Chapter VI · Derivatives and integrals

Substitution and rationalisation

§§135–136 35 min

Changing the variable of integration, seen as a picture and as the chain rule; and the strategy of choosing a new variable that turns an algebraical integrand into a rational one.

Why this matters

Rational functions can always be integrated (§133). The next class, algebraical functions such as ax2+2bx+c\sqrt{ax^2 + 2bx + c}, cannot in general, and here Hardy introduces the main tool of practical integration: substitution. Put x=f(t)x = f(t) for a cleverly chosen ff, and a hard integral in xx may become an easy one in tt. The most important special case is rationalisation: choosing tt so that the new integrand is a rational function, which §133 then finishes off. The next two lessons apply this idea to square roots of quadratics.

The idea

Take ∫dx1+x\displaystyle\int \frac{dx}{1 + \sqrt x}. The square root is the obstacle. If we write x=t2x = t^2 (with t≥0t \ge 0), then x=t\sqrt x = t, and the integrand becomes 1/(1+t)1/(1+t), which is rational. But the integral is not simply ∫dt/(1+t)\int dt/(1+t): changing the variable changes the scale along the axis. A small step dtdt in tt corresponds to a step dx=2t dtdx = 2t\,dt in xx, so the integrand must be multiplied by 2t2t:

∫dx1+x=∫2t dt1+t=2t−2log⁡(1+t)=2x−2log⁡(1+x).\int \frac{dx}{1+\sqrt x} = \int \frac{2t\,dt}{1+t} = 2t - 2\log(1+t) = 2\sqrt{x} - 2\log(1 + \sqrt{x}) .

Why the factor f′(t)f'(t)? Think of the integral as an area, as Hardy will justify in §148. Cut the tt-axis into equal strips. The map x=f(t)x = f(t) sends them to strips of the xx-axis that are not equal: each is stretched by roughly f′(t)f'(t). For the areas to match, the height in the tt-picture must be the height in the xx-picture times that stretch.

Explore Substitution as matching strips

Try this. The lower picture is ∫dx/(1+x)\int dx/(1+\sqrt x) from x=0x = 0 to 44; the upper is the same integral after x=t2x = t^2. Move the highlighted strip kk from left to right and read the stretch factor. Increase nn. Then try the example 1−x2\sqrt{1-x^2} with x=sin⁡tx = \sin t.

∫11+x dx\int \dfrac{1}{1+\sqrt x}\,dx with x=t2x = t^2 becomes ∫2t1+t dt\int \dfrac{2t}{1+t}\,dt
tψ(f(t)) f′(t)
xψ(x)
  • strips (alternate colours)
  • …and their partners
  • strip k in both pictures
Strip 1: tt from 00 to 0.3330.333 (width 0.3330.333) becomes xx from 00 to 0.1110.111 (width 0.1110.111, stretched by 0.3330.333). Its area is 0.09130.0913 in both pictures (computed separately: 0.09130.0913). Totals: ∫2t1+t dt=1.8028\int \dfrac{2t}{1+t}\,dt = 1.8028 and ∫11+x dx=1.8028\int \dfrac{1}{1+\sqrt x}\,dx = 1.8028.
What to notice

Equal strips in tt become wider and wider strips in xx (stretch ≈2t\approx 2t), and the upper curve 2t/(1+t)2t/(1+t) is correspondingly taller than 1/(1+x)1/(1+\sqrt{x}) at the matching place. Each pair of matching strips has exactly the same area, so the totals agree. With x=sin⁡tx = \sin t the stretch is cos⁡t\cos t, which shrinks towards t=12πt = \tfrac12\pi, where the strips in xx crowd together.

Precise statements

Theorem Integration by substitution Hardy: §136 (1)

Let ϕ\phi be an integral of ψ\psi, and let ff be differentiable, with values where ψ\psi is defined. Then

∫ψ{f(t)} f′(t) dt=ϕ{f(t)}.\int \psi\{f(t)\}\,f'(t)\,dt = \phi\{f(t)\} .

If moreover ff is strictly monotonic on an interval, with inverse t=g(x)t = g(x), then ϕ(x)\phi(x) is recovered as the integral in tt evaluated at t=g(x)t = g(x).

Hardy states the practical rule like this: put x=f(t)x = f(t), multiply by f′(t)f'(t), integrate the resulting function of tt if you can, and express the result in terms of xx. The theorem is the same rule as the chain rule backwards of §133, read with the roles of the variables exchanged.

