Convergence Tests: Which Series Test to Use and When

What Are Convergence Tests?

Convergence tests are rules that decide whether an infinite series $\sum a_n$ converges (adds up to a finite number) or diverges, usually without finding the sum itself. Each test looks at the terms $a_n$, compares them with a series whose behaviour is already known, or compares the series with an integral. The nine you meet in a first calculus course are the nth-term test, the geometric series test, the $p$-series test, the integral test, the direct comparison test, the limit comparison test, the ratio test, the root test and the alternating series test.

A series converges when its sequence of partial sums settles on a limit. Write the $N$-th partial sum as

$$S_N = a_1 + a_2 + a_3 + \cdots + a_N = \sum_{n=1}^{N} a_n.$$

Then

$$\sum_{n=1}^{\infty} a_n = S \quad \text{means} \quad \lim_{N \to \infty} S_N = S,$$

where $S$ is a finite real number. If the partial sums grow without bound, oscillate, or fail to settle for any other reason, the series diverges. The ideas behind this definition, including absolute and conditional convergence, are set out on the convergence and divergence of series page; this guide is about choosing and applying the right test.

Which Convergence Test Should You Use First?

Try the tests in a fixed order, cheapest first, and stop at the first one that gives a verdict. A Quora user asks "How to know which test I should use (Gauss's test, Raabe's test, D'Alembert's test, or any other test) when I test the convergence or divergence of a series", and the honest answer is that the form of $a_n$ tells you, if you look in this order:

1. Nth-term test. Find $\lim_{n \to \infty} a_n$. If it is not 0 (or does not exist), the series diverges and you are done.

2. Known families. Is it geometric, $\sum a r^{n}$? Is it a $p$-series, $\sum \frac{1}{n^{p}}$? Each has a one-line verdict.

3. Alternating signs. If the terms alternate, $(-1)^{n} b_n$, check the alternating series test, then test $\sum b_n$ to see whether the convergence is absolute.

4. Factorials or $n$-th powers. Terms with $n$ factorial, $c^{n}$, or $n^{n}$ go to the ratio test. A whole term raised to the power $n$ goes to the root test.

5. Rational or root expressions. Terms built from powers of $n$, such as $\frac{2n+1}{n^{3}+4}$, go to the limit comparison test against a $p$-series that matches the highest powers.

6. Easy inequalities. If the term is plainly smaller than a convergent benchmark, or plainly larger than a divergent one, the direct comparison test is quickest.

7. A function you can integrate. If $a_n = f(n)$ for a positive, decreasing $f$ with an easy antiderivative (logarithms in the denominator are the classic case), use the integral test.

Paul's Online Notes, a widely used course reference, makes the same point: there is no single recipe that always works, so the order is a default, not a law. The order still saves time, because step 1 settles a surprising share of homework series in one line.

What Does Each Convergence Test Say?

Each test below lists its exact statement, the conditions it needs, the series it suits, and when it gives no answer. Throughout, $a_n$ is the $n$-th term and $L$ is a limit computed from the terms.

The Nth-Term (Divergence) Test

$$\text{If } \lim_{n \to \infty} a_n \neq 0 \text{ or the limit does not exist, then } \sum a_n \text{ diverges.}$$

If the terms do not shrink to 0, the partial sums keep receiving a non-vanishing amount and cannot settle. The divergence test can only prove divergence. A limit of 0 proves nothing, as the harmonic series shows.

The Geometric Series Test

$$\sum_{n=0}^{\infty} a r^{n} = \frac{a}{1 - r} \quad \text{when } |r| < 1, \qquad \text{and it diverges when } |r| \geq 1 \text{ (with } a \neq 0).$$

Here $a$ is the first term and $r$ is the common ratio. This is the only test in the list that also hands you the sum, which is why the geometric series works as the benchmark for the ratio and root tests.

The $p$-Series Test

$$\sum_{n=1}^{\infty} \frac{1}{n^{p}} \text{ converges if } p > 1 \text{ and diverges if } p \leq 1.$$

The case $p = 1$ is the harmonic series, the most famous divergent series with terms that go to 0. The $p$-series family is the standard comparison partner for any term built from powers of $n$.

The Integral Test

If $f$ is positive, continuous and decreasing on $[1, \infty)$ and $a_n = f(n)$, then

$$\sum_{n=1}^{\infty} a_n \text{ and } \int_{1}^{\infty} f(x),dx \text{ either both converge or both diverge.}$$

The picture is a staircase of rectangles of width 1 sitting on or under the curve $y = f(x)$, so the sum and the area are tied together. The integral test is the natural choice when $f$ has an easy antiderivative. The three conditions are not decoration: drop "decreasing" and the test can give the wrong answer.

