The geometric sequence xn is the yardstick of analysis. Almost every convergence test in Chapter VIII works by comparing something with xn, the geometric series of §78 is built from it, and its behaviour decides that of nkxn, xn/n! and many others. §72 settles xn for every real x, and shows the monotone theorem of §69 earning its keep: it gives the limit without any estimate of n0.
The idea
Multiplying by x again and again does one of a few things. If x>1, each step makes the term larger by a fixed factor, and the terms run off to infinity. If 0<x<1, each step shrinks the term by a fixed factor, and the terms die away to 0. A negative x does the same in size but flips the sign every time. The borderline cases x=1 and x=−1 are exactly where “larger” or “smaller by a fixed factor” fails.
ExploreFive regimes of xn
Try this.Drag x slowly from 1.6 down to −1.6. Stop at 1.05, 1, 0.95, −0.95, −1 and −1.05. Watch which part of the strip under the plot lights up.
ϕ(n)=xn,x=0.9
nxⁿ
xⁿ (points beyond ±4.5 are off the plot)
xⁿ → 0. Here ∣x∣n<0.01 for every n≥44; the closer ∣x∣ is to 1, the later this happens, but it always does.
What to notice
The behaviour changes only at x=1 and x=−1. Just inside, ∣x∣=0.95, the terms still reach 0, only later (the read-out says when ∣x∣n<0.01); just outside, ∣x∣=1.05, they still escape, only slowly. There is no gradual transition in the limit, only in the speed.
The surprise in Hardy’s proof is that we never have to compute when xn becomes small or large. For x>1 the sequence increases, so (§69) it tends to a limit or to +∞. A limit l would have to satisfy l=xl, because xn+1 is both x⋅xn and a shifted copy of xn. With x>1 and l≥1 that is impossible, so xn→+∞.
Precise statements
TheoremThe limit of xnHardy: §72
As n→∞,
xn→+∞(x>1);xn→1(x=1);xn→0(−1<x<1);
xn oscillates finitely when x=−1, and oscillates infinitely when x<−1.
Two small facts are used in the proof.
LemmaShifting does not change a limitHardy: Examples XXV, 7
If ϕ(n)→l, then ϕ(n+1)→l (and more generally ϕ(n+p)→l for any fixed integer p).
LemmaBernoulli’s inequalityHardy: §72, Example
If δ>0 and n is a positive integer, then (1+δ)n≥1+nδ.
Proofs
ProofShifting does not change a limit
Strategy
The same n0 works, because n≥n0 implies n+1≥n0.
Use the threshold for ϕ
Given δ>0, there is n0 with ∣ϕ(m)−l∣<δ for all m≥n0. If n≥n0, then m=n+1≥n0, so ∣ϕ(n+1)−l∣<δ. (For ϕ(n+p) with p negative, use n0−p instead.)
∎
ProofThe limit of xn (Hardy's proof)
Strategy
For x>0 the sequence is monotone, so it has a limit or tends to +∞; any limit l must satisfy l=xl. Negative x reduces to positive x through ∣xn∣=∣x∣n.
x=0 and x=1
The sequences are constant (0 and 1), so they tend to 0 and 1.
x>1
xn+1=x⋅xn>xn, so the sequence increases. By §69 it tends to a limit l or to +∞. If xn→l, then xn+1→l (shifting), but also xn+1=x⋅xn→xl (Theorem II). By uniqueness of limits, l=xl, so (x−1)l=0 and l=0. But every term is greater than 1, and by Cor. 3 of §69 the limit is at least each term: l≥x>1. This contradiction leaves xn→+∞.
0<x<1
Now the sequence decreases and is positive, so it is bounded below by 0 and tends to a limit l≥0. As before l=xl, and since x=1, l=0.
−1<x<0
∣xn−0∣=∣x∣n and 0<∣x∣<1, so ∣xn∣<δ whenever ∣x∣n<δ, which holds for large n by the previous step. Hence xn→0.
x=−1 and x<−1
(−1)n oscillates finitely (Oscillating sequences). If x<−1, then ∣xn∣=∣x∣n→+∞, so xn is unbounded; its terms alternate in sign, so it is positive for infinitely many n and negative for infinitely many n, and so it tends neither to +∞ nor to −∞. It oscillates infinitely.
