Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Why this matters

How does sin⁡x\sin x compare with xx when xx is small? A calculator set to radians gives sin⁡0.1≈0.0998334\sin 0.1 \approx 0.0998334, sin⁡0.01≈0.00999983\sin 0.01 \approx 0.00999983 and sin⁡0.001≈0.000999999833\sin 0.001 \approx 0.000999999833: each is almost exactly xx. Physicists use this all the time. For a pendulum that swings through small angles they replace sin⁡θ\sin\theta by θ\theta, which turns an equation they cannot solve into one they can. Why does that work, and how good is the replacement?

The precise question is about the ratio sin⁡xx\frac{\sin x}{x}. Its values at the same points are

xx0.10.010.001
sin⁡xx\frac{\sin x}{x}0.9983340.9999830.9999998

(to the digits shown, and the same at -0.1, -0.01, -0.001). Does the ratio approach 1 as x→0x \to 0? The limit laws cannot tell us. At x=0x = 0 the ratio is “00\frac{0}{0}”, where the quotient law says nothing (When the laws do not apply), and sin⁡x\sin x is not a polynomial, so there is nothing to factor or cancel.

What we do have is a comparison. The lemma on sine, angle and tangent says that sin⁡θ<θ<tan⁡θ\sin\theta < \theta < \tan\theta for 0<θ<π20 < \theta < \frac{\pi}{2}, and we will see that this traps sin⁡xx\frac{\sin x}{x} between cos⁡x\cos x and 1. If both of these approach 1, the trapped function has nowhere else to go. This page makes that idea a theorem, the squeeze theorem, and uses it to prove that sin⁡xx\frac{\sin x}{x} approaches 1.

Trapping a function

Suppose that, near a point aa, a function ff lies between two functions gg and hh, and that g(x)g(x) and h(x)h(x) both approach the same number LL as x→ax \to a. Then f(x)f(x), which is caught between them, must approach LL too: the two bounds close in on LL and leave ff no room. This works even when ff itself is hard to handle, as long as the bounds are easy.

Here is a function that is hard to handle. As xx approaches 0, the factor sin⁡(1/x)\sin(1/x) swings between −1{-1} and 1 faster and faster, and xsin⁡(1/x)x \sin(1/x) swings with it. But the factor xx shrinks the swings: by part (a) of the proposition on the bounds of sine and cosine, −1≤sin⁡(1/x)≤1-1 \le \sin(1/x) \le 1, so the values of xsin⁡(1/x)x\sin(1/x) stay between −∣x∣-\abs{x} and ∣x∣\abs{x}.

Figure 1:Graph, for −0.5≤x≤0.5-0.5 \le x \le 0.5 and x≠0x \ne 0, of one of three functions, chosen by a slider for kk, which takes the values −1{-1}, 0 and 1 and starts at 0. At k=−1k = -1 it is the lower bound g(x)=−∣x∣g(x) = -\abs{x}, at k=0k = 0 the function f(x)=xsin⁡(1/x)f(x) = x \sin(1/x), and at k=1k = 1 the upper bound h(x)=∣x∣h(x) = \abs{x} (the widget plots k∣x∣+(1−∣k∣) xsin⁡(1/x)k\abs{x} + (1 - \abs{k})\,x\sin(1/x)). Taken together, the graphs at k=−1k = -1 and k=1k = 1 are the edges of a band shaped like a bow tie, which narrows to the origin. At k=0k = 0, the graph of ff swings between the heights −∣x∣-\abs{x} and ∣x∣\abs{x}, faster and faster as xx approaches 0; close to 0 the drawing is a jumble of lines, because the plotted curve joins finitely many computed points. An open circle marks the origin, where the formula is undefined. A table lists the values at x=0.1,0.01,0.001,−0.001,−0.01,−0.1x = 0.1, 0.01, 0.001, -0.001, -0.01, -0.1. For k=0k = 0 they are -0.0544021, -0.00506366, 0.00082688, 0.00082688, -0.00506366, -0.0544021; for k=−1k = -1 they are −∣x∣-\abs{x}, and for k=1k = 1 they are ∣x∣\abs{x}. At each of these six points the value of ff lies between those of gg and hh. A second slider moves a point along the graph, starting at x=0.25x = 0.25, with a readout of xx and the value there.

