Chapter I · Real variables

Points of accumulation

§18 30 min

When points of a set crowd arbitrarily close to a given point, and why "arbitrarily close" forces "infinitely many".

Why this matters

A limit is a point that the terms of a sequence crowd towards. Before Hardy defines limits (Chapter IV), he isolates the purely geometric version of that idea for an arbitrary set of numbers: a point of accumulation is a point that the set crowds towards, whether or not the point belongs to the set. The definition is the first one in the book of the shape “however small δ\delta is, …”, which every definition of analysis will share.

The idea

Take a set SS of points on the line and a point ξ\xi. Draw a small interval around ξ\xi, from ξ−δ\xi - \delta to ξ+δ\xi + \delta, and ask whether it contains any point of SS other than ξ\xi itself. Two things can happen.

  • For some δ\delta the answer is no: a small enough interval isolates ξ\xi from the rest of SS. Then ξ\xi is not a point of accumulation.
  • For every δ\delta, however small, the answer is yes. Then ξ\xi is a point of accumulation of SS.

The phrase “other than ξ\xi itself” matters: a member of SS does not count as a crowd around itself.

Explore Probe a point with a shrinking interval

Try this. For S={1/n}S = \{1/n\}, test ξ=1/2\xi = 1/2, 1/41/4 and 0.40.4, and shrink δ\delta until the read-out says the point is isolated, if you can. Then test ξ=0\xi = 0. Switch to the positive integers, to 1/m+1/n1/m + 1/n (try 1/31/3 and 0.30.3), and to the positive rationals (try 00 and 2\sqrt2).

S={1/n:n=1,2,3,… }S = \{1/n : n = 1, 2, 3, \dots\}, probe (ξ−δ, ξ+δ)(\xi - \delta,\ \xi + \delta)
−0.4−0.3−0.2−0.100.10.20.30.4ξ = 0
  • points of S
  • the interval (ξ − δ, ξ + δ)
  • ξ (hollow if not in S)
Infinitely many points of S lie in the interval, and however small you make δ that stays true: ξ is a point of accumulation, although it is not itself in S.
What to notice

Points of {1/n}\{1/n\} other than 00 are isolated, whether in SS (1/21/2) or not (0.40.4): a small enough δ\delta empties the interval. At 00 no δ\delta works; infinitely many 1/n1/n are always inside, although 00 is not in SS. In {1/m+1/n}\{1/m + 1/n\} the point 1/31/3 is a member and a point of accumulation, while 0.3=15+1100.3 = \tfrac15 + \tfrac1{10} is an isolated member. For the positive rationals every ξ≥0\xi \ge 0 is a point of accumulation, members or not.

Notice something in the probe: whenever the interval around a point of accumulation contains some points, it contains infinitely many. That is not a coincidence, and it is the first theorem below.

Precise statements

Definition Set (aggregate) Hardy: §18

A set (Hardy: aggregate) of numbers, or of the corresponding points on the line, is any collection of real numbers defined in any way whatever.

Definition Point of accumulation Hardy: §18

A point ξ\xi is a point of accumulation of the set SS if every interval (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) with δ>0\delta > 0 contains at least one point of SS other than ξ\xi. The point ξ\xi may or may not belong to SS.

Theorem Infinitely many points near a point of accumulation Hardy: §18

If ξ\xi is a point of accumulation of SS, then every interval (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains infinitely many points of SS.

Examples Four model sets Hardy: Examples IX

(1) The positive integers, or all the integers, have no points of accumulation. (2) For the set of all rationals, every real number is a point of accumulation. (3) The set 1,12,13,…1, \tfrac12, \tfrac13, \dots has exactly one point of accumulation, 00. (4) For the positive rationals, the points of accumulation are 00 and all positive numbers.

Proofs

Proof Every interval round a point of accumulation contains infinitely many points of SS
Strategy

Find one point of SS in the interval. Then shrink the interval so that it no longer reaches that point, and find another; repeat. Each new point is distinct from all earlier ones, because the intervals keep shrinking past them.

A first point

By definition the interval (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains a point ξ1\xi_1 of SS with ξ1≠ξ\xi_1 \ne \xi.

