Unit 9, lesson 106

Infinite Geometric Series and Convergence

When adding all the terms of a geometric sequence gives a finite total, why the ratio must be less than 1 in absolute value, and how the sum formula turns repeating decimals into fractions.

  • Theory part 211
  • Practice part 212
  • Series

Walk halfway to a wall that is 1 meter away, then half of the distance that is left, then half of that again. You take infinitely many steps, yet you never pass the wall, and the step lengths 12,14,18,…\frac{1}{2}, \frac{1}{4}, \frac{1}{8}, \dots add up to exactly 1 meter. Adding infinitely many numbers can give a finite total, but only when the numbers shrink in the right way. For geometric series there is a clean test for when this happens and an exact formula for the total.

Key terms

Partial sums

Definition The infinite geometric series with first term a1≠0a_1 \ne 0 and common ratio r≠0r \ne 0 is

∑k=1∞a1rk−1=a1+a1r+a1r2+⋯ .\begin{aligned} &\sum_{k=1}^{\infty} a_1 r^{k-1} \\[0.4em] &\quad = a_1 + a_1 r + a_1 r^2 + \cdots. \end{aligned}

Its nnth partial sum is the sum of its first nn terms,

Sn=∑k=1na1rk−1={a1(1−rn)1−r,r≠1,n a1,r=1.\begin{aligned} S_n &= \sum_{k=1}^{n} a_1 r^{k-1} \\[0.8em] &= \begin{cases} \dfrac{a_1(1 - r^n)}{1 - r}, & r \ne 1, \\[2mm] n\,a_1, & r = 1. \end{cases} \end{aligned}

The value of the infinite series, if it has one, is defined through the sequence S1,S2,S3,…S_1, S_2, S_3, \dots of partial sums.

In plain words Nobody can add infinitely many numbers one at a time. Instead you keep a running total and watch what it does as more and more terms come in.

Example For 12+14+18+⋯\frac{1}{2} + \frac{1}{4} + \frac{1}{8} + \cdots both the first term and the ratio are 12\frac{1}{2}, so

Sn=12(1−(12)n)1−12=1−12n.S_n = \frac{\frac{1}{2}\left(1 - \left(\frac{1}{2}\right)^n\right)}{1 - \frac{1}{2}} = 1 - \frac{1}{2^n}.
nn112233441010
SnS_n0.50.50.750.750.8750.8750.93750.93750.99902343750.9990234375

Each partial sum falls short of 11 by exactly the last term added.

Convergence

Definition A series a1+a2+a3+⋯a_1 + a_2 + a_3 + \cdots converges to the number SS, called its sum, if its partial sums Sn=a1+a2+⋯+anS_n = a_1 + a_2 + \cdots + a_n approach SS as nn grows. Precisely, for every positive number ε\varepsilon there is an integer NN such that

∣Sn−S∣<εfor every n>N.\lvert S_n - S \rvert < \varepsilon \quad \text{for every } n > N.

In that case one writes a1+a2+a3+⋯=Sa_1 + a_2 + a_3 + \cdots = S. For a geometric series the terms are an=a1rn−1a_n = a_1 r^{n-1}.

In plain words Pick any tolerance, however small. From some point on, every running total is within that tolerance of SS and stays there.

Example For 12+14+⋯\frac{1}{2} + \frac{1}{4} + \cdots the gap between the partial sum and 11 is

∣Sn−1∣=12n.\lvert S_n - 1 \rvert = \frac{1}{2^n}.

To be within 0.0010.001 of 11 you need 2n>10002^n > 1000, which holds for every n≥10n \ge 10 because 210=10242^{10} = 1024. Any tolerance ε>0\varepsilon > 0 works the same way, because 12n<ε\frac{1}{2^n} < \varepsilon once nn is large enough and stays true for every larger nn. So the series converges and its sum is 11.

Divergence

Definition A series that does not converge diverges. Its partial sums may tend to ∞\infty or to −∞-\infty, or they may keep moving around without settling near any single number. Since each term after the first is the difference an=Sn−Sn−1a_n = S_n - S_{n-1} of two partial sums, a series whose partial sums approach SS has terms that approach S−S=0S - S = 0. So a series whose terms do not approach 00 diverges. The converse is false. Terms that approach 00 do not on their own make a series converge.

In plain words A divergent series has no sum. Its running total either runs off forever or never makes up its mind.

Example For 1+2+4+8+⋯1 + 2 + 4 + 8 + \cdots the partial sums are Sn=2n−1S_n = 2^n - 1, which grow without bound, so the series diverges. The terms 2n−12^{n-1} do not approach 00, and that alone already settles it. Shrinking terms are not enough on their own. The series 1+12+13+14+⋯1 + \frac{1}{2} + \frac{1}{3} + \frac{1}{4} + \cdots has terms that approach 00, yet grouping them as 13+14>12\frac{1}{3} + \frac{1}{4} > \frac{1}{2}, then 15+⋯+18>12\frac{1}{5} + \cdots + \frac{1}{8} > \frac{1}{2} and so on adds more than 12\frac{1}{2} again and again. Its partial sums pass every number, as the calculus lesson on the harmonic series shows in detail.

