Unit 2, chapter 8

Trigonometric substitution for a circle inside a square root

How trigonometric substitution with a sine turns the square root of 4 minus x squared into a cosine, using a circle parameterization, angle bounds, the nonnegative square root and the inverse sine.

  • Chapter 8
  • Trigonometric substitution, partial fractions and completing the square

The integral of 1/4−x21/\sqrt{4 - x^2} from 00 to 11 has a square root in the denominator and no matching inner derivative in the numerator. The expression under the root is 22−x22^2 - x^2, a squared radius minus x2x^2, and its square root is the height of a circle of radius 22 above the point xx. Putting x=2sin⁡θx = 2\sin\theta makes the root equal to 2cos⁡θ2\cos\theta and the differential equal to 2cos⁡θ dθ2\cos\theta\,d\theta, so the integrand reduces to 11 in the angle variable. This page explains that trigonometric substitution, how to choose angle bounds on which the square root keeps its sign, and how the inverse sine confirms the value π/6\pi/6.

Key terms

Trigonometric substitution

Definition Trigonometric substitution is the substitution rule with xx written as a trigonometric function of a new variable θ\theta. The rule says that if gg has a continuous derivative on [α,β][\alpha, \beta] and ff is continuous on an interval containing every value g(θ)g(\theta) with θ\theta in [α,β][\alpha, \beta], then

∫g(α)g(β)f(x) dx=∫αβf(g(θ)) g′(θ) dθ.\begin{aligned} &\int_{g(\alpha)}^{g(\beta)} f(x)\,dx \\[0.4em] &\quad = \int_{\alpha}^{\beta} f(g(\theta))\,g'(\theta)\,d\theta. \end{aligned}

For a constant a>0a > 0, three standard choices clear a square root with a Pythagorean identity, each on an angle range where the root has a known sign. For a2−x2\sqrt{a^2 - x^2} with −a≤x≤a-a \le x \le a, take x=asin⁡θx = a\sin\theta with −π/2≤θ≤π/2-\pi/2 \le \theta \le \pi/2, where cos⁡θ≥0\cos\theta \ge 0 and the root equals acos⁡θa\cos\theta. For a2+x2\sqrt{a^2 + x^2}, take x=atan⁡θx = a\tan\theta with −π/2<θ<π/2-\pi/2 < \theta < \pi/2, where sec⁡θ>0\sec\theta > 0 and the root equals asec⁡θa\sec\theta. For x2−a2\sqrt{x^2 - a^2} with x≥ax \ge a, take x=asec⁡θx = a\sec\theta with 0≤θ<π/20 \le \theta < \pi/2, where tan⁡θ≥0\tan\theta \ge 0 and the root equals atan⁡θa\tan\theta. When x≤−ax \le -a, the range π/2<θ≤π\pi/2 < \theta \le \pi is used instead, and there the root equals −atan⁡θ-a\tan\theta. On each of these ranges the substitution is one-to-one, so θ\theta can be recovered from xx when an antiderivative has to be written back in xx.

In plain words A constant square plus or minus x2x^2 under a root has the shape of a Pythagorean identity. Choose xx as a sine, tangent or secant so that the expression under the root becomes a single square, keep the angle in a range where the root of that square has a known sign, and change dxdx and the bounds along with the integrand.

Example For the integral of this page, the integrand is continuous on [0,1][0, 1], since 4−x2≥34 - x^2 \ge 3 there. Take x=2sin⁡θx = 2\sin\theta. As θ\theta runs from 00 to π/6\pi/6, the value of xx runs from 00 to 11, the differential is dx=2cos⁡θ dθdx = 2\cos\theta\,d\theta, and cos⁡θ≥3/2>0\cos\theta \ge \sqrt{3}/2 > 0, so 4−4sin⁡2θ=2cos⁡θ\sqrt{4 - 4\sin^2\theta} = 2\cos\theta. Hence

∫01dx4−x2=∫0π/62cos⁡θ2cos⁡θ dθ=∫0π/61 dθ=π6,\begin{aligned} &\int_0^1 \frac{dx}{\sqrt{4 - x^2}} \\[0.4em] &\quad = \int_0^{\pi/6} \frac{2\cos\theta}{2\cos\theta}\,d\theta \\[0.4em] &\quad = \int_0^{\pi/6} 1\,d\theta = \frac{\pi}{6}, \end{aligned}

which is about 0.52360.5236. The factor 2cos⁡θ2\cos\theta is never 00 on [0,π/6][0, \pi/6], so the cancellation is valid. The circle identity simplified the denominator, and the derivative of 2sin⁡θ2\sin\theta supplied the same factor in the numerator.