Now the strategy that gives the method its power. Hardy considers integrals ∫R(x,y) dx\int R(x, y)\,dx, where RR is a rational function of two variables and yy is an algebraical function of xx (§135). For example

px+q+ax2+2bx+cpx+q−ax2+2bx+c\frac{px + q + \sqrt{ax^2+2bx+c}}{px + q - \sqrt{ax^2+2bx+c}}

is a rational function of xx and y=ax2+2bx+cy = \sqrt{ax^2 + 2bx + c}. This form is no more general than “an algebraical function of xx” (a rational function of xx and yy is itself algebraical), but it is far more convenient.

Theorem Rationalisation Hardy: §136

If there is a variable tt such that both xx and yy are rational functions of tt, say x=R1(t)x = R_1(t) and y=R2(t)y = R_2(t), then

∫R(x,y) dx=∫R{R1(t),R2(t)} R1′(t) dt,\int R(x, y)\,dx = \int R\{R_1(t), R_2(t)\}\,R_1'(t)\,dt ,

and the right-hand side is the integral of a rational function of tt, which can be found by partial fractions.

For example, ∫R(x,x) dx\int R(x, \sqrt x)\,dx is rationalised by x=t2x = t^2, x=t\sqrt{x} = t, and becomes ∫2t R(t2,t) dt\int 2t\,R(t^2, t)\,dt. When can such a tt be found? In general that is a deep question (it leads to the genus of an algebraic curve) and Hardy does not pursue it. He treats the most useful case: y2y^2 a quadratic in xx, where the graph of yy is a conic. That is the next lesson.

Proofs

Proof Integration by substitution
Strategy

Differentiate ϕ{f(t)}\phi\{f(t)\} with respect to tt. For the second part, substitute t=g(x)t = g(x) back and use the chain rule once more.

Chain rule

ddtϕ{f(t)}=ϕ′{f(t)} f′(t)=ψ{f(t)} f′(t)\dfrac{d}{dt}\phi\{f(t)\} = \phi'\{f(t)\}\,f'(t) = \psi\{f(t)\}\,f'(t), so ϕ{f(t)}\phi\{f(t)\} is an integral of ψ{f(t)}f′(t)\psi\{f(t)\}f'(t).

Any integral in tt has this form

If Θ(t)\Theta(t) is any integral of ψ{f(t)}f′(t)\psi\{f(t)\}f'(t) on the interval, then Θ(t)=ϕ{f(t)}+C\Theta(t) = \phi\{f(t)\} + C, by the uniqueness theorem of §130.

Return to xx

If ff is strictly monotonic with inverse gg, then f{g(x)}=xf\{g(x)\} = x, so Θ{g(x)}=ϕ(x)+C\Theta\{g(x)\} = \phi(x) + C, an integral of ψ\psi. This is why the substitution must be invertible on the interval used: otherwise we cannot express the result in terms of xx.

The proof of the rationalisation theorem is the same with ψ(x)=R(x,y(x))\psi(x) = R(x, y(x)) and f=R1f = R_1, since then ψ{R1(t)}=R{R1(t),R2(t)}\psi\{R_1(t)\} = R\{R_1(t), R_2(t)\}.

Worked examples and exercises

Exercise A first rationalisation ●○○ Hardy: §136

Find ∫dxxx−1\displaystyle\int \frac{dx}{x\sqrt{x-1}} for x>1x > 1 by putting x=1+t2x = 1 + t^2 with t>0t > 0. (This is the third example in the widget.)

Hint
x−1=t\sqrt{x-1} = t and dx=2t dtdx = 2t\,dt.

Solution

The integral becomes ∫2t dt(1+t2) t=∫2 dt1+t2=2arctan⁡t\displaystyle\int \frac{2t\,dt}{(1+t^2)\,t} = \int\frac{2\,dt}{1+t^2} = 2\arctan t. Since t=x−1t = \sqrt{x-1},

∫dxxx−1=2arctan⁡x−1.\int \frac{dx}{x\sqrt{x-1}} = 2\arctan\sqrt{x-1} .

Check: ddx2arctan⁡x−1=21+(x−1)⋅12x−1=1xx−1\dfrac{d}{dx}2\arctan\sqrt{x-1} = \dfrac{2}{1 + (x-1)}\cdot\dfrac{1}{2\sqrt{x-1}} = \dfrac{1}{x\sqrt{x-1}}.

