Everything in analysis (continuity, derivatives, integrals, infinite series) is built on one idea: a quantity getting as close as we like to a value. In §58 Hardy turns the vague phrase “ϕ(n) is nearly l when n is large” into a definition precise enough to prove things with. This lesson is about that definition: how to read it, how to use it, and how it can fail.
The idea
Take the sequence ϕ(n)=1/n. Everyone agrees that it “tends to 0”, but what exactly is being claimed? Not that some term equals0: none does. Not that the terms are small: 1/1000 is small for one purpose and huge for another.
Hardy’s answer, prepared in §56, is a challenge. Someone names a tolerance δ, say 0.01. We reply with a place n0 in the sequence from which on every term is within δ of 0: here n0=101 will do, because 1/n<0.01 as soon as n>100. The challenger names a smaller δ; we find a later n0. The claim “1/n tends to 0” means that we can always answer, however small δ is.
ExploreAnswer the challenge for 1/n
Try this.Set δ to about 0.1 and find the smallest n0 for which no term from n0 on is orange. Then make δ ten times smaller and do it again. (Use the menu to show more terms when you need them.)
ϕ(n)=n1 , candidate limit l=0
nφ(n)
band l ± δ
l
n₀ (drag)
n ≥ n₀, good
n ≥ n₀, bad
10 terms with n ≥ n₀ fall outside the band: n = 1, 2, 3, 4, 5, ….
What to notice
Each smaller δ needs a larger n0, roughly 1/δ of them. But some n0 always works: the green band never stays empty to the right. That is what “tends to 0” means.
Two features of the game matter, and both are easy to miss.
The order of the moves. First δ is chosen, thenn0. Our n0 may depend on δ, which is why Hardy writes it as n0(δ). A single n0 that worked for every δ at once would be a much stronger (and usually false) claim.
“From n0 on”, not “at n0”. It is not enough that one term is inside the band. Every term with n≥n0 must be.
Precise statements
Here is Hardy’s definition in modern dress. The notation ∣ϕ(n)−l∣ is the distance between ϕ(n) and l; Hardy calls it their difference “taken positively”.
DefinitionLimit of a sequenceHardy: Definition I, §58
The sequence ϕ(n)tends to the limit l as n→∞ if, for every positive number δ, however small, there is a number n0(δ) such that
∣ϕ(n)−l∣<δfor all n≥n0(δ).
We then write limn→∞ϕ(n)=l, or simply ϕ(n)→l.
In §59 Hardy reads the definition as a picture, and the widget above draws exactly that picture. Plot the points (n,ϕ(n)) and draw the lines y=l−δ and y=l+δ: this is the green band. The definition says that, however narrow the band, we can draw a vertical line x=n0 (the purple line) such that every point on or to the right of it lies inside the band.
Sequences like n2 do not approach any number, but they do something just as definite: they eventually exceed any bound. §60 gives this its own definition, with a large number Δ in place of the small δ.
DefinitionTending to +∞Hardy: Definition II, §60
The sequence ϕ(n)tends to +∞ if, for every number Δ, however large, there is a number n0(Δ) such that
ϕ(n)>Δfor all n≥n0(Δ).
We write ϕ(n)→+∞ (or ϕ(n)→∞).
The symbol ∞ here is not a number. “ϕ(n)→+∞” is a single phrase, defined as a whole by the sentence above, just as “ϕ(n)→l” is.
Our first result is the first of Hardy’s “points concerning the definitions” (§61): only the tail of a sequence matters.
TheoremFinitely many terms do not matterHardy: §61 (1)
Let ψ(n)=ϕ(n) for every n except finitely many. If ϕ(n)→l then ψ(n)→l, and if ϕ(n)→+∞ then ψ(n)→+∞.
The other points in §61 are warnings, and each one corrects a natural misreading.
Infinitely many changes can matter. Change 1/n to 1 whenever n is a multiple of 100. However far out we put n0, there are later multiples of 100, and for δ≤1 they all lie outside the band. The new sequence has no limit.
The test is for every n≥n0. For that same sequence we can certainly find a single n0 (say n0=101) at which ∣ϕ(n0)∣<δ. That proves nothing; the multiples of 100 beyond it still fail.