Try this: with k=0k = 0, look at the table: do the values of ff settle as xx gets closer to 0? Now set k=−1k = -1 and k=1k = 1 and compare the three tables row by row. How wide is the band at x=0.001x = 0.001, and what does that say about f(0.001)f(0.001)? Then move the traced point towards 0 and switch kk between −1{-1}, 0 and 1 at each position.

The table of ff alone looks irregular: its values change sign, and their sizes do not shrink at a steady rate. What the tables show, at the six points they list, is that the value of ff lies between those of gg and hh, in a band whose width 2∣x∣2\abs{x} shrinks to 0. A picture and a table only check finitely many points; the squeeze theorem turns the band into a proof that xsin⁡(1/x)x\sin(1/x) approaches 0 (Example 1).

Main results

The squeeze theorem

Proof (Rigorous track)

We play the ε–δ game of the definition of a limit for gg and hh with the same tolerance ε\eps: near aa, both g(x)g(x) and h(x)h(x) lie strictly between L−εL - \eps and L+εL + \eps, and f(x)f(x), caught between them, lies there too.

Let ε>0\eps > 0. Since lim⁡x→ag(x)=L\lim_{x \to a} g(x) = L, there is a δ1>0\delta_1 > 0 such that gg is defined and ∣g(x)−L∣<ε\abs{g(x) - L} < \eps at every xx with 0<∣x−a∣<δ10 < \abs{x - a} < \delta_1. Since lim⁡x→ah(x)=L\lim_{x \to a} h(x) = L, there is a δ2>0\delta_2 > 0 such that hh is defined and ∣h(x)−L∣<ε\abs{h(x) - L} < \eps at every xx with 0<∣x−a∣<δ20 < \abs{x - a} < \delta_2.

Let δ=min⁡(δ1,δ2,r)\delta = \min(\delta_1, \delta_2, r), which is positive, and let 0<∣x−a∣<δ0 < \abs{x - a} < \delta. Since δ\delta is at most each of δ1\delta_1, δ2\delta_2 and rr, also ∣x−a∣<δ1\abs{x - a} < \delta_1, ∣x−a∣<δ2\abs{x - a} < \delta_2 and ∣x−a∣<r\abs{x - a} < r (property 2 of the order rules, transitivity). So xx lies in both windows and in the punctured interval of the hypothesis: f(x)f(x), g(x)g(x) and h(x)h(x) are defined, g(x)≤f(x)≤h(x)g(x) \le f(x) \le h(x), ∣g(x)−L∣<ε\abs{g(x) - L} < \eps and ∣h(x)−L∣<ε\abs{h(x) - L} < \eps. By part (a) of the proposition on distance inequalities, applied to the number g(x)g(x) with centre LL and radius ε\eps, the inequality ∣g(x)−L∣<ε\abs{g(x) - L} < \eps says L−ε<g(x)<L+εL - \eps < g(x) < L + \eps; in the same way, L−ε<h(x)<L+εL - \eps < h(x) < L + \eps. Hence

L−ε<g(x)≤f(x)≤h(x)<L+ε.L - \eps < g(x) \le f(x) \le h(x) < L + \eps .

By transitivity (property 2 of the order rules: u<v≤wu < v \le w implies u<wu < w, and u≤v<wu \le v < w implies u<wu < w), L−ε<f(x)L - \eps < f(x) and f(x)<L+εf(x) < L + \eps. By part (a) of the proposition on distance inequalities again, this says ∣f(x)−L∣<ε\abs{f(x) - L} < \eps.

So ff is defined at every xx with 0<∣x−a∣<r0 < \abs{x - a} < r, which by part (e) of the proposition on distance inequalities is the open interval (a−r,a+r)(a - r, a + r) with aa removed; the definition of a limit applies to ff, and δ\delta wins the round ε\eps. Since ε>0\eps > 0 was arbitrary, lim⁡x→af(x)=L\lim_{x \to a} f(x) = L.

