Infinite Sequences and Series

Study Sheet

Infinite Sequences and Series

Everything you need for AP Calculus BC, Topic 10

Sequences: Convergence, Divergence & Limits

Concept
What a sequence limit means

A sequence {an}\{a_n\} is an ordered list a1,a2,a3,a_1,a_2,a_3,\dots We say {an}\{a_n\} converges to LL if limnan=L\displaystyle\lim_{n\to\infty}a_n=L (a finite number); otherwise it diverges. To find the limit, treat nn as a continuous variable and use the same tools as limits at infinity: dominant terms, dividing by the highest power, or L'Hôpital's Rule.

Reminder — L'Hôpital's Rule:limxcf(x)g(x)=limxcf(x)g(x)when 00 or \lim_{x\to c}\frac{f(x)}{g(x)} = \lim_{x\to c}\frac{f'(x)}{g'(x)} \quad\text{when } \tfrac{0}{0} \text{ or } \tfrac{\infty}{\infty}
Example
Rational and root sequences
limn3n2n5n2+7=35,limnlnnn=L’Hlimn1/n1=0,limn(1+2n)n=e2.\lim_{n\to\infty}\frac{3n^2-n}{5n^2+7}=\frac35,\qquad \lim_{n\to\infty}\frac{\ln n}{n}\overset{\text{L'H}}{=}\lim_{n\to\infty}\frac{1/n}{1}=0,\qquad \lim_{n\to\infty}\Big(1+\tfrac{2}{n}\Big)^{n}=e^{2}.

The first divides by n2n^2; the middle uses L'Hôpital; the last is the definition of exe^x.

Tip

Tip: Useful facts --- limnrn=0\displaystyle\lim_{n\to\infty}r^{n}=0 if r<1|r|<1; n1/n1n^{1/n}\to1; nkn!0\dfrac{n^k}{n!}\to0 and rnn!0\dfrac{r^n}{n!}\to0 for any fixed k,rk,r (factorials beat everything). A bounded, monotonic sequence always converges.

Series, Partial Sums, Geometric Series & the nnth-Term Test

Concept
Series and partial sums

A series is the sum n=1an\displaystyle\sum_{n=1}^{\infty}a_n. Its value is the limit of the partial sums SN=n=1NanS_N=\sum_{n=1}^{N}a_n: the series converges to SS if limNSN=S\displaystyle\lim_{N\to\infty}S_N=S. Do not confuse the sequence of terms ana_n with the sequence of partial sums SNS_N.

Concept
Geometric series

n=0arn=a+ar+ar2+\displaystyle\sum_{n=0}^{\infty}ar^{n}=a+ar+ar^2+\cdots converges if and only if r<1|r|<1, and then

n=0arn=a1r(a is the first term, r the common ratio).\sum_{n=0}^{\infty}ar^{n}=\frac{a}{1-r}\qquad(\text{$a$ is the \emph{first} term, $r$ the common ratio}).

If r1|r|\ge 1 it diverges.

Example
Summing a geometric series
n=123n=2/3113=2/32/3=1,n=05(12)n=51(12)=53/2=103.\sum_{n=1}^{\infty}\frac{2}{3^{n}}=\frac{2/3}{1-\frac13}=\frac{2/3}{2/3}=1, \qquad \sum_{n=0}^{\infty}5\Big(-\tfrac{1}{2}\Big)^{n}=\frac{5}{1-(-\frac12)}=\frac{5}{3/2}=\frac{10}{3}.

For the first, the first term (at n=1n=1) is 23\frac23 and r=13r=\frac13.

Concept
The nnth-Term Test (Test for Divergence)

If limnan0\displaystyle\lim_{n\to\infty}a_n\neq 0 (or does not exist), then an\displaystyle\sum a_n diverges. Warning: if liman=0\lim a_n=0, the test is inconclusive --- the series may converge or diverge. It can never prove convergence.

Example
Divergence test

n=1n2n+1\displaystyle\sum_{n=1}^{\infty}\frac{n}{2n+1} diverges because limnn2n+1=120\displaystyle\lim_{n\to\infty}\frac{n}{2n+1}=\frac12\neq0. But for 1n\sum\frac1n, lim1n=0\lim\frac1n=0, so the test says nothing (in fact it diverges --- see the pp-series).