The Direct Comparison Test

A Brainly prompt says "Use the comparison test to determine if the following series converges or diverges", and the test is short to state. If $0 \leq a_n \leq b_n$ for all large $n$, then

$$\sum b_n \text{ converges} \Rightarrow \sum a_n \text{ converges}, \qquad \sum a_n \text{ diverges} \Rightarrow \sum b_n \text{ diverges.}$$

Smaller than convergent proves convergence, and larger than divergent proves divergence. The other two directions prove nothing. The comparison test suits terms with an obvious inequality, such as $\frac{1}{n^{2}+1} < \frac{1}{n^{2}}$.

The Limit Comparison Test

If $a_n > 0$ and $b_n > 0$ and

$$\lim_{n \to \infty} \frac{a_n}{b_n} = c, \quad \text{where } 0 < c < \infty,$$

then $\sum a_n$ and $\sum b_n$ both converge or both diverge. The limit comparison test removes the need for a neat inequality. You choose $b_n$ by keeping only the highest power of $n$ in the numerator and the denominator, so $\frac{2n+1}{n^{3}+4}$ is compared with $\frac{n}{n^{3}} = \frac{1}{n^{2}}$.

The Ratio Test

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

If $L < 1$ the series converges absolutely, if $L > 1$ (or $L = \infty$) it diverges, and if $L = 1$ the test gives no answer.

A Quora reader asks, "Why does the ratio test for series convergence work?" Because the ratio test is a geometric series in disguise.

If $L < 1$, pick any number $r$ with $L < r < 1$; from some point on, each term is at most $r$ times the one before, so the tail is bounded by a convergent geometric series $a_N\left(r + r^{2} + r^{3} + \cdots\right)$. If $L > 1$, the terms eventually grow, so they cannot go to 0. When $L = 1$, the comparison with a geometric series is too blunt to decide. The ratio test is the first choice for $n$ factorial, $c^{n}$ and their products.

The Root Test

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

The verdicts match the ratio test: $L < 1$ converges absolutely, $L > 1$ diverges, $L = 1$ is inconclusive. The root test is the natural choice when the whole term is raised to the power $n$, as in $\left(\frac{n}{2n+1}\right)^{n}$. It is also the stronger of the two: whenever the ratio test settles a series, the root test does too, but not the other way round.

The Alternating Series Test

For $\sum (-1)^{n+1} b_n$ with $b_n > 0$, if

$$b_{n+1} \leq b_n \text{ for all large } n \quad \text{and} \quad \lim_{n \to \infty} b_n = 0,$$

then the series converges. The alternating series test also gives an error bound: stopping after $N$ terms leaves an error no bigger than the first omitted term, $b_{N+1}$. It never proves divergence; if $b_n$ does not tend to 0, use the nth-term test instead.

What Does Each Test Need, And When Is It Inconclusive? A Summary Table

The table puts all nine convergence tests side by side. Read across a row to see what a test needs before you use it and the case where it stays silent.

$L = 1$ is not a verdict. That one line saves more marks than any other entry in the table. The ratio and root tests return $L = 1$ for every $p$-series and every rational function of $n$, which is exactly where the comparison tests take over.

How Do You Recognise Which Test A Series Wants?

The shape of $a_n$ usually names the test before any work starts. This pattern table is the skill the flowchart assumes:

The memorizer who learns the flowchart as a chant tends to run every test in order on every series. The pattern table is faster, because one glance at $a_n$ skips most of the list.

Which Convergence Test Works When A Natural Log Is In The Denominator?

A Quora reader asks, "What convergence test would I use for an infinite series in which a natural log is in the denominator?" The answer depends on what sits next to the logarithm:

• $\frac{1}{\ln n}$ alone: direct comparison. For $n \geq 2$, $\ln n < n$, so $\frac{1}{\ln n} > \frac{1}{n}$, and the series diverges because the harmonic series does.

• $\frac{1}{n \ln n}$: the integral test. The antiderivative of $\frac{1}{x \ln x}$ is $\ln(\ln x)$, which grows without bound, so the series diverges (Example 5 works it in full).

• $\frac{1}{n (\ln n)^{2}}$: the integral test again. Now $\int_{2}^{\infty} \frac{dx}{x (\ln x)^{2}} = \frac{1}{\ln 2} \approx 1.4427$, so the series converges.

