This is a long set of 20 corrected exercises devoted to a single topic: integrating the trigonometric functions and their inverses, in Integral Calculus, 201-SN3-RE (formerly 201-NYB). It is the chapter of integral calculus where students lose the most marks, not because the computations are hard, but because about ten families look alike and no question ever says which one it belongs to. The set covers them ALL, including the ones textbooks dispatch in a single line: the cosecant, the cotangent, the inverse cotangent and the inverse secant with its absolute value.
It is meant to be worked in four sessions of about 75 minutes, one per part, not in one sitting. Part A, powers and products: the three cases of sine and cosine, the antiderivative of the secant and that of the cosecant, the tangent-secant family with its reduction formula, its mirror image the cotangent-cosecant family, and the identities that replace an entire technique. Part B, the three trigonometric substitutions, with completing the square and, above all, the cases where a trigonometric substitution is a bad idea. Part C, the six inverse functions: the three basic forms to recognize, integration by parts with the invisible factor 1, products with x, the inverse secant and its absolute value, then compositions and identities. Part D, definite and improper integrals, antiderivatives that look different, geometric applications and a final decision tree.
Nearly every exercise ends with a question of method rather than of computation, because that is where the mark is decided. The thread of the whole set fits in one sentence: before choosing a technique, look at the NUMERATOR, then at the parity of the exponents, then at the sign under the root. Trigonometric substitution is the last resort of the chapter, never the first reflex.
Self-checking setType your answers and check them question by question: the page tells you right or wrong before you open the solution. With an account, every answer right on the first try earns points, and your right answers stay filled in when you come back.
•Powers of sine and cosine: if one exponent is ODD, split off one factor from it and substitute the other function; if both are EVEN, reduce the powers with sin2x=21−cos2x and cos2x=21+cos2x. Shortcut: sinxcosx=2sin2x.
•Tangent-secant family: if the power of sec is EVEN, split off sec2x and set u=tanx; if the power of tan is ODD, split off secxtanx and set u=secx. Otherwise, convert with tan2x=sec2x−1. The cotangent-cosecant family is identical up to sign, every name that starts with “co” carrying a minus sign.
•The four antiderivatives to know by heart: ∫tanxdx=−ln∣cosx∣, ∫cotxdx=ln∣sinx∣, ∫secxdx=ln∣secx+tanx∣, ∫cscxdx=−ln∣cscx+cotx∣. And the cyclic case ∫sec3xdx=21(secxtanx+ln∣secx+tanx∣).
•Substitutions: a2−x2 calls for x=asinθ; a2+x2 calls for x=atanθ; x2−a2 calls for x=asecθ, with two branches and x2−a2=a∣tanθ∣. A quadratic always comes down to one of the three by completing the square.
•The three inverse forms: ∫a2−u2du=arcsinau; ∫a2+u2du=a1arctanau; ∫uu2−a2du=a1arcsecau. The absolute value in the third is not decoration: without it, the antiderivative is wrong on the negative branch.
•Integrating an inverse function: parts with dv=dx, the invisible factor 1. ∫arcsinxdx=xarcsinx+1−x2; ∫arctanxdx=xarctanx−21ln(1+x2); the arccos and arccot versions change sign, and the sum of the two antiderivatives of a pair is always 2πx.
Part A: Powers and products of trigonometric functions (/50)
Exercise 1: Powers of sine and cosine: the three cases
Every product sinmxcosnx is handled according to the PARITY of the two exponents, and there are only three situations. You split off one factor from the ODD power, because that factor will supply the du; if both exponents are even, no factor can be split off and the powers must be reduced.
a) Compute ∫sin5xcos2xdx, stating which power you attack and why.
b) Compute ∫cos5xdx.
c) Compute ∫sin4xdx by power reduction. How many times must the power be lowered?
d) Compute ∫sin2xcos2xdx in the shortest possible way.
e) Compute ∫0π/2sin6xdx and state the decision rule of the whole exercise in two sentences.
Show the solution
Answers
a)−3cos3x+52cos5x−7cos7x+C
b)sinx−32sin3x+5sin5x+C
c)83x−4sin2x+32sin4x+C, two reductions
d)8x−32sin4x+C
e)325π≈0.4909; odd: split off, even: reduce
a)
The sine carries the odd power: split off one factor from it. ∫sin4xcos2x⋅sinxdx=∫(1−cos2x)2cos2xsinxdx. With u=cosx and du=−sinxdx: −∫(1−u2)2u2du=−∫(u2−2u4+u6)du, that is −3cos3x+52cos5x−7cos7x+C. You attack the odd power because, once one factor has been taken from it, what remains is an EVEN power, the only kind the identity sin2=1−cos2 can convert.
b)
∫cos4x⋅cosxdx=∫(1−sin2x)2cosxdx. With u=sinx: ∫(1−u2)2du=u−32u3+5u5, that is sinx−32sin3x+5sin5x+C. A cosine alone, to an odd power, is handled exactly like the mixed case: it is the case n=0.
c)
No odd power: the power must be reduced with sin2x=21−cos2x. ∫sin4xdx=41∫(1−2cos2x+cos22x)dx. The term cos22x forces a SECOND reduction: cos22x=21+cos4x. This gives 41[x−sin2x+2x+8sin4x]=83x−4sin2x+32sin4x+C. Rule: a power 2k calls for k successive reductions, and the term in x that comes out of them explains why the integral of an even power is never periodic.
d)
The short route goes through the double angle BEFORE reducing: sin2xcos2x=(sinxcosx)2=4sin22x=81−cos4x. Hence ∫sin2xcos2xdx=8x−32sin4x+C, in one line instead of three. Recognizing sinxcosx=2sin2x always saves one reduction.
e)
An even power on a standard interval: Wallis's formula gives ∫0π/2sinnxdx=nn−1⋅n−2n−3⋯21⋅2π for even n, here 65⋅43⋅21⋅2π=325π≈0.4909. Decision rule: if one of the two exponents is ODD, split off one factor from it and substitute the other function; if BOTH are even, reduce the powers with the double-angle formulas, as many times as needed.
Tick the exercises you have done or want to review: a free account, no password, keeps your ticks and your right answers from one visit to the next, tells you which chapter to tackle next and lets you ask for the exercise you are missing. Create your space, an email is enough.
Exercise 2: Secant and cosecant: the two antiderivatives you cannot guess
∫secxdx does not follow from any formula in the table: it takes an artificial multiplication, found in the 17th century for the needs of Mercator's map projection. The same trick, transposed, gives the cosecant.
a) Prove that ∫secxdx=ln∣secx+tanx∣+C by multiplying the integrand by secx+tanxsecx+tanx. What structure then appears?
b) Obtain ∫cscxdx the same way and write the result in two equivalent forms.
c) Show that ln∣secx+tanx∣ and −ln∣secx−tanx∣ are the same function, without differentiating.
d) Compute ∫0π/4secxdx and give a decimal value.
e) A student writes ∫secxdx=2sec2x+C by analogy with ∫xdx. Refute this, then say which of the four integrals secx, sec2x, secxtanx, sec3x are immediate and why.
