Corrected exercises: the substitution rule (MATH 141)
This is the corrected exercise set for the substitution chapter of MATH 141, Calculus 2, the second calculus course at McGill University. It follows section 5.5 of Stewart, the first technique of integration of the course and the one every later technique leans on. Every number is chosen to be done by hand: the bounds land on ln2, 12π or 37, and the solutions write the line u=…, du=…, and the line of new bounds, because that is where the method marks are.
The thread running through the whole set: a substitution is finished only when the integral contains NO x at all, not in the integrand, not in the differential, not in the bounds. A constant missing from du is supplied by a constant; a variable factor never is. An x left over after du is used is rewritten through u. The bounds travel through u=g(x), even when they come out in decreasing order. And on a symmetric interval the substitution u=−x shows that an odd integrand integrates to 0, which settles integrals no antiderivative could reach.
The traps named explicitly in the solutions: dividing by 3x2 to manufacture a du, keeping the bounds of x in an integral in u, reordering reversed bounds while keeping the minus sign, pulling a stray x out of the integral as if it were a constant, the sign of ∫tanxdx, dropping the absolute value of ln∣u∣ and declaring a finite integral undefined, answering ln(1+x2) for ∫1+x2dx, quoting symmetry on an interval that is not symmetric or across a discontinuity, and believing that u=sinx is illegal on [0,π].
•Substitution rule: ∫f(g(x))g′(x)dx=∫f(u)du with u=g(x), du=g′(x)dx. It is the chain rule read backwards.
•Definite integrals: ∫abf(g(x))g′(x)dx=∫g(a)g(b)f(u)du, valid when g′ is continuous on [a,b] and f is continuous on the values of g. No one-to-one condition.
•A constant missing from du is supplied: xdx=21d(x2). A factor containing x cannot be moved through the integral sign.
•Logarithms: ∫g(x)g′(x)dx=ln∣g(x)∣+C. So ∫tanxdx=−ln∣cosx∣+C=ln∣secx∣+C and ∫cotxdx=ln∣sinx∣+C.
•Table forms after a linear substitution: ∫x2+a2dx=a1arctanax+C and ∫a2−x2dx=arcsinax+C for a>0.
•Symmetry, f continuous on [−a,a]: odd f gives ∫−aaf=0; even f gives ∫−aaf=2∫0af. Reflection: ∫0af(a−x)dx=∫0af(x)dx.
Part A: the basics (/50)
Exercise 1: Choosing u: the inside function whose derivative is already there
The substitution rule reads ∫f(g(x))g′(x)dx=∫f(u)du with u=g(x) and du=g′(x)dx. It is the chain rule read backwards, and it applies when the integrand contains an inside function g(x) AND, as a factor, its derivative g′(x), possibly up to a constant multiple.
The whole question is the choice of u. After the substitution the integral must be written in u alone: no x in the integrand, none in the differential. Every answer below must be checked by differentiating it.
a) ∫x2(x3+1)5dx
b) ∫xcosxdx
c) ∫x(lnx)3dx
d) ∫9−x2xdx. Then say why ∫9−x2dx needs a different u, and compute it.
e) Compute ∫sinxcosxdx twice, once with u=sinx and once with u=cosx. The two answers look different: are they both right?
Show the solution
Answers
a)18(x3+1)6+C
b)2sinx+C
c)4(lnx)4+C
d)−9−x2+C; ∫9−x2dx=arcsin3x+C
e)2sin2x+C1 and −2cos2x+C2: both right, they differ by the constant 21.
a) The inside function is x3+1, raised to the fifth power, and its derivative 3x2 is present up to the constant 3. Let u=x3+1, so du=3x2dx and x2dx=31du. Then ∫x2(x3+1)5dx=31∫u5du=31⋅6u6+C=18(x3+1)6+C. Check: dxd18(x3+1)6=186(x3+1)5⋅3x2=x2(x3+1)5. The constant 31 is legitimate because it is a NUMBER: it can be moved through the integral sign. What cannot be moved is a factor containing x, and this is exactly why the x2 outside the bracket is the whole reason the substitution works. Expanding (x3+1)5 would also succeed, but costs six terms and ten minutes.
b) The inside function is x, and its derivative 2x1 sits in the integrand up to the constant 21. Let u=x, so du=2xdx and xdx=2du. Then ∫xcosxdx=2∫cosudu=2sinu+C=2sinx+C. Check: dxd2sinx=2cosx⋅2x1, the integrand. The reflex u=cosx goes nowhere: its derivative −sinx⋅2x1 is not in the integrand. The rule is to choose the INSIDE of the composition, not the outside function.
c) The inside function is lnx, and its derivative x1 is the other factor. Let u=lnx, du=xdx: ∫x(lnx)3dx=∫u3du=4u4+C=4(lnx)4+C. Check: dxd4(lnx)4=(lnx)3⋅x1. A x1 standing next to a function of lnx is the most reliable signal of the chapter: it is du already written for you.