• $\frac{\ln n}{n^{2}}$ (log on top): limit comparison with $\frac{1}{n^{3/2}}$. The ratio of the two is $\frac{\ln n}{\sqrt{n}} \to 0$, and the extended limit comparison rule (limit 0, benchmark convergent) gives convergence.

Logarithms grow more slowly than any positive power of $n$, so a log rarely changes the verdict of the power it sits beside. The exception is the borderline power $n^{1}$, which is exactly where the integral test earns its place. (Cauchy's condensation test is another route for these, but it sits outside most first courses.)

What Do You Do When The Ratio Test Gives 1?

Switch tests. $L = 1$ means the terms shrink slower than any geometric series, so the answer lies in the comparison or integral family:

• For terms built from powers of $n$, use the $p$-series test or limit comparison.

• For logarithms, use the integral test.

• For alternating terms, use the alternating series test.

The finer tests named in that Quora question exist for this exact gap. Raabe's test computes

$$R = \lim_{n \to \infty} n\left(\frac{a_n}{a_{n+1}} - 1\right),$$

and the series of positive terms converges if $R > 1$ and diverges if $R < 1$. For the $p$-series, $R = p$, which recovers the $p$-series rule. Gauss's test refines Raabe's in the same spirit. These are rarely needed in a first course, but they explain why "$L = 1$" is the start of a second question, not the end of the first.

Which Tests Prove Absolute Convergence And Which Prove Conditional?

A series converges absolutely when $\sum |a_n|$ converges, and conditionally when $\sum a_n$ converges but $\sum |a_n|$ does not. The tests split cleanly:

• Absolute convergence: the ratio and root tests prove it directly, and any positive-term test (comparison, limit comparison, integral, $p$-series) applied to $|a_n|$ proves it too.

• Conditional convergence: the alternating series test is the usual source. It proves $\sum a_n$ converges, and a separate test on $\sum |a_n|$ shows whether the convergence is only conditional.

Absolute convergence implies convergence, so it is the stronger result to aim for first.

Examples Of Convergence Tests

Example 1

Does $\sum_{n=1}^{\infty} \frac{n}{n+1}$ converge or diverge?

Test choice: the term is a ratio of equal powers, so try the nth-term test first.

$$\lim_{n \to \infty} \frac{n}{n+1} = \lim_{n \to \infty} \frac{1}{1 + \frac{1}{n}} = 1$$

$$1 \neq 0$$

Final answer: the series diverges by the nth-term test.

Example 2

Does $\sum_{n=1}^{\infty} \frac{1}{n^{2}}$ converge or diverge?

A tempting first attempt uses the ratio test:

$$L = \lim_{n \to \infty} \frac{\frac{1}{(n+1)^{2}}}{\frac{1}{n^{2}}} = \lim_{n \to \infty} \left(\frac{n}{n+1}\right)^{2} = 1$$

"$L$ is not less than 1, so the series diverges."

That conclusion breaks at once. The same calculation for $\sum \frac{1}{n}$ also gives $L = 1$, and the partial sums of $\sum \frac{1}{n^{2}}$ never pass 2 (each $S_N < 2 - \frac{1}{N}$ for $N \geq 2$). A test that returns the same number for a divergent series and a bounded one cannot be deciding anything. $L = 1$ is the inconclusive case, not a verdict of divergence.

The rescue is the test built for this form. The term is $\frac{1}{n^{p}}$ with $p = 2$:

$$p = 2 > 1$$

Final answer: the series converges by the $p$-series test. Its sum is $\frac{\pi^{2}}{6} \approx 1.6449$, a value no convergence test supplies (Euler found it by other means).

Example 3

Does $\sum_{n=1}^{\infty} \frac{2n+1}{n^{3}+4}$ converge or diverge?

Test choice: powers of $n$ only, so use limit comparison. Keep the highest powers: $\frac{2n}{n^{3}} = \frac{2}{n^{2}}$, so compare with $b_n = \frac{1}{n^{2}}$.

$$\lim_{n \to \infty} \frac{a_n}{b_n} = \lim_{n \to \infty} \frac{(2n+1),n^{2}}{n^{3}+4}$$

$$= \lim_{n \to \infty} \frac{2n^{3} + n^{2}}{n^{3} + 4} = 2$$

$$0 < 2 < \infty, \quad \text{and } \sum \frac{1}{n^{2}} \text{ converges } (p = 2)$$

Final answer: the series converges by the limit comparison test.

Students who reach for the ratio test on a rational term like this one get $L = 1$ every time and stall; the habit that fixes it is reading off the highest powers before choosing a test, which points straight to a $p$-series partner.