Tip

Tip: Always run the nnth-term test first. If the terms don't shrink to 00, you're done --- it diverges. If they do go to 00, you must try another test.

The Integral Test & pp-Series

Concept
Integral Test

If ff is positive, continuous, and decreasing for x1x\ge 1 and an=f(n)a_n=f(n), then

n=1an and 1f(x)dx either both converge or both diverge.\sum_{n=1}^{\infty}a_n \text{ and } \int_{1}^{\infty}f(x)\,dx \text{ either both converge or both diverge.}

The integral's value is not the sum; it only decides convergence.

Concept
pp-Series

n=11np\displaystyle\sum_{n=1}^{\infty}\frac{1}{n^{p}} converges if p>1p>1 and diverges if p1p\le 1. The case p=1p=1, 1n\sum\frac1n, is the harmonic series, which diverges.

Example
Integral test in action

Test n=11n2+1\displaystyle\sum_{n=1}^{\infty}\frac{1}{n^2+1}. With f(x)=1x2+1f(x)=\frac{1}{x^2+1} (positive, decreasing),

1dxx2+1=limb[arctanx]1b=π2π4=π4 (finite),\int_{1}^{\infty}\frac{dx}{x^2+1}=\lim_{b\to\infty}\big[\arctan x\big]_1^{b}=\frac{\pi}{2}-\frac{\pi}{4}=\frac{\pi}{4}\ (\text{finite}),

so the series converges. (The sum is not π4\frac{\pi}{4}; the integral only certifies convergence.)

Tip

Tip: 1n2\displaystyle\sum\frac{1}{n^{2}}, 1n3\sum\frac{1}{n^{3}} converge; 1n=1n1/2\sum\frac{1}{\sqrt n}=\sum\frac{1}{n^{1/2}} and 1n\sum\frac1n diverge. Memorize the p>1p>1 threshold --- it powers most comparison tests.

Comparison & Limit Comparison Tests

Concept
Direct Comparison Test

Suppose 0anbn0\le a_n\le b_n for all large nn.

  • [leftmargin=*,topsep=1pt]
  • If bn\sum b_n converges, then an\sum a_n converges (smaller than convergent).
  • If an\sum a_n diverges, then bn\sum b_n diverges (bigger than divergent).

Compare against a known geometric or pp-series.

Concept
Limit Comparison Test

If an,bn>0a_n,b_n>0 and limnanbn=L\displaystyle\lim_{n\to\infty}\frac{a_n}{b_n}=L with 0<L<0<L<\infty (a finite positive number), then an\sum a_n and bn\sum b_n both converge or both diverge. Choose bnb_n from the dominant-term behavior of ana_n.

Example
Direct comparison

n=11n2+n\displaystyle\sum_{n=1}^{\infty}\frac{1}{n^2+n} converges: since 1n2+n1n2\frac{1}{n^2+n}\le\frac{1}{n^2} and 1n2\sum\frac1{n^2} (p=2p=2) converges.

Example
Limit comparison

Test n=12n+1n3n+4\displaystyle\sum_{n=1}^{\infty}\frac{2n+1}{n^3-n+4}. For large nn the terms behave like 2nn3=2n2\frac{2n}{n^3}=\frac{2}{n^2}, so compare with bn=1n2b_n=\frac{1}{n^2}:

limn2n+1n3n+41n2=limn2n3+n2n3n+4=2  (0<2<).\lim_{n\to\infty}\frac{\frac{2n+1}{n^3-n+4}}{\frac{1}{n^2}}=\lim_{n\to\infty}\frac{2n^3+n^2}{n^3-n+4}=2\ \ (0<2<\infty).

Since 1n2\sum\frac1{n^2} converges, the series converges.

Tip

Tip: Pick the comparison series by keeping only the highest-degree term of numerator and denominator. Limit comparison is more forgiving than direct comparison when the inequality direction is awkward.

The Ratio Test & the Root Test

Concept
Ratio Test

Let L=limnan+1an\displaystyle L=\lim_{n\to\infty}\left|\frac{a_{n+1}}{a_n}\right|.

