Skip to content
Topos Uranos

Trigonometric and hyperbolic substitutions

Summary

This lesson finds the antiderivatives of the functions that contain one of the roots a2−x2\sqrt{a^2 - x^2}, a2+x2\sqrt{a^2 + x^2} or x2−a2\sqrt{x^2 - a^2}, with a>0a > 0, by replacing xx by a trigonometric or hyperbolic function whose square, by a fundamental identity, turns the radicand into a perfect square. Each substitution is a case of the change of variable by the inverse function: the interval of the new variable is chosen so that the substitution is invertible, its derivative does not vanish and the root simplifies with a known sign, without an absolute value or with one that is justified. This yields the antiderivatives of the three roots, first with trigonometric functions and then with hyperbolic ones, and the two routes are compared. The lesson closes with the method of completing the square, which reduces the root of any polynomial of degree two to one of the three above.

Learning objectives

  1. Choose, for each of the roots a2−x2\sqrt{a^2 - x^2}, a2+x2\sqrt{a^2 + x^2} and x2−a2\sqrt{x^2 - a^2}, the trigonometric substitution and the interval of the new variable on which it is invertible, and simplify the root with the correct sign.
  2. Prove the antiderivatives of the three roots, with the interval on which each one holds, including the endpoints at which the root vanishes.
  3. Apply the hyperbolic substitutions, compare them with the trigonometric ones and check that the antiderivatives given by the two routes differ by a constant.
  4. Complete the square to reduce ax2+bx+c\sqrt{ax^2 + bx + c} to one of the three roots, and recognize the cases in which the domain is the real line, two rays, a bounded interval or a set containing no intervals.

Three roots and three trigonometric substitutions

Consider the functions that contain a root of the form a2−x2\sqrt{a^2 - x^2}, a2+x2\sqrt{a^2 + x^2} or x2−a2\sqrt{x^2 - a^2}, with a>0a > 0. No rule of the table of antiderivatives handles them in general, but each one becomes a trigonometric function without roots if xx is replaced by asin⁡ta \sin t, atan⁡ta \tan t or asec⁡ta \sec t: the Pythagorean identities 1−sin⁡2t=cos⁡2t1 - \sin^2 t = \cos^2 t, 1+tan⁡2t=sec⁡2t1 + \tan^2 t = \sec^2 t and sec⁡2t−1=tan⁡2t\sec^2 t - 1 = \tan^2 t turn the radicand into a square. The theorem on trigonometric substitutions, from the lesson on the fundamental trigonometric identities, already did that computation with a=1a = 1. Integrating requires something more: the theorem of change of variable by the inverse function, from the lesson on integration by change of variable, requires an interval TT on which the substitution x=h(t)x = h(t) is differentiable with h′(t)≠0h'(t) \neq 0, and the integrand to be defined on I=h(T)I = h(T). That is why the intervals of tt are open: at their endpoints, the derivative of the substitution vanishes or does not exist.

PropositionThe three trigonometric substitutions

Let a>0a > 0. In each case, hh is differentiable on the interval TT with h′(t)≠0h'(t) \neq 0, the image I=h(T)I = h(T) and the inverse are as indicated, and the root simplifies as follows:

  • h(t)=asin⁡th(t) = a \sin t, on T=(−π2,π2)T = \ioo{-\frac{\pi}{2}}{\frac{\pi}{2}}: I=(−a,a)I = \ioo{-a}{a}, h−1(x)=arcsin⁡xah^{-1}(x) = \arcsin \frac{x}{a}, and a2−h(t)2=acos⁡t\sqrt{a^2 - h(t)^2} = a \cos t;
  • h(t)=atan⁡th(t) = a \tan t, on T=(−π2,π2)T = \ioo{-\frac{\pi}{2}}{\frac{\pi}{2}}: I=RI = \mathbb{R}, h−1(x)=arctan⁡xah^{-1}(x) = \arctan \frac{x}{a}, and a2+h(t)2=asec⁡t\sqrt{a^2 + h(t)^2} = a \sec t;
  • h(t)=asec⁡th(t) = a \sec t, on T+=(0,π2)T_+ = \ioo{0}{\frac{\pi}{2}} and on T−=(π2,π)T_- = \ioo{\frac{\pi}{2}}{\pi}: h(T+)=(a,∞)h(T_+) = \ioo{a}{\infty}, h(T−)=(−∞,−a)h(T_-) = \ioo{-\infty}{-a}, in both cases h−1(x)=arccos⁡axh^{-1}(x) = \arccos \frac{a}{x}, and h(t)2−a2\sqrt{h(t)^2 - a^2} equals atan⁡ta \tan t on T+T_+ and −atan⁡t-a \tan t on T−T_-.