d) The inside function is 9−x2, whose derivative −2x is present up to the constant −2. Let u=9−x2, du=−2xdx, so xdx=−21du and ∫9−x2xdx=−21∫u−1/2du=−21⋅2u1/2+C=−9−x2+C. Check: dxd(−9−x2)=−29−x2−2x=9−x2x. Without the factor x on top, u=9−x2 fails: du would need an x that is not there, and dividing by it is forbidden. The integral ∫9−x2dx is instead a table integral in disguise. Factor 9: 9−x2=31−(x/3)2, then w=3x, dx=3dw, gives ∫31−w23dw=arcsinw+C=arcsin3x+C. One factor x changes the answer from a square root to an inverse sine.
e) With u=sinx, du=cosxdx: ∫sinxcosxdx=∫udu=2sin2x+C1. With u=cosx, du=−sinxdx: ∫sinxcosxdx=−∫udu=−2cos2x+C2. Both are right, since both differentiate back to sinxcosx. They are not contradictory, because 2sin2x−(−2cos2x)=2sin2x+cos2x=21, a constant. Two antiderivatives of the same function on an interval differ by a constant, and the +C absorbs it. A third route, sinxcosx=21sin2x with w=2x, gives −4cos2x+C3, again off by a constant. On an exam, any of the three earns full marks; what loses the mark is dropping the +C, which is the only thing that makes the three answers the same answer.
Exercise 2: Definite integrals: change the bounds, or go back to x, never half of each
A definite integral can be computed by substitution along two correct routes. Route 1: find an antiderivative in x (substitute, integrate, substitute back), then evaluate it at the ORIGINAL bounds. Route 2: carry the bounds through u=g(x), so that x=a becomes u=g(a) and x=b becomes u=g(b), and never come back to x.
Both routes give the same number. What gives a wrong number is the mixture: an integrand in u with bounds in x. The figure shows the integrand of part c).
a) Compute ∫03x1+x2dx by route 1, then by route 2.
b) Compute ∫1e2xlnxdx.
c) Compute ∫0π/2sinxecosxdx. The new bounds come out in decreasing order: explain what happens to the minus sign of du, and use the figure to check the sign of your answer.
d) Compute ∫14x(1+x)2dx.
e) A student writes ∫03x1+x2dx=21∫03udu. Find the number she obtains, say exactly what is wrong, and name the other mixture that is just as wrong.
Show the solution
Answers
a)37 by both routes
b)342
c)e−1 (positive, as the shaded area requires)
d)31
e)She gets 333/4, not 37: u runs from 1 to 4, not from 0 to 3.
a) Let u=1+x2, du=2xdx, so xdx=21du. Route 1: ∫x1+x2dx=21∫u1/2du=31u3/2+C=31(1+x2)3/2+C, then [31(1+x2)3/2]03=31(43/2−13/2)=31(8−1)=37. Route 2: when x=0, u=1; when x=3, u=4. So ∫03x1+x2dx=21∫14u1/2du=31[u3/2]14=37. The same number, as it must be. Route 2 is shorter and is the one to use on an exam, provided the bound line (when x=…, u=…) is WRITTEN: it is the line the marker looks for.
b) Let u=lnx, du=xdx. When x=1, u=ln1=0; when x=e2, u=2. Then ∫1e2xlnxdx=∫02u1/2du=32[u3/2]02=32⋅22=342. The bounds 1 and e2 were chosen for the logarithm: without a calculator, ln(e2)=2 is the kind of bound an exam uses, and it only becomes simple AFTER the substitution. Order of magnitude: 2≈1.41, so the answer is about 1.89.
c) Let u=cosx, du=−sinxdx, so sinxdx=−du. When x=0, u=cos0=1; when x=2π, u=0. Then ∫0π/2sinxecosxdx=∫10eu(−du)=−∫10eudu=∫01eudu=e−1. The new lower bound 1 is LARGER than the new upper bound 0, and that is fine: keep the bounds in the order the substitution gives them. The minus sign of du and the reversed order cancel, since −∫10=∫01. The classic slip is to use one of them and not the other: reorder the bounds to ∫01 AND keep the minus, and the answer becomes 1−e. The figure refutes it at once: the integrand is positive on (0,2π), the shaded area is positive, so the integral is positive. With e≈2.72, e−1≈1.72, plausible for a region about 1.57 wide and at most about 1.5 high.
d) The inside function is 1+x, and its derivative 2x1 is present up to the constant 21. Let u=1+x, du=2xdx, so xdx=2du. When x=1, u=2; when x=4, u=3. Then ∫14x(1+x)2dx=2∫23u−2du=2[−u1]23=2(−31+21)=31. The new bounds are 1+1 and 1+4, not 1 and 4: a bound is always the value of the WHOLE u, constant included.
e) She has substituted in the integrand and the differential but not in the bounds: 21∫03u1/2du=31[u3/2]03=31(3)3/2=333/4, about 0.76, instead of 37≈2.33. The numbers 0 and 3 are values of x; the variable of her new integral is u, which runs from 1 to 4. Her integral still contains an x, hidden in the bounds, and that is exactly the unfinished substitution. The symmetric mixture is just as wrong: change the bounds to 1 and 4, then substitute back to x and evaluate 31[(1+x2)3/2]14=31(173/2−8), which is about 20.7. Choose a route and follow it to the end: bounds in x with an antiderivative in x, or bounds in u with an antiderivative in u.
Exercise 3: When an x survives: rewrite it in terms of u
Sometimes, once u=g(x) is chosen and du has absorbed what it can, an x is left in the integrand. It is NOT a constant and it cannot be pulled out of the integral. It must be rewritten: solve u=g(x) for x (or for the power of x that appears) and replace it.