L<1converges (absolutely);L>1 (or )diverges;L=1inconclusive.L<1\Rightarrow \text{converges (absolutely)};\quad L>1\ (\text{or }\infty)\Rightarrow\text{diverges};\quad L=1\Rightarrow\text{inconclusive.}

Best when ana_n contains factorials or nnth powers.

Concept
Root Test

Let L=limnann=limnan1/n\displaystyle L=\lim_{n\to\infty}\sqrt[n]{|a_n|}=\lim_{n\to\infty}|a_n|^{1/n}. Same conclusions as the ratio test (L<1L<1 converges, L>1L>1 diverges, L=1L=1 inconclusive). Best when ana_n is raised to the nnth power, e.g. (n2n+1)n\left(\frac{n}{2n+1}\right)^{n}.

Example
Ratio test with a factorial

Test n=12nn!\displaystyle\sum_{n=1}^{\infty}\frac{2^{n}}{n!}:

L=limn2n+1(n+1)!n!2n=limn2n+1=0<1,L=\lim_{n\to\infty}\frac{2^{n+1}}{(n+1)!}\cdot\frac{n!}{2^{n}} =\lim_{n\to\infty}\frac{2}{n+1}=0<1,

so it converges. (In fact it sums to e21e^2-1.)

Example
Root test

Test n=1(3n+14n1)n\displaystyle\sum_{n=1}^{\infty}\Big(\frac{3n+1}{4n-1}\Big)^{n}:

L=limn(3n+14n1)=34<1  converges.L=\lim_{n\to\infty}\Big(\frac{3n+1}{4n-1}\Big)=\frac34<1\ \Rightarrow\ \textbf{converges}.
Tip

Tip: If the ratio/root test gives L=1L=1, switch tests (often a pp-series or comparison). Ratio-test L=1L=1 happens for every pp-series, so it can never settle those.

Alternating Series, Error Bound & Absolute vs. Conditional Convergence

Concept
Alternating Series Test (AST)

A series (1)nbn\displaystyle\sum(-1)^{n}b_n with bn>0b_n>0 converges if both hold:

  • [label=(),leftmargin=*,topsep=1pt]
  • bnb_n is decreasing (bn+1bnb_{n+1}\le b_n), and
  • limnbn=0\displaystyle\lim_{n\to\infty}b_n=0.
Concept
Alternating Series Error Bound

If an alternating series meets the AST conditions and converges to SS, then the error after NN terms is bounded by the first omitted term:

SSNbN+1.|S-S_N|\le b_{N+1}.
Concept
Absolute vs. conditional convergence

an\sum a_n converges absolutely if an\sum|a_n| converges (this forces an\sum a_n to converge). It converges conditionally if an\sum a_n converges but an\sum|a_n| diverges.

Example
Conditional convergence & error

The alternating harmonic series n=1(1)n+1n\displaystyle\sum_{n=1}^{\infty}\frac{(-1)^{n+1}}{n} converges by the AST (1n0\frac1n\downarrow0), but 1n\sum\frac1n diverges, so it converges conditionally. To approximate its sum within 0.010.01, we need bN+1=1N+10.01b_{N+1}=\frac{1}{N+1}\le0.01, i.e. N99N\ge99 terms.

Tip

Tip: To classify: test an\sum|a_n| first (ratio/comparison/pp-series). Converges \Rightarrow absolute. Diverges but the alternating series itself passes the AST \Rightarrow conditional.

Power Series: Radius & Interval of Convergence

Concept
Power series and its radius

A power series centered at cc is n=0an(xc)n\displaystyle\sum_{n=0}^{\infty}a_n(x-c)^{n}. There is a radius of convergence RR so it converges for xc<R|x-c|<R and diverges for xc>R|x-c|>R. Find RR with the ratio test: solve limnan+1(xc)n+1an(xc)n<1\displaystyle\lim_{n\to\infty}\left|\frac{a_{n+1}(x-c)^{n+1}}{a_n(x-c)^{n}}\right|<1.

Concept
Interval of convergence --- check the endpoints!