Show the solution
Answers
a)ln∣secx+tanx∣+C, form ff′
b)−ln∣cscx+cotx∣+C=ln∣cscx−cotx∣+C
c)Product equal to 1: opposite logarithms
d)ln(2+1)≈0.8814
e)False; sec2x and secxtanx are immediate
a)
∫secx⋅secx+tanxsecx+tanxdx=∫secx+tanxsec2x+secxtanxdx. Now the numerator is exactly the DERIVATIVE of the denominator, since (secx+tanx)′=secxtanx+sec2x. This is the form ∫ff′, hence ln∣secx+tanx∣+C. The whole trick is to manufacture a denominator whose derivative is the numerator.
b)
Multiply by cscx+cotxcscx+cotx: the numerator becomes csc2x+cscxcotx, that is, the NEGATIVE of the derivative of cscx+cotx, since (cscx+cotx)′=−cscxcotx−csc2x. Hence ∫cscxdx=−ln∣cscx+cotx∣+C, also written ln∣cscx−cotx∣+C, or lntan2x+C, this last form being the most convenient for evaluating at bounds.
c)
Because (secx+tanx)(secx−tanx)=sec2x−tan2x=1: the two quantities are RECIPROCALS of each other. Since the logarithm of a reciprocal is the negative of the logarithm, ln∣secx−tanx∣=−ln∣secx+tanx∣, which gives the stated equality. Two formulas that look different in two different textbooks can therefore denote exactly the same antiderivative.
The refutation is immediate by differentiation: (2sec2x)′=secx⋅secxtanx=sec2xtanx, which is not secx. The analogy with ∫xdx fails because secx is not the variable of integration: the power rule applies only to x, never to a function. Of the four, TWO are immediate and can be read straight off the table of derivatives, backwards: ∫sec2xdx=tanx+C and ∫secxtanxdx=secx+C. The other two are not: secx requires the trick of a), and sec3x a cyclic integration by parts.
Exercise 3: Tangent and secant: the two cases and the stubborn case
The family tanmxsecnx obeys two rules, based on the two derivatives (tanx)′=sec2x and (secx)′=secxtanx, and on the identity sec2x=1+tan2x. Only one case escapes both rules.
a) Compute ∫sec4xdx. Which factor do you split off, and which substitution follows?
b) Compute ∫tan3xsec3xdx. Which factor do you split off this time?
c) Compute ∫tan4xdx. Why does neither rule apply directly, and what move replaces them?
d) Compute ∫sec3xdx by parts, setting u=secx and dv=sec2xdx.
e) Derive the reduction formula ∫secnxdx=n−1secn−2xtanx+n−1n−2∫secn−2xdx and check that it gives back the result of d).
Show the solution
Answers
a)tanx+3tan3x+C
b)5sec5x−3sec3x+C
c)3tan3x−tanx+x+C
d)21(secxtanx+ln∣secx+tanx∣)+C
e)Reduction formula for secn; n=3 gives back d)
a)
The power of the secant is EVEN: split off sec2x, which will be the du. ∫sec2x⋅sec2xdx=∫(1+tan2x)sec2xdx. With u=tanx and du=sec2xdx: ∫(1+u2)du=u+3u3, that is tanx+3tan3x+C.
b)
The power of the tangent is ODD: this time split off the block secxtanx, the derivative of the secant. ∫tan2xsec2x⋅secxtanxdx=∫(sec2x−1)sec2x⋅secxtanxdx. With u=secx: ∫(u2−1)u2du=5u5−3u3, that is 5sec5x−3sec3x+C.
c)
Here n=0: there is no secant from which to split off a sec2x, and the power of the tangent is even, so secxtanx cannot be split off either. The move is to MANUFACTURE the missing secant with tan2x=sec2x−1: ∫tan2x(sec2x−1)dx=∫tan2xsec2xdx−∫tan2xdx=3tan3x−(tanx−x)+C. You go down two powers at a time this way, until you reach ∫tan2x=tanx−x or ∫tanx=−ln∣cosx∣.
d)
With u=secx, du=secxtanxdx, dv=sec2xdx and v=tanx: I=secxtanx−∫secxtan2xdx=secxtanx−∫secx(sec2x−1)dx=secxtanx−I+∫secxdx. The integral you are looking for reappears on both sides: 2I=secxtanx+ln∣secx+tanx∣, hence I=21(secxtanx+ln∣secx+tanx∣)+C. This is the CYCLIC case, the only one in the family that does not reduce to a substitution.
e)
Generalize d) by setting u=secn−2x and dv=sec2xdx: ∫secnxdx=secn−2xtanx−(n−2)∫secn−2xtan2xdx. Replacing tan2x with sec2x−1, the original integral comes back with the coefficient −(n−2): (n−1)∫secnxdx=secn−2xtanx+(n−2)∫secn−2xdx, which gives the stated formula. For n=3: 2secxtanx+21∫secxdx, exactly the result of d). For n=4, it also gives back a), which provides an independent check.
Exercise 4: Cotangent and cosecant: the mirror image, and its signs
The family cotmxcscnx is the mirror image of the previous one, with (cotx)′=−csc2x and (cscx)′=−cscxcotx and the identity csc2x=1+cot2x. The two minus signs are the only difficulty, and they are enough to cost a lot of marks.
a) Compute ∫cotxdx and compare with ∫tanxdx.
b) Compute ∫cot3xdx.
c) Compute ∫cot4xcsc4xdx.
d) Compute ∫csc3xdx by parts, then check that the result is indeed the mirror image of ∫sec3xdx.
e) Draw up the table of the six basic trigonometric derivatives and of the matching antiderivatives, showing exactly where the minus signs are.
Show the solution
Answers
a)ln∣sinx∣+C versus −ln∣cosx∣+C
b)−2cot2x−ln∣sinx∣+C
c)−5cot5x−7cot7x+C
d)21(−cscxcotx−ln∣cscx+cotx∣)+C
e)The co functions carry the minus sign
a)
∫cotxdx=∫sinxcosxdx=ln∣sinx∣+C, a direct ff′ form. For the tangent, ∫tanxdx=∫cosxsinxdx=−ln∣cosx∣+C, also written ln∣secx∣+C. The minus sign comes from the derivative of the cosine being negative: it is the prototype of every sign error in the cotangent-cosecant family.
b)
∫cotx(csc2x−1)dx=∫cotxcsc2xdx−∫cotxdx. For the first one, u=cotx and du=−csc2xdx give −2cot2x. Hence −2cot2x−ln∣sinx∣+C. Note the two minus signs, one from the substitution, the other from a).
c)
The power of the cosecant is even: split off csc2x. ∫cot4x(1+cot2x)csc2xdx; with u=cotx and du=−csc2xdx: −∫(u4+u6)du=−5u5−7u7, that is −5cot5x−7cot7x+C. The rule is the one for the secant, up to an overall sign.
d)
With u=cscx and dv=csc2xdx, so v=−cotx: I=−cscxcotx−∫cscxcot2xdx=−cscxcotx−∫cscx(csc2x−1)dx=−cscxcotx−I+∫cscxdx. Hence I=21(−cscxcotx−ln∣cscx+cotx∣)+C. This is indeed the exact mirror image of ∫sec3xdx=21(secxtanx+ln∣secx+tanx∣), with the opposite overall sign: every time you go from a function to its cofunction, a minus sign appears.
e)
Derivatives: (sin)′=cos; (cos)′=−sin; (tan)′=sec2; (cot)′=−csc2; (sec)′=sectan; (csc)′=−csccot. Matching antiderivatives: ∫cos=sin; ∫sin=−cos; ∫sec2=tan; ∫csc2=−cot; ∫sectan=sec; ∫csccot=−csc. Then the ones that are not immediate: ∫tan=−ln∣cos∣ and ∫cot=+ln∣sin∣, as well as ∫sec=ln∣sec+tan∣ and ∫csc=−ln∣csc+cot∣. The memory rule fits in one sentence: EVERY function whose name starts with “co” (cosine, cotangent, cosecant) carries a minus sign in its derivative, and therefore in the antiderivative of its cofunction.