Only then is the integral written in u alone, and only then can it be integrated.
a) ∫xx−1dx
b) ∫03x+1xdx
c) ∫x3x2+1dx. Explain how the factor x3 is split.
d) A student writes ∫xx−1dx=x∫udu=32x(x−1)3/2+C. Show by differentiation that this is wrong, and identify the term that betrays the error.
e) ∫041+xdx, with u=1+x. Here it is dx itself that must be rewritten.
Show the solution
Answers
a)52(x−1)5/2+32(x−1)3/2+C
b)38
c)51(x2+1)5/2−31(x2+1)3/2+C
d)Its derivative is xx−1+32(x−1)3/2: one term too many.
e)4−2ln3
a) Let u=x−1, du=dx. The factor x−1 becomes u, but a factor x remains. From u=x−1: x=u+1. So ∫xx−1dx=∫(u+1)u1/2du=∫(u3/2+u1/2)du=52u5/2+32u3/2+C=52(x−1)5/2+32(x−1)3/2+C. Check: the derivative is (x−1)3/2+(x−1)1/2=(x−1)1/2[(x−1)+1]=xx−1. The substitution does not simplify the root, it moves the difficulty to a place where the power rule applies: (u+1)u expands, xx−1 does not.
b) Let u=x+1, du=dx, x=u−1. When x=0, u=1; when x=3, u=4. Then ∫03x+1xdx=∫14uu−1du=∫14(u1/2−u−1/2)du=[32u3/2−2u1/2]14=(316−4)−(32−2)=34+34=38. The division uu−1 is term by term, uu=u1/2 and u1=u−1/2. Sanity check: the integrand grows from 0 to 23 on an interval of length 3, so the integral lies between 0 and 4.5; 38≈2.67 fits.
c) The inside function is x2+1, with derivative 2x. Write x3dx=x2⋅xdx: ONE factor x goes with dx to form du, the other two form x2, which is rewritten. Let u=x2+1, du=2xdx, xdx=21du, x2=u−1. Then ∫x3x2+1dx=21∫(u−1)u1/2du=21(52u5/2−32u3/2)+C=51(x2+1)5/2−31(x2+1)3/2+C. Check: the derivative is x(x2+1)3/2−x(x2+1)1/2=x(x2+1)1/2[(x2+1)−1]=x3x2+1. The split works because the leftover power is EVEN: x2 is a function of x2+1. With x2x2+1 the leftover would be a single x=u−1, and that integral belongs to the trigonometric substitution chapter.
d) Differentiate her answer with the product rule: dxd[32x(x−1)3/2]=32(x−1)3/2+32x⋅23(x−1)1/2=32(x−1)3/2+xx−1. The second term is the integrand; the first, 32(x−1)3/2, is extra, and it is precisely the derivative of the x she froze. At x=2 her derivative is 32+2=38 where the integrand is 2. Only a CONSTANT passes through the integral sign; x varies, so ∫xh(x)dx=x∫h(x)dx. The correct answer is that of part a).
e) Let u=1+x. Here du=2xdx does not appear in the integrand, so solve for x instead: x=u−1, x=(u−1)2 and dx=2(u−1)du. When x=0, u=1; when x=4, u=3. Then ∫041+xdx=∫13u2(u−1)du=∫13(2−u2)du=[2u−2lnu]13=(6−2ln3)−(2−0)=4−2ln3. With ln3≈1.10 that is about 1.80, and the integrand falls from 1 to 31 over a length 4, so the value must lie between 34 and 4: it does. Rewriting dx through u is the same gesture as rewriting a stray x: whatever is left in x is expressed in u.
Exercise 4: Numerator equals derivative of the denominator: logarithms and their absolute value
When the numerator of a fraction is the derivative of its denominator, up to a constant, the substitution u= denominator turns the integral into ∫udu=ln∣u∣+C. So ∫g(x)g′(x)dx=ln∣g(x)∣+C.
The absolute value is not decoration: ln∣u∣ is an antiderivative of u1 on each interval where u=0, including where u<0. The figure shows y=tanx on [0,3π].
a) Compute ∫tanxdx and ∫cotxdx. Give both usual forms of the first answer.
b) Compute ∫0π/3tanxdx and ∫π/6π/2cotxdx, and check the sign of the first against the figure.
c) Compute ∫x2+3x+52x+3dx and ∫ex+3exdx. Is an absolute value needed in either answer?
d) Compute ∫−10x2−42xdx. A student stops, saying that ln(−4) does not exist. Answer him.
e) Compute ∫ee3xlnxdx. Then explain why ∫1+x2dx is NOT ln(1+x2)+C, and give its value.
c)ln(x2+3x+5)+C and ln(ex+3)+C: no absolute value needed, both denominators are positive.
d)ln34: with ln∣x2−4∣ nothing is undefined.
e)ln3; ∫1+x2dx=arctanx+C
a) Write tanx=cosxsinx. The derivative of the denominator is −sinx, the numerator up to the constant −1. Let u=cosx, du=−sinxdx: ∫tanxdx=−∫udu=−ln∣cosx∣+C. Since −ln∣cosx∣=ln∣cosx∣1, the same answer reads ln∣secx∣+C, the form of Stewart's table. For the cotangent, cotx=sinxcosx and u=sinx, du=cosxdx give ∫cotxdx=ln∣sinx∣+C, with no minus sign. The two signs are opposite because the derivatives of cos and sin have opposite signs; forgetting the minus of d(cosx) is the most common error on this integral, and part b) shows how to catch it.
