exponentials make comparison awkward, but they shrink in a very regular way. The ratio and root tests measure that shrinkage with a single limit and give the cleanest verdicts in the subject.

Measuring how fast terms shrink

A series converges when its terms die away fast enough. For p-series you measured the speed against a benchmark, but for series like \(\sum \frac{n^2}{3^n}\) or \(\sum \frac{7^n}{n!}\) the comparison algebra gets ugly, and factorials barely respond to the integral test at all. These series, though, have a strength: each term is a fairly regular multiple of the one before. The ratio test measures exactly that: how much do the terms shrink, term to term? One limit answers the question without any need to sum anything or find a benchmark series.

The ratio test

For a series \(\sum a_n\) with nonzero terms, compute

L = \lim_{n \to \infty} \left|\frac{a_{n+1}}{a_n}\right|

Then: if \(L < 1\) the series converges absolutely; if \(L > 1\), or the ratio grows without bound, the series diverges; and if \(L = 1\) the test is silent, meaning it gives no verdict and you must try something else. The absolute values mean the test works even for series with mixed signs, and the verdict \(L < 1\) is the strong one: absolute convergence. When \(L > 1\) the terms themselves fail to approach 0, since each is eventually a fixed factor larger than its predecessor, so the divergence is instant.

Why the ratio test works

Suppose the ratio settles below some number \(r < 1\), say the ratio is at most \(r\) for every \(n\) beyond \(N\). Then each term from \(N\) onward is at most \(r\) times the one before, so the tail satisfies \(|a_{N+k}| \le |a_N|\, r^k\). The whole tail is therefore bounded by a geometric series with ratio \(r\):

\sum_{k=0}^{\infty} |a_N| r^k = \frac{|a_N|}{1 - r}

a finite number. A convergent ceiling of geometric series pins the tail down, so the original series converges absolutely. The test is a comparison test in disguise, with the geometric series as its built-in benchmark, chosen automatically by the data.

Worked example: powers and factorials

First, \(\sum \frac{n^2}{3^n}\). The ratio is \(\left|\frac{a_{n+1}}{a_n}\right| = \frac{(n+1)^2}{3^{n+1}} \cdot \frac{3^n}{n^2} = \frac{1}{3}\left(\frac{n+1}{n}\right)^2 \to \frac{1}{3}\). Since \(\frac{1}{3} < 1\), the series converges absolutely. Second, \(\sum \frac{7^n}{n!}\). The ratio is \(\frac{7^{n+1}}{(n+1)!} \cdot \frac{n!}{7^n} = \frac{7}{n+1}\), which shrinks to 0. Any \(L < 1\) converges, and \(L = 0\) is the strongest possible verdict: factorials crush exponentials eventually, because dividing a factorial by its successor cancels almost everything. In the other direction, \(\sum \frac{n!}{2^n}\) has ratio \(\frac{n+1}{2}\), which grows past every bound, so it diverges.

The root test

The root test measures shrinkage from a different angle: how large is the typical term when spread over \(n\) factors? Compute

L = \lim_{n \to \infty} \sqrt[n]{|a_n|}

with exactly the same thresholds: \(L < 1\) converges absolutely, \(L > 1\) diverges, \(L = 1\) silent. The root test shines when the whole term is raised to the \(n\)th power, because the \(n\)th root undoes it. For \(\sum \left(\frac{n}{2n+1}\right)^n\), the root is \(\frac{n}{2n+1} \to \frac{1}{2} < 1\), so the series converges. The ratio test on the same series would wallow in binomial expansions; the root test finishes it in one line. When a series suits both tests, they always agree.

Worked example: a series with mixed signs

Does \(\sum (-1)^{n+1} \frac{4^n}{n^{2n}}\) converge? The signs alternate, so the alternating series test is a candidate, but checking that the sizes decrease means comparing \(4^{n+1}n^{2n}\) with \(4^n(n+1)^{2n+2}\), which is unpleasant. The root test sidesteps signs entirely:

\sqrt[n]{\left|(-1)^{n+1} \frac{4^n}{n^{2n}}\right|} = \frac{4}{n^2} \to 0

Since \(L = 0 < 1\), the series converges absolutely, which is stronger than what the alternating test would have given and settles any rearrangement questions in passing. When a term combines an \(n\)th power with a factorial or a sign pattern, try the root test first.

The dead zone at L = 1

Both tests fail in the same place: whenever the limit comes out 1, there is no verdict. This is not a defect but a genuine boundary, because the p-series all sit exactly there. For \(\sum \frac{1}{n^2}\), the ratio is \(\left(\frac{n}{n+1}\right)^2 \to 1\) and the root is \(\frac{1}{\sqrt[n]{n^2}} \to 1\), yet \(\sum \frac{1}{n^2}\) converges. For \(\sum \frac{1}{n}\) the same limits come out 1, and it diverges. Series at the border between dying fast enough and not can only be separated by other tools: the integral test, the comparison tests, or the alternating series test. A practical warning follows: reaching for the ratio test on anything that looks like a p-series wastes time, since the answer is always \(L = 1\).

Choosing which test

Use the shape of the terms as the guide. Factorials anywhere, or exponentials \(c^n\) sitting over polynomials: ratio test, because the ratios cancel cleanly, as in \(\frac{7}{n+1}\) or \(\frac{1}{3}(\frac{n+1}{n})^2\). A whole expression raised to the \(n\)th power: root test, because the root undoes it in one step. Anything built from powers of \(n\) alone: skip both, because \(L = 1\) is guaranteed, and go to comparison or the integral test. And always start with the divergence test: if the terms do not approach 0, no ratio or root limit is needed.

Common mistakes to avoid

The most expensive habit is treating \(L = 1\) as a verdict, usually as divergence: it is silence, not a no. Second, do not conclude convergence from the ratio of the first few terms; the limit is what matters, and early terms can mislead. Third, remember the ratio test needs the absolute values: \(\frac{a_{n+1}}{a_n}\) without them can converge to a negative number and misdirect you on series with alternating signs. Fourth, \(L > 1\) means divergence, not oscillation: the terms are growing, so the series cannot converge, whatever its signs. Fifth, avoid the root test on factorials: \(\sqrt[n]{n!}\) does have a limit, but computing it needs tools this course has not built, while the ratio is one line.

Key takeaways

The ratio test computes \(L = \lim \left|\frac{a_{n+1}}{a_n}\right|\) and the root test computes \(L = \lim \sqrt[n]{|a_n|}\); both give absolute convergence for \(L < 1\), divergence for \(L > 1\), and silence for \(L = 1\). The ratio test suits factorials and exponentials over polynomials; the root test suits terms raised to the \(n\)th power. Both work through mixed signs, and \(L < 1\) even settles rearrangement questions by giving absolute convergence. The dead zone \(L = 1\) contains every p-series, so series built from powers of \(n\) need comparison, the integral test, or the alternating series test instead.

Flashcards

Flashcard

Ratio test

L = lim |aₙ₊₁/aₙ|

Measures term-to-term shrinkage.

Test your understanding

Question 1 of 5

For the series Σ n²/3ⁿ, what is lim |aₙ₊₁/aₙ|?

Question 2 of 5

What does the ratio test say about Σ n!/2ⁿ?

Question 3 of 5

What does the ratio test conclude for the p-series Σ 1/n²?

Question 4 of 5

Using the root test on Σ (n/(2n+1))ⁿ, what is the verdict?

Question 5 of 5

Why does Σ 7ⁿ/n! converge?

Standards: 10 (BC)

← Previous topicNext topic →