A series with x in it
A power series is a series whose terms are powers of x − a, each multiplied by a coefficient: . The number a is its center. Each value of x gives a different series of numbers, and some of those series converge while others do not.
Take . At the terms are 0.5, 0.125, 0.041667, 0.015625, 0.00625, …, they shrink quickly, and the series converges (its sum is ln 2 = 0.693147). At x = 2 the terms are 2, 2, 2.666667, 4, 6.4, …, they grow, and the series diverges.
At the center itself every term after the first is 0, so a power series always converges at x = a. The question is how far from a it keeps converging.
The ratio test, with x inside
Apply the ratio test to the terms . Dividing one term by the one before leaves one factor of x − a, so the size of the ratio is . The first factor does not depend on n. If approaches a number k as n grows, then L = k|x − a|.
The series converges when L < 1, which is when . That number is the radius of convergence, . When |x − a| > R, L is above 1 and the series diverges.
An interval on the number line
|x − a| < R says that x is less than R away from a. On the number line that is every number from a − R to a + R, not counting the two ends: one stretch, centered on a, with a length of 2R.
A center at 2 and a radius of 3. The series converges for every x less than 3 away from 2, which is every x between −1 and 5.
The radius of
The terms are . Dividing the next term by this one gives , whose size is . As n grows, tends to 1, so L = |x|.
L < 1 means |x| < 1. The center is 0 and the radius is R = 1, so the series converges for −1 < x < 1 and diverges for |x| > 1. That agrees with the two tries: is inside, and x = 2 is outside.
The two endpoints
At x = 1 and x = −1, L = |x| = 1, and the ratio test decides nothing. You can see it in the ratio itself: at an endpoint it is , which is 0.5 at n = 1, 0.909091 at n = 10 and 0.990099 at n = 100, creeping up to 1. So each endpoint needs its own test.
At x = 1 the series is , the harmonic series. It is a p-series with p = 1, so it diverges, and x = 1 is not in the interval.
At x = −1 the series is . The signs alternate, and the sizes 1, , , … decrease and tend to 0, so the alternating series test says it converges (its sum is −ln 2 = −0.693147). So x = −1 is in the interval.
The interval of convergence is : the left end is included and the right end is not. In bracket notation it is written [−1, 1), with a square bracket at an included end and a round one at an excluded end.
The interval of convergence of . The filled circle at −1 is included, because the alternating series there converges. The hollow circle at 1 is not, because the harmonic series diverges.
Four ways to end
The radius fixes where the interval stops, but not whether its ends belong to it. Four series with center 0 and radius 1 show every possibility.
is the geometric series. At x = 1 its terms are all 1, and at x = −1 they are 1, −1, 1, −1, …; neither tends to 0, so both ends diverge and the interval is −1 < x < 1.
converges at x = 1, where it is the p-series with p = 2, and at x = −1, where the sizes are the same . Both ends are included: .
gives , as above. Replace x by −x, and swaps the two ends: it alternates at x = 1 and is the harmonic series at x = −1, so its interval is .
A center away from 0
Take the series whose nth term is divided by . Dividing the next term by this one gives , and tends to 1, so L is |x − 2| divided by 3.
L < 1 means |x − 2| < 3. The center is 2 and the radius is 3, so the series converges for −1 < x < 5.
Now the ends. At x = 5, x − 2 = 3, and the terms are divided by , which is : a p-series with p = 2, which converges. At x = −1, x − 2 = −3, and the terms are , whose sizes are again , so this converges too. Both ends are included, and the interval is .
The interval of convergence for the series centered at 2 with radius 3. Both circles are filled, because at each end the sizes of the terms are , which add up to a finite total.
An infinite radius, and a radius of 0
For , the ratio is , whose size is |x| divided by n + 1. Whatever x is, this tends to 0 as n grows, so L = 0 for every x. The series converges for every real number, and the radius is said to be infinite. This is the series for .
For , the ratio is (n + 1)x, whose size grows without limit for every x except 0. The series converges only at its center, and the radius is 0.
The usual mistakes
Stopping at the radius. R = 1 for does not make the interval −1 < x < 1 or . Each end must be tested on its own, and here one is in and one is out.
Testing an endpoint with the ratio test. At an endpoint L = 1 exactly, which is the one value the ratio test cannot decide. Use the nth term test, a p-series, a comparison or the alternating series test there.
Forgetting the center. |x − 2| < 3 means −1 < x < 5, from 2 − 3 to 2 + 3, not −3 < x < 3.
A model with a limited range
In the application below, the energy in a stretched spring is given as a power series. The ratio test gives its radius, the two ends are tested with the nth term test, and the series is then summed at a point inside the interval and refused at a point outside it.
Worked example: A Spring Model Written as a Power Series: How Far It May Be Trusted
Question A laboratory models the energy stored in a spring stretched x cm as E(x) = ∑n=1∞ n xn4n joules, for a stretch of x centimeters. (a) Find the radius of convergence and the interval of x on which the series converges, with each end included or excluded as it falls. (b) Find the energy the model gives at an extension of 2 cm, and say what it gives at an extension of 5 cm.
1.Apply the ratio test to the sizes of the terms: |an+1an| = (n+1)|x|n+14n+1 × 4nn|x|n = n+1n × |x|4, and n+1n → 1, so the limit is |x|4.
The ratio of one term to the last is n+1n × |x|4, which tends to |x|4. 2.The series converges while that limit is below 1, which is |x| < 4, so the radius of convergence is R = 4.
The series converges while |x|4 < 1, that is while |x| < 4. 3.Test the two ends, which the ratio test leaves undecided. At x = 4 the terms are n × 4n4n = n, which do not tend to 0. At x = −4, a compression of 4 cm, they are (−1)n n, which do not tend to 0 either. By the nth term test both ends diverge.
At x = 4 the terms are n and at x = −4 they are (−1)n n: neither tends to 0. 4.(a) The radius is 4 and the interval of convergence is −4 < x < 4, with both ends excluded. The model says nothing at all about a stretch of 4 cm or more.
(a) R = 4, and the interval is −4 < x < 4 with both ends excluded. 5.(b) At x = 2 the series is ∑n=1∞ n(12)n, and ∑n=1∞ nrn = r(1−r)2 for |r| < 1, so the energy is 12(12)2 = 2 joules. At x = 5 the ratio is 54, which is above 1, so the terms grow without bound and the series diverges: the model gives nothing for a 5 cm extension, and a different model is needed there.
(b) E(2) = 2 joules, while at x = 5 the series diverges and gives nothing.
Answer: (a) R = 4, and the interval is −4 < x < 4 with both ends excluded; (b) 2 joules at 2 cm, and nothing at 5 cm, where the series diverges
Common mistakes
- Stopping at the radius and writing the interval as −4 ≤ x ≤ 4. The ratio test is silent at the ends, so each one must be tested on its own; here both fail the nth term test and both are therefore excluded.
- Reading a divergent series as an energy of zero, or as an infinite energy. A divergent series has no sum, so the model makes no claim at 5 cm. That is a limit on the model, not a statement about the spring.