∎
Hardy’s alternative for x>1 is more direct and gives an explicit n0.
ProofBernoulli’s inequality, and xn→+∞ again
Strategy
Expand (1+δ)n by the binomial theorem and throw away all terms after the second; they are positive.
The inequality
(1+δ)n=1+nδ+(2n)δ2+⋯+δn≥1+nδ, since every omitted term is positive (and for n=1 there is equality).
Apply it
Write x=1+δ with δ=x−1>0. Given Δ, if n>Δ/δ then xn≥1+nδ>Δ. So n0=[Δ/δ]+1 works, and xn→+∞. (For 0<x<1, apply this to 1/x>1: (1/x)n→+∞, so xn→0.)
∎
The explorer’s “Bernoulli” toggle draws the line 1+n(x−1) under the points. The bound is very generous (it grows linearly, the true terms exponentially), but it is enough.
Ratios: the idea that grows out of xn
If each term is at least K times the previous one, with K>1, the sequence grows at least as fast as Kn. If each term is at most K times the previous one in size, with K<1, it shrinks at least as fast as Kn. This is the germ of the ratio test for series. Try it on nxn with x=0.9.
Explorenxn: a rise before the fall
Try this.With x=0.9, find an n0 for δ=0.5 (show 250 terms). Then move x to 0.98.
ϕ(n)=nxn , candidate limit l=0
nφ(n)
band l ± δ
l
n₀ (drag)
n ≥ n₀, good
n ≥ n₀, bad
42 terms with n ≥ n₀ fall outside the band: n = 1, 2, 3, 4, 5, ….
What to notice
The terms first grow (the factor n wins early), then fall (the factor xn wins). The ratio of consecutive terms is nn+1x, which tends to x<1; once it is below some K<1 the terms shrink geometrically. The closer x is to 1, the later and higher the peak, but the limit is still 0.
Worked examples and exercises
ExerciseRatio at least K>1●○○Hardy: Examples XXVII, 1–2
Suppose ϕ(n)>0 and ϕ(n+1)≥Kϕ(n) for all n≥n0, where K>1. Show that ϕ(n)→+∞.
Solution
By induction, ϕ(n)≥Kn−n0ϕ(n0) for n≥n0. Here ϕ(n0)>0 is a fixed number and Kn−n0=Kn/Kn0→+∞ by the theorem. So given Δ, Kn−n0>Δ/ϕ(n0) for large n, and then ϕ(n)>Δ.
ExerciseRatio tending to l with ∣l∣<1●●○Hardy: Examples XXVII, 4, 6
Show that if ϕ(n+1)/ϕ(n)→l with −1<l<1, then ϕ(n)→0.
Hint
Pick K with ∣l∣<K<1, say K=21(1+∣l∣). For large n, ∣ϕ(n+1)∣<K∣ϕ(n)∣.
Solution
Let K=21(1+∣l∣), so ∣l∣<K<1. Applying the definition with δ=K−∣l∣>0, there is n0 with ∣ϕ(n+1)/ϕ(n)∣<∣l∣+δ=K for n≥n0, that is, ∣ϕ(n+1)∣<K∣ϕ(n)∣. By induction ∣ϕ(n)∣≤Kn−n0∣ϕ(n0)∣ for n≥n0. Since Kn→0, the right side tends to 0; so ∣ϕ(n)∣<δ′ for large n, for any δ′>0. Hence ϕ(n)→0.
Exercisenrxn●●○Hardy: Examples XXVII, 7
Let r be a positive integer. Find the behaviour of nrxn for every real x.
Hint
For x=0 the ratio of consecutive terms is (nn+1)rx→x.