b) ∫0π/3tanxdx=[−ln∣cosx∣]0π/3=−ln21+ln1=ln2. On (0,3π) the tangent is positive, the shaded region of the figure lies above the axis, so the integral must be positive: ln2≈0.69 is, and the region, about 1.05 wide and at most 3≈1.73 high but mostly much lower, is consistent with it. The wrong sign convention, ln∣cosx∣, gives −ln2, negative for a positive integrand: reject it on sight. Then ∫π/6π/2cotxdx=[ln∣sinx∣]π/6π/2=ln1−ln21=ln2. The two answers agree, and they should: the substitution x↦2π−x exchanges tan and cot and maps [0,3π] onto [6π,2π].
c) The derivative of x2+3x+5 is exactly 2x+3. With u=x2+3x+5, du=(2x+3)dx and ∫x2+3x+52x+3dx=ln∣x2+3x+5∣+C. The discriminant of the denominator is 9−20=−11<0 and its leading coefficient is positive, so x2+3x+5>0 for every x (it equals (x+23)2+411), and the answer may be written ln(x2+3x+5)+C. Likewise u=ex+3, du=exdx give ∫ex+3exdx=ln(ex+3)+C, and ex+3>3 needs no absolute value. Dropping the bars is allowed only after a sentence that proves the argument is positive; without that sentence, keep them.
d) Let u=x2−4, du=2xdx. When x=−1, u=−3; when x=0, u=−4. Then ∫−10x2−42xdx=∫−3−4udu=[ln∣u∣]−3−4=ln4−ln3=ln34. The student used lnu where the antiderivative of u1 is ln∣u∣: on the interval [−4,−3], where u is negative, ln∣u∣=ln(−u) is perfectly defined and its derivative is −u−1=u1. The integrand is continuous on [−1,0] (the denominator vanishes only at ±2), so the integral certainly exists. Sign check: on (−1,0) the numerator 2x and the denominator x2−4 are both negative, the integrand is positive, and ln34>0.
e) The numerator is not visibly the derivative of the denominator, until the fraction is read as lnx1/x: the derivative of lnx is x1. Let u=lnx, du=xdx; when x=e, u=1; when x=e3, u=3. Then ∫ee3xlnxdx=∫13udu=ln3−ln1=ln3. As for ∫1+x2dx: the derivative of ln(1+x2) is 1+x22x, not 1+x21. The rule ln∣g∣ needs g′ ON TOP, and here the numerator 1 is not a constant multiple of 2x. No substitution is needed at all: ∫1+x2dx=arctanx+C is in the table. The same denominator gives a logarithm with x on top and an inverse tangent with 1 on top, and exercise 6 turns that difference into a method.
Exercise 5: Symmetric intervals: odd integrands vanish, even ones double
Let f be continuous on [−a,a]. If f is ODD, that is f(−x)=−f(x), then ∫−aaf(x)dx=0. If f is EVEN, f(−x)=f(x), then ∫−aaf(x)dx=2∫0af(x)dx. Both facts are proved by one substitution, u=−x, and they settle integrals that no technique of the course could compute directly.
The figure shows the odd function y=x4−x2 on [−2,2].
a) Prove the odd case: split ∫−aa at 0 and use u=−x on ∫−a0f(x)dx. Point to where continuity on [−a,a] is used.
b) Give ∫−22x4−x2dx without computing. Then compute ∫02x4−x2dx and deduce the total area of the shaded region.
c) Compute ∫−11(1+x2sinx+x2cos(x3))dx.
d) Two claims: ∫−11x3dx=0 because x31 is odd, and ∫−12x3dx=0 because x3 is odd. Say what is wrong with each, and compute the second integral correctly.
e) Compute ∫02[(x−1)3e(x−1)2+3(x−1)2]dx with the substitution u=x−1 first.
Show the solution
Answers
a)∫−a0f(x)dx=∫0af(−u)du=−∫0af(u)du, which cancels ∫0af.
b)0; ∫02x4−x2dx=38; shaded area 316.
c)0+32sin1=32sin1
d)First: not continuous at 0, the theorem does not apply. Second: [−1,2] is not symmetric; the value is 415.
e)0+2=2
a) Split: ∫−aaf(x)dx=∫−a0f(x)dx+∫0af(x)dx. In the first integral let u=−x, du=−dx; when x=−a, u=a; when x=0, u=0. So ∫−a0f(x)dx=∫a0f(−u)(−du)=∫0af(−u)du=−∫0af(u)du, using f(−u)=−f(u). A variable of integration is a dummy letter, so ∫0af(u)du=∫0af(x)dx, and the two pieces cancel: the total is 0. For an even f, the same lines with f(−u)=f(u) give ∫−a0f=∫0af, hence the factor 2. Continuity on the WHOLE of [−a,a] is what makes each of these integrals exist in the first place; the substitution itself only reflects the interval.