Exercise 5: The identities that replace a technique
Some trigonometric integrals seem to call for a heavy change of variable when an identity makes them immediate. We will use 1+cosx=2cos22x, 1−cosx=2sin22x, and multiplication by the conjugate.
a) Compute ∫1+cosxdx with the half-angle identity.
b) Compute ∫1−cosxdx.
c) Compute ∫1+sinxdx by multiplying by the conjugate 1−sinx.
d) Compute ∫0π1−cosxdx.
e) A student computes ∫02π1+cosxdx and finds zero. Where is the error, and what is the true value?
Show the solution
Answers
a)tan2x+C
b)−cot2x+C
c)tanx−secx+C
d)22≈2.828
e)Absolute value forgotten; 42≈5.657
a)
1+cosx=2cos22x, so ∫2cos22xdx=21∫sec22xdx=tan2x+C, the factor 2 from the inner derivative cancelling the 21. Three lines, where the Weierstrass substitution would have taken ten.
∫(1+sinx)(1−sinx)1−sinxdx=∫cos2x1−sinxdx=∫(sec2x−secxtanx)dx=tanx−secx+C. The conjugate turns a stubborn denominator into cos2x, and both pieces are then antiderivatives from the table. It is the same idea as the denominator secx+tanx in the integral of the secant.
d)
1−cosx=2sin2x, and on [0,π] the argument 2x stays in [0,2π], where the sine is positive: the absolute value drops. ∫0π2sin2xdx=2[−2cos2x]0π=2(0+2)=22≈2.828.
e)
The error is writing 1+cosx=2cos2x and forgetting the ABSOLUTE VALUE. On [0,2π], the argument 2x runs over [0,π] and the cosine changes sign there at 2π, that is, at x=π. Without the absolute value, the second half of the integral exactly cancels the first and you get zero, which is absurd for a positive function. You must split: ∫02π2cos2xdx=2∫0π2cos2xdx=22[2sin2x]0π=42≈5.657. General rule: u2=∣u∣, never u, and you must ALWAYS check the sign on the interval before simplifying a root.
Part B: Trigonometric substitutions (/50)
Exercise 6: The root of a squared minus x squared: the sine
In front of a2−x2, set x=asinθ with θ∈[−2π,2π]: the root becomes acosθ, positive on that whole interval, and the radical disappears. The return to the variable x is read off a right triangle with hypotenuse a.
a) Compute ∫9−x2dx and say why the result contains no root.
b) Compute ∫4−x2x2dx, showing the return to x through the triangle in detail.
c) Compute ∫x29−x2dx.
d) Compute ∫024−x2dx by changing the bounds, then recover the result by geometry.
e) Why do we require θ∈[−2π,2π] and not just any interval? What would happen with θ in the second quadrant?
Show the solution
Answers
a)arcsin3x+C
b)2arcsin2x−2x4−x2+C
c)−x9−x2−arcsin3x+C
d)π, a quarter of the disk of radius 2
e)Inverse defined, cosθ≥0, x runs over [−a,a]
a)
x=3sinθ, dx=3cosθdθ and 9−x2=3cosθ. The integral becomes ∫3cosθ3cosθdθ=∫dθ=θ+C=arcsin3x+C. The result contains no root because the substitution was used precisely to eliminate it: the inverse sine is what keeps track of it, and that is the deep reason why inverse trigonometric functions show up everywhere in this chapter.
b)
x=2sinθ: ∫2cosθ4sin2θ⋅2cosθdθ=4∫sin2θdθ=2θ−2sinθcosθ+C. The right triangle with angle θ, opposite side x and hypotenuse 2 gives sinθ=2x, cosθ=24−x2 and θ=arcsin2x. Hence 2arcsin2x−2x4−x2+C.
c)
x=3sinθ: ∫9sin2θ3cosθ⋅3cosθdθ=∫sin2θcos2θdθ=∫cot2θdθ=∫(csc2θ−1)dθ=−cotθ−θ+C. The triangle gives cotθ=x9−x2, hence −x9−x2−arcsin3x+C. Notice that Part A of the set has just been put to use: a trigonometric substitution almost always ends with an integral of trigonometric powers.
d)
With x=2sinθ, the bound x=0 becomes θ=0 and the bound x=2 becomes θ=2π. ∫0π/22cosθ⋅2cosθdθ=4∫0π/2cos2θdθ=4⋅4π=π. Geometrically, y=4−x2 is the upper half of the circle of radius 2, and the interval [0,2] cuts out a QUARTER of the disk: its area is 4π(2)2=π, which confirms the computation without writing a single antiderivative.
e)
For three reasons that go together. First, the inverse sine is defined as an inverse only on that interval: without this restriction, θ could not be recovered uniquely from x. Next, the cosine is POSITIVE or zero there, which lets you write a2−x2=acosθ without an absolute value; in the second quadrant the cosine is negative and you would have to write −acosθ, with a sign that would spread through the whole computation. Finally, x=asinθ runs there over exactly [−a,a], that is, the whole domain of the integrand, once and only once.
Exercise 7: The root of a squared plus x squared: the tangent
In front of a2+x2, set x=atanθ with θ∈(−2π,2π): the identity 1+tan2θ=sec2θ turns the root into asecθ, and dx=asec2θdθ.
a) Compute ∫x2x2+4dx.
b) Compute ∫(x2+9)3/2dx and comment on how unexpectedly simple the result is.
c) Compute ∫x2+9dx. Which integral from Part A shows up along the way?
d) Evaluate ∫03x2+9dx and give a decimal value.
e) The same integral can be done with x=3sinhu. Write the first two lines of that computation and say in which case this route is preferable.
Show the solution
Answers
a)−4xx2+4+C
b)9x2+9x+C
c)2xx2+9+29ln∣x+x2+9∣+C, via ∫sec3
d)≈10.33
e)9∫cosh2udu: a power reduction
a)
x=2tanθ, dx=2sec2θdθ, x2+4=2secθ: ∫4tan2θ⋅2secθ2sec2θdθ=41∫tan2θsecθdθ=41∫sin2θcosθdθ=−4sinθ1+C. The triangle with legs x and 2 and hypotenuse x2+4 gives sinθ=x2+4x, hence −4xx2+4+C.
b)
x=3tanθ: (x2+9)3/2=27sec3θ, so ∫27sec3θ3sec2θdθ=91∫cosθdθ=9sinθ+C=9x2+9x+C. The result is ALGEBRAIC, with no logarithm and no inverse tangent, although the integrand looked worse than that of ∫x2+9dx, which does give an inverse tangent. The lesson: a higher power in the denominator often simplifies the computation rather than complicating it.
c)
∫3secθ⋅3sec2θdθ=9∫sec3θdθ, that is, exactly the cyclic integral of Part A. Using its result: 29(secθtanθ+ln∣secθ+tanθ∣). With tanθ=3x and secθ=3x2+9: 2xx2+9+29lnx+x2+9+C, the constant −29ln3 having been absorbed into C.
d)
At x=3: 2318+29ln(3+18)≈6.364+8.910=15.274. At x=0: 0+29ln3≈4.944. Difference: about 10.33.
e)
x=3sinhu gives dx=3coshudu and x2+9=3coshu, since cosh2u−sinh2u=1 and the hyperbolic cosine is always positive. The integral becomes 9∫cosh2udu=29∫(1+cosh2u)du, that is, an elementary POWER REDUCTION, where the trigonometric route forced the cyclic integral of sec3. The hyperbolic route is therefore clearly preferable as soon as the root sits in the numerator, and it gives the logarithm directly since arcsinhx=ln(x+x2+1). The trigonometric route remains preferable when the question imposes a triangle or when the other factors of the integrand are themselves trigonometric.
