Chapter V · Limits of functions of a continuous variable. Continuous and discontinuous functions

Limits as xx tends to a point

§§93–95 35 min

What lim⁡x→aϕ(x)=l\lim_{x\to a}\phi(x) = l means, the limits from the right and from the left, and why a monotone function always has both.

Why this matters

“As xx approaches aa, ϕ(x)\phi(x) approaches ll” is the sentence on which continuity, derivatives and most of calculus rest. In §§93–95 Hardy makes it precise. He reaches it from the last lesson by a change of variable, ends with the two-quantifier definition that every analysis course now uses, and adds the one-sided limits ϕ(a−0)\phi(a - 0) and ϕ(a+0)\phi(a + 0), which we need to describe jumps.

The idea

Hardy’s route is a substitution. If ϕ(x)→l\phi(x) \to l as x→∞x \to \infty, put y=1/xy = 1/x. Large positive xx means small positive yy, so the statement becomes: by taking yy positive and small enough, we can make ψ(y)=ϕ(1/y)\psi(y) = \phi(1/y) as close to ll as we please. That is a statement about yy approaching 00 from the right, and once it is written out we can forget xx and use it for any function ψ\psi. Shifting by aa (put x=a+yx = a + y) then gives limits at any point aa.

The game has the same shape as before, but the second move is different. The opponent names a tolerance ε>0\varepsilon > 0, a horizontal band l±εl \pm \varepsilon. We answer with a distance δ>0\delta > 0, a vertical strip around aa. We win if every xx in the strip other than aa itself gives a point of the graph inside the band.

Explore Find δ\delta for x2x^2 near 11

Try this. Set ε=0.5\varepsilon = 0.5 and drag the purple handle (or use the δ\delta slider) until the verdict is green. Halve ε\varepsilon and repeat a few times. Compare each working δ\delta with ε/3\varepsilon/3.

f(x)=x2f(x) = x^2, claim lim⁡x→1f(x)=1\lim_{x\to 1} f(x) = 1 at a=1a=1
xy
  • y = f(x)
  • band L ± ε
  • strip of width δ (drag)
  • escapes the band
This δ fails. For example x=1.5x = 1.5 gives ∣f(x)−1∣=1.25≥ε|f(x) - 1| = 1.25 \ge \varepsilon. Try a smaller δ — or ask whether any δ can work.
What to notice

Each smaller ε\varepsilon needs a smaller δ\delta, and δ=ε/3\delta = \varepsilon/3 always works once ε≤3\varepsilon \le 3. That is no accident: ∣x2−1∣=∣x−1∣ ∣x+1∣|x^2 - 1| = |x - 1|\,|x + 1|, and near x=1x = 1 the factor ∣x+1∣|x + 1| stays below 33, so a strip of half-width ε/3\varepsilon/3 cannot climb out of a band of half-width ε\varepsilon. The proof below turns this into an argument.

Two details of the game are essential.

  • x=ax = a is excluded. The strip is 0<∣x−a∣<δ0 < |x - a| < \delta. The limit describes the values near aa, never the value at aa, which may be different or may not exist at all. The next lesson is about exactly this.
  • δ\delta may depend on ε\varepsilon (and on aa), just as n0n_0 depended on δ\delta in Chapter IV.

Precise statements

From here on we write the tolerance as ε\varepsilon and the closeness of xx to aa as δ\delta, which is the modern habit. Hardy uses the same two letters the other way round; see “Hardy’s language” below.

Definition Limit as x→ax \to a Hardy: §94

ϕ(x)\phi(x) tends to the limit ll as xx tends to aa if for every ε>0\varepsilon > 0 there is a δ>0\delta > 0 such that

∣ϕ(x)−l∣<εwhenever 0<∣x−a∣<δ.|\phi(x) - l| < \varepsilon \quad \text{whenever } 0 < |x - a| < \delta.

We write lim⁡x→aϕ(x)=l\lim_{x \to a} \phi(x) = l, or ϕ(x)→l\phi(x) \to l as x→ax \to a. The case a=0a = 0 is Hardy’s definition in §93.

Definition One-sided limits Hardy: §§93–94

ϕ(x)→l\phi(x) \to l as x→ax \to a from the right (Hardy: “by values greater than aa”, written x→a+0x \to a + 0) if for every ε>0\varepsilon > 0 there is a δ>0\delta > 0 with ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon whenever a<x<a+δa < x < a + \delta. The limit from the left uses a−δ<x<aa - \delta < x < a. The two one-sided limits, when they exist, are written

ϕ(a+0)=lim⁡x→a+0ϕ(x),ϕ(a−0)=lim⁡x→a−0ϕ(x).\phi(a + 0) = \lim_{x \to a+0} \phi(x), \qquad \phi(a - 0) = \lim_{x \to a-0} \phi(x).