Exercise An nnth root of a linear function ●○○ Hardy: Examples XLIX, 9

Show that ∫R{x,ax+bn} dx\displaystyle\int R\{x, \sqrt[n]{ax+b}\}\,dx, with RR rational and a≠0a \ne 0, is reduced to the integral of a rational function by the substitution ax+b=ynax + b = y^n.

Hint
Express xx and dx/dydx/dy in terms of yy.

Solution

With ax+b=ynax + b = y^n we have ax+bn=y\sqrt[n]{ax+b} = y, x=(yn−b)/ax = (y^n - b)/a and dx/dy=nyn−1/adx/dy = ny^{n-1}/a. All three are rational in yy, so

∫R{x,ax+bn} dx=∫R{yn−ba, y}nyn−1a dy,\int R\{x, \sqrt[n]{ax+b}\}\,dx = \int R\left\{\frac{y^n - b}{a},\ y\right\}\frac{n y^{n-1}}{a}\,dy ,

the integral of a rational function of yy. (For even nn we take y≥0y \ge 0 and ax+b≥0ax + b \ge 0, so that the substitution is invertible.)

Exercise Two different roots ●●○ Hardy: §136

Find ∫dxx+x3\displaystyle\int \frac{dx}{\sqrt{x} + \sqrt[3]{x}} for x>0x > 0.

Hint
Choose x=tkx = t^k so that both x\sqrt x and x3\sqrt[3]x become powers of tt: kk must be divisible by 22 and by 33.
Hint
With x=t6x = t^6 the integrand becomes 6t5t3+t2=6t3t+1\dfrac{6t^5}{t^3 + t^2} = \dfrac{6t^3}{t+1}. Divide.

Solution

Put x=t6x = t^6, t>0t > 0. Then x=t3\sqrt x = t^3, x3=t2\sqrt[3]x = t^2, dx=6t5 dtdx = 6t^5\,dt, and

∫6t5 dtt3+t2=6∫t3t+1 dt=6∫(t2−t+1−1t+1)dt=2t3−3t2+6t−6log⁡(1+t).\int \frac{6t^5\,dt}{t^3+t^2} = 6\int\frac{t^3}{t+1}\,dt = 6\int\left(t^2 - t + 1 - \frac{1}{t+1}\right)dt = 2t^3 - 3t^2 + 6t - 6\log(1+t) .

With t=x1/6t = x^{1/6}: the answer is 2x−3x3+6x6−6log⁡(1+x6)2\sqrt x - 3\sqrt[3]x + 6\sqrt[6]x - 6\log(1 + \sqrt[6]x).

Exercise The binomial integral ∫(1+x)pxq dx\int (1+x)^p x^q\,dx ●●● Hardy: Examples XLIX, 11

Let p,qp, q be rational. Show that ∫(1+x)pxq dx\int (1+x)^p x^q\,dx can be rationalised if qq is an integer (put 1+x=ts1 + x = t^s, where ss is the denominator of pp), and if p+qp + q is an integer (put 1+x=xts1 + x = xt^s).

Hint
In the first case (1+x)p=tsp(1+x)^p = t^{sp} and spsp is an integer.
Hint
In the second case solve for xx: x=1/(ts−1)x = 1/(t^s - 1). Then 1+x=ts/(ts−1)1 + x = t^s/(t^s - 1).

Solution

qq an integer. With 1+x=ts1 + x = t^s: (1+x)p=tsp(1+x)^p = t^{sp} with spsp an integer, xq=(ts−1)qx^q = (t^s - 1)^q is rational in tt because qq is an integer, and dx=sts−1dtdx = st^{s-1}dt. The integrand becomes s tsp+s−1(ts−1)qs\,t^{sp + s - 1}(t^s - 1)^q, a rational function.

p+qp + q an integer. With 1+x=xts1 + x = xt^s we get x=1/(ts−1)x = 1/(t^s - 1) and 1+x=ts/(ts−1)1 + x = t^s/(t^s - 1). Then

(1+x)pxq=tsp(ts−1)p⋅1(ts−1)q=tsp(ts−1)−(p+q),(1+x)^p x^q = \frac{t^{sp}}{(t^s-1)^{p}}\cdot\frac{1}{(t^s-1)^q} = t^{sp}(t^s - 1)^{-(p+q)},

rational because spsp and p+qp + q are integers; and dx/dt=−sts−1/(ts−1)2dx/dt = -st^{s-1}/(t^s-1)^2 is rational. (The third case, pp an integer, is like the first with x=usx = u^s, ss the denominator of qq.) A theorem of Chebyshev says these three cases are the only ones in which the integral is elementary.