Exercise 8: The root of x squared minus a squared: the secant and its two branches
In front of x2−a2, set x=asecθ: the root becomes a∣tanθ∣. It is the only one of the three substitutions whose domain splits into TWO pieces, x≥a and x≤−a, and that is where nearly all the errors hide.
a) Compute ∫x2−4dx for x>2, and express the result without an inverse trigonometric function.
b) Compute ∫xx2−9dx for x>3.
c) Why do we write x2−a2=a∣tanθ∣ and not atanθ? Give the interval of θ chosen for each of the two branches.
d) Check that the antiderivative of a) remains valid on the negative branch by computing ∫−4−3x2−4dx, then by computing the same integral again with the substitution x=−u.
e) Summarize the three substitutions of the set in a table: which root, which change of variable, which interval, and what the root turns into.
Show the solution
Answers
a)ln∣x+x2−4∣+C
b)x2−9−3arccosx3+C
c)u2=∣u∣: tanθ≤0 on the negative branch
d)≈0.3545 by both computations
e)asinθ, atanθ, asecθ and their intervals
a)
x=2secθ, dx=2secθtanθdθ and, for x>2 with θ∈[0,2π), x2−4=2tanθ without an absolute value, since the tangent is positive there. ∫2tanθ2secθtanθdθ=∫secθdθ=ln∣secθ+tanθ∣+C. The triangle gives secθ=2x and tanθ=2x2−4, hence ln2x+x2−4+C=lnx+x2−4+C′, the constant −ln2 being absorbed.
b)
x=3secθ: ∫3secθ3tanθ⋅3secθtanθdθ=3∫tan2θdθ=3∫(sec2θ−1)dθ=3tanθ−3θ+C. With tanθ=3x2−9 and θ=arcsec3x=arccosx3: x2−9−3arccosx3+C.
c)
Because u2=∣u∣ without exception, and the tangent CHANGES sign from one branch to the other. On the branch x≥a, take θ∈[0,2π), where secθ≥1 and tanθ≥0: the absolute value drops. On the branch x≤−a, take θ∈(2π,π], where secθ≤−1, as needed to reach the negative values of x, but where tanθ≤0: the root is then −atanθ, and a minus sign runs through the whole computation. Forgetting this sign is the classic error, and it produces an antiderivative whose derivative is the negative of the integrand.
d)
With the antiderivative lnx+x2−4: at x=−3, −3+5≈0.7639 and at x=−4, −4+12≈0.5359; the difference of the logarithms is about −0.2693+0.6238=0.3545. With the substitution x=−u, the integral becomes ∫34u2−4du=ln(4+12)−ln(3+5)≈2.0101−1.6556=0.3545. The two values agree: the logarithmic form, unlike the inverse secant form, crosses both branches without any adjustment, which makes it the antiderivative to remember.
e)
Root a2−x2: set x=asinθ, with θ∈[−2π,2π], the root becomes acosθ. Root a2+x2: set x=atanθ, with θ∈(−2π,2π), the root becomes asecθ. Root x2−a2: set x=asecθ, with θ∈[0,2π) if x≥a and θ∈(2π,π] if x≤−a, the root becomes a∣tanθ∣. In all three cases, the interval is chosen so that the inverse function exists AND the root is positive without an absolute value.
Exercise 9: Complete the square first
A root or a denominator of degree two almost never comes in the standard form a2±x2. Completing the square is the step that brings it there, and on its own it decides which of the three substitutions, or which of the three inverse functions, will come out.
a) Compute ∫x2+2x+5dx.
b) Compute ∫3−2x−x2dx. Which detail of the sign of the x2 term changes everything?
c) Compute ∫x2+6x+13dx.
d) Compute ∫3−2x−x2x+1dx. Why is this computation much shorter than the one in b)?
e) Compute ∫x2+6x+132x+5dx, then state the general rule for splitting a first-degree numerator.
Show the solution
Answers
a)ln∣x+1+x2+2x+5∣+C
b)arcsin2x+1+C: the sign of x2 decides
c)21arctan2x+3+C
d)−3−2x−x2+C: the numerator is the derivative of the radicand
e)ln(x2+6x+13)−21arctan2x+3+C
a)
x2+2x+5=(x+1)2+4: this is the form u2+a2, so the root calls for the tangent, or directly for the formula ∫u2+a2du=lnu+u2+a2. With u=x+1: lnx+1+x2+2x+5+C, which can also be written arcsinh2x+1+C.
b)
The coefficient of x2 is NEGATIVE: it must be factored out first, 3−2x−x2=3−(x2+2x)=3+1−(x+1)2=4−(x+1)2. This time you get the form a2−u2, which calls for the sine and gives an inverse sine: arcsin2x+1+C. The sign of the square term therefore decides everything: positive, you get a logarithm or an inverse hyperbolic sine; negative, an inverse sine. A sign error at this step leads to a function of an entirely different type.
c)
x2+6x+13=(x+3)2+4, the form u2+a2 in a denominator with no root: this is the inverse tangent. ∫u2+4du=21arctan2u, that is 21arctan2x+3+C. The factor 21 comes from the a in the denominator and is very often forgotten.
d)
Because the numerator is proportional to the DERIVATIVE of the radicand: (3−2x−x2)′=−2−2x=−2(x+1). The integral can therefore be written −21∫3−2x−x2(3−2x−x2)′dx=−21⋅23−2x−x2=−3−2x−x2+C. No completing the square, no trigonometric substitution: having the right factor in the numerator short-circuits the whole machinery. So ALWAYS look at the numerator before completing the square.
e)
Split the numerator into two pieces: the one proportional to the derivative of the denominator, and a constant. Here 2x+5=(2x+6)−1, and 2x+6 is exactly (x2+6x+13)′. Hence ∫x2+6x+132x+6dx−∫x2+6x+13dx=ln(x2+6x+13)−21arctan2x+3+C; the absolute value is unnecessary, the quadratic being always strictly positive. General rule: in front of quadraticpx+q, write the numerator in the form α×(derivative of the denominator)+β. The first term gives a logarithm, the second an inverse tangent after completing the square. Under a root, the first gives a root and the second an inverse sine or a logarithm.
Exercise 10: When a trigonometric substitution is a bad idea
A root of the type a2±x2 does not automatically call for a trigonometric substitution. In half the cases met on exams, a simple algebraic substitution is enough, and pulling out the triangle costs ten minutes.
a) Compute ∫4−x2xdx, then compare with ∫4−x2dx. What changes everything?
b) Compute ∫xx2−9dx.
c) Compute ∫1+x2x3dx. Is an ODD power in the numerator enough?
d) Compute ∫x2+4xdx and ∫x2+4dx: two neighbouring integrals, two answers of a different nature.
e) State the test that decides, in five seconds, between an algebraic substitution and a trigonometric substitution.
