MATH 141 Calculus 2 • McGill University, Montreal

Corrected exercises: integration by parts (MATH 141)

This is the corrected exercise set for integration by parts in MATH 141, Calculus 2, the second calculus course at McGill University. It follows section 7.1 of Stewart, Clegg and Watson, and it is written for the exam as it is set: no calculator, so every answer is exact (π\pi, ee, ln⁡2\ln 2) and every solution names its uu, dudu, dvdv and vv before using the formula, because that is where the method marks are.

The thread running through the whole set: integration by parts does not compute an integral, it TRADES it. The formula ∫u dv=uv−∫v du\int u\,dv = uv - \int v\,du replaces the integral you have by a new one, and the only question is whether the new one is easier. The power that goes down, the logarithm that becomes 1x\frac{1}{x}, the integral that comes back and is solved like an equation, the bracket [uv]ab\left[uv\right]_a^b that is a number to evaluate at both bounds: every exercise is that same trade, judged the same way.

The traps named explicitly in the solutions: the product of two integrals taken for the integral of a product, the choice of uu that raises the power, the minus sign of v=−cos⁡xv = -\cos x lost in −∫v du-\int v\,du, table signs that do not alternate, the cyclic integral that turns into I=II = I because the roles were swapped, the new integral of a definite computation left without its bounds, the lower bound forgotten because ln⁡1=0\ln 1 = 0, and the sign of the derivative of an integral with a variable lower bound.

10 corrected exercises • 100 points • 150 minutes

Revision sheet for this chapter → Every MATH 141 chapter →

Course recap

  • • Integration by parts: ∫u dv=uv−∫v du\int u\,dv = uv - \int v\,du, the product rule (uv)′=u′v+uv′(uv)' = u'v + uv' integrated and rearranged.
  • • Definite form: ∫abu dv=[uv]ab−∫abv du\int_a^b u\,dv = \left[uv\right]_a^b - \int_a^b v\,du, both halves evaluated at both bounds.
  • • Choose uu to simplify when differentiated and dvdv to be integrable. Guide (LIATE): logarithm, inverse trigonometric, algebraic, trigonometric, exponential, in order of preference for uu.
  • • Lone factor: ∫ln⁡x dx=xln⁡x−x+C\int \ln x\,dx = x\ln x - x + C, with dv=dxdv = dx; same gesture for arctan⁡x\arctan x and arcsin⁡x\arcsin x.
  • • Polynomial of degree nn times eaxe^{ax}, sin⁡bx\sin bx or cos⁡bx\cos bx: nn trades, or the tabular method with alternating signs +,−,+,…+, -, +, \dots.
  • • Cyclic (eaxe^{ax} with sin⁡bx\sin bx or cos⁡bx\cos bx): two trades with the SAME type as uu, then solve for the integral; the factor a2+b2a^2 + b^2 appears.
  • • Any antiderivative of dvdv is a valid vv: a well chosen constant can simplify ∫v du\int v\,du.

Part A: the basics (/50)

Exercise 1: Choosing u and dv: the new integral is the judge

Integration by parts is the product rule read backwards: from (uv)′=u′v+uv′(uv)' = u'v + uv' comes ∫u dv=uv−∫v du\int u\,dv = uv - \int v\,du. The formula does not compute anything. It TRADES the integral you have for another one, ∫v du\int v\,du, and the whole skill is to choose uu so that the new integral is easier than the old one.

Two requirements decide the choice: dvdv must be something you can integrate, and uu should be the factor that becomes SIMPLER when differentiated. In every part, write the four pieces uu, dudu, dvdv, vv before the formula.

  • a) Compute ∫xe2x dx\int x e^{2x}\,dx with u=xu = x and dv=e2x dxdv = e^{2x}\,dx.
  • b) Redo the first step with the opposite choice, u=e2xu = e^{2x} and dv=x dxdv = x\,dx. Is the resulting equation false? Why is it useless?
  • c) Compute ∫x3ln⁡x dx\int x^3 \ln x\,dx. Explain why here the polynomial must be dvdv and not uu.
  • d) Compute ∫x 2x dx\int x\, 2^x\,dx, and name the two wrong antiderivatives of 2x2^x that students write for vv.
  • e) Check the answers to a) and c) by differentiating them.
Show the solution

Answers

  • a) ∫xe2x dx=xe2x2−e2x4+C=e2x(2x−1)4+C\int x e^{2x}\,dx = \frac{x e^{2x}}{2} - \frac{e^{2x}}{4} + C = \frac{e^{2x}(2x - 1)}{4} + C
  • b) x2e2x2−∫x2e2x dx\frac{x^2 e^{2x}}{2} - \int x^2 e^{2x}\,dx: true, but the power went UP.
  • c) ∫x3ln⁡x dx=x4ln⁡x4−x416+C\int x^3 \ln x\,dx = \frac{x^4 \ln x}{4} - \frac{x^4}{16} + C
  • d) ∫x 2x dx=x 2xln⁡2−2x(ln⁡2)2+C\int x\,2^x\,dx = \frac{x\,2^x}{\ln 2} - \frac{2^x}{(\ln 2)^2} + C; wrong vv: 2x2^x and 2x+1x+1\frac{2^{x+1}}{x+1}.
  • e) Both derivatives give back the integrand: xe2xx e^{2x} and x3ln⁡xx^3 \ln x.

a) u=xu = x, du=dxdu = dx; dv=e2x dxdv = e^{2x}\,dx, v=12e2xv = \frac{1}{2}e^{2x} (a substitution you do at sight: the derivative of 12e2x\frac{1}{2}e^{2x} is e2xe^{2x}). Then ∫xe2x dx=xe2x2−∫12e2x dx=xe2x2−e2x4+C=e2x(2x−1)4+C\int x e^{2x}\,dx = \frac{x e^{2x}}{2} - \int \frac{1}{2}e^{2x}\,dx = \frac{x e^{2x}}{2} - \frac{e^{2x}}{4} + C = \frac{e^{2x}(2x - 1)}{4} + C. The trade was good: the new integral, ∫12e2x dx\int \frac{1}{2}e^{2x}\,dx, has lost the factor xx and is immediate. That is the test of a choice, and it can be run BEFORE finishing the computation: differentiating u=xu = x gives 11, and the power of xx disappears.

b) u=e2xu = e^{2x}, du=2e2x dxdu = 2e^{2x}\,dx; dv=x dxdv = x\,dx, v=x22v = \frac{x^2}{2}. The formula gives ∫xe2x dx=x2e2x2−∫x22⋅2e2x dx=x2e2x2−∫x2e2x dx\int x e^{2x}\,dx = \frac{x^2 e^{2x}}{2} - \int \frac{x^2}{2} \cdot 2e^{2x}\,dx = \frac{x^2 e^{2x}}{2} - \int x^2 e^{2x}\,dx. This equation is TRUE: parts never produces a false identity. It is useless because the new integral carries x2x^2 where the old one carried xx: the trade went the wrong way, and applying parts again with the same choice would give x3x^3, then x4x^4, forever. When the integral you get is worse than the one you had, stop after one line and swap uu and dvdv: that decision, written down, is worth marks on its own.