Each Pythagorean identity gives the square of the root, and the sign of the trigonometric function on the chosen interval decides which of the two square roots it is.

Proof

  1. a2−a2sin⁡2t=acos⁡2t=a∣cos⁡t∣=acos⁡t,t∈T\sqrt{a^2 - a^2 \sin^2 t} = a \sqrt{\cos^2 t} = a \abs{\cos t} = \resaltar{a \cos t}, \qquad t \in T

    On TT, cos⁡t>0\cos t > 0, and the derivative h′(t)=acos⁡th'(t) = a \cos t is positive. By the theorem on bijective restrictions, from the lesson on the inverse trigonometric functions, sine is a bijection from [−π2,π2]\icc{-\frac{\pi}{2}}{\frac{\pi}{2}} onto [−1,1]\icc{-1}{1}, and it is increasing, by the theorem on monotonicity on the principal intervals; without the endpoints, h(T)=(−a,a)h(T) = \ioo{-a}{a}, and t=arcsin⁡xat = \arcsin \frac{x}{a}. Since y2=∣y∣\sqrt{y^2} = \abs{y} for every real yy:

  2. a2+a2tan⁡2t=asec⁡2t=a∣sec⁡t∣=asec⁡t,t∈T\sqrt{a^2 + a^2 \tan^2 t} = a \sqrt{\sec^2 t} = a \abs{\sec t} = \resaltar{a \sec t}, \qquad t \in T

    The derivative h′(t)=asec⁡2th'(t) = a \sec^2 t is positive on TT, tangent is a bijection from TT onto R\mathbb{R} by the same theorem, and t=arctan⁡xat = \arctan \frac{x}{a}. On TT, sec⁡t>0\sec t > 0.

  3. a2sec⁡2t−a2=a∣tan⁡t∣=atan⁡t,t∈T+,a∣tan⁡t∣=−atan⁡t,t∈T−\sqrt{a^2 \sec^2 t - a^2} = a \abs{\tan t} = \resaltar{a \tan t}, \quad t \in T_+, \qquad a \abs{\tan t} = \resaltar{-a \tan t}, \quad t \in T_-

    The derivative h′(t)=asec⁡ttan⁡t=asin⁡tcos⁡2th'(t) = a \sec t \tan t = \frac{a \sin t}{\cos^2 t} is positive on T+T_+ and on T−T_-, where sin⁡t>0\sin t > 0. If ∣x∣>a\abs{x} > a, then ax\frac{a}{x} lies in [−1,1]\icc{-1}{1} and is not zero; by the theorem on bijective restrictions, t=arccos⁡axt = \arccos \frac{a}{x} is the only tt in [0,π]\icc{0}{\pi} with cos⁡t=ax\cos t = \frac{a}{x}, that is, with asec⁡t=xa \sec t = x, and it lies in T+T_+ if x>ax > a and in T−T_- if x<−ax < -a. Tangent is positive on T+T_+ and negative on T−T_-.

The figure shows a device for returning to the variable xx: in each case, a right triangle with an angle tt, whose sides are aa, xx and the root. From it one reads off the other trigonometric functions of tt; for example, if x=asin⁡tx = a \sin t, then cos⁡t=a2−x2a\cos t = \frac{\sqrt{a^2 - x^2}}{a} and tan⁡t=xa2−x2\tan t = \frac{x}{\sqrt{a^2 - x^2}}. The triangle only works if tt lies between 00 and π2\frac{\pi}{2}; for the other values, the same equalities are proved with the theorem on cross compositions, from the lesson on the inverse trigonometric functions, according to which cos⁡(arcsin⁡y)=1−y2\cos(\arcsin y) = \sqrt{1 - y^2} on [−1,1]\icc{-1}{1}, or directly with the sign fixed in the proposition.

In the third case, the sign on T−T_- is unavoidable if one wants the same inverse arccos⁡ax\arccos \frac{a}{x} on both branches. The interval (π,3π2)\ioo{\pi}{\frac{3\pi}{2}}, where tangent is positive, avoids the sign, but then the inverse is 2π−arccos⁡ax2\pi - \arccos \frac{a}{x}; the two choices give the same antiderivative, up to a constant. The theorem on trigonometric substitutions only dealt with x≥1x \geq 1; the proposition extends it to the negative branch.