Show the solution
Answers
a)−4−x2+C versus arcsin2x+C
b)3(x2−9)3/2+C
c)3(1+x2)3/2−1+x2+C
d)21ln(x2+4)+C; 21arctan2x+C
e)The factor x in the numerator decides
a)
The numerator x is, up to the factor −21, the derivative of 4−x2. With u=4−x2 and du=−2xdx: −21∫u−1/2du=−u=−4−x2+C. The second integral does NOT have this factor x: no algebraic substitution works, you need x=2sinθ, and you get arcsin2x+C. A single factor x separates an algebraic answer from an inverse sine.
b)
Same structure: u=x2−9, du=2xdx, so 21∫u1/2du=3u3/2=3(x2−9)3/2+C. A trigonometric substitution would reach the same result after three times as many lines and a return through the triangle.
c)
Yes, provided you adjust. With u=1+x2, you have x2=u−1 and xdx=2du, so ∫1+x2x2⋅xdx=21∫uu−1du=21∫(u1/2−u−1/2)du=3u3/2−u1/2, that is 3(1+x2)3/2−1+x2+C. An odd power always lets you split off an xdx and convert the rest, exactly like the odd powers of sine in Part A: it is the same idea in different clothing.
d)
∫x2+4xdx=21ln(x2+4)+C, the form ff′, a LOGARITHMIC answer. ∫x2+4dx=21arctan2x+C, an INVERSE TANGENT answer. Two integrands that differ only by a factor x, two families of functions with nothing in common. This is the example to keep in mind so that you never answer by reflex.
e)
Look at the NUMERATOR, and at it alone, before anything else. If it contains a factor proportional to the derivative of the expression under the root or in the denominator, that is, a lone x when the expression is in x2, set u= that expression and the computation is over in two lines. If it does not, or if it carries an EVEN power of x, no algebraic substitution can work and the trigonometric substitution becomes necessary. In one sentence: the missing factor x is exactly what costs you the triangle.
Part C: Inverse trigonometric functions (/50)
Exercise 11: The three basic antiderivatives and how to recognize them
Only three forms produce an inverse trigonometric function: ∫a2−u2du=arcsinau, ∫a2+u2du=a1arctanau and ∫uu2−a2du=a1arcsecau. The whole art is to bring the integrand back to one of the three.
a) Tell the three forms apart by their silhouette alone: root in the denominator or not, sign in front of u2, presence of a factor u outside the root. Illustrate with ∫25−x2dx, ∫25+x2dx and ∫xx2−25dx.
b) Compute ∫8+2x−x2dx.
c) Compute ∫(x+1)x2+2xdx.
d) Compute ∫02x2−2x+5dx and give a decimal value.
e) Compute ∫1+e2xexdx and ∫x1−(lnx)2dx. What common observation lets you handle both integrals at a glance?
Show the solution
Answers
a)arcsin5x, 51arctan5x, 51arcsec5x
b)arcsin3x−1+C
c)arcsec∣x+1∣+C
d)arctan21≈0.4636
e)arctan(ex)+C; arcsin(lnx)+C
a)
Root in the denominator with a MINUS SIGN in front of u2: inverse sine, here arcsin5x+C. No root, PLUS sign: inverse tangent, here 51arctan5x+C. Root with a minus sign in front of a2 AND a factor u ALONE outside the root: inverse secant, here 51arcsec5x+C. The silhouette is enough, provided you check the three points in this order: the root, the sign, the outside factor.
b)
Complete the square under the root, factoring out the minus sign first: 8+2x−x2=8−(x2−2x)=8+1−(x−1)2=9−(x−1)2. With u=x−1 and a=3, this is the first form: arcsin3x−1+C.
c)
Here x2+2x=(x+1)2−1, and the factor x+1 outside the root is exactly the u of the third form: ∫uu2−1du with u=x+1 and a=1. Hence arcsec∣x+1∣+C. Without the factor x+1 in the denominator, the integral would have given a logarithm and not an inverse secant: that factor, which looks like a nuisance, is in fact what makes the formula applicable.
d)
x2−2x+5=(x−1)2+4, the second form with u=x−1 and a=2. The antiderivative is 21arctan2x−1, so [21arctan2x−1]02=21(arctan21−arctan(−21))=arctan21≈0.4636, the inverse tangent being odd.
e)
In both cases, an INNER substitution brings you back to a basic form. For the first, u=ex gives du=exdx and ∫1+u2du=arctanu=arctan(ex)+C. For the second, u=lnx gives du=xdx and ∫1−u2du=arcsinu=arcsin(lnx)+C. The common observation is that the factor that seems to clutter the integrand, ex on one side and x1 on the other, is precisely the DERIVATIVE of the expression that plays the role of u. The three basic forms almost never apply as they stand: they apply to a u, and spotting that u is all the work.
Exercise 12: The invisible factor 1: integrating an inverse function
An inverse trigonometric function has no obvious antiderivative, but it has a simple DERIVATIVE. That is exactly the situation where integration by parts with dv=dx settles everything: differentiate the inverse function and integrate the factor 1, which had stayed invisible.
a) Compute ∫arcsinxdx by setting u=arcsinx and dv=dx.
b) Compute ∫arccosxdx, then check that the two results are consistent using the identity arcsinx+arccosx=2π.
c) Compute ∫arctanxdx and ∫arccotxdx.
d) Compute ∫arcsecxdx for x>1.
e) Evaluate ∫01arctanxdx and give a decimal value.
Show the solution
Answers
a)xarcsinx+1−x2+C
b)xarccosx−1−x2+C
c)xarctanx∓21ln(1+x2)+C
d)xarcsecx−ln∣x+x2−1∣+C
e)4π−2ln2≈0.4388
a)
u=arcsinx gives du=1−x2dx, and dv=dx gives v=x. Hence ∫arcsinxdx=xarcsinx−∫1−x2xdx. The second integral is algebraic, the numerator being proportional to the derivative of the radicand: it equals −1−x2. Result: xarcsinx+1−x2+C.
b)
The same computation, with du=−1−x2dx, gives xarccosx−1−x2+C. Check: the sum of the two antiderivatives is x(arcsinx+arccosx)=2πx, the two roots cancelling. And that is indeed an antiderivative of arcsinx+arccosx=2π, a constant function. The two computations therefore check each other, and the fact that the roots cancel exactly is no accident: it is the identity, translated.
c)
∫arctanxdx=xarctanx−∫1+x2xdx=xarctanx−21ln(1+x2)+C. For the inverse cotangent, the derivative changes sign: ∫arccotxdx=xarccotx+21ln(1+x2)+C. Here again, the sum of the two antiderivatives is 2πx, since arctanx+arccotx=2π.
d)
u=arcsecx gives du=∣x∣x2−1dx, which equals xx2−1dx since x>1. Hence ∫arcsecxdx=xarcsecx−∫xx2−1xdx=xarcsecx−∫x2−1dx. The factor x cancels, and what remains is the integral from Part B: xarcsecx−lnx+x2−1+C.
Exercise 13: Products with x: the division that follows integration by parts
When an inverse function is multiplied by a power of x, you still integrate by parts, but the remaining integral is no longer immediate: it requires a polynomial division or an identity. That second step is what separates a careful computation from a sloppy one.
a) Compute ∫xarctanxdx, showing in detail the division that follows the integration by parts.
b) Compute ∫xarcsinxdx.
c) Compute ∫x2arctanxdx.
d) Evaluate ∫01xarctanxdx and give a decimal value.
e) Why is the inverse function always chosen as u, and never as dv? Connect your answer to the mnemonic rule for choosing u.