The set of xx where the series converges is the interval of convergence, (cR,c+R)(c-R,\,c+R) possibly including one or both endpoints. The ratio test never decides the endpoints (L=1L=1 there): substitute each endpoint and test the resulting numeric series separately.

Example
Full interval of convergence

Find the interval for n=1(x2)nn\displaystyle\sum_{n=1}^{\infty}\frac{(x-2)^{n}}{n}. Ratio test:

limn(x2)n+1n+1n(x2)n=x2<1  R=1, 1<x<3.\lim_{n\to\infty}\left|\frac{(x-2)^{n+1}}{n+1}\cdot\frac{n}{(x-2)^{n}}\right|=|x-2|<1\ \Rightarrow\ R=1,\ 1<x<3.

Endpoints: at x=3x=3, 1n\sum\frac{1}{n} diverges (harmonic). At x=1x=1, (1)nn\sum\frac{(-1)^n}{n} converges (AST). Interval: [1,3)\boxed{[1,3)}.

Concept
Differentiating & integrating a power series

Within the open interval of convergence you may differentiate or integrate term by term, and RR stays the same (endpoints may change):

f(x)=an(xc)n  f(x)=nan(xc)n1,fdx=C+ann+1(xc)n+1.f(x)=\sum a_n(x-c)^n \ \Rightarrow\ f'(x)=\sum n\,a_n(x-c)^{n-1},\quad \int f\,dx=C+\sum \frac{a_n}{n+1}(x-c)^{n+1}.
Tip

Tip: R=R=\infty means it converges for all xx (e.g. ex,sinx,cosxe^x,\sin x,\cos x); R=0R=0 means only at x=cx=c. Endpoints require a fresh test every time --- never assume.

Taylor & Maclaurin Series; the Lagrange Error Bound

Concept
Taylor and Maclaurin series

The Taylor series of ff centered at cc is

f(x)=n=0f(n)(c)n!(xc)n=f(c)+f(c)(xc)+f(c)2!(xc)2+f(x)=\sum_{n=0}^{\infty}\frac{f^{(n)}(c)}{n!}(x-c)^{n}=f(c)+f'(c)(x-c)+\frac{f''(c)}{2!}(x-c)^2+\cdots

A Maclaurin series is the special case c=0c=0.

Concept
Common Maclaurin series (memorize)
ex=n=0xnn!=1+x+x22!+(all x)sinx=n=0(1)nx2n+1(2n+1)!=xx33!+x55!(all x)cosx=n=0(1)nx2n(2n)!=1x22!+x44!(all x)11x=n=0xn=1+x+x2+(x<1)\begin{aligned} e^{x}&=\sum_{n=0}^{\infty}\frac{x^{n}}{n!}=1+x+\frac{x^2}{2!}+\cdots &&(\text{all }x)\\ \sin x&=\sum_{n=0}^{\infty}\frac{(-1)^{n}x^{2n+1}}{(2n+1)!}=x-\frac{x^3}{3!}+\frac{x^5}{5!}-\cdots &&(\text{all }x)\\ \cos x&=\sum_{n=0}^{\infty}\frac{(-1)^{n}x^{2n}}{(2n)!}=1-\frac{x^2}{2!}+\frac{x^4}{4!}-\cdots &&(\text{all }x)\\ \frac{1}{1-x}&=\sum_{n=0}^{\infty}x^{n}=1+x+x^2+\cdots &&(|x|<1) \end{aligned}
Example
Building a Taylor polynomial

Third-degree Maclaurin polynomial of f(x)=ln(1+x)f(x)=\ln(1+x): f(0)=0f(0)=0, f=11+xf'=\frac{1}{1+x} so f(0)=1f'(0)=1, f=1(1+x)2f''=-\frac{1}{(1+x)^2} so f(0)=1f''(0)=-1, f=2(1+x)3f'''=\frac{2}{(1+x)^3} so f(0)=2f'''(0)=2. Thus

P3(x)=0+1x+12!x2+23!x3=xx22+x33.P_3(x)=0+1\cdot x+\frac{-1}{2!}x^2+\frac{2}{3!}x^3=x-\frac{x^2}{2}+\frac{x^3}{3}.
Concept
Lagrange Error Bound

If Pn(x)P_n(x) is the nnth-degree Taylor polynomial about cc, the remainder satisfies

Rn(x)=f(x)Pn(x)M(n+1)!xcn+1,|R_n(x)|=|f(x)-P_n(x)|\le \frac{M}{(n+1)!}\,|x-c|^{\,n+1},

where MM is any bound on f(n+1)|f^{(n+1)}| between cc and xx. This measures how good the polynomial approximation is.