b) f(x)=x4−x2 satisfies f(−x)=−x4−x2=−f(x) and is continuous on [−2,2]: it is odd, so ∫−22x4−x2dx=0, with no computation. For the right half, let u=4−x2, du=−2xdx; when x=0, u=4; when x=2, u=0. Then ∫02x4−x2dx=−21∫40u1/2du=21∫04u1/2du=31[u3/2]04=38. The shaded region has two lobes, the right one above the axis and the left one below, each of area 38: the INTEGRAL is 38−38=0 but the AREA is 316. The symmetry theorem computes signed integrals; an area question asks for 2∫02.
c) Split the integral. The function 1+x2sinx is odd (odd numerator, even denominator) and continuous on [−1,1], so its integral is 0. This is the point of the theorem: no technique of the course finds an antiderivative of 1+x2sinx, and none is needed. The function x2cos(x3) is even, since (−x)2=x2 and cos((−x)3)=cos(−x3)=cos(x3), so its integral is 2∫01x2cos(x3)dx. With u=x3, du=3x2dx, bounds 0 and 1: 2⋅31∫01cosudu=32sin1. Total: 32sin1, an exact answer; sin1 is a number like π and is left as it is.
d) The first claim fails the hypothesis: x31 is not defined at 0, so it is not continuous on [−1,1], and the theorem says nothing. The two halves are not finite numbers that cancel; ∫−11x−3dx is not even a definite integral in the sense of this chapter, and such integrals are treated, and shown to diverge, in the chapter on improper integrals. The second claim fails the other hypothesis: [−1,2] is not symmetric about 0. Split it into the symmetric part and the rest: ∫−12x3dx=∫−11x3dx+∫12x3dx=0+[4x4]12=4−41=415. Check directly: [4x4]−12=4−41. Before quoting symmetry, write the two checks: the interval is [−a,a], and the integrand is continuous on it.
e) Let u=x−1, du=dx; when x=0, u=−1; when x=2, u=1. The integral becomes ∫−11(u3eu2+3u2)du on a symmetric interval. The function u3eu2 is odd and continuous, so its integral is 0; the function 3u2 is even and gives [u3]−11=2. Total: 2. The shift u=x−1 reveals that the original integrand is odd ABOUT x=1, the midpoint of [0,2]: symmetry is a property of the integrand relative to the centre of the interval, and a translation brings that centre to 0. The odd part could also be done by w=u2, which gives 21∫11wewdw=0 because the new bounds coincide.
Part B: problems and reasoning (/50)
Exercise 6: Integrals that look alike: logarithm, inverse tangent or inverse sine
The five integrals below differ by a power of x or by where the square sits. They do not have the same kind of answer: one is a logarithm, one an inverse tangent, one an inverse sine, one a plain power. What decides is not the look of the denominator but what du can absorb.
For each one, name u, write du, and say which formula of the table the integral in u is.
a) ∫1+x2xdx and ∫1+x2dx.
b) ∫1+x4xdx
c) ∫1+x4x3dx
d) ∫0ln31+e2xexdx
e) ∫012−x4xdx, and, for contrast, ∫2−x2xdx.
Show the solution
Answers
a)21ln(1+x2)+C and arctanx+C
b)21arctan(x2)+C
c)41ln(1+x4)+C
d)3π−4π=12π
e)8π; ∫2−x2xdx=−2−x2+C
a) In 1+x2x the numerator is half the derivative of the denominator: u=1+x2, du=2xdx, and ∫1+x2xdx=21∫udu=21ln(1+x2)+C, no absolute value since 1+x2>0. In 1+x21 there is no x on top to build du=2xdx, so u=1+x2 is impossible; and no substitution is needed, since ∫1+x2dx=arctanx+C is the table. The answer ln(1+x2) for the second integral is wrong, and its derivative, 1+x22x, shows it at once.
b) Here u=1+x4 would need du=4x3dx, and only x is on top. But x4=(x2)2, and xdx is half of d(x2). Let u=x2, du=2xdx: ∫1+x4xdx=21∫1+u2du=21arctanu+C=21arctan(x2)+C. Check: dxd21arctan(x2)=21⋅1+x42x. The substitution does not simplify the denominator, it turns 1+x4 into 1+u2, the denominator of the table's inverse tangent.
c) Now the numerator x3 is a quarter of the derivative of 1+x4. Let u=1+x4, du=4x3dx: ∫1+x4x3dx=41∫udu=41ln(1+x4)+C. Compare with b): the same denominator, a numerator x3 instead of x, and the answer switches from an inverse tangent to a logarithm. The test is always the same: which function of x, differentiated, gives the numerator up to a constant? 1+x4 gives 4x3; x2 gives 2x.
d) Since e2x=(ex)2 and exdx=d(ex), let u=ex, du=exdx. When x=0, u=1; when x=ln3, u=3. Then ∫0ln31+e2xexdx=∫131+u2du=arctan3−arctan1=3π−4π=12π. The bound ln3=21ln3 looks awkward in x and becomes the tidy 3 in u: on a no-calculator exam, bounds like this are a strong hint that the substitution u=ex is the intended route. Choosing u=1+e2x fails: du=2e2xdx, and the numerator is only ex.
e) As in b), x4=(x2)2 and xdx=21d(x2). Let u=x2, du=2xdx, bounds 0 and 1: ∫012−x4xdx=21∫012−u2du. Factor 2 under the root, 2−u2=21−(u/2)2, and let w=2u, du=2dw, bounds 0 and 21: 21∫01/21−w2dw=21arcsin21=21⋅4π=8π. For contrast, in 2−x2x the root contains x2, whose derivative −2x IS the numerator up to a constant: u=2−x2 gives −21∫u−1/2du=−2−x2+C, a plain power with no inverse sine anywhere. The same x on top plays two different roles: with x2 under the root it completes the du of the whole root, with x4 it completes the du of the square x2.