Signs need the absolute value. For (−1)n/n the terms fall on both sides of the limit 0, so the condition must be ∣ϕ(n)∣<δ, not ϕ(n)<δ.
A limit need not be a value, and a value need not be the limit.1/n never equals its limit 0. The constant sequence 0,0,0,… equals its limit at every term. The sequence (sin21nπ)/n equals its limit 0 at every even n and at no odd n.
Large is not the same as tending to infinity.(−1)nn is enormous in size for large n, but it changes sign at every step, so it never stays above Δ=0. A sequence that tends to +∞ is eventually positive.
ExploreInfinitely many exceptions
Try this.This is 1/n with every hundredth term changed to 1. Find an n0 that works for δ=0.5 with only the first 40 terms shown. Then show 250 terms.
ϕ(n)={11/n100∣notherwise , candidate limit l=0
nφ(n)
band l ± δ
l
n₀ (drag)
n ≥ n₀, good
n ≥ n₀, bad
2 terms with n ≥ n₀ fall outside the band: n = 1, 2.
What to notice
Every apparent success is undone by the next multiple of 100. The terms in between do tend to 0, but the definition demands every term from n0 on, so no n0 works for any δ≤1.
Proofs
A proof that ϕ(n)→l is a strategy for the game: a rule that produces a winning n0 from any δ the challenger names. Here is the simplest example, which Hardy works out at the end of §58.
Proof1/n→0
Strategy
Given an arbitrary δ>0, we must name an n0 and show that 1/n<δ for every n≥n0. Since 1/n gets smaller as n grows, it is enough to make 1/n0<δ, that is, n0>1/δ.
Let δ>0 be given
The proof must work for every positive δ, so we may not assume anything about it except δ>0.
Choose n0=[1/δ]+1
Here [x] is the integral part of x (the greatest integer not exceeding x). Since [x]>x−1, we have n0>1/δ. This is the smallest choice that works, but any larger integer would do as well.
Check every n≥n0
If n≥n0 then n>1/δ, and since both sides are positive, 1/n<δ. Also 1/n>0, so ∣1/n−0∣=1/n<δ.
Conclude
Every δ>0 has a matching n0(δ), which is exactly what Definition I requires. Hence lim1/n=0.
∎
The rule n0(δ)=[1/δ]+1 is a function of δ. Drawing its graph shows Hardy’s remark that “the smaller δ is, the larger, as a rule, will be n0(δ)”. It grows without bound as δ shrinks, but it is finite for every δ>0.
ExploreThe winning rule n0(δ) for 1/n
Try this.Drag δ from 1 down towards 0.05 and watch the step function climb.
δn₀(δ)
smallest n₀ for 1/n
1/δ
your δ
For δ=0.15 the smallest choice is n0=[1/δ]+1=7. Halve δ and n₀ roughly doubles; but it is always finite.
What to notice
n0(δ) jumps at δ=1/2,1/3,1/4,… and stays close to the dashed curve 1/δ. A proof only needs some valid n0, not the smallest one, so a cruder rule such as n0=[2/δ]+1 would prove the same thing.
The proof of our theorem uses the same strategy, with one new idea: push n0 past every term that was changed.
ProofFinitely many terms do not matter
Strategy
Use the n0 that works for ϕ, but if necessary move it to the right of every altered term. Beyond that point, ψ and ϕ are the same sequence.
Name the altered terms
Only finitely many n have ψ(n)=ϕ(n), so they have a largest member; call it N. (If there are none, take N=0.) For every n>N we have ψ(n)=ϕ(n).
Take any δ>0 and use the limit of ϕ
Since ϕ(n)→l, there is an n0 with ∣ϕ(n)−l∣<δ for all n≥n0.
Let n1=max(n0,N+1)
If n≥n1 then n≥n0, so ∣ϕ(n)−l∣<δ; and n>N, so ψ(n)=ϕ(n). Together, ∣ψ(n)−l∣<δ for all n≥n1.
The case +∞ is the same
Replace “∣ϕ(n)−l∣<δ” by “ϕ(n)>Δ” throughout. The argument does not change.