Show the solution
Answers
a)2x2+1arctanx−2x+C
b)42x2−1arcsinx+4x1−x2+C
c)3x3arctanx−6x2+61ln(1+x2)+C
d)4π−21≈0.2854
e)Differentiate the inverse function: the I of LIATE
a)
u=arctanx, dv=xdx, so v=2x2: 2x2arctanx−21∫1+x2x2dx. The degree of the numerator equals that of the denominator: the division is COMPULSORY before anything else, and it gives 1+x2x2=1−1+x21. Hence 2x2arctanx−21(x−arctanx)=2x2+1arctanx−2x+C.
b)
u=arcsinx, v=2x2: 2x2arcsinx−21∫1−x2x2dx. The remaining integral is the one from Part B, with a=1: ∫1−x2x2dx=21arcsinx−2x1−x2. Hence 2x2arcsinx−41arcsinx+4x1−x2=42x2−1arcsinx+4x1−x2+C.
c)
u=arctanx, v=3x3: 3x3arctanx−31∫1+x2x3dx. Division: 1+x2x3=x−1+x2x, whose integral is 2x2−21ln(1+x2). Result: 3x3arctanx−6x2+61ln(1+x2)+C.
Because integration by parts only pays off if you DIFFERENTIATE the difficult factor. An inverse trigonometric function has a very simple algebraic derivative, a rational function or a root, whereas none of its antiderivatives is known when you start: taking it as dv would force you to integrate what you cannot integrate, and the computation would stop at the first line. This is exactly the order of priority summed up by the acronym LIATE, where the I of Inverse functions comes right after the L of Logarithms: those two families are the ones you always differentiate, and never the ones you integrate.
Exercise 14: The inverse secant and the absolute value
The inverse secant is the only one of the six inverse trigonometric functions whose derivative carries an absolute value: (arcsecx)′=∣x∣x2−11. This absolute value is not a matter of careful notation, it is what makes the formula true on both branches of the domain.
a) The domain of arcsec is ∣x∣≥1, so it comes in two pieces. Justify the presence of ∣x∣ by differentiating arcsecx=arccosx1 in both sign cases.
b) Compute ∫xx2−9dx and write the antiderivative in a form valid on BOTH branches.
c) A student suggests 31arcsec3x, without an absolute value. Test the suggestion at x=−4 by differentiating, and say exactly what goes wrong.
d) Evaluate ∫−6−4xx2−9dx. Could the sign of your answer be predicted without computing?
e) Some textbooks define the inverse secant on [0,2π)∪[π,23π) so as to write the derivative without an absolute value. Explain the mechanism and say what drawback this convention brings.
Show the solution
Answers
a)x2=∣x∣: derivative positive everywhere
b)31arcsec3x+C
c)At x=−4: +0.0945 versus −0.0945
d)≈−0.108, negative like the integrand
e)Derivative without absolute value, but arcsecx=arccosx1
a)
Differentiating arccosx1: (arccosx1)′=−1−x211⋅(−x21)=x2x2x2−11=x2⋅∣x∣x2−11=∣x∣x2−11. It is the simplification x2=∣x∣ that produces the absolute value, and it cannot be avoided: for x>1 it gives xx2−11, but for x<−1 it gives −xx2−11, a POSITIVE quantity even though x is negative. The derivative of the inverse secant is therefore positive everywhere, which is consistent with a function that increases on each of its branches.
b)
With u=3x, the third basic form gives 31arcsec3x+C, also written 31arccos∣x∣3+C. The absolute value INSIDE is what makes the expression valid for x>3 as well as for x<−3, without changing formula from one branch to the other.
c)
Without the absolute value, the suggestion reads 31arccosx3. At x=−4, its derivative is 31⋅∣x∣x2−93=∣x∣x2−91=471≈+0.0945, whereas the integrand there is xx2−91=−471≈−0.0945. The SIGN is wrong: the suggestion is an antiderivative of the negative of the integrand on the whole negative branch. It remains perfectly correct for x>3, which makes it all the more dangerous: it passes the check if you only test a positive point.
d)
With the correct antiderivative: at x=−4, 31arccos43≈30.7227≈0.2409; at x=−6, 31arccos21=9π≈0.3491. The integral is 0.2409−0.3491≈−0.108. The negative sign was entirely predictable: on [−6,−4], the root is positive but x is negative, so the integrand is negative on the whole interval, and the integral of a negative function can only be negative. This is the check to make every time before accepting an answer.
e)
The mechanism is as follows: on [π,23π), the tangent is POSITIVE while the secant is negative, so the relation sec2θ−1=tanθ holds without an absolute value and the derivative becomes xx2−11 on both branches. The notation gains from it, and the trigonometric substitutions of Part B become lighter. The drawback is real: this convention breaks the relation arcsecx=arccosx1, which is the one calculators and computer algebra systems implement. A result obtained with one convention and checked with the other will look wrong when it is not, and that is exactly why, in a written solution, you must state the convention you use or, more simply, prefer the logarithmic form lnx+x2−a2 whenever it is available.
Exercise 15: Compositions and identities: two antiderivatives for one integral
An inverse function composed with something else is handled by substituting first, then integrating by parts. And since inverse functions obey identities, the same integral often has two answers that look incompatible.
a) Compute ∫arctanxdx by setting t=x.
b) Compute ∫exarcsin(ex)dx.
c) Compute ∫1+x2arctanxdx and ∫(1+x2)arctanxdx. What structure do they share?
d) Given that arctanx+arctanx1=2π for x>0, compute ∫arctanx1dx without any integration by parts.
e) Two students compute ∫xx2−1dx for x>1. One answers arcsecx, the other arctanx2−1. Settle the matter.
Show the solution
Answers
a)(x+1)arctanx−x+C
b)exarcsin(ex)+1−e2x+C
c)2(arctanx)2+C; ln∣arctanx∣+C
d)2πx−xarctanx+21ln(1+x2)+C
e)Both: the same function on (1,+∞)
a)
t=x gives x=t2 and dx=2tdt: ∫arctant⋅2tdt=2∫tarctantdt. This is the integral of exercise 13: 2(2t2+1arctant−2t)=(t2+1)arctant−t. Going back to x: (x+1)arctanx−x+C. Substituting first turned a composition into a product already handled.
b)
The factor ex is the derivative of the argument: that is the signal to set u=ex, hence ∫arcsinudu=uarcsinu+1−u2, that is exarcsin(ex)+1−e2x+C, valid for ex≤1, that is, x≤0.
c)
In both cases, 1+x2dx is exactly d(arctanx): setting u=arctanx, the first becomes ∫udu=2u2, that is 2(arctanx)2+C, and the second ∫udu=ln∣u∣, that is ln∣arctanx∣+C. The shared structure is that the inverse tangent appears TWICE, once as a function and once, in disguise, as a differential. Spotting 1+x2dx as a ready-made du brings these integrals down in one line.
d)
∫arctanx1dx=∫(2π−arctanx)dx=2πx−xarctanx+21ln(1+x2)+C, reusing the antiderivative of the inverse tangent. The identity has saved a whole integration by parts, and it is a reminder that arctanx1 is not a new function but the inverse tangent turned around. Be careful, though: for x<0, the identity becomes arctanx+arctanx1=−2π, and the constant changes from one branch to the other.
e)
Both are right. Differentiating the second suggestion: (arctanx2−1)′=1+(x2−1)1⋅x2−1x=xx2−11, which is indeed the integrand. The two functions are therefore two antiderivatives of the same function on the interval (1,+∞), which is connected: they differ by a constant, and since both are zero at x=1, that constant is ZERO, and they are exactly equal. It is in fact only a disguised identity: if θ=arcsecx, then secθ=x and tanθ=x2−1, so θ=arctanx2−1. The lesson holds for the whole chapter: two answers that look very different are often the same, and the only judge is differentiation.