Exercise 7: An unknown function: the substitution must carry the bounds
A function f is continuous on [0,4], and all that is known about it is ∫04f(x)dx=10 and ∫02f(x)dx=3. No formula is given, so no antiderivative can be written: the only way to use the substitution rule is route 2, carrying the bounds, until the integral in u is one of the two given ones.
The figure shows one possible f (solid) and the graph of y=f(2x) on [0,2] (dashed), with the region under f(2x) shaded.
a) Compute ∫02f(2x)dx.
b) Compute ∫02xf(x2)dx.
c) Compute ∫0π/2cosxf(4sinx)dx.
d) Compute ∫12f(2x)dx, using BOTH pieces of data.
e) A student finds ∫02f(2x)dx=23. Find her error, and explain on the figure why the answer to a) must be half of 10.
Show the solution
Answers
a)5
b)23
c)25
d)21(10−3)=27
e)She kept the bounds 0 and 2; u=2x runs over [0,4], and squeezing by 2 halves the area of 10.
a) Let u=2x, du=2dx, so dx=21du. When x=0, u=0; when x=2, u=4. Then ∫02f(2x)dx=21∫04f(u)du=21⋅10=5. The last step uses that the letter of integration is a dummy: ∫04f(u)du and ∫04f(x)dx are the same number. Without a formula for f there was no other route: an antiderivative cannot be written, but a change of bounds can always be made.
b) The inside function is x2 and the factor x is half its derivative. Let u=x2, du=2xdx; when x=0, u=0; when x=2, u=2. Then ∫02xf(x2)dx=21∫02f(u)du=21⋅3=23. The bound 2 was chosen so that u lands exactly on the given interval [0,2]: with an unknown f, the new bounds must be those of a known integral, which is a useful check that the substitution is the intended one.
c) The inside function is 4sinx, with derivative 4cosx, present up to the constant 4. Let u=4sinx, du=4cosxdx; when x=0, u=0; when x=2π, u=4. Then ∫0π/2cosxf(4sinx)dx=41∫04f(u)du=410=25. The trigonometric look is irrelevant: the gesture is the one of b), inside function, its derivative outside, bounds carried.
d) With u=2x as in a), x=1 gives u=2 and x=2 gives u=4: ∫12f(2x)dx=21∫24f(u)du. The integral over [2,4] is not given, but additivity gives it: ∫24f=∫04f−∫02f=10−3=7. So ∫12f(2x)dx=27. As a consistency check, ∫01f(2x)dx=21∫02f=23, and 23+27=5, the answer of a).
e) She substituted u=2x and dx=21du correctly, but kept the bounds of x: 21∫02f(u)du=23. That is the value of ∫01f(2x)dx, as d) showed: her integral covers only half the interval. On the figure, the dashed graph of f(2x) on [0,2] is the solid graph of f on [0,4] squeezed horizontally by a factor 2: every height of f appears once, and every width is halved. Areas are multiplied by 21, so the shaded region has half the area under f on [0,4], that is 5. This is the geometric meaning of the factor 21 in dx=21du, and it only works if the whole graph of f, from u=0 to u=4, is included.
Exercise 8: Five statements to correct
Each statement below was written in a MATH 141 tutorial, and each is false. Say what is wrong, give the correct statement or the correct value, and settle it with a differentiation or a computation.
a) With u=x3+1 and du=3x2dx, ∫(x3+1)5dx=∫u53x2du=18x2(x3+1)6+C.
b) With u=x2, ∫02xex2dx=21∫02eudu=2e2−1.
c) ∫x2+9dx=91arctanx+C, since x2+91=91⋅1+x21.
d) ∫−ππx2sinxdx=2∫0πx2sinxdx, because x2 is even.
e) On [0,π] the substitution u=sinx is illegal, since sin is not one-to-one there, so ∫0πcosxesinxdx cannot be computed by substitution.
Show the solution
Answers
a)False: x2 is not a constant. Expand: 16x16+135x13+x10+710x7+45x4+x+C.
b)False: u runs from 0 to 4; the value is 2e4−1.
c)False: x2+91=91⋅1+(x/3)21; the answer is 31arctan3x+C.
d)False: x2sinx is odd; the integral is 0.
e)False: the rule only needs g′ continuous; u=sinx gives ∫00eudu=0, which is correct.
a) FALSE. du=3x2dx does not appear in the integrand, and the student forced it by dividing by 3x2, then treated x2 as a constant. Differentiate the result to see it: with F(x)=181(x3+1)6x−2, F′(x)=(x3+1)5−9x3(x3+1)6. The first term is the integrand; the second is the price of the frozen x2, and it is not zero (at x=1, F′(1)=32−964=32). No substitution applies here, because no inside function has its derivative present. Expand instead with the binomial coefficients 1,5,10,10,5,1: (x3+1)5=x15+5x12+10x9+10x6+5x3+1, so ∫(x3+1)5dx=16x16+135x13+x10+710x7+45x4+x+C. Correct statement: a missing CONSTANT factor of du can be supplied; a missing VARIABLE factor cannot, and then this u is the wrong tool.
