To decide whether an infinite series converges, first check that its terms go to 0 (if not, it diverges), then look at the shape of the terms: geometric and p-series you can read off, factorials point to the ratio test, nth powers to the root test, rational expressions to limit comparison with a p-series, alternating signs to the alternating series test. The flowchart below puts those checks in order, and the seven worked examples show each branch in use, including the cases where a test tells you nothing.
The flowchart, as eight questions
Work down the list and stop at the first question that settles the series. Below, a(n) is the nth term.
- Does a(n) → 0? If the limit is not 0 (or doesn't exist), the series diverges. Done. If it is 0, keep going: this test can never prove convergence.
- Is it geometric, ∑ a·rⁿ? It converges exactly when |r| < 1, to (first term)/(1 − r).
- Is it a p-series, ∑ 1/nᵖ? It converges exactly when p > 1.
- Does it telescope? Split into partial fractions and look for cancelling terms; the partial sums then give the exact sum.
- Are the signs alternating? Use the alternating series test (terms decreasing in size, eventually, and tending to 0). Then test ∑ |a(n)| separately to decide absolute or conditional convergence.
- Are there factorials, or a constant to the power n? Use the ratio test. If the whole term is something to the power n, the root test is usually quicker.
- Is it a ratio of powers of n (polynomials, roots)? Use limit comparison with ∑ 1/nᵖ, where p is (degree of bottom) − (degree of top).
- Is a(n) = f(n) for a positive, decreasing f that you can integrate? Use the integral test. This is the branch for logs, as in 1/(n (ln n)²).
If a bounded factor such as sin n, cos n or (−1)ⁿ sits on top of an otherwise simple term, a direct comparison usually ends the question fastest.
Series and convergence are the hardest unit of Calculus II for many students we work with; our college maths tutoring page for US students explains how online sessions with Ajay Vatsyayan support that unit.
Cheat sheet: each test and its blind spot
| test | use when the terms look like | conclusion | blind spot |
|---|---|---|---|
| divergence (nth term) | anything; always check first | lim a(n) ≠ 0 means diverges | lim = 0 tells you nothing |
| geometric | a·rⁿ | converges if and only if r is between −1 and 1 | none |
| p-series | 1/nᵖ | converges if and only if p > 1 | none |
| direct comparison | a term you can bound above or below | smaller than a convergent series converges; bigger than a divergent one diverges | the inequality must point the useful way |
| limit comparison | rational or algebraic in n | same behaviour as ∑ b(n) if lim a(n)/b(n) is finite and positive | needs positive terms |
| ratio | factorials, kⁿ, products | limit L < 1 converges, L > 1 diverges | L = 1: no information (every p-series gives 1) |
| root | (expression)ⁿ | limit L < 1 converges, L > 1 diverges | L = 1: no information |
| integral | f(n) with f positive, continuous, decreasing | series and integral behave the same | f must be eventually decreasing |
| alternating series | (−1)ⁿ b(n) | converges if b(n) decreases to 0 | says nothing about absolute convergence |
Worked examples, one per branch
Example 1: divergence test
∑ (3n² + 1)/(5n² − n)
The terms tend to 3/5, not 0, so the series diverges. No other test is needed. Students sometimes spend ten minutes on a ratio test here (which gives 1 and decides nothing).
Example 2: limit comparison
∑ (2n + 7)/(n³ + 4n)
Top degree 1, bottom degree 3, so compare with b(n) = 1/n². Then
a(n)/b(n) = (2n³ + 7n²)/(n³ + 4n) → 2
The limit is finite and positive and ∑ 1/n² converges (p = 2), so the series converges.
Example 3: ratio test with a factorial
∑ n³ 2ⁿ / n!
a(n+1)/a(n) = [(n + 1)³ 2ⁿ⁺¹ / (n + 1)!] × [n! / (n³ 2ⁿ)] = ((n + 1)/n)³ × 2/(n + 1)
The first factor tends to 1 and the second to 0, so the limit is 0 < 1. The series converges (absolutely). Factorials grow faster than any power or exponential, which is why the limit is 0.
Example 4: root test
∑ ((2n + 1)/(3n − 2))ⁿ
The nth root of a(n) is (2n + 1)/(3n − 2), which tends to 2/3 < 1. The series converges. A ratio test would work too, but with much messier algebra.
Example 5: integral test
∑ from n = 2 of 1/(n (ln n)²)
f(x) = 1/(x (ln x)²) is positive and decreasing for x ≥ 2. Substitute w = ln x:
∫₂^∞ dx/(x (ln x)²) = [−1/ln x]₂^∞ = 1/ln 2 ≈ 1.443
The integral is finite, so the series converges. (The number 1.443 is the value of the integral, not the sum of the series.) Compare: ∑ 1/(n ln n) diverges, because its integral is ln(ln x), which grows without bound.
Example 6: alternating, then absolute or conditional
∑ (−1)ⁿ n/(n² + 3)
Let b(n) = n/(n² + 3). It tends to 0. Is it decreasing? The derivative of x/(x² + 3) is (3 − x²)/(x² + 3)², negative for x > √3, so b(n) decreases from n = 2 onwards. (It does not decrease at the start: b(1) = 1/4 but b(2) = 2/7.) "Eventually decreasing" is enough, so the series converges by the alternating series test.