Part D: Definite integrals, pitfalls and synthesis (/50)
Exercise 16: The bounds you forget to change
In a DEFINITE integral, a change of variable must come with a change of bounds. Two methods are correct: transform the bounds and never come back to x, or keep the bounds in x and return to the original variable before evaluating. Mixing the two is the most costly error of the chapter.
a) Compute ∫014−x2dx by trigonometric substitution, transforming the bounds.
b) Compute ∫02x24−x2dx.
c) Compute ∫031+x2dx and interpret the angle obtained geometrically.
d) A student writes ∫014−x2dx=∫01dθ=1. Diagnose the error precisely and assess its size.
e) State the two correct methods and say which one you would recommend on an exam, with your reason.
Show the solution
Answers
a)6π≈0.5236
b)π
c)3π, that is 60°
d)Bounds not changed: 1 instead of 6π, a 91% error
e)Transform the bounds and stay in θ
a)
Set x=2sinθ, so dx=2cosθdθ and 4−x2=2cosθ. The bounds follow: x=0 gives θ=0, and x=1 gives sinθ=21, that is θ=6π. The integral becomes ∫0π/6dθ=6π≈0.5236. Direct check: [arcsin2x]01=arcsin21=6π.
b)
Same substitution and bounds, except that x=2 gives θ=2π: ∫0π/24sin2θ⋅2cosθ⋅2cosθdθ=16∫0π/2sin2θcos2θdθ. Part A showed that sin2θcos2θ=81−cos4θ, hence 2[θ−4sin4θ]0π/2=2⋅2π=π≈3.1416.
c)
[arctanx]03=arctan3=3π≈1.0472. Geometrically, the substitution x=tanθ makes θ the angle whose tangent is x: integrating 1+x2dx amounts to accumulating dθ, so the integral measures exactly the angle swept as x goes from 0 to 3, that is, 60 degrees. This is also what makes arctan the most convenient tool for computing π numerically.
d)
The student did substitute dx and the root, but kept the bounds in x while writing the integral in θ: the upper bound x=1 was copied as θ=1. Yet θ=1 radian corresponds to x=2sin1≈1.68, which is not the requested bound. The result, 1, exceeds the true value 6π≈0.524 by 91%. The symptom is easy to spot: an answer without π for an integral that should have produced one.
e)
First method: transform the bounds along with the variable, then evaluate directly in θ without ever going back to x. Second method: first work out the INDEFINITE integral, go back to x through the triangle, then evaluate between the original bounds. On an exam, I recommend the first: it avoids the return through the triangle entirely, and that return is the longest and most error-prone step of any trigonometric substitution. It also has the advantage of producing simple numerical bounds, often 0, 6π, 4π or 2π, which make the evaluation immediate. The only absolute rule is never to leave on the same line a variable and bounds that do not match.
Exercise 17: Two antiderivatives that look different
The same integral can receive answers that look incompatible depending on the method used. The theorem that settles the question is well known, but it carries a hypothesis that the chapter on inverse functions specifically violates.
a) Compute ∫sinxcosxdx in THREE ways: with u=sinx, with u=cosx, then with the double angle.
b) Show that the three answers differ pairwise by a constant, and compute these constants.
c) State the theorem that guarantees this result, stressing its hypothesis.
d) On R∖{0}, ln∣x∣+3 and the function equal to ln∣x∣ for x>0 and to ln∣x∣+7 for x<0 are both antiderivatives of x1. Is the hypothesis of c) satisfied?
e) Connect this phenomenon to the question about the inverse secant handled earlier, and say what you must write in a solution to be beyond reproach.
Show the solution
Answers
a)2sin2x, −2cos2x, −4cos2x
b)Differences 21, 41, −41
c)F−G constant on an INTERVAL
d)No: one constant per piece
e)State the working interval
a)
With u=sinx: ∫udu=2sin2x+C. With u=cosx, where du=−sinxdx: −∫udu=−2cos2x+C. With the double angle, sinxcosx=2sin2x: 21∫sin2xdx=−4cos2x+C.
b)
2sin2x−(−2cos2x)=2sin2x+cos2x=21, a constant. 2sin2x−(−4cos2x)=2sin2x+41−2sin2x=41, a constant as well. And by subtracting the previous two, the third pair differs by −41. The three answers are therefore the same antiderivative up to a constant, which is exactly what the +C allows: three correct methods cannot give essentially different functions.
c)
Theorem: if F and G are two antiderivatives of the same function on an INTERVAL, then F−G is constant on that interval. The proof takes one line, (F−G)′=0, but it relies on the Mean Value Theorem, which requires an interval, that is, a CONNECTED domain, all in one piece. This is the hypothesis people forget to check, because it is almost always satisfied.
d)
No, it is not: R∖{0} is not an interval but the union of two disjoint intervals. The difference of the two proposed functions is −3 on the positive half-line and +4 on the negative one: it is LOCALLY constant without being constant. Yet both functions really are antiderivatives of x1 on each of the two pieces. The correct conclusion is therefore that on a domain made of several pieces, an antiderivative carries one constant PER PIECE, and the notation ln∣x∣+C is a convenient abuse that you must know how to defuse when you work on both sides of zero.
e)
It is exactly the same situation. The domain of xx2−91 is (−∞,−3)∪(3,+∞), also made of two pieces: nothing forces the antiderivative to be written in one piece, and that is why a formula that is correct on one branch can be wrong on the other, like the version of the inverse secant without an absolute value. To be beyond reproach in a solution, do three things: state the working interval before computing, check that the proposed antiderivative is defined and differentiable on it, and, when the domain comes in several pieces, say so explicitly rather than writing a global +C that hides the problem. A definite integral is also never computed across a discontinuity: it must be split, or recognized as improper.
Exercise 18: Improper integrals with a trigonometric answer
The inverse trigonometric functions are bounded: that is what lets some integrals over an infinite interval converge, and it is also what makes π appear in results where nothing announced it.
a) Compute ∫0+∞1+x2dx by going back to the definition with a limit.
b) Compute ∫011−x2dx. In what sense is this integral improper, even though its bounds are finite?
c) Compute ∫−∞+∞x2+2x+5dx.
d) Compute ∫0+∞(1+x2)2dx.
e) Compute ∫1+∞xx2−1dx and explain why this integral, improper at both ends, still converges.