Ratios with absolute value less than 1

Definition For a real number rr, the powers rnr^n behave as follows as nn grows.

  • If ∣r∣<1\lvert r \rvert < 1, then rn→0r^n \to 0.
  • If r=1r = 1, then rn=1r^n = 1 for every nn.
  • If r=−1r = -1, then rnr^n alternates between −1-1 and 11.
  • If ∣r∣>1\lvert r \rvert > 1, then ∣rn∣\lvert r^n \rvert grows without bound.

For 0<∣r∣<10 < \lvert r \rvert < 1 the first fact has a direct reason. Given a tolerance ε\varepsilon with 0<ε<10 < \varepsilon < 1, take logarithms in ∣r∣n<ε\lvert r \rvert^n < \varepsilon and divide by the negative number ln⁡∣r∣\ln \lvert r \rvert, which reverses the inequality. So ∣rn∣<ε\lvert r^n \rvert < \varepsilon exactly when

n>ln⁡εln⁡∣r∣.n > \frac{\ln \varepsilon}{\ln \lvert r \rvert}.

In plain words Multiplying again and again by a number between −1-1 and 11 shrinks the result toward 00, however close that number is to 11. A ratio near 11 just takes more steps.

Example Powers of 0.90.9 shrink slowly but surely.

nn10105050100100
0.9n0.9^n0.34870.34870.005150.005150.00002660.0000266

Getting below 0.0010.001 needs

n>ln⁡0.001ln⁡0.9≈65.6,n > \frac{\ln 0.001}{\ln 0.9} \approx 65.6,

so 0.9n<0.0010.9^n < 0.001 for every n≥66n \ge 66.

Infinite geometric sum

Definition Let a1≠0a_1 \ne 0. The geometric series a1+a1r+a1r2+⋯a_1 + a_1 r + a_1 r^2 + \cdots converges if and only if ∣r∣<1\lvert r \rvert < 1, and in that case

∑k=1∞a1rk−1=a11−r.\sum_{k=1}^{\infty} a_1 r^{k-1} = \frac{a_1}{1 - r}.

The reason is the partial sum formula. For ∣r∣<1\lvert r \rvert < 1 it can be written

Sn=a11−r−a11−r rn,S_n = \frac{a_1}{1 - r} - \frac{a_1}{1 - r}\,r^n,

and the last term approaches 00 because rn→0r^n \to 0. For ∣r∣≥1\lvert r \rvert \ge 1 every term has size ∣a1∣∣r∣n−1≥∣a1∣>0\lvert a_1 \rvert \lvert r \rvert^{n-1} \ge \lvert a_1 \rvert > 0, so the terms do not approach 00 and the series diverges. The calculus lessons start the index at 00 and write the same series as ∑k=0∞a1rk\sum_{k=0}^{\infty} a_1 r^k, with the same sum.

In plain words When the ratio is strictly between −1-1 and 11, the infinite sum is the first term divided by one minus the ratio. For every other ratio there is no sum.

Example The series 18+6+2+23+⋯18 + 6 + 2 + \frac{2}{3} + \cdots has a1=18a_1 = 18 and r=13r = \frac{1}{3}, so its sum is

181−13=27.\frac{18}{1 - \frac{1}{3}} = 27.

A negative ratio works the same way. The series 10−5+52−⋯10 - 5 + \frac{5}{2} - \cdots has r=−12r = -\frac{1}{2} and sum 103/2=203\frac{10}{3/2} = \frac{20}{3}. Always find the first term by putting in the starting index. In

∑k=1∞5 (0.2)k=1+0.2+0.04+⋯\sum_{k=1}^{\infty} 5\,(0.2)^k = 1 + 0.2 + 0.04 + \cdots

the first term is 5⋅0.2=15 \cdot 0.2 = 1, so the sum is 11−0.2=1.25\frac{1}{1 - 0.2} = 1.25, not 50.8=6.25\frac{5}{0.8} = 6.25.

The ratios 1 and negative 1

Definition Let a1≠0a_1 \ne 0. For r=1r = 1 the series is a1+a1+a1+⋯a_1 + a_1 + a_1 + \cdots with partial sums Sn=n a1S_n = n\,a_1, which tend to ∞\infty if a1>0a_1 > 0 and to −∞-\infty if a1<0a_1 < 0. For r=−1r = -1 the series is a1−a1+a1−⋯a_1 - a_1 + a_1 - \cdots and its partial sums alternate,

Sn={a1,n odd,0,n even.S_n = \begin{cases} a_1, & n \text{ odd}, \\ 0, & n \text{ even}. \end{cases}

Both series diverge, and the formula a11−r\frac{a_1}{1 - r} applies to neither of them.