Sine substitution

Definition Let a>0a > 0. The sine substitution is x=asin⁡θx = a\sin\theta with −π/2≤θ≤π/2-\pi/2 \le \theta \le \pi/2. On this interval the sine increases from −1-1 to 11, so each xx in [−a,a][-a, a] comes from exactly one angle, θ=arcsin⁡(x/a)\theta = \arcsin(x/a). Since 1−sin⁡2θ=cos⁡2θ1 - \sin^2\theta = \cos^2\theta and cos⁡θ≥0\cos\theta \ge 0 on [−π/2,π/2][-\pi/2, \pi/2], the substitution gives

a2−x2=a2cos⁡2θ,a2−x2=acos⁡θ,dx=acos⁡θ dθ.\begin{gathered} a^2 - x^2 = a^2\cos^2\theta, \\[0.4em] \sqrt{a^2 - x^2} = a\cos\theta, \\[0.4em] dx = a\cos\theta\,d\theta. \end{gathered}

In plain words The root and the differential produce the same factor acos⁡θa\cos\theta. That is why dx/a2−x2dx/\sqrt{a^2 - x^2} reduces to dθd\theta, while a2−x2 dx\sqrt{a^2 - x^2}\,dx becomes a2cos⁡2θ dθa^2\cos^2\theta\,d\theta, which the half-angle identity integrates.

Example The integrand 9−x2\sqrt{9 - x^2} is continuous on [0,3][0, 3]. With x=3sin⁡θx = 3\sin\theta, the bounds 00 and 33 become 00 and π/2\pi/2, the root becomes 3cos⁡θ3\cos\theta, and the identity cos⁡2θ=12(1+cos⁡2θ)\cos^2\theta = \tfrac12(1 + \cos 2\theta) gives

∫039−x2 dx=∫0π/29cos⁡2θ dθ=92[θ+12sin⁡2θ]0π/2=9π4≈7.0686.\begin{aligned} &\int_0^3 \sqrt{9 - x^2}\,dx \\[0.4em] &\quad = \int_0^{\pi/2} 9\cos^2\theta\,d\theta \\[0.4em] &\quad = \frac{9}{2}\Bigl[\theta + \tfrac12\sin 2\theta\Bigr]_0^{\pi/2} \\[0.4em] &\quad = \frac{9\pi}{4} \approx 7.0686. \end{aligned}

The graph of 9−x2\sqrt{9 - x^2} is the upper half of the circle of radius 33, and 9π/49\pi/4 is the area of a quarter of that disk, as it should be. At θ=π/2\theta = \pi/2 the root is 00, which is still nonnegative, so the endpoint causes no trouble.

Circle parameterization

Definition Let a>0a > 0. On the circle x2+y2=a2x^2 + y^2 = a^2, measure the angle θ\theta from the positive yy-axis, turning toward the positive xx-axis. The point at angle θ\theta is

(x,y)=(asin⁡θ, acos⁡θ).(x, y) = (a\sin\theta,\ a\cos\theta).

As θ\theta runs over [−π/2,π/2][-\pi/2, \pi/2], the point runs once along the upper half of the circle, from (−a,0)(-a, 0) through (0,a)(0, a) to (a,0)(a, 0), and its height is y=acos⁡θ=a2−x2y = a\cos\theta = \sqrt{a^2 - x^2}. The more familiar parameterization (acos⁡t,asin⁡t)(a\cos t, a\sin t) measures tt from the positive xx-axis. It gives the same points with t=π/2−θt = \pi/2 - \theta, and it swaps the roles of sine and cosine.