In words. If, near aa (but not necessarily at aa), ff lies between a lower bound gg and an upper bound hh that approach the same number LL, then ff approaches LL too. The bounds need to hold only on some punctured interval around aa, however small, and they say nothing about f(a)f(a), which may be undefined. The theorem is also called the sandwich theorem or the pinching theorem.

Example. Let f(x)=x∣x∣f(x) = x\abs{x}. By the definition of the absolute value, f(x)=x2f(x) = x^2 for x≥0x \ge 0 and f(x)=−x2f(x) = -x^2 for x<0x < 0. Since x2≥0x^2 \ge 0, also −x2≤0≤x2-x^2 \le 0 \le x^2, so in both cases

−x2≤f(x)≤x2-x^2 \le f(x) \le x^2

for every real xx. The bounds are polynomials, so by direct substitution both approach 0 as x→0x \to 0, and the theorem (with any rr, say r=1r = 1) gives lim⁡x→0x∣x∣=0\lim_{x \to 0} x\abs{x} = 0. The function ff is given by two formulas, and the bounds are single polynomials: the theorem is useful exactly when ff is the hard part and the bounds are easy.

Non-example. For x≠0x \ne 0, −1≤sin⁡(1/x)≤1-1 \le \sin(1/x) \le 1, but the bounds −1{-1} and 1 approach different numbers, so the theorem says nothing about lim⁡x→0sin⁡(1/x)\lim_{x \to 0} \sin(1/x). In fact that limit does not exist (the example on sin⁡(1/x)\sin(1/x) in The Limit of a Function): the bounds leave room for the values to swing between them.

The hypotheses are all needed: Remark 2 gives a counterexample for each. First, a form of the theorem that is often quicker to use. Most squeezes bound the distance from f(x)f(x) to LL.

With g=hg = h, the squeeze theorem also says that two functions that are equal on a punctured interval around aa have the same limit at aa: if f(x)=g(x)f(x) = g(x) for every xx with 0<∣x−a∣<r0 < \abs{x - a} < r, then g(x)≤f(x)≤g(x)g(x) \le f(x) \le g(x) there, and lim⁡x→ag(x)=L\lim_{x \to a} g(x) = L gives lim⁡x→af(x)=L\lim_{x \to a} f(x) = L. We use this when two formulas agree near aa but not everywhere.

Sine and cosine near 0

To use the squeeze theorem for sin⁡xx\frac{\sin x}{x}, we need bounds for sin⁡x\sin x and cos⁡x\cos x near 0, and their limits there. Everything rests on the lemma on sine, angle and tangent of Trigonometric Functions, which compares areas and uses no limits, and on the properties of sin⁡\sin and cos⁡\cos proved on that page. The lemma is about 0<θ<π20 < \theta < \frac{\pi}{2} only. For negative numbers we use the symmetries of property 4 of the point P(t)P(t): cos⁡(−t)=cos⁡t\cos(-t) = \cos t and sin⁡(−t)=−sin⁡t\sin(-t) = -\sin t.

The inequalities of the lemma are what the squeeze theorem needs. With them we get the limits of sin⁡\sin and cos⁡\cos at 0, without knowing that sin⁡\sin and cos⁡\cos are continuous.

So the limits of sin⁡\sin and cos⁡\cos at 0 are their values there, sin⁡0=0\sin 0 = 0 and cos⁡0=1\cos 0 = 1. That is no surprise, but it needed a proof: neither function is a polynomial, so direct substitution does not apply to them.

The limit of sine x over x

The chain of reasoning, from the start: areas of a triangle, a sector and a larger triangle give sin⁡θ<θ<tan⁡θ\sin\theta < \theta < \tan\theta (Lemma 1); dividing by θ\theta and using the symmetries of sin⁡\sin and cos⁡\cos traps sin⁡xx\frac{\sin x}{x} between cos⁡x\cos x and 1 (Lemma 1); the double-angle formula squeezes cos⁡x\cos x to 1 (Lemma 2); and the squeeze theorem finishes. No step uses a derivative, a series, or the continuity of sin⁡\sin or cos⁡\cos.