A second, different point

Put δ1=∣ξ1−ξ∣\delta_1 = |\xi_1 - \xi|, which is positive and at most δ\delta. The interval (ξ−δ1,ξ+δ1)(\xi - \delta_1, \xi + \delta_1) does not contain ξ1\xi_1, but by the definition it contains some ξ2∈S\xi_2 \in S with ξ2≠ξ\xi_2 \ne \xi. Then ∣ξ2−ξ∣<δ1=∣ξ1−ξ∣|\xi_2 - \xi| < \delta_1 = |\xi_1 - \xi|, so ξ2≠ξ1\xi_2 \ne \xi_1, and ξ2\xi_2 lies in the original interval.

Repeat

With δ2=∣ξ2−ξ∣\delta_2 = |\xi_2 - \xi| we find ξ3\xi_3 strictly nearer to ξ\xi than ξ2\xi_2, and so on. The distances ∣ξk−ξ∣|\xi_k - \xi| strictly decrease, so the points ξ1,ξ2,ξ3,…\xi_1, \xi_2, \xi_3, \dots are all different, and all lie in (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta).

Conclude

We can produce as many distinct points of SS in the interval as we please, which is what “infinitely many” means (§3).

Proof Examples IX, 3: the only point of accumulation of {1/n}\{1/n\} is 00
Strategy

Show that 00 passes the test for every δ\delta, and that every other point fails it for some δ\delta.

00 is a point of accumulation

Given δ>0\delta > 0, choose an integer n>1/δn > 1/\delta. Then 0<1/n<δ0 < 1/n < \delta, a point of SS other than 00 in (−δ,δ)(-\delta, \delta).

A negative ξ\xi is not

Take δ=∣ξ∣\delta = |\xi|. The interval (2ξ,0)(2\xi, 0) contains no positive number, so no point of SS.

A positive ξ\xi is not

Only finitely many points of SS are at least 12ξ\tfrac12\xi (those with n≤2/ξn \le 2/\xi). Let δ\delta be smaller than 12ξ\tfrac12\xi and smaller than the distance from ξ\xi to each of those finitely many points other than ξ\xi. Then (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains no point of SS other than possibly ξ\xi itself: the large ones are too far away, and the small ones (below 12ξ\tfrac12\xi) are below ξ−δ\xi - \delta.

Worked examples and exercises

Exercise ●○○ Hardy: Examples IX, 1–2

Show that the integers have no point of accumulation, and that every real number is a point of accumulation of the rationals.

Hint
For the integers, δ=12\delta = \tfrac12 works almost everywhere; what about at an integer, and half-way between two?
Hint
For the rationals, use density (§3).

Solution

Integers. Given ξ\xi, let δ\delta be less than the distance from ξ\xi to the nearest integer different from ξ\xi (this distance is positive: at most one integer is within 12\tfrac12 of ξ\xi, so take δ=12\delta = \tfrac12 if ξ\xi is an integer or half-way between two, and otherwise half the distance to the nearest integer). Then (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains no integer except perhaps ξ\xi.

Rationals. Any interval (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains infinitely many rationals (§3), so at least one different from ξ\xi. So every ξ\xi is a point of accumulation.

Exercise ●○○ Hardy: Examples IX, 4

Show that the points of accumulation of the set of positive rationals are 00 and the positive numbers.

Hint
Split into ξ<0\xi < 0, ξ=0\xi = 0 and ξ>0\xi > 0.

Solution

If ξ>0\xi > 0, any interval round ξ\xi contains the interval (ξ,ξ+δ)(\xi, \xi + \delta) of positive numbers, which contains rationals. If ξ=0\xi = 0, the interval (0,δ)(0, \delta) contains positive rationals (such as 1/n1/n with n>1/δn > 1/\delta). If ξ<0\xi < 0, the interval (ξ−∣ξ∣,0)(\xi - |\xi|, 0) contains no positive number at all. So the points of accumulation are exactly the ξ≥0\xi \ge 0; the point 00 is one of them although it is not in the set.

Exercise A set with infinitely many points of accumulation ●●○ Hardy: §18

Find all points of accumulation of S={1/m+1/n:m,n=1,2,3,… }S = \{1/m + 1/n : m, n = 1, 2, 3, \dots\}.