Show the solution
Answers
a)2π
b)2π, integrand infinite at 1
c)2π
d)4π
e)2π: singularity like x−11, decrease like x21
a)
By definition, ∫0+∞1+x2dx=limb→+∞∫0b1+x2dx=limb→+∞arctanb=2π≈1.5708. Convergence rests entirely on the inverse tangent being BOUNDED: the area keeps accumulating but stays capped, the integrand decreasing like x21.
b)
Its bounds are finite, but the INTEGRAND is not: 1−x21 tends to infinity as x→1−. It is an improper integral of the second kind, and you must write limb→1−[arcsinx]0b=limb→1−arcsinb=2π. An infinitely tall yet finite area: the graph shoots up vertically at x=1 while enclosing only 2π of area. Notice the unsettling result: it is the same value as in a), obtained on a bounded interval.
c)
Complete the square: x2+2x+5=(x+1)2+4, so the antiderivative is 21arctan2x+1. The integral, improper at both bounds, is split in two and equals 21(2π−(−2π))=2π≈1.5708. Each of the two halves converges separately, an essential condition for being allowed to conclude.
d)
The substitution x=tanθ turns the bounds 0 and +∞ into 0 and 2π, which makes the integral PROPER: ∫0π/2sec4θsec2θdθ=∫0π/2cos2θdθ=4π≈0.7854. This is one of the great services of the trigonometric substitution: it brings an infinite interval down to a bounded one, and the question of convergence settles itself.
e)
The integral is improper at 1, where the denominator vanishes, and at +∞. The antiderivative is arcsecx, so the value is limb→+∞arcsecb−lima→1+arcseca=2π−0=2π. It converges at both ends for two distinct reasons: at 1, the singularity behaves like x−11, with exponent 21<1, hence integrable; at infinity, the integrand decreases like x21, with exponent 2>1, hence integrable as well. The fact that the inverse secant is bounded by 2π sums up both at once, and the same value comes back for the fourth time, after a), b) and c): this is no coincidence, the four integrands are images of one another under the substitutions of the chapter.
Exercise 19: Applications: area, arc length and volume
The integrals of this chapter are not algebra drills: they come up naturally in the three great applications of integral calculus, and that is where you meet them, in physics as well as in geometry.
a) Prove the formula for the area of a disk by computing 4∫0aa2−x2dx.
b) Compute the length of the curve y=ln(cosx) for x∈[0,4π]. Which integral from Part A do you recognize?
c) Compute the length of the parabola y=2x2 for x∈[0,1].
d) Compute the volume generated by rotating about the x-axis the region under y=1+x21 for x∈[0,1].
e) A part has the shape of the region under y=1+x21 for x≥0. Is its area finite? And the volume generated by rotating it about the x-axis?
Show the solution
Answers
a)4⋅4πa2=πa2
b)∫0π/4secxdx≈0.8814
c)22+21ln(1+2)≈1.148
d)4π2≈2.467
e)Area 2π, volume 4π2: both finite
a)
By symmetry, the quarter disk in the first quadrant has area ∫0aa2−x2dx. With x=asinθ, the bounds become 0 and 2π: ∫0π/2acosθ⋅acosθdθ=a2∫0π/2cos2θdθ=a2⋅4π. Multiplying by 4: πa2. The formula for the area of a disk, learned in high school without proof, thus comes out of a trigonometric substitution and the power reduction of cos2.
b)
y′=−tanx, so 1+y′2=1+tan2x=secx, the secant being positive on [0,4π]. The length is ∫0π/4secxdx=ln(2+1)≈0.8814. This is exactly the integral of the secant computed in Part A, and the coincidence is no coincidence: the identity 1+tan2=sec2 brings out the secant in every arc length of a curve whose slope is a tangent.
c)
y′=x, so the length is ∫011+x2dx, the integral from Part B with a=1: [2x1+x2+21ln(x+1+x2)]01=22+21ln(1+2)≈0.7071+0.4407≈1.148. Compare with the chord, of length 1+0.25≈1.118: the parabola is only 3% longer than the segment joining its endpoints.
d)
By the disk method, V=π∫011+x2dx=π[arctanx]01=4π2≈2.467. A volume equal to 4π2: the square of π appears because π comes once from the rotation and once from the inverse tangent, two completely independent origins.
e)
The area is ∫0+∞1+x2dx=2π, so it is FINITE. The volume of revolution is π∫0+∞(1+x2)2dx=π⋅4π=4π2≈2.467, finite as well. Both converge because the integrand decreases fast enough, like x21 for the area and like x41 for the volume: squaring speeds up the decrease and makes convergence even more comfortable. This is not always the case, and the comparison is worth making with Gabriel's horn, generated by y=x1 on [1,+∞), whose volume is finite while the area under the curve is infinite.
Exercise 20: Synthesis problem: eight integrals, one decision tree
Here are eight mixed integrals, with no hint about the method, as on an exam. For each one, the difficulty is not the computation but the DIAGNOSIS: identifying in a few seconds which family it belongs to.
a) Compute ∫sin3xcos4xdx and ∫sec4xtan2xdx.
b) Compute ∫x2−16dx and ∫x2−16xdx.
c) Compute ∫x2+4x+8dx and ∫x2x2+1dx.
d) Compute ∫arctan(3x)dx and ∫xarcsin(x2)dx.
e) Write out the complete decision tree of the set: in front of a trigonometric integral or one with a quadratic under a root, which questions should you ask, and in what order?
e)Derivative in the numerator, inverse function, parity, square, triangle
a)
Odd power of sine: split off sinx and set u=cosx, hence −∫(1−u2)u4du=−5cos5x+7cos7x+C. For the second, EVEN power of secant: split off sec2x and set u=tanx, hence ∫(1+u2)u2du=3tan3x+5tan5x+C.
b)
The first has no factor x in the numerator: it is the substitution x=4secθ, or the formula directly, and you get lnx+x2−16+C. The second has one, and it is proportional to the derivative of the radicand: algebraic substitution u=x2−16, hence x2−16+C. Two neighbouring integrals, two opposite families, and the factor x is the only judge.
c)
x2+4x+8=(x+2)2+4: complete the square, then inverse tangent, 21arctan2x+2+C. For the second, a root of the form x2+a2 with no useful factor in the numerator: substitution x=tanθ, which gives −xx2+1+C.
d)
An inverse function alone: integration by parts with the invisible factor 1. ∫arctan3xdx=xarctan3x−∫1+9x23xdx=xarctan3x−61ln(1+9x2)+C. For the second, the factor x is the derivative of x2: first substitute u=x2, which gives 21∫arcsinudu=21(uarcsinu+1−u2), that is 2x2arcsin(x2)+1−x4+C.
e)
First question: is there, in the numerator, a factor proportional to the DERIVATIVE of what is under the root or in the denominator? If so, algebraic substitution, and the work is done. Second question: is the integrand an inverse trigonometric function, alone or multiplied by a power of x? If so, integration by parts, differentiating the inverse function. Third question: is it a product of trigonometric powers? If so, apply the parity rule, splitting off one factor from the odd power or reducing the powers if both are even. Fourth question: is the denominator or the radicand a quadratic? If so, complete the square, which brings you back to one of the three standard forms. Only then, fifth question: is there still a root of the type a2−x2, a2+x2 or x2−a2? Then, and only then, bring out the trigonometric substitution and its triangle. The order matters as much as the list: the trigonometric substitution is the last resort, not the first reflex, and half the marks lost on exams come from drawing it too early.
Something missing from this set?A type of exercise your teacher gives, a problem you are stuck on, a mistake you spotted: tell me, I read every request.
Sending goes through your free space: that is how I can answer you and tell you when the exercise is added. An email is enough, no password. Create your space.
Rather not create an account? Write to me directly.
Looking for an integral calculus tutor in Montreal?
Get in touch for a first session. We go through the chapter family by family, with a method for spotting in thirty seconds which of the ten techniques applies, instead of trying them all.