In plain words The substitution x=asin⁡θx = a\sin\theta says that xx is the horizontal coordinate of a point on a circle of radius aa. The root a2−x2\sqrt{a^2 - x^2} is then the height of that point, and the identity a2−a2sin⁡2θ=a2cos⁡2θa^2 - a^2\sin^2\theta = a^2\cos^2\theta is the Pythagorean theorem in the right triangle with hypotenuse aa. The circle helps choose the variable. It need not be the graph whose area is computed.

Example For a=2a = 2, the point with x=1x = 1 has sin⁡θ=1/2\sin\theta = 1/2, so θ=π/6\theta = \pi/6, and its height is 2cos⁡(π/6)=32\cos(\pi/6) = \sqrt{3}. In the integral of this page the integrand 1/4−x21/\sqrt{4 - x^2} equals 1/y1/y, the reciprocal of the height of the circle, so the region being measured is not the region under the semicircle. The value π/6\pi/6 equals the angle swept from θ=0\theta = 0 to θ=π/6\theta = \pi/6, in radians. It is not the area of the circular sector with that angle, which is 12⋅22⋅π6=π3\tfrac12 \cdot 2^2 \cdot \tfrac{\pi}{6} = \tfrac{\pi}{3}.

Angle bounds

Definition When a definite integral in xx is rewritten with x=g(θ)x = g(\theta), the new bounds are angles α\alpha and β\beta such that g(α)g(\alpha) is the old lower bound and g(β)g(\beta) is the old upper bound, chosen inside the range on which the substitution is one-to-one and the root has a known sign. For the sine substitution x=asin⁡θx = a\sin\theta and old bounds p≤qp \le q in [−a,a][-a, a], they are

α=arcsin⁡pa,β=arcsin⁡qa,\begin{gathered} \alpha = \arcsin\frac{p}{a}, \\[0.4em] \beta = \arcsin\frac{q}{a}, \end{gathered}

both in [−π/2,π/2][-\pi/2, \pi/2], and as θ\theta increases from α\alpha to β\beta, the value of xx increases from pp to qq. The angles are measured in radians.

In plain words Change the bounds together with the variable, and there is no need to substitute back at the end. Finding each new bound with the inverse sine puts it in the allowed range automatically. Another angle with the same sine gives the same endpoint, but the path between the bounds would then leave the range where the root equals acos⁡θa\cos\theta.

Example For the integral of 1/4−x21/\sqrt{4 - x^2} from −1-1 to 11, the new bounds are arcsin⁡(−1/2)=−π/6\arcsin(-1/2) = -\pi/6 and arcsin⁡(1/2)=π/6\arcsin(1/2) = \pi/6. A negative angle is allowed, and cos⁡θ>0\cos\theta > 0 on [−π/6,π/6][-\pi/6, \pi/6], so

∫−11dx4−x2=∫−π/6π/61 dθ=π3.\begin{aligned} &\int_{-1}^{1} \frac{dx}{\sqrt{4 - x^2}} \\[0.4em] &\quad = \int_{-\pi/6}^{\pi/6} 1\,d\theta = \frac{\pi}{3}. \end{aligned}

The bounds must be radians. The angle π/6\pi/6 is 3030 degrees, and bounds of −30-30 and 3030 in the θ\theta integral would give a value about 5757 times too large.

Nonnegative square root

Definition For a number s≥0s \ge 0, the symbol s\sqrt{s} denotes the nonnegative number whose square is ss. Consequently t2=∣t∣\sqrt{t^2} = \lvert t \rvert for every real tt, and for a>0a > 0

a2cos⁡2θ=a∣cos⁡θ∣,\sqrt{a^2\cos^2\theta} = a\lvert\cos\theta\rvert,

which equals acos⁡θa\cos\theta exactly when cos⁡θ≥0\cos\theta \ge 0, in particular for every θ\theta in [−π/2,π/2][-\pi/2, \pi/2].