Solution
If x=0 every term is 0. Otherwise the ratio is (1+1/n)rx→x (Theorem II, r times). For 0<x<1 or −1<x<0, the previous exercise gives nrxn→0. For x>1, the ratio is eventually above K=21(1+x)>1, so the terms tend to +∞ (Examples XXVII, 1). For x=1 the sequence is nr→+∞. For x≤−1, ∣nrxn∣=nr∣x∣n≥nr→+∞ and the signs alternate, so it oscillates infinitely. In short: the factor xn decides everything, and nr matters only when ∣x∣=1.
Exercisenx→1●●○Hardy: Examples XXVII, 10
Show that x1/n→1 for every x>0.
Hint
For x>1, the sequence x1/n decreases and stays above 1. If its limit l were greater than 1, what would x≥ln say?
Solution
If x=1 there is nothing to prove. Let x>1. Then x1/n>1, and x1/(n+1)<x1/n (raise both sides to the power n(n+1): xn<xn+1). So x1/n decreases and is bounded below by 1; it tends to a limit l≥1, and by the decreasing form of Cor. 3, x1/n≥l for every n, that is, x≥ln for every n. If l>1 then ln→+∞, which contradicts x≥ln. So l=1. If 0<x<1, then x1/n=1/(1/x)1/n→1/1=1 by Theorem III.
Exercisenn→1●●○Hardy: Examples XXVII, 11
Show that n1/n→1.
Hint
Write n1/n=1+hn with hn≥0, and keep only the third term of the binomial expansion of (1+hn)n.
Solution
For n≥2, n1/n>1, so hn>0, and n=(1+hn)n≥(2n)hn2=21n(n−1)hn2. Hence hn2≤2/(n−1) and 0<hn≤2/(n−1). Given δ>0, this is less than δ as soon as n−1>2/δ2. So hn→0 and n1/n→1. (Hardy’s proof is different: he shows that n1/n decreases from n=3 on, using (1+1/n)n<3 from the next lesson, and then argues as in the previous exercise.)
Exercisexn/n!→0●○○Hardy: Examples XXVII, 12–13
Show that xn/n!→0 for every real x, and deduce that nn!→+∞.
Solution
With un=xn/n! (and x=0), un+1/un=x/(n+1)→0, so by the ratio exercise un→0. For the second claim, let Δ>0. Taking x=Δ, we get Δn/n!<1 for large n, that is, n!>Δn, that is, nn!>Δ. So nn!→+∞.
Check yourself
Check yourselfChoose one
How does (−1)n, that is xn with x=−1, behave?
Why
It takes the values −1,1,−1,1,…: bounded, with no limit. Only for x<−1 do the sizes grow, giving infinite oscillation; only for ∣x∣<1 do the terms tend to 0.
Check yourselfTrue or false?
(0.9999)n does not tend to 0, because it is still greater than 0.9 after a thousand terms.
Why
False. (0.9999)1000≈0.905, but (0.9999)n<0.01 once n>log0.01/log0.9999≈46050. A slow start says nothing about the limit; the theorem covers every x with ∣x∣<1.
Check yourselfEnter a number
What is the smallest n with (0.9)n<0.01?
Why
(0.9)n<0.01 means n>log0.01/log0.9=43.7…, so n=44. Since (0.9)n decreases, every later n also works.
Check yourselfChoose one
In Hardy’s proof for x>1, the equation l=xl comes from:
Why
xn+1→l because shifting does not change a limit, and xn+1=x⋅xn→xl by Theorem II. A sequence has only one limit, so l=xl. Monotonicity is used separately, to know that a limit or +∞ is the only possibility.
Hardy’s language
Summary
Summary
xn→+∞ for x>1, →1 for x=1, →0 for ∣x∣<1; it oscillates finitely for x=−1 and infinitely for x<−1.
Hardy’s proof: monotone, so a limit or +∞; a limit would satisfy l=xl. Bernoulli’s (1+δ)n≥1+nδ gives an explicit alternative.
Shifting a sequence does not change its limit; this, with uniqueness, is how "l=xl" arises.
Ratio principle: if ∣ϕ(n+1)/ϕ(n)∣→l<1 then ϕ(n)→0; if the ratio tends to l>1 (positive terms) then ϕ(n)→+∞. So nrxn→0 and xn/n!→0 for ∣x∣<1 and all x respectively.