Tending to ±∞\pm\infty is defined by the same recipe with a large Δ\Delta: ϕ(x)→∞\phi(x) \to \infty as x→ax \to a if for every Δ\Delta there is a δ>0\delta > 0 with ϕ(x)>Δ\phi(x) > \Delta whenever 0<∣x−a∣<δ0 < |x - a| < \delta; and similarly on one side. If none of these happens, ϕ(x)\phi(x) oscillates as x→ax \to a (or as x→a+0x \to a + 0), finitely or infinitely according as it is or is not bounded near aa.

Theorem Two-sided means both one-sided, and equal Hardy: §94

ϕ(x)→l\phi(x) \to l as x→ax \to a if and only if ϕ(a−0)\phi(a - 0) and ϕ(a+0)\phi(a + 0) both exist and both equal ll.

Hardy closes the section with monotone functions, which can never oscillate near a point.

Theorem Monotone functions have one-sided limits Hardy: §95

If ϕ(x)\phi(x) increases steadily in an interval (a−h,a+h)(a - h, a + h), then ϕ(a−0)\phi(a - 0) and ϕ(a+0)\phi(a + 0) exist, and

ϕ(a−0)≤ϕ(a)≤ϕ(a+0).\phi(a - 0) \le \phi(a) \le \phi(a + 0).

The same holds, with the inequalities reversed, for a steadily decreasing function.

A jump of a monotone function is therefore always a clean gap between ϕ(a−0)\phi(a - 0) and ϕ(a+0)\phi(a + 0), with ϕ(a)\phi(a) somewhere in the gap.

Explore One side at a time

Try this. Here is [x][x] near a=1a = 1. Make ε\varepsilon small (say 0.20.2). With the claimed limit L=1L = 1, can any δ\delta win? Move LL to 00 and try again. Then look for an LL that works.

f(x)=[x]f(x) = [x], claim lim⁡x→1f(x)=1\lim_{x\to 1} f(x) = 1 at a=1a=1
xy
  • y = f(x)
  • band L ± ε
  • strip of width δ (drag)
  • escapes the band
This δ fails. For example x=0.5x = 0.5 gives ∣f(x)−1∣=1≥ε|f(x) - 1| = 1 \ge \varepsilon. Try a smaller δ — or ask whether any δ can work.
What to notice

With L=1L = 1 every strip fails on its left half, where [x]=0[x] = 0; with L=0L = 0 it fails on the right half, where [x]=1[x] = 1. No LL survives a small ε\varepsilon, so lim⁡x→1[x]\lim_{x\to1}[x] does not exist. Yet each half on its own is perfectly behaved: ϕ(1−0)=0\phi(1 - 0) = 0 and ϕ(1+0)=1\phi(1 + 0) = 1, and [x][x], being steadily increasing, obeys ϕ(1−0)≤ϕ(1)=1≤ϕ(1+0)\phi(1-0) \le \phi(1) = 1 \le \phi(1+0).

Proofs

Here is a model ε\varepsilon–δ\delta proof. It is a rule that produces a winning δ\delta from any ε\varepsilon, like the rule n0=[1/δ]+1n_0 = [1/\delta] + 1 in Chapter IV.

Proof x2→1x^2 \to 1 as x→1x \to 1
Strategy

Factor the distance: ∣x2−1∣=∣x−1∣ ∣x+1∣|x^2 - 1| = |x - 1|\,|x + 1|. The first factor is what we control. The second is not small, but it is bounded near 11; bound it first, then choose δ\delta.

Let ε>0\varepsilon > 0 be given; first insist on δ≤1\delta \le 1

If ∣x−1∣<1|x - 1| < 1 then 0<x<20 < x < 2, so ∣x+1∣<3|x + 1| < 3.

Choose δ=min⁡(1,ε/3)\delta = \min(1, \varepsilon/3)

Then for 0<∣x−1∣<δ0 < |x - 1| < \delta we have both ∣x+1∣<3|x + 1| < 3 and ∣x−1∣<ε/3|x - 1| < \varepsilon/3.

Check

∣x2−1∣=∣x−1∣ ∣x+1∣<(ε/3)⋅3=ε|x^2 - 1| = |x - 1|\,|x + 1| < (\varepsilon/3) \cdot 3 = \varepsilon. Every ε\varepsilon has its δ\delta, so x2→1x^2 \to 1.