In plain words Replacing a2cos⁡2θ\sqrt{a^2\cos^2\theta} by acos⁡θa\cos\theta is a claim about the sign of the cosine. It is true on the angle range you chose and false wherever the cosine is negative, so check the range before removing the absolute value.

Example On [0,π/6][0, \pi/6] the cosine lies between 3/2\sqrt{3}/2 and 11, so 4cos⁡2θ=2cos⁡θ\sqrt{4\cos^2\theta} = 2\cos\theta there. At θ=2π/3\theta = 2\pi/3 the cosine is −1/2-1/2, and 4cos⁡2θ=1\sqrt{4\cos^2\theta} = 1 while 2cos⁡θ=−12\cos\theta = -1. A substitution that let θ\theta go past π/2\pi/2 would need the root written as −2cos⁡θ-2\cos\theta wherever the cosine is negative.

Inverse sine

Definition For −1≤y≤1-1 \le y \le 1, the inverse sine arcsin⁡y\arcsin y, also written sin⁡−1y\sin^{-1} y, is the unique angle θ\theta in [−π/2,π/2][-\pi/2, \pi/2] with sin⁡θ=y\sin\theta = y. It is the inverse function of the sine restricted to [−π/2,π/2][-\pi/2, \pi/2], not the reciprocal 1/sin⁡y1/\sin y. For −1<y<1-1 < y < 1 it is differentiable, with

ddyarcsin⁡y=11−y2,\frac{d}{dy}\arcsin y = \frac{1}{\sqrt{1 - y^2}},

because differentiating sin⁡θ=y\sin\theta = y gives cos⁡θ⋅dθ/dy=1\cos\theta \cdot d\theta/dy = 1, and cos⁡θ=1−y2\cos\theta = \sqrt{1 - y^2} since cos⁡θ>0\cos\theta > 0 for −π/2<θ<π/2-\pi/2 < \theta < \pi/2. Hence for a>0a > 0, on the interval −a<x<a-a < x < a,

∫dxa2−x2=arcsin⁡xa+C.\begin{aligned} &\int \frac{dx}{\sqrt{a^2 - x^2}} \\[0.4em] &\quad = \arcsin\frac{x}{a} + C. \end{aligned}

In plain words The inverse sine reads the angle off from its sine. In a sine substitution, arcsin⁡(x/a)\arcsin(x/a) is exactly the angle θ\theta, so the antiderivative θ+C\theta + C of the reduced integrand becomes arcsin⁡(x/a)+C\arcsin(x/a) + C when it is written back in xx.

Example Let F(x)=arcsin⁡(x/2)F(x) = \arcsin(x/2) for −2<x<2-2 < x < 2. The chain rule and the nonnegative root 1−x2/4=124−x2\sqrt{1 - x^2/4} = \tfrac12\sqrt{4 - x^2} give

F′(x)=1/21−x2/4=14−x2.\begin{aligned} F'(x) &= \frac{1/2}{\sqrt{1 - x^2/4}} \\[0.4em] &= \frac{1}{\sqrt{4 - x^2}}. \end{aligned}

The integrand is continuous on [0,1][0, 1], so the fundamental theorem gives F(1)−F(0)=π/6−0=π/6F(1) - F(0) = \pi/6 - 0 = \pi/6, the value found above. A rough bound agrees. On [0,1][0, 1] the integrand increases from 1/21/2 to 1/3≈0.57741/\sqrt{3} \approx 0.5774, so over this interval of length 11 the integral lies between 0.50.5 and 0.57740.5774, and π/6≈0.5236\pi/6 \approx 0.5236 does.