c) u=ln⁡xu = \ln x, du=dxxdu = \frac{dx}{x}; dv=x3 dxdv = x^3\,dx, v=x44v = \frac{x^4}{4}. Then ∫x3ln⁡x dx=x4ln⁡x4−∫x44⋅1x dx=x4ln⁡x4−∫x34 dx=x4ln⁡x4−x416+C\int x^3 \ln x\,dx = \frac{x^4 \ln x}{4} - \int \frac{x^4}{4} \cdot \frac{1}{x}\,dx = \frac{x^4 \ln x}{4} - \int \frac{x^3}{4}\,dx = \frac{x^4 \ln x}{4} - \frac{x^4}{16} + C. Here the polynomial must be dvdv, against the habit built in part a). Taking dv=ln⁡x dxdv = \ln x\,dx would require an antiderivative of ln⁡x\ln x just to START, and would leave ∫3x2(xln⁡x−x) dx\int 3x^2(x \ln x - x)\,dx, which contains x3ln⁡xx^3 \ln x again. Differentiating ln⁡x\ln x, on the other hand, turns it into 1x\frac{1}{x}, which cancels against x4x^4: the logarithm is the factor that simplifies, so it is uu. The rule of thumb LIATE (logarithm, inverse trigonometric, algebraic, trigonometric, exponential, in that order of preference for uu) says the same thing, but the reason is the new integral, not the mnemonic.

d) u=xu = x, du=dxdu = dx; dv=2x dxdv = 2^x\,dx, v=2xln⁡2v = \frac{2^x}{\ln 2}, because ddx2x=2xln⁡2\frac{d}{dx}2^x = 2^x \ln 2. Then ∫x 2x dx=x 2xln⁡2−1ln⁡2∫2x dx=x 2xln⁡2−2x(ln⁡2)2+C\int x\,2^x\,dx = \frac{x\,2^x}{\ln 2} - \frac{1}{\ln 2}\int 2^x\,dx = \frac{x\,2^x}{\ln 2} - \frac{2^x}{(\ln 2)^2} + C. The two classic wrong values of vv: v=2xv = 2^x, copied from exe^x (its derivative is 2xln⁡22^x \ln 2, off by the factor ln⁡2≈0.69\ln 2 \approx 0.69), and v=2x+1x+1v = \frac{2^{x+1}}{x + 1}, the power rule applied to a variable EXPONENT, which is not a power function at all. Both errors propagate: vv appears twice, once in uvuv and once in ∫v du\int v\,du, which is why the correct answer carries ln⁡2\ln 2 to the first power AND to the second.

e) For a): ddx[e2x(2x−1)4]=2e2x(2x−1)+2e2x4=4xe2x4=xe2x\frac{d}{dx}\left[\frac{e^{2x}(2x - 1)}{4}\right] = \frac{2e^{2x}(2x - 1) + 2e^{2x}}{4} = \frac{4x e^{2x}}{4} = x e^{2x}. For c): ddx[x4ln⁡x4−x416]=x3ln⁡x+x44⋅1x−x34=x3ln⁡x\frac{d}{dx}\left[\frac{x^4 \ln x}{4} - \frac{x^4}{16}\right] = x^3 \ln x + \frac{x^4}{4} \cdot \frac{1}{x} - \frac{x^3}{4} = x^3 \ln x. Both are confirmed. This check costs thirty seconds and is the only way, on an exam without a calculator, to know that an antiderivative is right; it catches a lost factor 12\frac{1}{2} or a sign in −∫v du-\int v\,du immediately. Notice that the product rule, used to check, is exactly the rule that parts runs backwards.

Exercise 2: One factor only: ln x, arctan x and arcsin x with dv = dx

Some integrals have a single factor, and it is a function whose DERIVATIVE you know but whose antiderivative is not in the table: ln⁡x\ln x, arctan⁡x\arctan x, arcsin⁡x\arcsin x. Parts still applies, with the invisible partner dv=dxdv = dx and v=xv = x: the trade replaces the function by its derivative, which is algebraic.

The figure shows the graph of y=ln⁡xy = \ln x and the region AA under it between x=1x = 1 and x=ex = e.

0.511.522.533.5-1-0.50.511.5Ax = ey = ln xx
  • a) Compute ∫ln⁡x dx\int \ln x\,dx.
  • b) Compute ∫arctan⁡x dx\int \arctan x\,dx, naming the substitution used for the new integral.
  • c) Compute ∫01/2arcsin⁡x dx\int_0^{1/2} \arcsin x\,dx exactly.
  • d) Compute the area of the region AA on the figure, and check that the value is plausible by comparing AA with a rectangle and a triangle that you can read on the figure.
  • e) Three students give ∫ln⁡x dx=1x+C\int \ln x\,dx = \frac{1}{x} + C, (ln⁡x)22+C\frac{(\ln x)^2}{2} + C and xln⁡x+Cx \ln x + C. Refute each one by differentiating, and name the error behind it.
Show the solution

Answers

  • a) ∫ln⁡x dx=xln⁡x−x+C\int \ln x\,dx = x \ln x - x + C
  • b) ∫arctan⁡x dx=xarctan⁡x−12ln⁡(1+x2)+C\int \arctan x\,dx = x \arctan x - \frac{1}{2}\ln(1 + x^2) + C
  • c) ∫01/2arcsin⁡x dx=π12+32−1\int_0^{1/2} \arcsin x\,dx = \frac{\pi}{12} + \frac{\sqrt 3}{2} - 1
  • d) A=1A = 1, between the triangle e−12\frac{e - 1}{2} and the rectangle e−1e - 1.
  • e) Derivatives: −1x2-\frac{1}{x^2}, ln⁡xx\frac{\ln x}{x}, ln⁡x+1\ln x + 1; none is ln⁡x\ln x.

a) u=ln⁡xu = \ln x, du=dxxdu = \frac{dx}{x}; dv=dxdv = dx, v=xv = x. Then ∫ln⁡x dx=xln⁡x−∫x⋅1x dx=xln⁡x−∫1 dx=xln⁡x−x+C\int \ln x\,dx = x \ln x - \int x \cdot \frac{1}{x}\,dx = x \ln x - \int 1\,dx = x \ln x - x + C. The trade replaced the logarithm, which we cannot integrate directly, by its derivative 1x\frac{1}{x}, and the v=xv = x swallowed it. Check: ddx(xln⁡x−x)=ln⁡x+1−1=ln⁡x\frac{d}{dx}(x \ln x - x) = \ln x + 1 - 1 = \ln x.