Now the absolute series ∑ n/(n² + 3). Limit comparison with 1/n gives limit 1, and ∑ 1/n diverges. So the series is conditionally convergent: it converges, but not absolutely.
Example 7: direct comparison with a bounded factor
∑ (2 + cos n)/n^(3/2) and ∑ (2 + cos n)/n
Since 1 ≤ 2 + cos n ≤ 3, the first series has terms at most 3/n^(3/2), a convergent p-series (p = 3/2), so it converges. The second has terms at least 1/n, and ∑ 1/n diverges, so it diverges. Same numerator, opposite answers: the comparison has to point the right way.
Absolute convergence and why it matters
A series converges absolutely if ∑ |a(n)| converges. Absolute convergence implies convergence, and an absolutely convergent series can be rearranged without changing its sum. A conditionally convergent series cannot: its terms can be reordered to add up to a different number. That is why exam questions so often ask "absolutely, conditionally, or divergent?" rather than just "does it converge?".
The ratio and root tests, when they give L < 1, prove absolute convergence directly.
Estimating the sum, not just yes or no
For an alternating series that passes the test, the error after stopping at the nth term is less than the next term's size. Take ∑ (−1)ⁿ⁺¹/n³:
S₄ = 1 − 1/8 + 1/27 − 1/64 = 1549/1728 ≈ 0.89641
The error is less than 1/5³ = 0.008. The true sum is about 0.90154 (it equals ¾ζ(3)), so the actual error is about 0.0051, inside the bound. Questions of the form "how many terms for error below 0.001?" use the same idea: find the first n with b(n+1) < 0.001.
Where marks are lost
- Using the divergence test to show convergence. a(n) → 0 is necessary, not sufficient: ∑ 1/n has terms tending to 0 and diverges.
- Ratio test on rational terms. For anything like (2n + 7)/(n³ + 4n) the ratio limit is 1. Go straight to limit comparison.
- Comparing the wrong way. Bounding a series above by a divergent one, or below by a convergent one, proves nothing (example 7).
- Skipping the decreasing check in the alternating test, or the positivity and decreasing conditions in the integral test. Write them down; markers look for them.
- Stopping at "converges" for an alternating series when the question asks absolute or conditional.
- Geometric sum with the wrong first term. ∑ from n = 1 of 4(−2/5)ⁿ⁻¹ starts at 4, so the sum is 4/(1 + 2/5) = 20/7. Check the first term by putting in the starting index.
- Confusing the integral's value with the series' sum. The integral test decides convergence only.
Practice set
Decide convergence, and for alternating series say absolute or conditional.
- ∑ n/√(n⁵ + 2)
- ∑ (−1)ⁿ/√(n + 1)
- ∑ 3ⁿ/(n² 2ⁿ)
- ∑ (n/(2n + 5))ⁿ
- ∑ from n = 1 of (2ⁿ + 5)/3ⁿ (find the sum)
- ∑ from n = 2 of 1/(n √(ln n))
Answers:
- Converges: limit comparison with 1/n^(3/2) gives limit 1.
- Conditionally convergent: the alternating test passes, but ∑ 1/√(n + 1) diverges (compare with p = 1/2).
- Diverges: the ratio limit is 3/2 (the terms actually grow without bound).
- Converges: the root test gives 1/2.
- Converges, sum 2 + 5/2 = 9/2 (two geometric series with r = 2/3 and r = 1/3).
- Diverges: the integral is 2√(ln x), which is unbounded.
Every limit, integral and sum here was checked in Python (sympy).
Where series lead next
Convergence tests are the entry ticket to power series: the ratio test finds the radius of convergence, and endpoint checks use the tests above. Taylor and Maclaurin series usually come next, then Fourier series in engineering and differential equations courses (our Fourier series guide shows how convergence questions reappear there). In real analysis these tests are proved from the ε-N definition; our post on the epsilon-delta definition of a limit explains the quantifier style involved.
Two Calculus II skills feed this unit: the integral test needs integration by parts and the tabular method, and limits of expressions like (2n + 1)/(3n − 2) are Calculus I limits. If you studied series in IB Maths, the IB Maths AA HL vs AP Calculus BC comparison shows where your background lines up.
How Maths Bodhi helps
The difficulty with series is choosing the test, not running it. In online sessions we give mixed sets with no headings, ask you to name the test before you start and then review your justification line by line. We explain and review your own work; we don't complete graded work. The university maths tutoring hub lists the later courses where series return.
Home tuition is available in Gurgaon (Gurugram). Students elsewhere in India and abroad learn with us online. You can book a free demo class.
Written and checked by the Maths Bodhi team (November 2026).
Frequently asked questions
- Which convergence test should I use first?
- Always start with the divergence test: if the terms don't go to 0, the series diverges and you're done.
- When do you use the ratio test vs the root test?
- Use the ratio test for factorials and products, and the root test when the whole term is raised to the power n.
- What does it mean if the ratio test equals 1?
- The test is inconclusive; switch to comparison, limit comparison or the integral test.
- What is the difference between absolute and conditional convergence?
- Absolute means ∑ |a(n)| converges; conditional means the series converges but ∑ |a(n)| diverges.
- Does the integral test give the sum of the series?
- No, it only tells you whether the series converges; the integral's value is generally different from the sum.
- Is there a series convergence tests cheat sheet for exams?
- Rules on notes vary by course, so learn the flowchart order and each test's blind spot from the table above until you can choose without it.