b) FALSE. The integrand and the differential were rewritten in u, the bounds were not. When x=0, u=0; when x=2, u=4. So ∫02xex2dx=21∫04eudu=2e4−1. Check by route 1: [21ex2]02=2e4−1. The student's value, 2e2−1≈3.2, is also refuted by size: on [1,2] alone the integrand is at least 1⋅e1≈2.7 and reaches 2e4≈109, so the integral is far larger than 3.2. Correct statement: a definite integral in u takes bounds in u.
c) FALSE. Factor the 9 correctly: x2+9=9(1+9x2)=9(1+(3x)2), so x2+91=91⋅1+(x/3)21, not 91⋅1+x21. Let u=3x, dx=3du: ∫x2+9dx=91∫1+u23du=31arctan3x+C. Check: dxd31arctan3x=31⋅1+x2/91/3=9+x21. The derivative of the student's answer is 9+9x21, a different function. Correct statement: ∫x2+a2dx=a1arctanax+C, and the factor a1 comes from the linear substitution u=ax.
d) FALSE. The parity that matters is that of the WHOLE integrand. x2 is even but sinx is odd, and (−x)2sin(−x)=−x2sinx: the product is odd. It is continuous on [−π,π], so ∫−ππx2sinxdx=0, with no technique needed. The doubling rule would require the product to be even. Correct statement: even times even and odd times odd are even; even times odd is odd.
e) FALSE. The substitution rule for definite integrals, ∫abf(g(x))g′(x)dx=∫g(a)g(b)f(u)du, requires g′ continuous on [a,b] and f continuous on the values of g; nothing requires g to be one-to-one. It is the chain rule, dxdF(g(x))=f(g(x))g′(x), followed by the fundamental theorem. Here u=sinx, du=cosxdx, and both bounds give u=0: ∫0πcosxesinxdx=∫00eudu=0. Route 1 confirms it: [esinx]0π=e0−e0=0. The figure of the solution shows why: on [0,2π] the integrand is positive with integral e−1, on [2π,π] it is negative with integral 1−e, and the two cancel. Correct statement: the substitution is legal; injectivity is a condition of the inverse substitution x=h(t), met in the trigonometric substitution chapter, not of u=g(x).
Exercise 9: Charge through three circuits: one substitution per current
The charge that passes through a wire between the times t=a and t=b is Q=∫abI(t)dt, where I is the current. With I in milliamperes and t in seconds, Q is in millicoulombs (mC). Three circuits are switched on at t=0, with currents IA(t)=t2+1610t, IB(t)=8te−t2 and IC(t)=t2+12t.
The figure shows the three currents; the charge of circuit A during the first 3 seconds is shaded. No calculator: give exact times.
a) Compute the charge QA that passes in circuit A during the first 3 seconds.
b) Find the time T at which a total of 30 mC has passed in circuit A.
c) Find the charge QB(T) passed in circuit B by time T. Show that it stays below 4 mC however long the circuit runs, and find when it reaches 3 mC.
d) Find QC(T). When has 4 mC passed in circuit C? Can any amount of charge be reached?
e) Currents B and C both fall back towards 0. Explain, from your formulas, why one circuit passes a bounded charge and the other does not.
Show the solution
Answers
a)QA=10(5−4)=10 mC
b)T=33 s
c)QB(T)=4(1−e−T2)<4; 3 mC at T=ln4 s
d)QC(T)=ln(T2+1); 4 mC at T=e4−1 s; any charge Q at T=eQ−1.
e)e−T2→0 keeps QB below 4; ln(T2+1) grows without bound.
a) Let u=t2+16, du=2tdt, so tdt=21du. When t=0, u=16; when t=3, u=25. Then QA=∫03t2+1610tdt=5∫1625u−1/2du=10[u1/2]1625=10(5−4)=10 mC. The numbers were built on the right triangle 3,4,5: 9+16=5. Sanity check against the figure: the current climbs from 0 to IA(3)=530=6 mA, so over 3 seconds the charge lies between 0 and 18 mC, and the shaded region, roughly a triangle of base 3 and height 6, is about 9 mC.
b) The same substitution with a variable upper bound: QA(T)=10[t2+16]0T=10(T2+16−4). Setting QA(T)=30: T2+16−4=3, T2+16=7, T2=49−16=33, T=33 s, a little under 6 s since 62=36. The first 10 mC took 3 seconds, the next 20 took about 2.7 more: the current is larger later, as the figure shows it approaching 10 mA.
c) Let u=t2, du=2tdt; the bounds 0 and T become 0 and T2. Then QB(T)=∫0T8te−t2dt=4∫0T2e−udu=4[−e−u]0T2=4(1−e−T2). Since e−T2>0 for every T, QB(T)<4: however long the circuit runs, less than 4 mC passes. Setting QB(T)=3: 1−e−T2=43, e−T2=41, T2=ln4, T=ln4=2ln2 s, about 1.18 s with ln2≈0.69. Three quarters of everything circuit B will ever deliver has passed in just over a second.