Check yourself

  1. Evaluate the integral below with a sine substitution, and check the result with the inverse sine.
∫02dx4−x2\int_0^{\sqrt{2}} \frac{dx}{\sqrt{4 - x^2}}
Show the answer

The integrand is continuous on [0,2][0, \sqrt{2}], since 4−x2≥24 - x^2 \ge 2 there. Put x=2sin⁡θx = 2\sin\theta. The bound x=0x = 0 gives θ=0\theta = 0, and x=2x = \sqrt{2} gives sin⁡θ=2/2\sin\theta = \sqrt{2}/2, so θ=π/4\theta = \pi/4. On [0,π/4][0, \pi/4] the cosine is at least 2/2>0\sqrt{2}/2 > 0, so the root is 2cos⁡θ2\cos\theta, and with dx=2cos⁡θ dθdx = 2\cos\theta\,d\theta

∫02dx4−x2=∫0π/41 dθ=π4.\begin{aligned} &\int_0^{\sqrt{2}} \frac{dx}{\sqrt{4 - x^2}} \\[0.4em] &\quad = \int_0^{\pi/4} 1\,d\theta = \frac{\pi}{4}. \end{aligned}

The inverse sine gives the same value, arcsin⁡(2/2)−arcsin⁡0=π/4≈0.7854\arcsin(\sqrt{2}/2) - \arcsin 0 = \pi/4 \approx 0.7854. As a rough check, the integrand rises from 1/21/2 to 1/21/\sqrt{2} over an interval of length 2\sqrt{2}, so the integral lies between 2/2≈0.7071\sqrt{2}/2 \approx 0.7071 and 11.

  1. Evaluate the integral below, and explain the answer with the geometry of the circle.
∫014−x2 dx\int_0^1 \sqrt{4 - x^2}\,dx
Show the answer

With x=2sin⁡θx = 2\sin\theta and θ\theta from 00 to π/6\pi/6, the root is 2cos⁡θ2\cos\theta and dx=2cos⁡θ dθdx = 2\cos\theta\,d\theta, so the new integrand is 4cos⁡2θ=2+2cos⁡2θ4\cos^2\theta = 2 + 2\cos 2\theta. Then

∫014−x2 dx=∫0π/6(2+2cos⁡2θ) dθ=[2θ+sin⁡2θ]0π/6=π3+32≈1.9132.\begin{aligned} &\int_0^1 \sqrt{4 - x^2}\,dx \\[0.4em] &\quad = \int_0^{\pi/6} (2 + 2\cos 2\theta)\,d\theta \\[0.4em] &\quad = \Bigl[2\theta + \sin 2\theta\Bigr]_0^{\pi/6} \\[0.4em] &\quad = \frac{\pi}{3} + \frac{\sqrt{3}}{2} \approx 1.9132. \end{aligned}

This time the graph is the upper half of the circle of radius 22, so the answer is an area in the circle picture. The radius to the point (1,3)(1, \sqrt{3}) splits the region into a sector with angle π/6\pi/6 and area π/3\pi/3, and a right triangle with legs 11 and 3\sqrt{3} and area 3/2\sqrt{3}/2. The integrand decreases from 22 to 3≈1.732\sqrt{3} \approx 1.732, and the value lies between those two numbers, as it must over an interval of length 11.

  1. In the integral of this page, someone takes the upper bound θ=5π/6\theta = 5\pi/6, since sin⁡(5π/6)=1/2\sin(5\pi/6) = 1/2 as well, replaces the root by 2cos⁡θ2\cos\theta, and gets 5π/6≈2.6185\pi/6 \approx 2.618. What went wrong?
Show the answer

The angle 5π/65\pi/6 lies outside [−π/2,π/2][-\pi/2, \pi/2], the range on which the sine substitution is one-to-one and the cosine is nonnegative, and two things fail. For π/2<θ≤5π/6\pi/2 < \theta \le 5\pi/6 the cosine is negative, so there 4cos⁡2θ=−2cos⁡θ\sqrt{4\cos^2\theta} = -2\cos\theta, not 2cos⁡θ2\cos\theta. Also, as θ\theta runs from 00 to 5π/65\pi/6, the point x=2sin⁡θx = 2\sin\theta climbs from 00 to 22 and comes back to 11, so it passes through x=2x = 2, where the integrand is undefined, and the substitution rule does not apply as stated. The inverse sine gives the correct bound arcsin⁡(1/2)=π/6\arcsin(1/2) = \pi/6 and the value π/6\pi/6. The rough bound catches the error as well, since the integral lies between 0.50.5 and 0.57740.5774.

How this lesson connects

It also uses circle coordinates and right-triangle trigonometry.