b) u=arctan⁡xu = \arctan x, du=dx1+x2du = \frac{dx}{1 + x^2}; dv=dxdv = dx, v=xv = x. Then ∫arctan⁡x dx=xarctan⁡x−∫x1+x2 dx\int \arctan x\,dx = x \arctan x - \int \frac{x}{1 + x^2}\,dx. The new integral is a substitution: w=1+x2w = 1 + x^2, dw=2x dxdw = 2x\,dx, so ∫x1+x2 dx=12∫dww=12ln⁡(1+x2)\int \frac{x}{1 + x^2}\,dx = \frac{1}{2}\int \frac{dw}{w} = \frac{1}{2}\ln(1 + x^2), with no absolute value since 1+x2>01 + x^2 > 0. Final answer: xarctan⁡x−12ln⁡(1+x2)+Cx \arctan x - \frac{1}{2}\ln(1 + x^2) + C. Check: arctan⁡x+x1+x2−12⋅2x1+x2=arctan⁡x\arctan x + \frac{x}{1 + x^2} - \frac{1}{2} \cdot \frac{2x}{1 + x^2} = \arctan x. The substitution is only a TOOL here; the structural move is dv=dxdv = dx.

c) u=arcsin⁡xu = \arcsin x, du=dx1−x2du = \frac{dx}{\sqrt{1 - x^2}}; dv=dxdv = dx, v=xv = x. Then ∫01/2arcsin⁡x dx=[xarcsin⁡x]01/2−∫01/2x1−x2 dx\int_0^{1/2} \arcsin x\,dx = \left[x \arcsin x\right]_0^{1/2} - \int_0^{1/2} \frac{x}{\sqrt{1 - x^2}}\,dx. The bracket is 12⋅π6−0=π12\frac{1}{2} \cdot \frac{\pi}{6} - 0 = \frac{\pi}{12}. For the integral, w=1−x2w = 1 - x^2, dw=−2x dxdw = -2x\,dx, so an antiderivative of x1−x2\frac{x}{\sqrt{1 - x^2}} is −1−x2-\sqrt{1 - x^2}, and ∫01/2x1−x2 dx=[−1−x2]01/2=−32+1\int_0^{1/2} \frac{x}{\sqrt{1 - x^2}}\,dx = \left[-\sqrt{1 - x^2}\right]_0^{1/2} = -\frac{\sqrt 3}{2} + 1. Hence ∫01/2arcsin⁡x dx=π12−1+32\int_0^{1/2} \arcsin x\,dx = \frac{\pi}{12} - 1 + \frac{\sqrt 3}{2}. Plausibility, with π≈3.14\pi \approx 3.14 and 3≈1.73\sqrt 3 \approx 1.73: 0.26+0.87−1≈0.130.26 + 0.87 - 1 \approx 0.13. The graph of arcsin⁡\arcsin is convex on [0,12][0, \frac{1}{2}], so it lies below its chord from (0,0)(0, 0) to (12,π6)(\frac{1}{2}, \frac{\pi}{6}), and the area must be less than that triangle, 12⋅12⋅π6=π24≈0.131\frac{1}{2} \cdot \frac{1}{2} \cdot \frac{\pi}{6} = \frac{\pi}{24} \approx 0.131. It is, just barely, as the near-straight graph suggests.

d) ln⁡x≥0\ln x \ge 0 on [1,e][1, e], so the area is the integral: A=∫1eln⁡x dx=[xln⁡x−x]1e=(e⋅1−e)−(1⋅0−1)=0+1=1A = \int_1^e \ln x\,dx = \left[x \ln x - x\right]_1^e = (e \cdot 1 - e) - (1 \cdot 0 - 1) = 0 + 1 = 1. Plausibility: the region sits inside the rectangle [1,e]×[0,1][1, e] \times [0, 1] of area e−1≈1.72e - 1 \approx 1.72, and since ln⁡\ln is concave its graph lies ABOVE the chord from (1,0)(1, 0) to (e,1)(e, 1), so AA exceeds the triangle of area e−12≈0.86\frac{e - 1}{2} \approx 0.86. The value 11 sits between the two. A result like e−1e - 1 or ee (the typical outcome of dropping the −x-x) is ruled out by the figure at once.

e) ddx1x=−1x2\frac{d}{dx}\frac{1}{x} = -\frac{1}{x^2}: that student DIFFERENTIATED ln⁡x\ln x instead of integrating it. ddx(ln⁡x)22=ln⁡xx\frac{d}{dx}\frac{(\ln x)^2}{2} = \frac{\ln x}{x}: the power rule was applied as if ln⁡x\ln x were the variable, forgetting the chain rule factor 1x\frac{1}{x}; that is the antiderivative of ln⁡xx\frac{\ln x}{x}, a substitution problem, not of ln⁡x\ln x. ddx(xln⁡x)=ln⁡x+1\frac{d}{dx}(x \ln x) = \ln x + 1: this student wrote uvuv and stopped, forgetting the trade's second half, −∫v du=−x-\int v\,du = -x. All three errors are caught by the same thirty-second derivative, which is why it belongs at the end of every answer.

Exercise 3: Repeated parts and the tabular method

When uu is a polynomial of degree nn, each integration by parts lowers its degree by one, so nn integrations finish the job. The tabular method organizes them: a column D of successive derivatives of uu down to 00, a column I of successive antiderivatives of dvdv, and signs that ALTERNATE, +,−,+,−,…+, -, +, -, \dots, starting with ++.

The table is only a bookkeeping of repeated parts: every line of it is one trade, and every minus sign is the minus of −∫v du-\int v\,du.

  • a) Compute ∫x2e−x dx\int x^2 e^{-x}\,dx by two explicit integrations by parts.
  • b) Compute ∫x3cos⁡(2x) dx\int x^3 \cos(2x)\,dx with the tabular method, showing the table.
  • c) Compute ∫0πx2cos⁡x dx\int_0^{\pi} x^2 \cos x\,dx and explain the sign of the result.
  • d) Explain why the table is useless for ∫x2ln⁡x dx\int x^2 \ln x\,dx and for ∫exsin⁡x dx\int e^x \sin x\,dx, and say what to do instead for the first one.
  • e) A student uses the table for ∫x2e−x dx\int x^2 e^{-x}\,dx with the signs +,+,++, +, + and gets −e−x(x2−2x+2)+C-e^{-x}(x^2 - 2x + 2) + C. Differentiate his answer and locate the error.
Show the solution

Answers

  • a) ∫x2e−x dx=−e−x(x2+2x+2)+C\int x^2 e^{-x}\,dx = -e^{-x}(x^2 + 2x + 2) + C
  • b) x3sin⁡2x2+3x2cos⁡2x4−3xsin⁡2x4−3cos⁡2x8+C\frac{x^3 \sin 2x}{2} + \frac{3x^2 \cos 2x}{4} - \frac{3x \sin 2x}{4} - \frac{3\cos 2x}{8} + C
  • c) ∫0πx2cos⁡x dx=−2π\int_0^{\pi} x^2 \cos x\,dx = -2\pi
  • d) The D column of ln⁡x\ln x never reaches 00; for exsin⁡xe^x \sin x neither column ends. ∫x2ln⁡x dx=x3ln⁡x3−x39+C\int x^2 \ln x\,dx = \frac{x^3 \ln x}{3} - \frac{x^3}{9} + C by one trade.
  • e) His derivative is e−x(x−2)2e^{-x}(x - 2)^2; the signs must alternate.