Proof Two-sided means both one-sided, and equal
Strategy

The punctured strip 0<∣x−a∣<δ0 < |x - a| < \delta is the union of a left strip and a right strip. A δ\delta that works for the whole strip works for each half; conversely, take the smaller of the two one-sided δ\delta’s.

If ϕ(x)→l\phi(x) \to l, both one-sided limits are ll

Given ε\varepsilon, the δ\delta for the two-sided limit gives ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon on a<x<a+δa < x < a + \delta and on a−δ<x<aa - \delta < x < a, which is what the one-sided definitions ask.

Conversely, suppose ϕ(a−0)=ϕ(a+0)=l\phi(a - 0) = \phi(a + 0) = l

Given ε\varepsilon, there are δ1>0\delta_1 > 0 with ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon on a−δ1<x<aa - \delta_1 < x < a, and δ2>0\delta_2 > 0 with the same on a<x<a+δ2a < x < a + \delta_2.

Take δ=min⁡(δ1,δ2)\delta = \min(\delta_1, \delta_2)

If 0<∣x−a∣<δ0 < |x - a| < \delta, then xx lies in one of the two half-strips, so ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon.

Proof Monotone functions have one-sided limits
Strategy

This is the monotone-limit theorem of the last lesson in a new setting. For xx approaching aa from the left, the values ϕ(x)\phi(x) climb and never pass ϕ(a)\phi(a), so they should tend to their upper bound. Hardy gets this from §92 by putting y=1/(a−x)y = 1/(a - x); here is the direct argument.

Let ll be the upper bound of ϕ(x)\phi(x) for a−h<x<aa - h < x < a

These values are bounded above by ϕ(a)\phi(a), since ϕ\phi increases steadily. So the upper bound ll exists and l≤ϕ(a)l \le \phi(a).

Given ε>0\varepsilon > 0, find x1<ax_1 < a with ϕ(x1)>l−ε\phi(x_1) > l - \varepsilon

l−εl - \varepsilon is less than the upper bound, so some value exceeds it.

Take δ=a−x1\delta = a - x_1

If x1<x<ax_1 < x < a then l−ε<ϕ(x1)≤ϕ(x)≤ll - \varepsilon < \phi(x_1) \le \phi(x) \le l, so ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon. Hence ϕ(a−0)=l≤ϕ(a)\phi(a - 0) = l \le \phi(a).

The right side is symmetric

For a<x<a+ha < x < a + h, the values are bounded below by ϕ(a)\phi(a); their lower bound is ϕ(a+0)\phi(a + 0), by the same argument, and it is at least ϕ(a)\phi(a).

Worked examples and exercises

Exercise ●○○ Hardy: Examples XXXV, 2–3

Let mm be an integer. Show that xm→0x^m \to 0 as x→0x \to 0 when m>0m > 0. When m<0m < 0, show that xm→+∞x^m \to +\infty as x→+0x \to +0, and that as x→−0x \to -0 it tends to −∞-\infty or +∞+\infty according as mm is odd or even.

Hint
For m>0m > 0: if ∣x∣<1|x| < 1 then ∣xm∣≤∣x∣|x^m| \le |x|.
Hint
For m=−k<0m = -k < 0: xm=1/xkx^m = 1/x^k, and 1/xk>Δ1/x^k > \Delta when 0<x<Δ−1/k0 < x < \Delta^{-1/k}.

Solution

m>0m > 0. Given ε>0\varepsilon > 0, take δ=min⁡(1,ε)\delta = \min(1, \varepsilon). If 0<∣x∣<δ0 < |x| < \delta then ∣x∣<1|x| < 1, so ∣xm∣=∣x∣m≤∣x∣<ε|x^m| = |x|^m \le |x| < \varepsilon.

m=−km = -k, k>0k > 0, from the right. Given Δ>0\Delta > 0, take δ=Δ−1/k\delta = \Delta^{-1/k}. If 0<x<δ0 < x < \delta then xk<δk=1/Δx^k < \delta^k = 1/\Delta, so xm=1/xk>Δx^m = 1/x^k > \Delta. (For Δ≤0\Delta \le 0 any δ\delta works.) So xm→+∞x^m \to +\infty as x→+0x \to +0.

From the left. Put x=−yx = -y with y>0y > 0: xm=(−1)k/yk=(−1)mymx^m = (-1)^k / y^k = (-1)^m y^m. As x→−0x \to -0, y→+0y \to +0 and ym→+∞y^m \to +\infty; so xm→+∞x^m \to +\infty if mm is even and −∞-\infty if mm is odd. For odd negative mm, then, xmx^m tends to +∞+\infty on one side and −∞-\infty on the other, so it tends to neither as x→0x \to 0.