Example
Lagrange error bound applied

Approximate cos(0.2)\cos(0.2) with P2(x)=1x22P_2(x)=1-\frac{x^2}{2}. Here c=0c=0, n=2n=2, and f(3)(x)=sinx1|f^{(3)}(x)|=|\sin x|\le 1, so M=1M=1:

R2(0.2)13!0.23=0.00860.00133.|R_2(0.2)|\le\frac{1}{3!}|0.2|^{3}=\frac{0.008}{6}\approx0.00133.

So P2(0.2)=10.02=0.98P_2(0.2)=1-0.02=0.98 is accurate to about 0.00130.0013.

Which Convergence Test Should I Use?

Concept
A decision checklist
  • [label=.,leftmargin=*,topsep=1pt,itemsep=1pt]
  • nnth-term test --- does an0a_n\to0? If not, it diverges. (Always start here.)
  • Geometric? form arnar^n: converges iff r<1|r|<1, sum a1r\frac{a}{1-r}.
  • pp-series? form 1np\frac{1}{n^p}: converges iff p>1p>1.
  • Factorials or nnth powers? Use the ratio (or root) test.
  • Looks like a pp-series/geometric? Use direct or limit comparison.
  • Positive, decreasing, easily integrated? Use the integral test.
  • Alternating (1)nbn(-1)^n b_n? Use the AST; then check an\sum|a_n| for absolute vs. conditional.

Going Deeper: Advanced Series

Concept
Building new series from known ones (substitution)

Substitute into a known Maclaurin series instead of computing derivatives. Replace xx by any expression:

ex2=n=0(x2)nn!=n=0(1)nx2nn!,11+x2=n=0(x2)n=n=0(1)nx2n (x<1).e^{-x^2}=\sum_{n=0}^{\infty}\frac{(-x^2)^{n}}{n!}=\sum_{n=0}^{\infty}\frac{(-1)^n x^{2n}}{n!},\qquad \frac{1}{1+x^2}=\sum_{n=0}^{\infty}(-x^2)^{n}=\sum_{n=0}^{\infty}(-1)^n x^{2n}\ (|x|<1).
Example
Building by differentiating / integrating

Integrate 11+x2=(1)nx2n\dfrac{1}{1+x^2}=\sum(-1)^n x^{2n} term by term to get arctanx\arctan x:

arctanx=n=0(1)nx2n+12n+1=xx33+x55 (x1).\arctan x=\sum_{n=0}^{\infty}\frac{(-1)^n x^{2n+1}}{2n+1}=x-\frac{x^3}{3}+\frac{x^5}{5}-\cdots\ (|x|\le1).

Similarly integrating 11x\frac{1}{1-x} gives ln(1x)=n=1xnn-\ln(1-x)=\sum_{n=1}^{\infty}\frac{x^n}{n}.

Example
Telescoping series

Use partial fractions so most terms cancel:

n=11n(n+1)=n=1(1n1n+1).\sum_{n=1}^{\infty}\frac{1}{n(n+1)}=\sum_{n=1}^{\infty}\Big(\frac1n-\frac1{n+1}\Big).

SN=11N+11S_N=1-\frac{1}{N+1}\to1, so the sum is 1\boxed{1}. The value comes straight from the partial-sum limit, not a named test.

Example
Using a series to evaluate a tricky limit
limx0sinxxx3=limx0(xx36+x5120)xx3=limx0(16+x2120)=16.\lim_{x\to0}\frac{\sin x - x}{x^3} =\lim_{x\to0}\frac{\big(x-\frac{x^3}{6}+\frac{x^5}{120}-\cdots\big)-x}{x^3} =\lim_{x\to0}\Big(-\frac16+\frac{x^2}{120}-\cdots\Big)=-\frac16.