Example 4

Does $\sum_{n=1}^{\infty} \frac{n^{2}}{3^{n}}$ converge or diverge?

Test choice: an exponential $3^{n}$ in the term, so use the ratio test.

$$\frac{a_{n+1}}{a_n} = \frac{(n+1)^{2}}{3^{n+1}} \cdot \frac{3^{n}}{n^{2}} = \frac{1}{3}\left(\frac{n+1}{n}\right)^{2}$$

$$L = \lim_{n \to \infty} \frac{1}{3}\left(1 + \frac{1}{n}\right)^{2} = \frac{1}{3}$$

$$\frac{1}{3} < 1$$

Final answer: the series converges (absolutely) by the ratio test. Its sum happens to be $\frac{3}{2}$.

Example 5

Does $\sum_{n=2}^{\infty} \frac{1}{n \ln n}$ converge or diverge?

Test choice: a logarithm next to $n$ in the denominator, so use the integral test with $f(x) = \frac{1}{x \ln x}$.

Check the conditions: on $[2, \infty)$, $f$ is positive and continuous, and it is decreasing because $x \ln x$ increases.

$$\int_{2}^{\infty} \frac{dx}{x \ln x} = \lim_{b \to \infty} \Big[\ln(\ln x)\Big]_{2}^{b}$$

$$= \lim_{b \to \infty} \left(\ln(\ln b) - \ln(\ln 2)\right) = \infty$$

Final answer: the series diverges by the integral test. It diverges very slowly: the first 999 terms (from $n = 2$ to $n = 1000$) add to only about 2.73.

Example 6

Does $\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{\sqrt{n}}$ converge absolutely, conditionally, or not at all?

Test choice: the sign alternates, so use the alternating series test with $b_n = \frac{1}{\sqrt{n}}$.

$$b_{n+1} = \frac{1}{\sqrt{n+1}} < \frac{1}{\sqrt{n}} = b_n$$

$$\lim_{n \to \infty} \frac{1}{\sqrt{n}} = 0$$

So the series converges. Now test the absolute values:

$$\sum_{n=1}^{\infty} \left|\frac{(-1)^{n+1}}{\sqrt{n}}\right| = \sum_{n=1}^{\infty} \frac{1}{n^{1/2}}, \quad p = \frac{1}{2} \leq 1$$

which diverges.

Final answer: the series converges conditionally (its sum is about 0.6049).

Why Do Convergence Tests Matter?

"You cannot add infinitely many numbers one at a time." That is the whole reason the tests exist. A computer can add a million terms, but a million terms of the harmonic series give only about 14.39, which looks like a series settling down when it is in fact heading to infinity. Partial sums alone mislead:

• Most series have no quick closed-form sum. The sum of $\frac{1}{n^{2}+1}$ does have one, $\frac{\pi \coth \pi - 1}{2} \approx 1.0767$, but finding it takes complex analysis, while the comparison test proves convergence in one line.

• Slow divergence hides. The series $\sum \frac{1}{n \ln n}$ crawls from about 2.73 at $n = 1000$ upward at a rate no table of values would flag.

• Power series need a guarantee. Calculators evaluate $\sin x$, $e^{x}$ and $\ln x$ from power series, and those series can only be trusted inside their interval of convergence, found with the ratio or root test.

• Rigour came late, and it came from this problem. Augustin-Louis Cauchy, in his 1821 Cours d'analyse, was the first to make a rigorous study of the conditions under which infinite series converge, according to his MacTutor biography. Before that, mathematicians often manipulated divergent series as if they had sums.

So a test is a guarantee bought in advance: once it says "converges", every later step that uses the sum is standing on solid ground.

Where Are Convergence Tests Used In Real Life?

Convergence tests are used wherever an answer is computed as an infinite sum and someone needs to know the sum is real before relying on it:

• Calculators and software libraries compute functions from truncated power series, and the ratio test fixes where each series can be used.

• Physics writes approximations as series in a small parameter, and the series is only trusted where it converges.

• Finance values a perpetuity (a payment that continues forever, discounted each year) as a geometric series, which has a finite value only when the discount ratio is below 1.

• Signal processing builds sounds and images from Fourier series, whose convergence is studied with refinements of these tests.

• Computer science bounds the running time of algorithms with sums whose growth is judged by the integral test.

Where Does Intuition Break With Convergence Tests?

Mistake 1: Reading "Terms Go To 0" As "Converges"

Where it slips in: Any series whose terms shrink, especially $\sum \frac{1}{n}$ and $\sum \frac{1}{\sqrt{n}}$.