Hint
Fix mm and let nn grow. Which points do you approach?
Hint
To show that nothing else is a point of accumulation, use the theorem: a point of accumulation has infinitely many points of SS near it, so both mm and nn cannot stay bounded.

Solution

Candidates. For fixed mm, the points 1/m+1/n1/m + 1/n approach 1/m1/m from above as nn grows, so each 1/m1/m is a point of accumulation; and since 1/m→01/m \to 0 as well, every interval round 00 contains points of SS (take m=nm = n large), so 00 is one too.

Nothing else. Let ξ\xi be a point of accumulation. By the theorem, every interval round ξ\xi contains infinitely many points 1/m+1/n1/m + 1/n, so there are such points arbitrarily close to ξ\xi with max⁡(m,n)\max(m, n) arbitrarily large (only finitely many pairs have both m,n≤Nm, n \le N). If both mm and nn are large, 1/m+1/n1/m + 1/n is near 00. If one of them, say mm, stays bounded, it takes some value infinitely often, and then 1/m+1/n1/m + 1/n is near 1/m1/m. So ξ\xi is 00 or some 1/m1/m. The points of accumulation are 0,1,12,13,…0, 1, \tfrac12, \tfrac13, \dots; note that this set has a point of accumulation of its own, 00.

Exercise Finite sets ●○○ Hardy: §18

Prove that a finite set has no point of accumulation.

Hint
Use the theorem, or choose δ\delta directly.

Solution

By the theorem a point of accumulation has infinitely many points of SS near it, which a finite set does not have. Directly: given ξ\xi, the finitely many points of SS other than ξ\xi are at positive distances from ξ\xi; let δ\delta be smaller than all of them. Then (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains no point of SS other than ξ\xi.

Explore Zoom into a point of accumulation

Try this. Choose “Zoom towards accumulation point 0”, run the zoom, and watch the count of points in view. Then zoom towards “the point 2” (=1+1= 1 + 1), and finally towards the accumulation point 1/21/2.

{1/m+1/n:m,n≥1}\{1/m+1/n : m, n \ge 1\}
00.20.40.60.81.01.21.41.61.82.0011/21/31/4
  • points of the set
  • points of accumulation
width of view: 2.3
1830 of the first 1830 points lie in this view. The view contains the point of accumulation 0: however far you zoom in on it, infinitely many points remain (only the first 1830 are drawn).
What to notice

Around 00 the points never run out, at any magnification: that is the theorem of this lesson. Around the isolated point 22 the view soon contains that point alone. Around 1/21/2 and 1/31/3, which are points of accumulation of {1/m+1/n}\{1/m + 1/n\}, points keep arriving from the right only.

Check yourself

Check yourself True or false?

A point of accumulation of SS must belong to SS.

Why

False. 00 is a point of accumulation of {1/n}\{1/n\} but is not in the set. Conversely a point of SS need not be a point of accumulation: every 1/n1/n is isolated.

Check yourself Choose one

Why does the definition say “a point of SS other than ξ\xi”?

Why

Without the clause, ξ=2\xi = 2 would be a “point of accumulation” of the positive integers, since every interval round 22 contains the point 22. The clause ensures that the crowd near ξ\xi consists of other points. Hardy notes that it is only needed when ξ∈S\xi \in S.

Check yourself Enter a number

How many points of accumulation does the set {(−1)n(1+1/n):n=1,2,… }\{(-1)^n(1 + 1/n) : n = 1, 2, \dots\} have?

Why

Two: 11 (approached by 1+12,1+14,…1 + \tfrac12, 1 + \tfrac14, \dots) and −1-1 (approached by −2,−(1+13),−(1+15),…-2, -(1 + \tfrac13), -(1 + \tfrac15), \dots). Neither is in the set, and every point of the set is isolated.

Hardy’s language

Summary

Summary
  • ξ\xi is a point of accumulation of SS if every interval (ξ−δ,ξ+δ)(\xi - \delta, \xi + \delta) contains a point of SS other than ξ\xi; it may or may not lie in SS.
  • Then every such interval contains infinitely many points of SS; so finite sets have none.
  • Model cases: the integers have none; {1/n}\{1/n\} has only 00; for the rationals every real number is one; for the positive rationals, 00 and every positive number.
  • To show ξ\xi is not a point of accumulation, exhibit one δ\delta that isolates it.