Calculus · Grade 12 and undergraduate

Alternating Series and Absolute Convergence

Quick answer

An alternating series switches sign from term to term. If the sizes of its terms shrink to 0, it converges: the partial sums step back and forth across the sum, each step shorter than the last. Stopping after n terms leaves an error no bigger than the first term left out. A series converges absolutely if the series of absolute values converges, and absolute convergence implies convergence. The alternating harmonic series converges but not absolutely, which is called conditional convergence.

What you'll learn

  • Apply the alternating series test
  • Bound the error of a partial sum of an alternating series
  • Classify a series as absolutely convergent, conditionally convergent or divergent
  • Use absolute convergence for series with irregular signs

Series whose signs alternate

The harmonic series 1+12+13+⋯1 + \tfrac{1}{2} + \tfrac{1}{3} + \cdots diverges. Flip every other sign and something changes:

1−12+13−14+⋯1 - \frac{1}{2} + \frac{1}{3} - \frac{1}{4} + \cdots

An alternating series has the form ∑(−1)nbn\sum (-1)^n b_n or ∑(−1)n+1bn\sum (-1)^{n+1} b_n with every bn>0b_n > 0.

The alternating series test

If bn>0b_n > 0, the bnb_n are decreasing, and lim⁡n→∞bn=0\lim_{n \to \infty} b_n = 0, then ∑(−1)n+1bn\sum (-1)^{n+1} b_n converges.

Why the partial sums close in

Follow the partial sums of 1−12+13−⋯1 - \tfrac{1}{2} + \tfrac{1}{3} - \cdots. Each step goes the opposite way from the last, and each is shorter. So the partial sums zigzag, and every zigzag is narrower than the one before.

Partial sums of 1 − 1/2 + 1/3 − 1/4 + ⋯ Dots at the partial sums for n = 1 to 8, joined in order: 1, 0.5, 0.83, 0.58, 0.78, 0.62, 0.76, 0.63. They zigzag above and below the dashed line at ln 2 ≈ 0.693, each swing smaller than the last. 24681nS
  • the sum, ln 2
Partial sums of 1 − 1/2 + 1/3 − 1/4 + ⋯

The odd partial sums come down, the even ones go up, and each odd one stays above every even one. With the step lengths bnb_n shrinking to 00, the two sides squeeze together on a single number. An alternating series converges when its steps shrink to zero, because each step undoes part of the last. That number, here, is ln⁡2\ln 2.

The error bound

The squeeze also measures accuracy. The sum always lies between two consecutive partial sums, so stopping after nn terms leaves an error smaller than the next step:

∣S−Sn∣≤bn+1\lvert S - S_n \rvert \le b_{n+1}

The sign of the error is known too: if the last term used was positive, SnS_n is too big.

Absolute and conditional convergence

For a series with terms of either sign, look at the sizes alone. If ∑∣an∣\sum \lvert a_n \rvert converges, the series converges absolutely, and an absolutely convergent series always converges: the signs can only cancel, not add. A series that converges while ∑∣an∣\sum \lvert a_n \rvert diverges converges conditionally. It converges only because of the cancellation.

SeriesAbsolute valuesVerdict
∑(−1)nn2\sum \tfrac{(-1)^n}{n^2}∑1n2\sum \tfrac{1}{n^2} convergesconverges absolutely
∑(−1)nn\sum \tfrac{(-1)^n}{n}∑1n\sum \tfrac{1}{n} divergesconverges conditionally
∑(−1)nnn+1\sum (-1)^n \tfrac{n}{n + 1}terms do not approach 00diverges

Worked examples

Common mistakes

Practice problems

  1. Classify ∑n=1∞(−1)nn+1\displaystyle\sum_{n=1}^{\infty} \frac{(-1)^n}{n + 1}.

    Answer

    Converges conditionally

    Full solution

    1n+1\tfrac{1}{n + 1} decreases to 00, so the series converges. The absolute values form a shifted harmonic series, which diverges.

  2. Classify ∑n=1∞(−1)nn3\displaystyle\sum_{n=1}^{\infty} \frac{(-1)^n}{n^3}.

    Answer

    Converges absolutely

    Full solution

    ∑1n3\sum \tfrac{1}{n^3} converges, a pp-series with p=3p = 3.