Don't do this: Compute $\lim a_n = 0$ and write "converges by the nth-term test".

The correct way: Treat a limit of 0 as "no verdict yet" and move to the next test in the decision order. The nth-term test only ever proves divergence.

The rusher lands here most often. The first instinct after seeing $\frac{1}{n} \to 0$ is to stop and declare convergence, and the habit that fixes it is writing the test's conclusion in full ("if the limit is not 0, then it diverges") before using it.

Mistake 2: Treating $L = 1$ As A Verdict

Where it slips in: The ratio or root test applied to $p$-series, rational terms and anything with only powers of $n$.

Don't do this: Conclude "diverges" (or "converges") because the ratio test returned exactly 1.

The correct way: Record "inconclusive", then switch to limit comparison, the $p$-series test or the integral test, as Example 2 shows.

Mistake 3: Comparing In The Wrong Direction

Where it slips in: Direct comparison with a known series, such as showing $\frac{1}{n^{2} - 1}$ against $\frac{1}{n^{2}}$.

Don't do this: Note that $\frac{1}{n^{2}-1} > \frac{1}{n^{2}}$ and conclude convergence because $\sum \frac{1}{n^{2}}$ converges.

The correct way: Bigger than convergent proves nothing. Use limit comparison instead: $\frac{1/(n^{2}-1)}{1/n^{2}} \to 1$, so both series converge (starting from $n = 2$).

Mistake 4: Forgetting That Convergence Says Nothing About Accuracy

Where it slips in: Questions that ask for an approximate sum, and any use of a few partial sums as "the answer".

Don't do this: Prove convergence, add the first few terms, and treat the result as the sum.

The correct way: Pair the test with an error bound. For an alternating series, the error after $N$ terms is at most $b_{N+1}$; the first 10 terms of $1 - \frac{1}{2} + \frac{1}{3} - \cdots$ give $0.6456$, while the true sum is $\ln 2 \approx 0.6931$.

The second-guesser often does all the hard work here and then distrusts it, adding more and more terms by hand to "check" a series the test has already settled. The error bound is what ends that loop: it says how many terms are enough.

Practice Problems On Convergence Tests

Name the test you would use first, then decide.

1. $\sum_{n=1}^{\infty} \frac{3}{2^{n}}$

2. $\sum_{n=1}^{\infty} \left(\frac{n}{2n+1}\right)^{n}$

3. $\sum_{n=1}^{\infty} \frac{1}{n^{2}+1}$

4. $\sum_{n=1}^{\infty} \frac{2^{n}}{1 \cdot 2 \cdot 3 \cdots n}$ (the denominator is $n$ factorial)

5. $\sum_{n=1}^{\infty} (-1)^{n} \frac{n}{n+1}$

6. $\sum_{n=1}^{\infty} \frac{1}{n^{1.01}}$

Answer to Question 1: Geometric with $r = \frac{1}{2}$; converges, sum $3$.

Answer to Question 2: Root test, $L = \frac{1}{2} < 1$; converges.

Answer to Question 3: Direct comparison with $\frac{1}{n^{2}}$; converges (sum about $1.0767$).

Answer to Question 4: Ratio test, $L = \lim \frac{2}{n+1} = 0 < 1$; converges (sum $e^{2} - 1 \approx 6.3891$).

Answer to Question 5: Nth-term test, $\left|a_n\right| \to 1$, so the terms do not go to 0; diverges.

Answer to Question 6: $p$-series with $p = 1.01 > 1$; converges, although the sum is large (about $100.58$).

Bottom Line

• Convergence tests decide whether an infinite series has a finite sum, usually without computing that sum.

• Start with the nth-term test, then match the form of $a_n$: geometric, $p$-series, alternating, ratio or root, limit comparison, direct comparison, integral.

• A limit of 0 in the nth-term test, or $L = 1$ in the ratio or root test, is never a verdict.

• Direct comparison only works below a convergent series or above a divergent one.

• Convergence proves a sum exists, and an error bound says how many terms you need.

A Practical Next Step

Work through Practice Problems 1 to 6, naming the test before you calculate anything; if Problem 2 or 4 trips you up, return to the ratio and root test subsections, and if Problem 3 does, reread Mistake 3 on comparison direction. Want a live Bhanzu trainer to walk through more convergence tests problems? Book a free demo class.

Read More

• Infinite series and their partial sums

• Alternating series and their behaviour

• Power series and where they converge

• Radius of convergence of a power series

• Taylor series for approximating functions

• Improper integrals behind the integral test