TheoremAntiderivatives of the three roots

Let a>0a > 0. On the closed interval [−a,a]\icc{-a}{a}:

∫a2−x2 dx=12xa2−x2+a22arcsin⁡xa+C\int \sqrt{a^2 - x^2} \, \mathrm{d}x = \frac{1}{2} x \sqrt{a^2 - x^2} + \frac{a^2}{2} \arcsin \frac{x}{a} + C

On R\mathbb{R}:

∫x2+a2 dx=12xx2+a2+a22ln⁡(x+x2+a2)+C\int \sqrt{x^2 + a^2} \, \mathrm{d}x = \frac{1}{2} x \sqrt{x^2 + a^2} + \frac{a^2}{2} \ln\left(x + \sqrt{x^2 + a^2}\right) + C

On each of the intervals (−∞,−a]\ioc{-\infty}{-a} and [a,∞)\ico{a}{\infty}, with one constant per interval:

∫x2−a2 dx=12xx2−a2−a22ln⁡∣x+x2−a2∣+C\int \sqrt{x^2 - a^2} \, \mathrm{d}x = \frac{1}{2} x \sqrt{x^2 - a^2} - \frac{a^2}{2} \ln \abs{x + \sqrt{x^2 - a^2}} + C

Each formula is obtained with the corresponding substitution of the preceding proposition, and it also holds at the endpoints at which the root vanishes.

Proof

  1. ∫a2cos⁡2t dt=a22(t+sin⁡tcos⁡t)+C→F1(x)=12xa2−x2+a22arcsin⁡xa\int a^2 \cos^2 t \, \mathrm{d}t = \frac{a^2}{2} \left(t + \sin t \cos t\right) + C \rightarrow \resaltar{F_1(x) = \frac{1}{2} x \sqrt{a^2 - x^2} + \frac{a^2}{2} \arcsin \frac{x}{a}}

    With x=asin⁡tx = a \sin t on T=(−π2,π2)T = \ioo{-\frac{\pi}{2}}{\frac{\pi}{2}}, the integrand in the variable tt is acos⁡t⋅acos⁡ta \cos t \cdot a \cos t. By the power-reducing formula cos⁡2t=1+cos⁡2t2\cos^2 t = \frac{1 + \cos 2t}{2} and the corollary of the linear change of variable, from the lesson on change of variable, one obtains GG. To go back, sin⁡t=xa\sin t = \frac{x}{a} and cos⁡t=a2−x2a\cos t = \frac{\sqrt{a^2 - x^2}}{a}, by the proposition; the inverse substitution theorem gives the antiderivative on (−a,a)\ioo{-a}{a}.

  2. lim⁡x→a−F1′(x)=lim⁡x→a−a2−x2=0=a2−a2\lim_{x \to a^-} F_1'(x) = \lim_{x \to a^-} \sqrt{a^2 - x^2} = \resaltar{0 = \sqrt{a^2 - a^2}}

    The function F1F_1 is continuous on [−a,a]\icc{-a}{a}: the root is, and so is the arcsine, by the corollary on the continuity of the inverse, from the lesson on the intermediate value theorem; its derivative on (−a,a)\ioo{-a}{a} is the root, which tends to 00 at the endpoints. By the corollary on the limit of the derivative, from the lesson on the theorems of Fermat, Rolle, Lagrange and Cauchy, F1F_1 has at aa a left derivative, and at −a-a a right derivative, both equal to the value of the root: F1F_1 is an antiderivative on [−a,a]\icc{-a}{a}.

  3. ∫a2sec⁡3t dt=a22(sec⁡ttan⁡t+ln⁡(sec⁡t+tan⁡t))+C→12xx2+a2+a22ln⁡(x+x2+a2)+C\int a^2 \sec^3 t \, \mathrm{d}t = \frac{a^2}{2} \left(\sec t \tan t + \ln\left(\sec t + \tan t\right)\right) + C \rightarrow \resaltar{\frac{1}{2} x \sqrt{x^2 + a^2} + \frac{a^2}{2} \ln\left(x + \sqrt{x^2 + a^2}\right) + C}