Exercise ●○○ Hardy: Examples XXXVI, 22

Let mm be an integer. Show that as x→m+0x \to m + 0, [x]→m[x] \to m and x−[x]→0x - [x] \to 0, while as x→m−0x \to m - 0, [x]→m−1[x] \to m - 1 and x−[x]→1x - [x] \to 1.

Hint
Which integer is [x][x] when m<x<m+1m < x < m + 1? When m−1<x<mm - 1 < x < m?

Solution

For m<x<m+1m < x < m + 1 we have [x]=m[x] = m exactly, so with δ=1\delta = 1 every such xx gives ∣[x]−m∣=0<ε|[x] - m| = 0 < \varepsilon; and x−[x]=x−mx - [x] = x - m, which is less than ε\varepsilon once x<m+εx < m + \varepsilon, so δ=min⁡(1,ε)\delta = \min(1, \varepsilon) works. For m−1<x<mm - 1 < x < m we have [x]=m−1[x] = m - 1 exactly, and x−[x]=x−m+1x - [x] = x - m + 1, which is within ε\varepsilon of 11 once x>m−εx > m - \varepsilon. So [m−0]=m−1≠m=[m+0][m - 0] = m - 1 \ne m = [m + 0]: the one-sided limits differ, and lim⁡x→m[x]\lim_{x \to m}[x] does not exist.

Exercise ●●○ Hardy: Examples XXXVI, 6

Let a,b>0a, b > 0. Show that as x→+0x \to +0

xa[bx]→ba,bx[xa]→0.\frac{x}{a}\left[\frac{b}{x}\right] \to \frac{b}{a}, \qquad \frac{b}{x}\left[\frac{x}{a}\right] \to 0.

How do the two functions behave as x→−0x \to -0?

Hint
Use t−1<[t]≤tt - 1 < [t] \le t for every real tt.
Hint
For the second function, what is [x/a][x/a] when 0<x<a0 < x < a? When −a<x<0-a < x < 0?

Solution

First function, x>0x > 0. From b/x−1<[b/x]≤b/xb/x - 1 < [b/x] \le b/x, multiplying by x/a>0x/a > 0,

ba−xa<xa[bx]≤ba.\frac{b}{a} - \frac{x}{a} < \frac{x}{a}\left[\frac{b}{x}\right] \le \frac{b}{a}.

So the distance from b/ab/a is less than x/ax/a, which is less than ε\varepsilon when 0<x<aε0 < x < a\varepsilon. Limit b/ab/a.

First function, x<0x < 0. Now x/a<0x/a < 0 and the inequalities reverse: b/a≤(x/a)[b/x]<b/a−x/a=b/a+∣x∣/ab/a \le (x/a)[b/x] < b/a - x/a = b/a + |x|/a. Again the limit is b/ab/a.

Second function. For 0<x<a0 < x < a, [x/a]=0[x/a] = 0, so the function is identically 00 there and tends to 00. For −a<x<0-a < x < 0, [x/a]=−1[x/a] = -1, so the function equals −b/x=b/∣x∣-b/x = b/|x|, which tends to +∞+\infty as x→−0x \to -0. So the second function has different behaviour on the two sides and no limit as x→0x \to 0.

Exercise A limit, if it exists, is unique ●●○ Hardy: §94

Show that ϕ(x)\phi(x) cannot tend to two different limits ll and l′l' as x→ax \to a. (Assume, as always, that ϕ\phi is defined at points arbitrarily close to aa on at least one side.)

Hint
Take ε=12∣l−l′∣\varepsilon = \tfrac12|l - l'| and use both definitions at one point xx.

Solution

Suppose l≠l′l \ne l' and put ε=12∣l−l′∣>0\varepsilon = \tfrac12|l - l'| > 0. There are δ1\delta_1 and δ2\delta_2 with ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon for 0<∣x−a∣<δ10 < |x - a| < \delta_1 and ∣ϕ(x)−l′∣<ε|\phi(x) - l'| < \varepsilon for 0<∣x−a∣<δ20 < |x - a| < \delta_2. Pick any xx with 0<∣x−a∣<min⁡(δ1,δ2)0 < |x - a| < \min(\delta_1, \delta_2) at which ϕ\phi is defined. Then ∣l−l′∣≤∣l−ϕ(x)∣+∣ϕ(x)−l′∣<2ε=∣l−l′∣|l - l'| \le |l - \phi(x)| + |\phi(x) - l'| < 2\varepsilon = |l - l'|, a contradiction. (This is why the hypothesis matters: if there were no such xx, every ll would satisfy the definition vacuously.)