Why radians. The lemma compares sin⁡θ\sin\theta with θ\theta, the length of an arc of the unit circle, and that length is the measure of the angle in radians (Definition 1). An angle of xx degrees is πx180\frac{\pi x}{180} radians, so in degrees the ratio is sin⁡(cx)x\frac{\sin(cx)}{x} with c=π180c = \frac{\pi}{180}. At x=0.01x = 0.01 it is about 0.0174533, close to π180≈0.0174533\frac{\pi}{180} \approx 0.0174533 rather than to 1. The lemmas prove that this is the limit, with no new tool.

Let c>0c > 0, and let 0<∣x∣<π2c0 < \abs{x} < \frac{\pi}{2c}. By property 4 of the absolute value, ∣cx∣=∣c∣ ∣x∣=c∣x∣\abs{cx} = \abs{c}\,\abs{x} = c\abs{x} (since c>0c > 0, ∣c∣=c\abs{c} = c by the definition of the absolute value), and multiplying 0<∣x∣<π2c0 < \abs{x} < \frac{\pi}{2c} by the positive number cc keeps both inequalities (property 5(a) of the order rules): 0<∣cx∣<π20 < \abs{cx} < \frac{\pi}{2}. So parts (b) and (c) of the lemma on sine and cosine near 0 apply to the number cxcx:

1−c2x22<cos⁡(cx)<sin⁡(cx)cx<1.1 - \frac{c^2x^2}{2} < \cos(cx) < \frac{\sin(cx)}{cx} < 1 .

By transitivity (property 2 of the order rules), 1−c2x22<sin⁡(cx)cx<11 - \frac{c^2x^2}{2} < \frac{\sin(cx)}{cx} < 1. Multiplying by the positive number cc keeps both inequalities (property 5(a)), and c⋅sin⁡(cx)cx=sin⁡(cx)xc \cdot \frac{\sin(cx)}{cx} = \frac{\sin(cx)}{x}, so

c−c3x22<sin⁡(cx)x<c,c - \frac{c^3x^2}{2} < \frac{\sin(cx)}{x} < c ,

and so the same with ≤\le. The lower bound is a polynomial in xx, so by part (a) of direct substitution it approaches c−0=cc - 0 = c; the upper bound is the constant cc, which approaches cc by part (a) of the limit laws. The bounds are defined for every real xx, and sin⁡(cx)x\frac{\sin(cx)}{x} for every x≠0x \ne 0. The number r=π2cr = \frac{\pi}{2c} is positive: it is the positive number π2\frac{\pi}{2} times 1c>0\frac{1}{c} > 0 (properties 6 and 5(a) of the order rules). So the squeeze theorem, with a=0a = 0, L=cL = c and this rr, gives

lim⁡x→0sin⁡(cx)x=c.\lim_{x \to 0} \frac{\sin(cx)}{x} = c .

With c=π180c = \frac{\pi}{180}: in degrees, the ratio approaches π180\frac{\pi}{180}, not 1.

The only law used with a quotient was the quotient law with M=2M = 2. The original quotient 1−cos⁡xx\frac{1 - \cos x}{x} has M=0M = 0 in its denominator, so the quotient law cannot be applied to it directly; the rewriting moves the “00\frac{0}{0}” into sin⁡xx\frac{\sin x}{x}, whose limit we know.

Worked examples

Common mistakes

Rigorous track

The proof of the squeeze theorem is in the dropdown under its statement. Two more remarks for readers who want the details.

Summary

Exercises

Where this leads

These topics build on this one:

Found an error on this page? Report it (the form opens with this page filled in).

The squeeze theorem is the tool for limits that the laws cannot reach: oscillating factors, and the limits of sin⁡\sin and cos⁡\cos, on which the calculus of the trigonometric functions rests.