∎
Notice why this argument cannot handle infinitely many changes: the step “they have a largest member” fails, and with it the choice of n1.
Worked examples and exercises
Try each one before opening the hints. Hardy’s examples are statements to be proved, so “show” is always implied.
ExerciseTending to −∞●○○Hardy: §60
Hardy leaves the definition of “ϕ(n)→−∞” to the reader. Write it down, and use it to prove that −n2→−∞.
Hint
Mirror Definition II: “however large Δ is, ϕ(n) is eventually below −Δ.”
Hint
For −n2<−Δ you need n2>Δ. When Δ>0 that means n>Δ.
Solution
Definition.ϕ(n)→−∞ if for every number Δ, however large, there is an n0(Δ) such that ϕ(n)<−Δ for all n≥n0(Δ). Equivalently, −ϕ(n)→+∞.
Proof for −n2. Let Δ be given. If Δ≤0 then n0=1 works, since −n2≤−1<0≤−Δ. If Δ>0, take n0=[Δ]+1, so that n0>Δ. For n≥n0 we get n>Δ, so n2>Δ and −n2<−Δ.
Exercise●○○Hardy: Examples XXIII, 1
Let k be a rational number. Show that nk→+∞ if k>0, that nk→0 if k<0, and that nk→1 if k=0. For k=4, how large must n be to make n4>10000?
Hint
For k>0: nk>Δ exactly when n>Δ1/k (with Δ>0).
Hint
For k<0 write nk=1/n∣k∣ and ask when this is less than δ.
Solution
k>0. Let Δ>0 (for Δ≤0 any n0 works, since nk>0). Since x↦xk is increasing on the positive numbers, nk>Δ if and only if n>Δ1/k. So n0=[Δ1/k]+1 works, and nk→+∞.
k<0. Let δ>0. Now 0<nk=1/n∣k∣, and 1/n∣k∣<δ if and only if n∣k∣>1/δ, that is, n>δ−1/∣k∣. So n0=[δ−1/∣k∣]+1 works, and nk→0.
k=0. Then nk=1 for every n, a constant sequence, and n0=1 works for every δ.
k=4.n4>104 exactly when n>10, so n≥11 is needed. In the same way n4>108 exactly when n≥101.
Exercise●○○Hardy: Examples XXIII, 5
Let ϕ(n)=1000000/n and ψ(n)=n/1000000. Show that ϕ(n)→0 and ψ(n)→+∞, even though ϕ(n)>ψ(n) for every n<1000000.
Hint
Copy the proof for 1/n, with 106/δ in place of 1/δ.
Solution
Given δ>0, take n0=[106/δ]+1. For n≥n0 we have n>106/δ, so 0<106/n<δ. Hence ϕ(n)→0.
Given Δ>0, take n0=[106Δ]+1. For n≥n0 we have n>106Δ, so n/106>Δ. Hence ψ(n)→+∞.
Finally, ϕ(n)>ψ(n) means 1012>n2, that is, n<106. The first million terms say nothing about the limit. That is the theorem on altering finitely many terms seen from another side: only the tail counts.
Exercise●●○Hardy: Examples XXIII, 6
Decide how each of these sequences behaves as n→∞, and prove it:
n−(−1)n1,n−(−1)n,n{1−(−1)n}.Hint
In the first two, n−(−1)n is either n−1 or n+1. So it is at least n−1.
Hint
Write out the first few terms of the third: 2,0,6,0,10,0,…. Which definition does it break, and how?
Solution
First. For n≥2 we have n−(−1)n≥n−1≥1, so 0<ϕ(n)≤1/(n−1). Given δ>0, take n0=[1/δ]+2. Then for n≥n0 we have n−1>1/δ, so ϕ(n)≤1/(n−1)<δ. Hence ϕ(n)→0.
Second.n−(−1)n≥n−1. Given Δ, take n0=max(1,[Δ]+2). For n≥n0, n−1>Δ, so the term exceeds Δ. Hence it tends to +∞.