a) First trade: u=x2u = x^2, du=2x dxdu = 2x\,dx; dv=e−x dxdv = e^{-x}\,dx, v=−e−xv = -e^{-x}. So ∫x2e−x dx=−x2e−x+∫2xe−x dx\int x^2 e^{-x}\,dx = -x^2 e^{-x} + \int 2x e^{-x}\,dx: the degree went from 22 to 11, the trade is good. Second trade on the new integral: u=2xu = 2x, du=2 dxdu = 2\,dx; dv=e−x dxdv = e^{-x}\,dx, v=−e−xv = -e^{-x}, so ∫2xe−x dx=−2xe−x+∫2e−x dx=−2xe−x−2e−x\int 2x e^{-x}\,dx = -2x e^{-x} + \int 2e^{-x}\,dx = -2x e^{-x} - 2e^{-x}. Total: −x2e−x−2xe−x−2e−x+C=−e−x(x2+2x+2)+C-x^2 e^{-x} - 2x e^{-x} - 2e^{-x} + C = -e^{-x}(x^2 + 2x + 2) + C. Check: e−x(x2+2x+2)−e−x(2x+2)=x2e−xe^{-x}(x^2 + 2x + 2) - e^{-x}(2x + 2) = x^2 e^{-x}. The sign of v=−e−xv = -e^{-x} is where marks go: it enters twice, and −(−e−x)-(-e^{-x}) becomes a PLUS in front of the new integral.

b) D column: x3x^3, 3x23x^2, 6x6x, 66, 00. I column: cos⁡2x\cos 2x, sin⁡2x2\frac{\sin 2x}{2}, −cos⁡2x4-\frac{\cos 2x}{4}, −sin⁡2x8-\frac{\sin 2x}{8}, cos⁡2x16\frac{\cos 2x}{16}. Multiply each D entry by the I entry ONE ROW BELOW, with the signs +,−,+,−+, -, +, -: +x3⋅sin⁡2x2+x^3 \cdot \frac{\sin 2x}{2}, −3x2⋅(−cos⁡2x4)-3x^2 \cdot \left(-\frac{\cos 2x}{4}\right), +6x⋅(−sin⁡2x8)+6x \cdot \left(-\frac{\sin 2x}{8}\right), −6⋅cos⁡2x16-6 \cdot \frac{\cos 2x}{16}. So ∫x3cos⁡2x dx=x3sin⁡2x2+3x2cos⁡2x4−3xsin⁡2x4−3cos⁡2x8+C\int x^3 \cos 2x\,dx = \frac{x^3 \sin 2x}{2} + \frac{3x^2 \cos 2x}{4} - \frac{3x \sin 2x}{4} - \frac{3\cos 2x}{8} + C. Three integrations by parts would have given the same four terms in three times the space. The diagonal pattern, D in row kk with I in row k+1k + 1, is the part to get right: pairing each D with the I of the SAME row gives the product x3cos⁡2xx^3 \cos 2x back, which is the integrand, not its antiderivative.

c) Table: D x2x^2, 2x2x, 22, 00; I cos⁡x\cos x, sin⁡x\sin x, −cos⁡x-\cos x, −sin⁡x-\sin x. Antiderivative: x2sin⁡x+2xcos⁡x−2sin⁡xx^2 \sin x + 2x \cos x - 2\sin x. At π\pi: 0+2π(−1)−0=−2π0 + 2\pi(-1) - 0 = -2\pi; at 00: 00. So ∫0πx2cos⁡x dx=−2π\int_0^{\pi} x^2 \cos x\,dx = -2\pi. The sign is right: cos⁡x>0\cos x > 0 on (0,π2)(0, \frac{\pi}{2}) and <0< 0 on (π2,π)(\frac{\pi}{2}, \pi), but the weight x2x^2 is much larger on the second half, so the negative part wins. A positive answer here would be a signal to recheck the table's signs, not a result to report.

d) For ∫x2ln⁡x dx\int x^2 \ln x\,dx, putting ln⁡x\ln x in the D column gives 1x\frac{1}{x}, −1x2-\frac{1}{x^2}, 2x3\frac{2}{x^3}, and so on: it never reaches 00. Putting x2x^2 in D instead requires antiderivatives of ln⁡x\ln x, then of xln⁡x−xx \ln x - x, and so on: worse at every row. The right move is ONE trade: u=ln⁡xu = \ln x, dv=x2 dxdv = x^2\,dx, v=x33v = \frac{x^3}{3}, giving x3ln⁡x3−∫x23 dx=x3ln⁡x3−x39+C\frac{x^3 \ln x}{3} - \int \frac{x^2}{3}\,dx = \frac{x^3 \ln x}{3} - \frac{x^3}{9} + C. For ∫exsin⁡x dx\int e^x \sin x\,dx, the derivatives of either factor cycle forever (sin⁡\sin, cos⁡\cos, −sin⁡-\sin, …\dots or exe^x, exe^x, …\dots): no column reaches 00. That integral is solved by making the original integral come back, which is the next exercise. The table is a tool for ONE form only: polynomial times a function you can integrate repeatedly.

e) ddx[−e−x(x2−2x+2)]=e−x(x2−2x+2)−e−x(2x−2)=e−x(x2−4x+4)=e−x(x−2)2\frac{d}{dx}\left[-e^{-x}(x^2 - 2x + 2)\right] = e^{-x}(x^2 - 2x + 2) - e^{-x}(2x - 2) = e^{-x}(x^2 - 4x + 4) = e^{-x}(x - 2)^2, which is not x2e−xx^2 e^{-x}. With the D column x2x^2, 2x2x, 22 and the I column e−xe^{-x}, −e−x-e^{-x}, e−xe^{-x}, −e−x-e^{-x}, his signs give x2(−e−x)+2x(e−x)+2(−e−x)x^2(-e^{-x}) + 2x(e^{-x}) + 2(-e^{-x}); the correct alternation +,−,++, -, + flips the middle term to −2xe−x-2x e^{-x}, which yields the answer of a). The minus signs of the table are the minus of −∫v du-\int v\,du, applied once per trade, so they cannot all be dropped. The derivative check located the error without redoing the table.

Exercise 4: Cyclic parts: the integral that comes back

For ∫exsin⁡x dx\int e^x \sin x\,dx no factor simplifies when differentiated: sin⁡\sin becomes cos⁡\cos, then −sin⁡-\sin, and exe^x stays exe^x. Two integrations by parts bring back the ORIGINAL integral, with a coefficient. You then stop integrating and solve for it, like an unknown in an equation.

The figure shows the graph of y=exsin⁡xy = e^x \sin x on [0,π][0, \pi] and the region AA under it.