  3. Classify ∑n=1∞(−1)nn2n+1\displaystyle\sum_{n=1}^{\infty} (-1)^n \frac{n}{2n + 1}.

    Answer

    Diverges

    Full solution

    The terms do not approach 00: their sizes approach 12\tfrac{1}{2}.

  4. Classify ∑n=2∞(−1)nln⁡n\displaystyle\sum_{n=2}^{\infty} \frac{(-1)^n}{\ln n}.

    Answer

    Converges conditionally

    Full solution

    1ln⁡n\tfrac{1}{\ln n} decreases to 00, so the series converges. Since ln⁡n<n\ln n < n, 1ln⁡n>1n\tfrac{1}{\ln n} > \tfrac{1}{n}, so the absolute values diverge by comparison with the harmonic series.

  5. Estimate ∑n=1∞(−1)n+1n3\displaystyle\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n^3} with three terms and bound the error.

    Answer

    About 0.9120.912, within 0.0160.016

    Full solution

    1−18+127≈0.9121 - \tfrac{1}{8} + \tfrac{1}{27} \approx 0.912, and the error is at most 143=164\tfrac{1}{4^3} = \tfrac{1}{64}.

  6. The series ∑n=0∞(−1)nn!\displaystyle\sum_{n=0}^{\infty} \frac{(-1)^n}{n!} equals 1e\tfrac{1}{e}. How many terms guarantee an error below 0.0010.001?

    Answer

    77 terms, n=0n = 0 to 66

    Full solution

    Stopping after the n=kn = k term leaves an error at most 1(k+1)!\tfrac{1}{(k + 1)!}. 16!=1720>0.001\tfrac{1}{6!} = \tfrac{1}{720} > 0.001 but 17!=15040<0.001\tfrac{1}{7!} = \tfrac{1}{5040} < 0.001, so stop after n=6n = 6.

  7. Classify ∑n=1∞(−1)nn+1n2\displaystyle\sum_{n=1}^{\infty} (-1)^n \frac{n + 1}{n^2}.

    Answer

    Converges conditionally

    Full solution

    n+1n2=1n+1n2\tfrac{n + 1}{n^2} = \tfrac{1}{n} + \tfrac{1}{n^2} decreases to 00, so the series converges. The absolute values are larger than 1n\tfrac{1}{n}, so they diverge.

  8. Classify ∑n=1∞cos⁡(nπ)n\displaystyle\sum_{n=1}^{\infty} \frac{\cos(n\pi)}{n}.

    Answer

    Converges conditionally

    Full solution

    cos⁡(nπ)=(−1)n\cos(n\pi) = (-1)^n, so this is ∑(−1)nn\sum \tfrac{(-1)^n}{n}, the alternating harmonic series with its signs flipped.

  9. Classify ∑n=1∞(−2)nn2\displaystyle\sum_{n=1}^{\infty} \frac{(-2)^n}{n^2}.

    Answer

    Diverges

    Full solution

    2nn2\tfrac{2^n}{n^2} grows without bound, so the terms do not approach 00.

  10. A student says ∑n=1∞(−1)nn\displaystyle\sum_{n=1}^{\infty} \frac{(-1)^n}{n} converges absolutely, because its terms approach 00. What went wrong?

    Hint

    What series do the absolute values form?

    Answer

    The absolute values form the harmonic series, which diverges. The series converges only conditionally.

    Full solution

    Terms approaching 00 is a condition for convergence of any kind, not a proof of absolute convergence. Absolute convergence means ∑∣an∣=∑1n\sum \lvert a_n \rvert = \sum \tfrac{1}{n} converges, and it does not.

    The series does converge, by the alternating series test, so it converges conditionally.

Frequently asked questions

What is the alternating series test?

If bₙ > 0, the bₙ are decreasing, and bₙ → 0, then the sum of (−1)ⁿbₙ converges.

How accurate is a partial sum of an alternating series?

The error after n terms is at most the size of the first omitted term, b₍ₙ₊₁₎, and the sum lies between any two consecutive partial sums.

What is absolute convergence?

A series converges absolutely if the series of absolute values of its terms converges. Absolutely convergent series always converge.

What is conditional convergence?

A series that converges but whose absolute values form a divergent series, such as 1 − 1/2 + 1/3 − 1/4 + ⋯.

Does the sum of an alternating series lie above or below a partial sum?

Between two consecutive partial sums. If the last term used was positive, the partial sum is too big; if negative, too small.

What to learn next