AP Calculus BC

Alternating Series Test and Its Error Bound

An alternating series whose terms have sizes b_n converges if b_n is decreasing, at least from some point on, and b_n approaches 0. If the sizes are decreasing from the first omitted term on, the error from stopping at a partial sum is at most the size of that term, with the same sign.

The test

An alternating series has terms that switch sign every step. Write it with the sign factor separated from a positive size bnb_n.

∑n=1∞(−1)n+1bn=b1−b2+b3−b4+⋯ ,bn>0\sum_{n=1}^{\infty} (-1)^{n+1} b_n = b_1 - b_2 + b_3 - b_4 + \cdots, \qquad b_n > 0

CED Topic 10.7, Alternating Series Test for Convergence, gives two conditions on bnb_n. If both hold, the series converges.

  1. The sizes decrease: bn+1≤bnb_{n+1} \le b_n for all nn from some point on.
  2. The sizes approach zero: lim⁡n→∞bn=0\lim_{n \to \infty} b_n = 0.

The sign factor may be written as (−1)n(-1)^n, (−1)n+1(-1)^{n+1} or cos⁡(nπ)\cos(n\pi). The test is the same in every case, because only the sizes bnb_n are checked.

Checking the two conditions

The limit condition is usually quick. The decreasing condition needs a reason, and there are two common ways to give one.

  • Compare consecutive terms directly. For bn=1n2b_n = \frac{1}{n^2}, a larger nn gives a larger denominator, so bn+1<bnb_{n+1} < b_n.
  • Replace nn with xx and show the derivative is negative. For f(x)=xx2+4f(x) = \frac{x}{x^2 + 4}, f′(x)=4−x2(x2+4)2f'(x) = \frac{4 - x^2}{(x^2 + 4)^2}, which is negative for x>2x > 2.

What a failed condition means depends on which one fails. If bnb_n does not approach 00, then the terms of the series do not approach 00 and the series diverges by the nth term test. If bnb_n approaches 00 and the sizes are not eventually decreasing, the alternating series test gives no conclusion and another method is needed.

The test never proves divergence

The alternating series test can only conclude that a series converges. A divergence conclusion always comes from a different test, most often the nth term test.

Absolute or conditional convergence

After an alternating series is shown to converge, the next question is how. CED Topic 10.9, Determining Absolute or Conditional Convergence, separates two cases by looking at ∑bn\sum b_n, the series with the signs removed.

Series∑bn\sum b_nClassification
∑(−1)n+1n2\sum \frac{(-1)^{n+1}}{n^2}∑1n2\sum \frac{1}{n^2} converges (p=2p = 2)Absolutely convergent
∑(−1)n+1n\sum \frac{(-1)^{n+1}}{n}∑1n\sum \frac{1}{n} diverges (harmonic)Conditionally convergent

If ∑bn\sum b_n converges, the alternating series converges absolutely, and the alternating series test was not even needed. If ∑bn\sum b_n diverges while the alternating series converges, the convergence is conditional: it depends on the cancellation between positive and negative terms.

The alternating series error bound

When a series satisfies the alternating series test, its partial sums give approximations with a built-in error estimate. This is CED Topic 10.10, Alternating Series Error Bound. Let SS be the sum and SnS_n the sum of the first nn terms.

∣S−Sn∣≤bn+1\left| S - S_n \right| \le b_{n+1}

In words, the error is at most the size of the first term left out. The bound requires the sizes to be decreasing from that term on. Two more facts come with it.

  • The error S−SnS - S_n has the same sign as the first omitted term. If that term is positive, SnS_n is an underestimate. If it is negative, SnS_n is an overestimate.
  • The true sum lies between any two consecutive partial sums SnS_n and Sn+1S_{n+1}.

To find how many terms are needed for a given accuracy, solve bn+1≤b_{n+1} \le the allowed error for the smallest nn. The term that meets the tolerance is the first one you leave out, so the number of terms used is one fewer than its position.

Common mistakes

  • Checking only that bn→0b_n \to 0 and skipping the decreasing condition.
  • Concluding that a series diverges because the alternating series test failed. Only the nth term test, or another test, can give divergence.
  • Using the last term included as the error bound. The bound is the first term omitted, bn+1b_{n+1}.
  • Reporting the bound as the exact error. It is an upper limit on the size of the error.
  • Using this error bound on a series that does not alternate, or whose terms are not decreasing.
  • Stopping at converges when the question asks for absolute or conditional convergence.

Worked examples

Worked example

Terms that decrease only eventually

Determine whether ∑n=1∞(−1)n+1nn2+4\sum_{n=1}^{\infty} (-1)^{n+1} \frac{n}{n^2 + 4} converges absolutely, converges conditionally, or diverges.