Exercise A curve rationalised by y=txy = tx ●●○ Hardy: Examples XLIX, 14

Let yy be defined by y2(x−y)=x2y^2(x - y) = x^2. Show that ∫R(x,y) dx\int R(x, y)\,dx can be reduced to the integral of a rational function.

Hint
Put y=txy = tx in the equation and solve for xx.

Solution

With y=txy = tx: t2x2(x−tx)=x2t^2x^2(x - tx) = x^2, so (for x≠0x \ne 0) t2(1−t)x=1t^2(1 - t)x = 1, that is

x=1t2(1−t),y=tx=1t(1−t).x = \frac{1}{t^2(1-t)},\qquad y = tx = \frac{1}{t(1-t)} .

Both are rational in tt, and so is dx/dtdx/dt. By the rationalisation theorem, ∫R(x,y) dx=∫R{x(t),y(t)} x′(t) dt\int R(x,y)\,dx = \int R\{x(t), y(t)\}\,x'(t)\,dt is the integral of a rational function. The trick works because every line y=txy = tx through the origin meets the curve in only one other point: the origin counts as a double point of this cubic.

Check yourself

Check yourself Choose one

After the substitution x=t2x = t^2 (t>0t > 0), the integral ∫sin⁡x dx\int \sin\sqrt{x}\,dx becomes

Why

Replace xx by t2t^2 and dxdx by f′(t) dt=2t dtf'(t)\,dt = 2t\,dt. Forgetting the factor f′(t)f'(t) (first option) is the most common error; dividing by it (third) is the second most common. The fourth substitutes in the wrong direction.

Check yourself Enter a number

Using ∫dx/(1+x)=2x−2log⁡(1+x)\int dx/(1+\sqrt x) = 2\sqrt x - 2\log(1+\sqrt x), compute the increase of this integral from x=0x = 0 to x=4x = 4 (to three decimals). Compare with the totals in the widget.

Why

At x=4x = 4: 4−2log⁡34 - 2\log 3; at x=0x = 0: 00. The difference is 4−2log⁡3≈1.80284 - 2\log 3 \approx 1.8028, the value the widget shows for both pictures with t1=2t_1 = 2.

Check yourself Choose one

Which substitution rationalises ∫x dx1+x4\displaystyle\int \frac{x\,dx}{1 + \sqrt[4]{x}}?

Why

With x=t4x = t^4 the root becomes tt, dx=4t3 dtdx = 4t^3\,dt, and the integrand 4t7/(1+t)4t^7/(1+t) is rational. With x=t2x = t^2 a square root t\sqrt t remains. The others leave a fourth root of something.

Check yourself True or false?

Any differentiable ff may be used in the substitution x=f(t)x = f(t), and the answer can always be converted back to xx.

Why

False. The first half of the theorem holds for any differentiable ff, but to express the result in xx we need tt as a function of xx, which requires ff to be invertible (for example strictly monotonic) on the interval used. With x=sin⁡tx = \sin t we restrict tt to [−12π,12π][-\tfrac12\pi, \tfrac12\pi] for exactly this reason.

Hardy’s language

Summary

Summary
  • Substitution: if x=f(t)x = f(t), then ∫ψ(x) dx=∫ψ{f(t)}f′(t) dt\int\psi(x)\,dx = \int\psi\{f(t)\}f'(t)\,dt. The factor f′(t)f'(t) is the stretch of the axis; never forget it.
  • To return to xx, the substitution must be invertible on the interval in question.
  • Rationalisation: find tt with xx and yy both rational in tt; then ∫R(x,y) dx\int R(x,y)\,dx is the integral of a rational function.
  • Roots of a linear function, ax+bn\sqrt[n]{ax+b}, are rationalised by ax+b=tnax + b = t^n; several roots of xx by a common power x=tkx = t^k.
  • Whether a rationalising tt exists depends on the curve y(x)y(x); for y2y^2 quadratic it always does (next lesson).