0.511.522.533.5-112345678Ay = e^x sin xπx
  • a) Let I=∫exsin⁡x dxI = \int e^x \sin x\,dx. Integrate by parts twice with u=sin⁡xu = \sin x, then u=cos⁡xu = \cos x, and solve for II.
  • b) A student does the first step as in a), then in the second step takes u=exu = e^x and dv=cos⁡x dxdv = \cos x\,dx. Show what he obtains, and state the rule that avoids it.
  • c) Compute the area of the region AA exactly, and check its size against the figure.
  • d) Compute ∫e2xcos⁡3x dx\int e^{2x} \cos 3x\,dx.
  • e) Compute ∫sin⁡(ln⁡x) dx\int \sin(\ln x)\,dx, for x>0x > 0, with dv=dxdv = dx in both steps.
Show the solution

Answers

  • a) ∫exsin⁡x dx=ex(sin⁡x−cos⁡x)2+C\int e^x \sin x\,dx = \frac{e^x(\sin x - \cos x)}{2} + C
  • b) He gets I=II = I, true and empty; keep the SAME type of function as uu in both steps.
  • c) A=eπ+12A = \frac{e^{\pi} + 1}{2}, more than 1010.
  • d) ∫e2xcos⁡3x dx=e2x(2cos⁡3x+3sin⁡3x)13+C\int e^{2x} \cos 3x\,dx = \frac{e^{2x}(2\cos 3x + 3\sin 3x)}{13} + C
  • e) ∫sin⁡(ln⁡x) dx=x(sin⁡(ln⁡x)−cos⁡(ln⁡x))2+C\int \sin(\ln x)\,dx = \frac{x\left(\sin(\ln x) - \cos(\ln x)\right)}{2} + C

a) First trade: u=sin⁡xu = \sin x, du=cos⁡x dxdu = \cos x\,dx; dv=ex dxdv = e^x\,dx, v=exv = e^x. So I=exsin⁡x−∫excos⁡x dxI = e^x \sin x - \int e^x \cos x\,dx. Second trade, on the new integral: u=cos⁡xu = \cos x, du=−sin⁡x dxdu = -\sin x\,dx; dv=ex dxdv = e^x\,dx, v=exv = e^x. So ∫excos⁡x dx=excos⁡x+∫exsin⁡x dx=excos⁡x+I\int e^x \cos x\,dx = e^x \cos x + \int e^x \sin x\,dx = e^x \cos x + I. Substituting: I=exsin⁡x−excos⁡x−II = e^x \sin x - e^x \cos x - I. The integral came back with coefficient −1-1, so move it to the left: 2I=ex(sin⁡x−cos⁡x)2I = e^x(\sin x - \cos x), and I=ex(sin⁡x−cos⁡x)2+CI = \frac{e^x(\sin x - \cos x)}{2} + C. The constant is added at the END, after dividing: writing +C+C inside the equation and then halving it changes nothing, but forgetting to divide the WHOLE right side by 22 is the classic loss. Check: 12[ex(sin⁡x−cos⁡x)+ex(cos⁡x+sin⁡x)]=exsin⁡x\frac{1}{2}\left[e^x(\sin x - \cos x) + e^x(\cos x + \sin x)\right] = e^x \sin x.

b) Second step with u=exu = e^x, du=ex dxdu = e^x\,dx; dv=cos⁡x dxdv = \cos x\,dx, v=sin⁡xv = \sin x: ∫excos⁡x dx=exsin⁡x−∫exsin⁡x dx=exsin⁡x−I\int e^x \cos x\,dx = e^x \sin x - \int e^x \sin x\,dx = e^x \sin x - I. Substituting into the first step: I=exsin⁡x−(exsin⁡x−I)=II = e^x \sin x - (e^x \sin x - I) = I. The equation is true and says nothing: the second trade simply UNDID the first, handing back the integral it had produced. The rule: in a cyclic integral, whichever type you chose as uu in the first step (here the trigonometric factor), keep the same type as uu in the second. Both consistent choices work, uu trigonometric twice or uu exponential twice; only the mixed one goes in a circle.

c) ex>0e^x > 0 and sin⁡x≥0\sin x \ge 0 on [0,π][0, \pi], so A=∫0πexsin⁡x dx=[ex(sin⁡x−cos⁡x)2]0π=eπ(0+1)2−1⋅(0−1)2=eπ+12A = \int_0^{\pi} e^x \sin x\,dx = \left[\frac{e^x(\sin x - \cos x)}{2}\right]_0^{\pi} = \frac{e^{\pi}(0 + 1)}{2} - \frac{1 \cdot (0 - 1)}{2} = \frac{e^{\pi} + 1}{2}. At the upper bound cos⁡π=−1\cos \pi = -1 turns −cos⁡x-\cos x into +1+1, and at the lower bound −cos⁡0=−1-\cos 0 = -1, SUBTRACTED, gives the second +12+\frac{1}{2}. Size, without a calculator: e3≈20e^3 \approx 20 (since e3=e⋅e2≈2.72⋅7.4e^3 = e \cdot e^2 \approx 2.72 \cdot 7.4), so eπ>20e^{\pi} > 20 and A>212A > \frac{21}{2}, more than 1010. On the figure, the curve rises to about 7.57.5 near x=3π4x = \frac{3\pi}{4} over a base of length π\pi; a triangle with that height and base has area about 11.711.7, the same order. A value like eπ−12\frac{e^{\pi} - 1}{2}, from a sign slip at x=0x = 0, would be within the same range, which is why the bracket must be written out at BOTH bounds.

d) Let J=∫e2xcos⁡3x dxJ = \int e^{2x} \cos 3x\,dx. First trade: u=cos⁡3xu = \cos 3x, du=−3sin⁡3x dxdu = -3\sin 3x\,dx; dv=e2x dxdv = e^{2x}\,dx, v=e2x2v = \frac{e^{2x}}{2}: J=e2xcos⁡3x2+32∫e2xsin⁡3x dxJ = \frac{e^{2x}\cos 3x}{2} + \frac{3}{2}\int e^{2x} \sin 3x\,dx. Second trade, SAME type as uu: u=sin⁡3xu = \sin 3x, du=3cos⁡3x dxdu = 3\cos 3x\,dx, v=e2x2v = \frac{e^{2x}}{2}: ∫e2xsin⁡3x dx=e2xsin⁡3x2−32J\int e^{2x} \sin 3x\,dx = \frac{e^{2x}\sin 3x}{2} - \frac{3}{2}J. So J=e2xcos⁡3x2+3e2xsin⁡3x4−94JJ = \frac{e^{2x}\cos 3x}{2} + \frac{3e^{2x}\sin 3x}{4} - \frac{9}{4}J, that is 134J=e2x(2cos⁡3x+3sin⁡3x)4\frac{13}{4}J = \frac{e^{2x}(2\cos 3x + 3\sin 3x)}{4} and J=e2x(2cos⁡3x+3sin⁡3x)13+CJ = \frac{e^{2x}(2\cos 3x + 3\sin 3x)}{13} + C. The coefficient 1+94=1341 + \frac{9}{4} = \frac{13}{4} is where the work is: the number 13=22+3213 = 2^2 + 3^2 always appears for eaxe^{ax} with cos⁡bx\cos bx, which is a free check on the arithmetic.