In plain words At the two edges of the interval −1<r<1-1 < r < 1 the series fails in two different ways. With ratio 11 the total runs away, and with ratio −1-1 it keeps jumping back and forth.

Example Here are the partial sums of 1−1+1−1+⋯1 - 1 + 1 - 1 + \cdots.

nn112233445566
SnS_n110011001100

They never settle near one number, so the series diverges. Putting r=−1r = -1 into the formula anyway would give 12\frac{1}{2}, a value the partial sums never take. Grouping the terms as (1−1)+(1−1)+⋯(1 - 1) + (1 - 1) + \cdots suggests 00, and grouping them as 1+(−1+1)+⋯1 + (-1 + 1) + \cdots suggests 11. Grouping is only safe in a series that already converges.

Repeating decimals

Definition When a block of kk digits, not all zero, repeats forever in a decimal, the repeating part is an infinite geometric series with ratio 10−k10^{-k}. If the block starts right after the decimal point and its digits form the whole number dd, then

0.d‾=d10k+d102k+⋯=d/10k1−10−k=d10k−1.\begin{aligned} 0.\overline{d} &= \frac{d}{10^k} + \frac{d}{10^{2k}} + \cdots \\[0.8em] &= \frac{d/10^k}{1 - 10^{-k}} \\[0.8em] &= \frac{d}{10^k - 1}. \end{aligned}

Since 0<10−k<10 < 10^{-k} < 1, the series converges. Leading zeros count as digits of the block, so the block 0707 has k=2k = 2 and d=7d = 7, and 0.070707…=7990.070707\ldots = \frac{7}{99}. A decimal whose repeating block starts later is a terminating decimal plus such a value divided by a power of 1010, so every repeating decimal is a rational number.

In plain words Each copy of the block is the previous copy moved kk places to the right, which is the same as multiplying by 10−k10^{-k}. The block divided by as many nines as it has digits gives the fraction.

Example For 0.363636…0.363636\ldots the block is 3636 with k=2k = 2, so the decimal equals 3699=411\frac{36}{99} = \frac{4}{11}. When the repeating starts later, split off the part before it. In 2.4151515…2.4151515\ldots the repeating part 0.0151515…0.0151515\ldots has first term 151000\frac{15}{1000} and ratio 1100\frac{1}{100}, so

2.4151515…=125+15/10001−1/100=125+166=797330.\begin{aligned} 2.4151515\ldots &= \frac{12}{5} + \frac{15/1000}{1 - 1/100} \\[0.8em] &= \frac{12}{5} + \frac{1}{66} \\[0.8em] &= \frac{797}{330}. \end{aligned}

The same rule gives 0.999…=99=10.999\ldots = \frac{9}{9} = 1, so 0.999…0.999\ldots and 11 are two names for the same number.

Check yourself

  1. Find the sum below, or explain why it does not exist.
∑k=1∞3(−12)k\sum_{k=1}^{\infty} 3\left(-\frac{1}{2}\right)^k
Show the answer

The first term is 3⋅(−12)=−323 \cdot \left(-\frac{1}{2}\right) = -\frac{3}{2} and the ratio is −12-\frac{1}{2}, whose absolute value is less than 11. So the series converges, and its sum is

−3/21−(−12)=−3/23/2=−1.\frac{-3/2}{1 - \left(-\frac{1}{2}\right)} = \frac{-3/2}{3/2} = -1.
  1. Write 0.2454545…0.2454545\ldots as a fraction in lowest terms.
Show the answer

Split off 0.2=150.2 = \frac{1}{5}. The rest is a geometric series with first term 451000\frac{45}{1000} and ratio 1100\frac{1}{100}, so

0.2454545…=15+45/10001−1/100=15+122=27110.\begin{aligned} 0.2454545\ldots &= \frac{1}{5} + \frac{45/1000}{1 - 1/100} \\[0.8em] &= \frac{1}{5} + \frac{1}{22} \\[0.8em] &= \frac{27}{110}. \end{aligned}
  1. A ball is dropped from a height of 9 meters. After each bounce it rises to two thirds of the height it fell from. If it keeps bouncing in this way forever, what total distance does it travel?
Show the answer

It falls 9 meters first. After that, each rebound height 6,4,83,…6, 4, \frac{8}{3}, \dots is traveled twice, once up and once down. These heights form a geometric series with first term 66 and ratio 23\frac{2}{3}, so the total is

9+2⋅61−23=9+36=45 meters.\begin{aligned} 9 + 2 \cdot \frac{6}{1 - \frac{2}{3}} &= 9 + 36 \\[0.8em] &= 45 \text{ meters.} \end{aligned}

How this lesson connects

Where this leads in Calculus