Substituting the Maclaurin series makes the 00\frac00 form transparent.

Example
Using a series to evaluate a non-elementary integral

ex2dx\int e^{-x^2}dx has no elementary antiderivative, but integrate the series term by term:

01ex2dx=01n=0(1)nx2nn!dx=n=0(1)nn!(2n+1)=113+110142+\int_0^{1}e^{-x^2}\,dx=\int_0^{1}\sum_{n=0}^{\infty}\frac{(-1)^n x^{2n}}{n!}\,dx =\sum_{n=0}^{\infty}\frac{(-1)^n}{n!\,(2n+1)}=1-\frac13+\frac{1}{10}-\frac{1}{42}+\cdots

The result is an alternating series, so its error bound is the first omitted term.

Concept
Endpoint analysis: the four possibilities

After finding RR by the ratio test, each endpoint is tested separately and independently. A single power series can produce any of (cR,c+R)(c-R,c+R), [cR,c+R)[c-R,c+R), (cR,c+R](c-R,c+R], or [cR,c+R][c-R,c+R]. Typical tools at an endpoint: the pp-series test, the nnth-term test, or the alternating series test. Never assume symmetry.

Example
Lagrange bound to guarantee accuracy

How many terms of the sinx\sin x series approximate sin(1)\sin(1) within 10410^{-4}? All derivatives of sin\sin are bounded by M=1M=1, so with c=0c=0, x=1x=1:

Rn(1)1(n+1)!.|R_n(1)|\le\frac{1}{(n+1)!}.

Since 17!=150400.0002\frac{1}{7!}=\frac{1}{5040}\approx0.0002 and 18!=1403200.0000248<104\frac{1}{8!}=\frac{1}{40320}\approx0.0000248<10^{-4}, taking n=7n=7 (through the x7x^7 term) guarantees the accuracy.

Tip

Big picture: Nearly every hard series problem reduces to a known one --- a geometric series, a pp-series, or one of the four Maclaurin series --- reached by algebra, substitution, differentiation, or integration. Recognize the skeleton, then transform.

Formulas, Proofs & Tips

Tip
L'Hôpital's Rule
limxcf(x)g(x)=limxcf(x)g(x)when 00 or \lim_{x\to c}\frac{f(x)}{g(x)} = \lim_{x\to c}\frac{f'(x)}{g'(x)} \quad\text{when } \tfrac{0}{0} \text{ or } \tfrac{\infty}{\infty}

What it means. If a limit gives the indeterminate form 00\tfrac{0}{0} or \tfrac{\infty}{\infty}, differentiate the top and bottom separately and try again.

Why it works. Near cc both functions are approximated by their tangent lines, f(x)f(c)(xc)f(x)\approx f'(c)(x-c) and g(x)g(c)(xc)g(x)\approx g'(c)(x-c). The common factor (xc)(x-c) cancels in the ratio, leaving f(c)g(c)\tfrac{f'(c)}{g'(c)}.

Tip. Check the form FIRST — applying it to a limit that isn't indeterminate gives a wrong answer. It is not the quotient rule: do not differentiate fg\tfrac{f}{g} as a whole.

Tip
Convergence tests
arn converges    r<1;1np converges    p>1\sum ar^{n} \text{ converges} \iff |r|<1; \qquad \sum \frac{1}{n^{p}} \text{ converges} \iff p>1

What it means. Geometric series need shrinking terms; pp-series need them to shrink fast enough.

Example. n1(12)n\sum_{n\ge1}\left(\tfrac12\right)^n converges since r=12<1|r|=\tfrac12<1, to 1/211/2=1\tfrac{1/2}{1-1/2}=1.

Why it works. The harmonic series p=1p=1 diverges: grouping terms as 12\tfrac12, 13+14>12\tfrac13+\tfrac14>\tfrac12, 15++18>12\tfrac15+\cdots+\tfrac18>\tfrac12, and so on adds 12\tfrac12 infinitely often. Larger pp shrinks the terms fast enough for the total to settle.

Tip. Terms tending to 00 is necessary but not sufficient — the harmonic series is the standard counterexample.