Third. The terms are 0 for even n and 2n for odd n. They do not tend to +∞: with Δ=1, every even n≥n0 gives a term 0<1, whatever n0 is. They do not tend to a limit l either: with δ=1, the odd terms 2n exceed l+1 as soon as n>(l+1)/2, so infinitely many terms lie outside the band, whatever n0 is. (In the language of the next lesson, the sequence oscillates infinitely.)
Exercise●○○Hardy: Examples XXIII, 7
Let θ be any real number. Show that (sinnθπ)/n→0.
Hint
You know nothing useful about sinnθπ except that it lies between −1 and 1.
Solution
Since ∣sinnθπ∣≤1, we have ∣ϕ(n)∣≤1/n for every n. Given δ>0, take n0=[1/δ]+1. For n≥n0, ∣ϕ(n)−0∣≤1/n<δ. The same n0 works for every θ: bounding the hard part by something simple is the main technique in this chapter.
ExerciseWhen one n0 serves every δ●●○Hardy: §58
Show that a single N works for everyδ>0 (that is, ∣ϕ(n)−l∣<δ for all n≥N and all δ>0) if and only if ϕ(n)=l for every n≥N.
Hint
One direction is immediate. For the other, fix n≥N: what do you know about the number ∣ϕ(n)−l∣?
Solution
If ϕ(n)=l for all n≥N, then ∣ϕ(n)−l∣=0<δ for every such n and every δ>0.
Conversely, fix n≥N and let d=∣ϕ(n)−l∣≥0. By hypothesis d<δ for every δ>0. If d>0, then taking δ=d gives d<d, which is impossible. So d=0, that is, ϕ(n)=l. Thus the only sequences for which n0 need not depend on δ are those that are eventually constant. For all other convergent sequences, n0 really is a function of δ.
Check yourself
Check yourselfChoose one
Which statement says that ϕ(n)→l?
Why
The order matters: δ comes first and n0 may depend on it. The first option fixes one n0 for all δ, which forces the sequence to be eventually constant (last exercise). The third only asks for one term near l; the sequence 1/n with 1 at multiples of 100 satisfies it for l=1 without tending to 1. The fourth says every term equals l.
Check yourselfEnter a number
For ϕ(n)=1/n, l=0 and δ=0.004, what is the smallest n0 that works?
Why
1/n<0.004 means n>250. The term 1/250=0.004 is not less than 0.004, so the smallest n0 is 251=[1/δ]+1.
Check yourselfTrue or false?
If ϕ(n)→l, then ϕ(n)=l for every n.
Why
False. A limit may be a value, even infinitely often: 1/n with every hundredth term replaced by 0 still tends to 0, and a constant sequence equals its limit everywhere. What is true is that the limit need not be a value at all, as 1/n shows.
Check yourselfTrue or false?
(−1)nn→+∞, because its terms become arbitrarily large.
Why
False. Take Δ=0. Every odd n gives a negative term, so no n0 makes all later terms exceed 0. The sizes ∣ϕ(n)∣=n do tend to +∞, but the sequence itself does neither that nor anything else definite: it oscillates.
Check yourselfChoose one
Let ϕ(n)=1 when n is a multiple of 100 and ϕ(n)=1/n otherwise. Which is correct?
Why
Whatever n0 is chosen, there is a multiple of 100 beyond it, where ∣ϕ(n)−0∣=1>21. The second option is Hardy’s point (3): finding one good n0-th term proves nothing. The last option fails for the same reason with the roles swapped: the terms 1/n stay far from 1.
Hardy’s language
A limit is not a value of the function: it is something quite distinct from these values, though it is defined by its relations to them and may possibly be equal to some of them.
— G. H. Hardy, §61
Summary
Summary
ϕ(n)→l means: for everyδ>0 there is an n0(δ) with ∣ϕ(n)−l∣<δ for alln≥n0(δ).
Picture: however narrow the band l±δ, all points from some vertical line onwards lie inside it.
ϕ(n)→+∞ means: for every Δ there is an n0(Δ) with ϕ(n)>Δ for all n≥n0(Δ). The symbol ∞ is only part of a phrase.
To prove a limit, give a rule for n0 in terms of δ and check it works; it need not be the smallest.
Changing finitely many terms never matters; changing infinitely many can. A limit need not be a value, and may be one.