e) Let K=∫sin⁡(ln⁡x) dxK = \int \sin(\ln x)\,dx. First trade: u=sin⁡(ln⁡x)u = \sin(\ln x), du=cos⁡(ln⁡x)x dxdu = \frac{\cos(\ln x)}{x}\,dx; dv=dxdv = dx, v=xv = x: K=xsin⁡(ln⁡x)−∫cos⁡(ln⁡x) dxK = x\sin(\ln x) - \int \cos(\ln x)\,dx, the xx of vv cancelling the 1x\frac{1}{x} of the chain rule. Second trade, same type: u=cos⁡(ln⁡x)u = \cos(\ln x), du=−sin⁡(ln⁡x)x dxdu = -\frac{\sin(\ln x)}{x}\,dx, v=xv = x: ∫cos⁡(ln⁡x) dx=xcos⁡(ln⁡x)+K\int \cos(\ln x)\,dx = x\cos(\ln x) + K. Hence K=xsin⁡(ln⁡x)−xcos⁡(ln⁡x)−KK = x\sin(\ln x) - x\cos(\ln x) - K and K=x(sin⁡(ln⁡x)−cos⁡(ln⁡x))2+CK = \frac{x\left(\sin(\ln x) - \cos(\ln x)\right)}{2} + C. This is the lone-factor gesture of exercise 2 and the cyclic gesture of this exercise at once: nothing in the integrand looks like a product, yet parts with dv=dxdv = dx produces the cycle.

Exercise 5: Parts on a definite integral: the bracket is evaluated at both bounds

On a definite integral the trade reads ∫abu dv=[uv]ab−∫abv du\int_a^b u\,dv = \left[uv\right]_a^b - \int_a^b v\,du. The bracket is a NUMBER, u(b)v(b)−u(a)v(a)u(b)v(b) - u(a)v(a), and the new integral keeps the same bounds. The two halves are evaluated separately, and each one at both ends.

The figure shows the graph of y=arcsin⁡xy = \arcsin x for 0≤x≤120 \le x \le \frac{1}{2}, the region AA under it and the region BB between the curve and the yy-axis, inside the rectangle of corners (0,0)(0, 0) and (12,π6)(\frac{1}{2}, \frac{\pi}{6}).

0.10.20.30.40.50.60.70.80.90.10.20.30.40.50.60.7AB(1/2, π/6)y = arcsin x
  • a) Compute ∫1exln⁡x dx\int_1^e x \ln x\,dx.
  • b) Compute ∫0π/2xcos⁡x dx\int_0^{\pi/2} x \cos x\,dx.
  • c) Compute ∫0πxsin⁡x dx\int_0^{\pi} x \sin x\,dx. Which half of the formula carries the value here?
  • d) Using the figure, express the area of BB as an integral in yy, then recover ∫01/2arcsin⁡x dx\int_0^{1/2} \arcsin x\,dx from A+BA + B without integrating by parts. Which term of the parts formula is the rectangle?
  • e) A student finds ∫12ln⁡x dx=2ln⁡2−2\int_1^2 \ln x\,dx = 2\ln 2 - 2. Find the correct value and explain how a size check rules out his answer before any recomputation.
Show the solution

Answers

  • a) ∫1exln⁡x dx=e2+14\int_1^e x \ln x\,dx = \frac{e^2 + 1}{4}
  • b) ∫0π/2xcos⁡x dx=π2−1\int_0^{\pi/2} x \cos x\,dx = \frac{\pi}{2} - 1
  • c) ∫0πxsin⁡x dx=π\int_0^{\pi} x \sin x\,dx = \pi, all of it from the bracket.
  • d) B=∫0π/6sin⁡y dy=1−32B = \int_0^{\pi/6} \sin y\,dy = 1 - \frac{\sqrt 3}{2}, so A=π12+32−1A = \frac{\pi}{12} + \frac{\sqrt 3}{2} - 1; the rectangle is [xarcsin⁡x]01/2\left[x \arcsin x\right]_0^{1/2}.
  • e) ∫12ln⁡x dx=2ln⁡2−1\int_1^2 \ln x\,dx = 2\ln 2 - 1; his value is negative for a positive integrand.

a) u=ln⁡xu = \ln x, du=dxxdu = \frac{dx}{x}; dv=x dxdv = x\,dx, v=x22v = \frac{x^2}{2}. Then ∫1exln⁡x dx=[x2ln⁡x2]1e−∫1ex2 dx\int_1^e x \ln x\,dx = \left[\frac{x^2 \ln x}{2}\right]_1^e - \int_1^e \frac{x}{2}\,dx. The bracket is e2⋅12−1⋅02=e22\frac{e^2 \cdot 1}{2} - \frac{1 \cdot 0}{2} = \frac{e^2}{2}. The new integral is [x24]1e=e2−14\left[\frac{x^2}{4}\right]_1^e = \frac{e^2 - 1}{4}. So the value is e22−e2−14=e2+14\frac{e^2}{2} - \frac{e^2 - 1}{4} = \frac{e^2 + 1}{4}. With e2≈7.4e^2 \approx 7.4 that is about 2.12.1; the integrand lies between 00 and ee on an interval of length e−1≈1.7e - 1 \approx 1.7, so the value must be below e(e−1)≈4.7e(e - 1) \approx 4.7, which it is. The lower bound contributed 00 to the bracket because ln⁡1=0\ln 1 = 0, but NOT to the new integral, where 14\frac{1}{4} survives: that 14\frac{1}{4} is exactly what gets lost when the lower bound is evaluated only once.

b) u=xu = x, du=dxdu = dx; dv=cos⁡x dxdv = \cos x\,dx, v=sin⁡xv = \sin x. Then ∫0π/2xcos⁡x dx=[xsin⁡x]0π/2−∫0π/2sin⁡x dx=π2⋅1−0−[−cos⁡x]0π/2=π2−(0+1)=π2−1\int_0^{\pi/2} x \cos x\,dx = \left[x \sin x\right]_0^{\pi/2} - \int_0^{\pi/2} \sin x\,dx = \frac{\pi}{2} \cdot 1 - 0 - \left[-\cos x\right]_0^{\pi/2} = \frac{\pi}{2} - (0 + 1) = \frac{\pi}{2} - 1. With π≈3.14\pi \approx 3.14, about 0.570.57: positive, as it must be since both factors are non-negative on [0,π2][0, \frac{\pi}{2}]. The step [−cos⁡x]0π/2=−cos⁡π2+cos⁡0=1\left[-\cos x\right]_0^{\pi/2} = -\cos\frac{\pi}{2} + \cos 0 = 1 carries two signs, and writing it on its own line is how they survive.