d) The numerator 2t is the derivative of t2+1: u=t2+1, du=2tdt, bounds 1 and T2+1. So QC(T)=∫1T2+1udu=ln(T2+1), no absolute value since u≥1. Setting ln(T2+1)=4: T2+1=e4, T=e4−1 s, about 7.3 s since e4≈54.6. For any target Q>0, the same computation gives T=eQ−1: every amount of charge is eventually reached, 10 mC after about 148 s, 20 mC after about 22,000 s. The logarithm grows without bound, just very slowly.
e) Both currents tend to 0, but at very different speeds. For large t, IB(t)=8te−t2 collapses, and its accumulated charge 4(1−e−T2) approaches 4 from below: after a few seconds, almost nothing more is added. IC(t)=t2+12t behaves like t2, and its accumulated charge is a logarithm, which never stops growing. So a current that dies out does not guarantee a bounded total charge: what decides is how fast it dies out, and the substitution made that visible by producing e−T2 in one case and ln(T2+1) in the other. Circuit A, whose current tends to 10 mA, passes an unbounded charge for a simpler reason: QA(T)=10(T2+16−4) grows roughly like 10T.
Exercise 10: A final exam question: the reflection x to a minus x
This is the shape of a long final exam question. One substitution, u=a−x, reflects the interval [0,a] onto itself, and combined with the original integral it computes values that no direct antiderivative gives. The figure shows y=sinx+cosxsinx (solid, region below it shaded) and y=sinx+cosxcosx (dashed) on [0,2π].
a) Let f be continuous on [0,a]. Prove that ∫0af(a−x)dx=∫0af(x)dx.
b) Let I=∫0π/2sinx+cosxsinxdx and J=∫0π/2sinx+cosxcosxdx. Show that I=J, compute I+J, and deduce I. Read the result on the figure.
c) Let K=∫0π1+cos2xxsinxdx. Using a) with a=π, show that K=2π∫0π1+cos2xsinxdx.
d) Compute ∫0π1+cos2xsinxdx by a second substitution, and deduce K.
e) Apply the method of c) to ∫0πxsinxdx and give its value.
Show the solution
Answers
a)u=a−x maps [0,a] onto [a,0] and dx=−du, so ∫0af(a−x)dx=∫0af(u)du.
b)I=J, I+J=2π, so I=4π.
c)K=π∫0π1+cos2xsinxdx−K, so K=2π∫0π1+cos2xsinxdx.
d)∫0π1+cos2xsinxdx=2π, so K=4π2.
e)∫0πxsinxdx=2π∫0πsinxdx=π
a) Let u=a−x, du=−dx. When x=0, u=a; when x=a, u=0. Then ∫0af(a−x)dx=∫a0f(u)(−du)=∫0af(u)du=∫0af(x)dx, the last equality because the letter of integration is a dummy. As in exercise 2 c), the reversed bounds and the minus sign of du cancel. Geometrically, the graph of f(a−x) is the graph of f flipped about the vertical line x=2a, and a flip does not change an area.
b) Apply a) with a=2π: since sin(2π−x)=cosx and cos(2π−x)=sinx, the integrand of I becomes cosx+sinxcosx, that of J. So I=J. Adding the two integrals, I+J=∫0π/2sinx+cosxsinx+cosxdx=∫0π/21dx=2π. The denominator never vanishes on [0,2π], since sinx+cosx≥1 there, so both integrals exist. Hence I=J=4π. On the figure, the two curves are mirror images about x=4π, where both equal 21, and at every x they add up to 1: the shaded region is exactly half of the rectangle [0,2π]×[0,1], whose area is 2π.
c) Apply a) with a=π: sin(π−x)=sinx and cos(π−x)=−cosx, so cos2(π−x)=cos2x. Then K=∫0π1+cos2x(π−x)sinxdx=π∫0π1+cos2xsinxdx−∫0π1+cos2xxsinxdx=π∫0π1+cos2xsinxdx−K. The unknown K appears on both sides: 2K=π∫0π1+cos2xsinxdx, which is the claim. The factor x, which blocked every substitution, has been traded for the constant π: that is the whole purpose of the reflection.
d) The inside function is cosx, whose derivative −sinx is on top up to a sign. Let w=cosx, dw=−sinxdx; when x=0, w=1; when x=π, w=−1. Then ∫0π1+cos2xsinxdx=∫1−11+w2−dw=∫−111+w2dw=arctan1−arctan(−1)=4π+4π=2π. So K=2π⋅2π=4π2. Sanity check: on [0,π] the integrand of K is non-negative and at most xsinx≤π, so 0≤K≤π2; with π2≈9.87, K≈2.47 is plausible. A marker expects the two substitutions to be named separately, with their bound lines.
e) The same steps with the integrand xsinx: ∫0πxsinxdx=∫0π(π−x)sinxdx, hence 2∫0πxsinxdx=π∫0πsinxdx=π[−cosx]0π=2π, and ∫0πxsinxdx=π. The method works for every integral of the form ∫0πxg(sinx)dx, since sin is unchanged by the reflection: the result is 2π∫0πg(sinx)dx. This particular integral will be met again in the next chapter, where integration by parts gives the same value π directly; here it came from substitution alone.