Check yourself

Explore Is there any limit at all?

Try this. This is sin⁡(1/x)\sin(1/x) near 00. The opponent names three tolerances. Move LL anywhere you like and try to survive the rounds.

f(x)=sin⁡1xf(x) = \sin\frac1x, claim lim⁡x→0f(x)=0\lim_{x\to 0} f(x) = 0 at a=0a=0Round 1 of 3
Opponent: “I name ε=0.5\varepsilon = 0.5. Find δ\delta so that ∣f(x)−L∣<ε|f(x)-L| < \varepsilon whenever 0<∣x−a∣<δ0 < |x-a| < \delta.”
xy
  • y = f(x)
  • band L ± ε
  • strip of width δ (drag)
  • escapes the band
What to notice

Every strip around 00, however thin, contains whole waves of the graph, so it contains points where sin⁡(1/x)=1\sin(1/x) = 1 and points where it is −1-1. No band narrower than 11 can hold both. The function oscillates finitely as x→0x \to 0, from either side.

Check yourself Choose one

Let ψ(x)=0\psi(x) = 0 for x≠0x \ne 0 and ψ(0)=1\psi(0) = 1. What is lim⁡x→0ψ(x)\lim_{x\to0} \psi(x)?

Why

The definition only looks at 0<∣x∣<δ0 < |x| < \delta, and there ψ(x)=0\psi(x) = 0 exactly. The value ψ(0)=1\psi(0) = 1 is invisible to the limit. Saying the limit fails confuses “has a limit” with “is continuous”, which is the subject of a later lesson.

Check yourself True or false?

If ϕ(a+0)\phi(a + 0) exists, then lim⁡x→aϕ(x)\lim_{x \to a} \phi(x) exists.

Why

False. Both one-sided limits must exist and be equal. [x][x] at a=1a = 1 has ϕ(1+0)=1\phi(1 + 0) = 1 but ϕ(1−0)=0\phi(1 - 0) = 0.

Check yourself Choose one

Which statement says that ϕ(x)→l\phi(x) \to l as x→ax \to a?

Why

The tolerance comes first and the strip answers it. The first option fixes one strip for all tolerances, which forces ϕ(x)=l\phi(x) = l on that whole punctured strip. The third only says ϕ\phi is bounded near aa (sin⁡(1/x)\sin(1/x) satisfies it with l=0l = 0). The fourth asks for a single good point.

Check yourself Enter a number

For ϕ(x)=3x+2\phi(x) = 3x + 2 at a=1a = 1 (limit 55) and ε=0.03\varepsilon = 0.03, what is the largest δ\delta that works?

Why

∣ϕ(x)−5∣=3∣x−1∣|\phi(x) - 5| = 3|x - 1|, which is less than 0.030.03 exactly when ∣x−1∣<0.01|x - 1| < 0.01. So δ=ε/3=0.01\delta = \varepsilon/3 = 0.01; any smaller δ\delta also works, and any larger one lets in points such as x=1+0.0101x = 1 + 0.0101, where the distance is 0.03030.0303.

Hardy’s language

It will of course be understood that ϕ(a − 0) has no meaning other than that of a conventional abbreviation for the limit on the left-hand side.

— G. H. Hardy, §95

Summary

Summary
  • lim⁡x→aϕ(x)=l\lim_{x\to a}\phi(x) = l: for every ε>0\varepsilon > 0 there is δ>0\delta > 0 with ∣ϕ(x)−l∣<ε|\phi(x) - l| < \varepsilon whenever 0<∣x−a∣<δ0 < |x - a| < \delta. The point x=ax = a is excluded.
  • One-sided limits use only a<x<a+δa < x < a + \delta or a−δ<x<aa - \delta < x < a; the two-sided limit exists exactly when ϕ(a−0)\phi(a - 0) and ϕ(a+0)\phi(a + 0) exist and agree.
  • To prove a limit, bound the awkward factor near aa first, then choose δ\delta in terms of ε\varepsilon (e.g. δ=min⁡(1,ε/3)\delta = \min(1, \varepsilon/3) for x2x^2 at 11).
  • A steadily increasing function has both one-sided limits at every point, and ϕ(a−0)≤ϕ(a)≤ϕ(a+0)\phi(a - 0) \le \phi(a) \le \phi(a + 0).
  • Hardy’s letters are swapped relative to modern usage: his δ\delta is the tolerance, his ϵ\epsilon the width of the strip.