c) u=xu = x, du=dxdu = dx; dv=sin⁡x dxdv = \sin x\,dx, v=−cos⁡xv = -\cos x. Then ∫0πxsin⁡x dx=[−xcos⁡x]0π+∫0πcos⁡x dx=(−π⋅(−1)−0)+[sin⁡x]0π=π+0=π\int_0^{\pi} x \sin x\,dx = \left[-x \cos x\right]_0^{\pi} + \int_0^{\pi} \cos x\,dx = (-\pi \cdot (-1) - 0) + \left[\sin x\right]_0^{\pi} = \pi + 0 = \pi. Here the new integral contributes NOTHING and the whole value is the bracket, the opposite of part b), where the bracket was π2\frac{\pi}{2} and the new integral took 11 away. Neither half can be skipped in advance: a student who assumes the bracket vanishes (it often does at x=0x = 0) would answer 00 here, for the area of a region that is visibly positive.

d) Slice BB horizontally: at height yy, with 0≤y≤π60 \le y \le \frac{\pi}{6}, the region runs from the yy-axis to the curve, where y=arcsin⁡xy = \arcsin x, that is x=sin⁡yx = \sin y. So B=∫0π/6sin⁡y dy=[−cos⁡y]0π/6=1−32B = \int_0^{\pi/6} \sin y\,dy = \left[-\cos y\right]_0^{\pi/6} = 1 - \frac{\sqrt 3}{2}. The two regions fill the rectangle: A+B=12⋅π6=π12A + B = \frac{1}{2} \cdot \frac{\pi}{6} = \frac{\pi}{12}. Hence A=∫01/2arcsin⁡x dx=π12−1+32A = \int_0^{1/2} \arcsin x\,dx = \frac{\pi}{12} - 1 + \frac{\sqrt 3}{2}, the value found by parts in exercise 2. The rectangle is the bracket [xarcsin⁡x]01/2\left[x \arcsin x\right]_0^{1/2}, and BB is the new integral ∫x du\int x\,du with u=arcsin⁡xu = \arcsin x, measured along the vertical axis. That is what integration by parts says geometrically: ∫u dv+∫v du\int u\,dv + \int v\,du is the difference of two rectangles, u(b)v(b)−u(a)v(a)u(b)v(b) - u(a)v(a).

e) ∫12ln⁡x dx=[xln⁡x−x]12=(2ln⁡2−2)−(1⋅0−1)=2ln⁡2−1\int_1^2 \ln x\,dx = \left[x \ln x - x\right]_1^2 = (2\ln 2 - 2) - (1 \cdot 0 - 1) = 2\ln 2 - 1. The student evaluated the antiderivative at the upper bound only, reasoning that ln⁡1=0\ln 1 = 0 kills the lower bound; it kills xln⁡xx \ln x but not −x-x, which contributes +1+1. Size check: on [1,2][1, 2], 0≤ln⁡x≤ln⁡20 \le \ln x \le \ln 2, so the integral lies between 00 and ln⁡2≈0.69\ln 2 \approx 0.69. His value 2ln⁡2−2≈−0.612\ln 2 - 2 \approx -0.61 is NEGATIVE for a non-negative integrand: impossible, and visible in five seconds. The correct 2ln⁡2−1≈0.392\ln 2 - 1 \approx 0.39 lies in the window.

Part B: problems and reasoning (/50)

Exercise 6: A reduction formula: the trade written once for every n

When the same integration by parts would be repeated nn times, it is written ONCE with a letter, and the result links the integral of index nn to the one of index n−1n - 1: a reduction formula. For n≥0n \ge 0, let In=∫01xnex dxI_n = \int_0^1 x^n e^x\,dx.

The figure shows the four integrands xnexx^n e^x for n=0,1,2,3n = 0, 1, 2, 3 on [0,1][0, 1]; from top to bottom, the curves are in the same order as the list on the left. They all meet at (1,e)(1, e).

0.250.50.7511.250.511.522.53n = 0n = 1n = 2n = 3(1, e)x
  • a) Prove that In=e−nIn−1I_n = e - n I_{n-1} for every n≥1n \ge 1. Where is the hypothesis n≥1n \ge 1 used?
  • b) Compute I0I_0, then I1I_1, I2I_2 and I3I_3 exactly.
  • c) Prove that 1n+1≤In≤en+1\frac{1}{n + 1} \le I_n \le \frac{e}{n + 1}, and check I2I_2 and I3I_3 against these bounds. What does the figure say about the order of the InI_n?
  • d) Let Jn=∫1e(ln⁡x)n dxJ_n = \int_1^e (\ln x)^n\,dx. Prove that Jn=e−nJn−1J_n = e - n J_{n-1} for n≥1n \ge 1, and deduce J2J_2 and J3J_3 without any new integration. Explain the coincidence with a substitution.
  • e) A student starts the recursion with I0=eI_0 = e. Compute what he obtains for I3I_3, and explain how an error in I0I_0 grows along the recursion.
Show the solution

Answers

  • a) u=xnu = x^n, dv=ex dxdv = e^x\,dx: In=[xnex]01−nIn−1=e−nIn−1I_n = \left[x^n e^x\right]_0^1 - n I_{n-1} = e - n I_{n-1}; n≥1n \ge 1 makes 0n=00^n = 0.
  • b) I0=e−1I_0 = e - 1, I1=1I_1 = 1, I2=e−2I_2 = e - 2, I3=6−2eI_3 = 6 - 2e
  • c) 1≤ex≤e1 \le e^x \le e on [0,1][0, 1]; 13≤e−2≤e3\frac{1}{3} \le e - 2 \le \frac{e}{3} and 14≤6−2e≤e4\frac{1}{4} \le 6 - 2e \le \frac{e}{4}; I0>I1>I2>I3I_0 > I_1 > I_2 > I_3.
  • d) Jn=InJ_n = I_n for all nn: J2=e−2J_2 = e - 2, J3=6−2eJ_3 = 6 - 2e (substitution x=etx = e^t).
  • e) He gets I3=−2e<0I_3 = -2e < 0; an error ε\varepsilon in I0I_0 becomes (−1)nn! ε(-1)^n n!\,\varepsilon in InI_n.

a) Integrate by parts with u=xnu = x^n, du=nxn−1 dxdu = n x^{n-1}\,dx; dv=ex dxdv = e^x\,dx, v=exv = e^x. Then In=[xnex]01−∫01nxn−1ex dx=(1ne1−0ne0)−nIn−1I_n = \left[x^n e^x\right]_0^1 - \int_0^1 n x^{n-1} e^x\,dx = (1^n e^1 - 0^n e^0) - n I_{n-1}. For n≥1n \ge 1, 0n=00^n = 0, so the bracket is ee and In=e−nIn−1I_n = e - n I_{n-1}. The hypothesis is used exactly there: for n=0n = 0 the lower bound would give 00e0=10^0 e^0 = 1, and the formula would not hold (it would not even make sense, since I−1I_{-1} is not defined). The choice u=xnu = x^n is the one of exercise 1, and it is right for the same reason: the power goes DOWN, from nn to n−1n - 1, which is what makes the formula a reduction.