  1. Identify the sizes: bn=nn2+4>0b_n = \frac{n}{n^2 + 4} > 0.
  2. Check the limit: lim⁡n→∞nn2+4=0\lim_{n \to \infty} \frac{n}{n^2 + 4} = 0, since the denominator has the higher degree.
  3. Check decreasing: with f(x)=xx2+4f(x) = \frac{x}{x^2 + 4}, f′(x)=(x2+4)−x⋅2x(x2+4)2=4−x2(x2+4)2f'(x) = \frac{(x^2 + 4) - x \cdot 2x}{(x^2 + 4)^2} = \frac{4 - x^2}{(x^2 + 4)^2}, which is negative for x>2x > 2. The first terms confirm it: b1=15b_1 = \frac{1}{5}, b2=14b_2 = \frac{1}{4}, b3=313b_3 = \frac{3}{13}, b4=15b_4 = \frac{1}{5}. The sizes rise once and then decrease from n=2n = 2 on, which is enough.
  4. Both conditions hold, so the series converges by the alternating series test.
  5. Test for absolute convergence with ∑nn2+4\sum \frac{n}{n^2 + 4}. Use limit comparison with 1n\frac{1}{n}: nn2+4⋅n=n2n2+4→1\frac{n}{n^2 + 4} \cdot n = \frac{n^2}{n^2 + 4} \to 1. The harmonic series diverges, so ∑nn2+4\sum \frac{n}{n^2 + 4} diverges.

The series converges conditionally: it converges by the alternating series test, and the series of absolute values diverges.

Worked example

Bounding the error of a partial sum

The series ∑n=1∞(−1)n+1n2\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n^2} converges to a sum SS. Approximate SS with the first four terms and bound the error.

  1. Confirm the conditions: bn=1n2b_n = \frac{1}{n^2} is positive, decreasing, and approaches 00, so the error bound applies.
  2. Compute the partial sum: S4=1−14+19−116=144−36+16−9144=115144≈0.7986S_4 = 1 - \frac{1}{4} + \frac{1}{9} - \frac{1}{16} = \frac{144 - 36 + 16 - 9}{144} = \frac{115}{144} \approx 0.7986.
  3. The first omitted term is the fifth: +125+\frac{1}{25}. So ∣S−S4∣≤125=0.04|S - S_4| \le \frac{1}{25} = 0.04.
  4. The omitted term is positive, so S4S_4 is an underestimate and SS lies between S4S_4 and S4+125S_4 + \frac{1}{25}: 0.7986<S<0.83860.7986 < S < 0.8386.
  5. As a check, the exact sum is known to be π212≈0.8225\frac{\pi^2}{12} \approx 0.8225, which lies in that interval. The actual error is about 0.0240.024, under the bound of 0.040.04.

S≈115144≈0.7986S \approx \frac{115}{144} \approx 0.7986, with error at most 125=0.04\frac{1}{25} = 0.04. The approximation is an underestimate.

Worked example

How many terms for a given accuracy

How many terms of ∑n=1∞(−1)n+1n3\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n^3} are needed to guarantee that the partial sum is within 0.0050.005 of the sum of the series?

  1. The sizes bn=1n3b_n = \frac{1}{n^3} are positive, decreasing, and approach 00, so ∣S−Sn∣≤bn+1=1(n+1)3|S - S_n| \le b_{n+1} = \frac{1}{(n+1)^3}.
  2. Require 1(n+1)3≤0.005\frac{1}{(n+1)^3} \le 0.005, which means (n+1)3≥200(n+1)^3 \ge 200.
  3. Test cubes: 53=1255^3 = 125 is too small and 63=2166^3 = 216 is large enough. So n+1=6n + 1 = 6 and n=5n = 5.
  4. Five terms are enough: S5=1−18+127−164+1125≈0.9044S_5 = 1 - \frac{1}{8} + \frac{1}{27} - \frac{1}{64} + \frac{1}{125} \approx 0.9044, with error at most 1216≈0.0046\frac{1}{216} \approx 0.0046.
  5. The first omitted term is −1216-\frac{1}{216}, which is negative, so S5S_5 is an overestimate.

Five terms. Then ∣S−S5∣≤1216≈0.0046<0.005|S - S_5| \le \frac{1}{216} \approx 0.0046 < 0.005.

Frequently asked questions

If the alternating series test fails, does the series diverge?

Not automatically. If the sizes bnb_n do not approach 00, the series diverges, and the reason to cite is the nth term test. If bnb_n approaches 00 and the sizes are not eventually decreasing, the alternating series test simply does not apply and you need another approach.

What is the difference between absolute and conditional convergence?

A series converges absolutely if it still converges after every term is replaced by its absolute value. It converges conditionally if it converges as written and the series of absolute values diverges. ∑(−1)n+1n2\sum \frac{(-1)^{n+1}}{n^2} is absolutely convergent, and ∑(−1)n+1n\sum \frac{(-1)^{n+1}}{n} is conditionally convergent.

Which term do I use for the alternating series error bound?

Use the first term you did not include. If the approximation uses the first nn terms, the error is at most bn+1b_{n+1}, the size of term number n+1n + 1. Using the last included term overstates the bound.