    With x=atan⁡tx = a \tan t on T=(−π2,π2)T = \ioo{-\frac{\pi}{2}}{\frac{\pi}{2}}, the integrand in the variable tt is asec⁡t⋅asec⁡2ta \sec t \cdot a \sec^2 t, and the antiderivative of sec⁡3t\sec^3 t is the one given by the proposition on the reduction of the powers of the secant, together with the theorem on the antiderivative of the secant, from the lesson on trigonometric integrals; on TT, sec⁡t+tan⁡t>0\sec t + \tan t > 0. Going back, with tan⁡t=xa\tan t = \frac{x}{a} and sec⁡t=x2+a2a\sec t = \frac{\sqrt{x^2 + a^2}}{a}, the logarithm is ln⁡(x+x2+a2)−ln⁡a\ln\left(x + \sqrt{x^2 + a^2}\right) - \ln a, and the constant −a22ln⁡a-\frac{a^2}{2} \ln a is absorbed into CC.

  4. G(t)=a22(sec⁡ttan⁡t−ln⁡∣sec⁡t+tan⁡t∣)→G(arccos⁡ax)=F3(x)+a22ln⁡aG(t) = \frac{a^2}{2} \left(\sec t \tan t - \ln \abs{\sec t + \tan t}\right) \rightarrow \resaltar{G\left(\arccos \frac{a}{x}\right) = F_3(x) + \frac{a^2}{2} \ln a}

    With x=asec⁡tx = a \sec t on T+T_+, the integrand in the variable tt is atan⁡t⋅asec⁡ttan⁡t=a2(sec⁡3t−sec⁡t)a \tan t \cdot a \sec t \tan t = a^2 \left(\sec^3 t - \sec t\right), by tan⁡2t=sec⁡2t−1\tan^2 t = \sec^2 t - 1. With the antiderivatives of sec⁡3t\sec^3 t and of sec⁡t\sec t, given by the proposition on the reduction of the powers of the secant and the theorem on the antiderivative of the secant, from the lesson on trigonometric integrals, one obtains GG; going back, with sec⁡t=xa\sec t = \frac{x}{a} and tan⁡t=x2−a2a\tan t = \frac{\sqrt{x^2 - a^2}}{a}, the result is the formula of the statement plus the constant a22ln⁡a\frac{a^2}{2} \ln a, on (a,∞)\ioo{a}{\infty}.

  5. −G(arccos⁡ax)=F3(x)+a22ln⁡a,x<−a-G\left(\arccos \frac{a}{x}\right) = \resaltar{F_3(x) + \frac{a^2}{2} \ln a}, \qquad x < -a

    On T−T_-, the root is −atan⁡t-a \tan t, the integrand in tt is −a2sec⁡ttan⁡2t-a^2 \sec t \tan^2 t and its antiderivative is −G-G; going back, tan⁡t=−x2−a2a\tan t = -\frac{\sqrt{x^2 - a^2}}{a}, and ln⁡∣x−x2−a2∣\ln \abs{x - \sqrt{x^2 - a^2}} appears. Since (x−x2−a2)(x+x2−a2)=a2\left(x - \sqrt{x^2 - a^2}\right)\left(x + \sqrt{x^2 - a^2}\right) = a^2, that logarithm equals 2ln⁡a−ln⁡∣x+x2−a2∣2 \ln a - \ln \abs{x + \sqrt{x^2 - a^2}}, and once again the result is F3(x)+a22ln⁡aF_3(x) + \frac{a^2}{2} \ln a, on (−∞,−a)\ioo{-\infty}{-a}. At ±a\pm a, as in the second stage, F3F_3 is continuous and its derivative tends to 00.

In the last two stages, F3F_3 is the function of the statement. The sign on the negative branch does not change the formula, because the absolute value in the logarithm absorbs it; but without the preceding proposition one would not know that the same expression holds on both rays. The same substitutions give the antiderivatives of the reciprocals of the roots, which the proposition on antiderivatives with a parameter, from the lesson on change of variable, obtained in another way: with x=atan⁡tx = a \tan t, for example, ∫1x2+a2 dx\int \frac{1}{\sqrt{x^2 + a^2}} \, \mathrm{d}x becomes ∫sec⁡t dt=ln⁡(sec⁡t+tan⁡t)+C\int \sec t \, \mathrm{d}t = \ln\left(\sec t + \tan t\right) + C, which after going back is ln⁡(x+x2+a2)−ln⁡a+C\ln\left(x + \sqrt{x^2 + a^2}\right) - \ln a + C.