Alternating series and absolute convergence ·… | StemPy Lab
Course 15.2·Lesson 3 of 3
Alternating series and absolute convergence
An alternating series changes sign at every term. If the terms shrink steadily to 0, the series converges, and the error after N terms is at most the first term left out. A series converges absolutely when the series of absolute values converges. If it converges only because of the signs, it converges conditionally. A closing guide tells you which test to try first.
about 35 min
Lesson 3 of 3Show all lessonsHide
Try the code and the simulations beside the lesson.
What you will be able to do
use the alternating series test and say what it cannot tell you;
bound the error after N terms, and find how many terms you need;
tell absolute convergence from conditional convergence;
choose the right test for a series, and avoid the trap in each test.
Terms that change sign
So far every series had positive terms. Now let the signs alternate:
1−21+31−41+⋯=n=1∑∞n(−1)n+1.
The positive series of the same numbers is the harmonic series, which diverges. But with the signs alternating, the partial sums go up and down, and each step is smaller than the one before. They zigzag towards a number. Here is the idea made exact.
The alternating series test
Write the series as n=1∑∞(−1)n+1bn with every bn>0. Suppose
the numbers bn are decreasing (at least from some point on), and
bn→0.
Then the series converges.
The reason is the zigzag. The partial sums S1,S3,S5,… go down and the sums S2,S4,S6,… go up. Each odd sum is above each even sum, and the gap between SN and SN+1 is bN+1, which tends to 0. So both groups squeeze onto one number.
The partial sums S1 to S12 of the alternating harmonic series plotted against the number of terms. The points jump up and down, alternately above and below a dashed green horizontal line at the height 0.693, the sum of the series. The jumps get smaller each time, so the points close in on the line from both sides. A shaded band between S9 and S10 is labelled as the interval that must contain the sum.
Example 1. The alternating harmonic series has bn=n1, which decreases to 0. So it converges. Its sum is ln2≈0.693, which you can prove with the Taylor series in the next course.
Example 2. In n=1∑∞n(−1)n we have bn=n1, which decreases to 0. The series converges.
What the test cannot do. The test says only "converges". If one of the two conditions fails, the test gives no verdict. Take ∑n+1(−1)nn. Here bn=n+1n→1, not 0. The terms do not tend to 0, so the series diverges by the nth-term test. Always check that bn→0 first.
The error bound
If a series satisfies the alternating series test, the sum S lies between any two consecutive partial sums. So the error from stopping at N terms is smaller than the first term that you leave out:
∣S−SN∣≤bN+1.
The sign of the error is the sign of the first omitted term. If you stopped just before a positive term, the sum is a little above SN.
Example 3. How many terms of n=1∑∞n2(−1)n+1 give the sum within 0.01? We need bN+1=(N+1)21≤0.01. That means (N+1)2≥100, so N+1≥10 and N=9. The sum of the first nine terms is 0.8280. The true sum is 12π2≈0.8225, which differs from S9 by 0.0055. That is less than 1001, as promised.
Example 4. Add 1−1+21−61+241−1201, the terms of n=0∑∞n!(−1)n up to n=5. The total is 3011≈0.3667. The next term is 7201, so the sum lies within 0.0014 of 0.3667. The true sum is e1=0.3679, which is indeed that close.
Absolute and conditional convergence
A series ∑anconverges absolutely if ∑∣an∣ converges. This is the stronger kind of convergence. The next fact is useful.
If ∑∣an∣ converges, then ∑an converges.
If the absolute values add up to a finite number, the signs cannot make things worse. A series that converges but does not converge absolutely converges conditionally.
∑n2(−1)n+1 converges absolutely, because ∑n21 is a convergent p-series.
∑n(−1)n+1 converges conditionally. It converges by the alternating series test, but ∑n1 diverges.
∑n2sinn converges absolutely, since n2sinn≤n21.
Why does the difference matter? An absolutely convergent series has the same sum in whatever order you add its terms. A conditionally convergent series does not. A famous theorem of Riemann says that you can rearrange its terms to make the sum any number you like. The sum depends on the signs balancing in the given order.
A routine for a series with mixed signs.
Check that an→0. If not, it diverges.
Test ∑∣an∣ with a test for positive series. If it converges, the series converges absolutely. Stop.
If ∑∣an∣ diverges, try the alternating series test. If it works, the series converges conditionally. If not, look further.
The ratio test gives a shortcut. If anan+1→L>1, the series diverges even with mixed signs, because the terms do not tend to 0.
Which test first, and the trap in each
Here is a guide. Read it from the top. Use the first test that fits the shape of the series.
nth-term test. Use it first, always. If an→0, the series diverges. Trap: if an→0, you have learned nothing. The harmonic series shows this.
Geometric series∑arn. Converges when ∣r∣<1. Trap: the sum 1−ra is for the first term a. Check where the sum starts.
p-series∑np1. Converges when p>1. Trap:p=1 diverges.
Integral test. Use it when ∫f is easy, as with nlnn1. Trap:f must be positive and decreasing, and the integral is not the sum.
Direct comparison. Use it when an inequality is obvious. Trap: an inequality in the wrong direction proves nothing.
Limit comparison. Use it for a fraction of powers of n. Trap:L must be finite and above 0, and the terms must be positive.
Ratio test. Use it for factorials and n in an exponent. Trap:L=1 means no answer.
Root test. Use it when the whole term is raised to the power n. Trap:L=1 means no answer.
Alternating series test. Use it when the signs alternate. Trap:bn must decrease and tend to 0, and failing the test does not prove divergence.
Absolute convergence. For mixed signs, test ∑∣an∣ first. Trap: a series that fails to converge absolutely may still converge conditionally.
Practise with new numbers
Try it in Python
The program adds the alternating harmonic series one term at a time. For each partial sum it prints the interval that traps the sum.
Common mistakes
Checking only that bn→0. The numbers bn must also decrease, at least eventually.
Saying that a failed alternating test means divergence. It means that the test is silent.
Taking the error bound as the last term that you kept. The bound is the first term that you left out.
Calling every convergent alternating series absolutely convergent. Test the absolute series separately.
Using the error bound for a series that does not alternate. The bound ∣S−SN∣≤bN+1 needs the alternating series test.
Remember these
Alternating series test: bn decreasing and bn→0 give convergence.
Error bound: ∣S−SN∣≤bN+1, and the sum lies between consecutive partial sums.
Absolute convergence implies convergence. Conditional convergence needs the signs.
Order of work: nth-term test, then the right test for the shape of the series.
Check yourself
Try each one on your own first. If you are stuck, open a hint. Then check with Show the answer.
Does n=1∑∞n+3(−1)n converge absolutely, converge conditionally, or diverge?
Does n=1∑∞2n+1(−1)nn converge?
The series n=1∑∞n(−1)n+1 is added up to 99 terms. What is the largest possible error?
Does n=1∑∞n3cos(nπ) converge absolutely?
Which test is the best first choice for n=1∑∞n!(−3)n?
References
OpenStax, Calculus Volume 2, Chapter 5, section 5.5 on alternating series and absolute and conditional convergence.
College Board, AP Calculus AB and BC Course and Exam Description, Unit 10, the topics on the alternating series test, absolute and conditional convergence, and the alternating series error bound.
Enroll free in Tests for Convergence
Every lesson is open to read. Enrolling also lets you:
Mark lessons done and carry on where you stopped.
Take the tests of this lesson and its course.
Count the course towards the certificate for its chapter.