b) I0=∫01ex dx=e−1I_0 = \int_0^1 e^x\,dx = e - 1. Then I1=e−1⋅I0=e−(e−1)=1I_1 = e - 1 \cdot I_0 = e - (e - 1) = 1, I2=e−2I1=e−2I_2 = e - 2I_1 = e - 2, I3=e−3I2=e−3(e−2)=6−2eI_3 = e - 3I_2 = e - 3(e - 2) = 6 - 2e. Each value costs one line; computing I3I_3 by the tabular method would give the antiderivative ex(x3−3x2+6x−6)e^x(x^3 - 3x^2 + 6x - 6), whose value at 11 minus its value at 00 is −2e+6-2e + 6, the same number. Size: with e≈2.72e \approx 2.72, I3≈6−5.44=0.56I_3 \approx 6 - 5.44 = 0.56.

c) On [0,1][0, 1], 1≤ex≤e1 \le e^x \le e, so xn≤xnex≤e xnx^n \le x^n e^x \le e\,x^n, and integrating (the comparison property of the definite integral) gives 1n+1≤In≤en+1\frac{1}{n + 1} \le I_n \le \frac{e}{n + 1}, since ∫01xn dx=1n+1\int_0^1 x^n\,dx = \frac{1}{n + 1}. For n=2n = 2: 13≈0.33≤e−2≈0.72≤e3≈0.91\frac{1}{3} \approx 0.33 \le e - 2 \approx 0.72 \le \frac{e}{3} \approx 0.91. For n=3n = 3: 14=0.25≤6−2e≈0.56≤e4≈0.68\frac{1}{4} = 0.25 \le 6 - 2e \approx 0.56 \le \frac{e}{4} \approx 0.68. Both hold. On the figure, for 0<x<10 < x < 1, xn+1<xnx^{n+1} < x^n, so each curve lies BELOW the previous one: the areas decrease, I0>I1>I2>I3>0I_0 > I_1 > I_2 > I_3 > 0, and the bounds show they tend to 00. These inequalities are the check that the recursion was run correctly, and they cost nothing.

d) u=(ln⁡x)nu = (\ln x)^n, du=n(ln⁡x)n−1⋅1x dxdu = n(\ln x)^{n-1} \cdot \frac{1}{x}\,dx; dv=dxdv = dx, v=xv = x. Then Jn=[x(ln⁡x)n]1e−∫1ex⋅n(ln⁡x)n−1⋅1x dx=(e⋅1−1⋅0)−nJn−1=e−nJn−1J_n = \left[x (\ln x)^n\right]_1^e - \int_1^e x \cdot n(\ln x)^{n-1} \cdot \frac{1}{x}\,dx = (e \cdot 1 - 1 \cdot 0) - n J_{n-1} = e - n J_{n-1}, where (ln⁡1)n=0(\ln 1)^n = 0 uses n≥1n \ge 1 again. The start is J0=∫1edx=e−1=I0J_0 = \int_1^e dx = e - 1 = I_0. Same recursion, same first term: Jn=InJ_n = I_n for every nn, so J2=e−2J_2 = e - 2 and J3=6−2eJ_3 = 6 - 2e with no new integration. The coincidence is a substitution: x=etx = e^t, dx=et dtdx = e^t\,dt, with tt from 00 to 11, turns JnJ_n into ∫01tnet dt=In\int_0^1 t^n e^t\,dt = I_n. The lone-logarithm trade of exercise 2 and the power-lowering trade of exercise 1 are the SAME trade, seen in two variables.

e) With I0=eI_0 = e: I1=e−e=0I_1 = e - e = 0, I2=e−0=eI_2 = e - 0 = e, I3=e−3e=−2eI_3 = e - 3e = -2e. A negative value for the integral of a positive function: impossible, and part c) catches it at once. The mechanism: if the value used for In−1I_{n-1} is off by ε\varepsilon, then In=e−nIn−1I_n = e - nI_{n-1} is off by −nε-n\varepsilon. Starting from an error ε=1\varepsilon = 1 in I0I_0, the errors are −1-1, +2+2, −6-6 in I1I_1, I2I_2, I3I_3: the error is multiplied by (−1)nn!(-1)^n n!. A reduction formula is exact, but it AMPLIFIES any slip in the first term, so I0I_0 is the value to double-check.

Exercise 7: Parts when the function is unknown: moving the derivative

Integration by parts does not need a formula for the functions involved. It moves a derivative from one factor to the other, and the bracket only needs VALUES at the bounds. This is how exam questions test the method on functions known only through a table.

A function ff has a continuous second derivative on [1,4][1, 4], and f(1)=2f(1) = 2, f(4)=7f(4) = 7, f′(1)=5f'(1) = 5, f′(4)=3f'(4) = 3. Moreover ∫14f(x) dx=12\int_1^4 f(x)\,dx = 12.

  • a) Compute ∫14xf′′(x) dx\int_1^4 x f''(x)\,dx.
  • b) Explain why the choice u=f′′(x)u = f''(x), dv=x dxdv = x\,dx is not available here.
  • c) Compute ∫14xf′(x) dx\int_1^4 x f'(x)\,dx.
  • d) Let F(x)=∫x1e−t2 dtF(x) = \int_x^1 e^{-t^2}\,dt. Compute ∫01F(x) dx\int_0^1 F(x)\,dx exactly, by integrating by parts with u=F(x)u = F(x).
  • e) A student gets −1−e−12-\frac{1 - e^{-1}}{2} in d). Locate the error, and give bounds for ∫01F(x) dx\int_0^1 F(x)\,dx that rule out his sign.
Show the solution

Answers

  • a) ∫14xf′′(x) dx=[xf′(x)]14−(f(4)−f(1))=7−5=2\int_1^4 x f''(x)\,dx = \left[x f'(x)\right]_1^4 - (f(4) - f(1)) = 7 - 5 = 2
  • b) It needs f′′′f''', which is not given and not even assumed to exist, and it raises the power of xx.
  • c) ∫14xf′(x) dx=28−2−12=14\int_1^4 x f'(x)\,dx = 28 - 2 - 12 = 14
  • d) ∫01F(x) dx=1−e−12\int_0^1 F(x)\,dx = \frac{1 - e^{-1}}{2}
  • e) He wrote F′(x)=+e−x2F'(x) = +e^{-x^2}; the value lies between e−12\frac{e^{-1}}{2} and 12\frac{1}{2}, so it is positive.

a) u=xu = x, du=dxdu = dx; dv=f′′(x) dxdv = f''(x)\,dx, v=f′(x)v = f'(x). Then ∫14xf′′(x) dx=[xf′(x)]14−∫14f′(x) dx\int_1^4 x f''(x)\,dx = \left[x f'(x)\right]_1^4 - \int_1^4 f'(x)\,dx. The bracket needs only values: 4f′(4)−1⋅f′(1)=12−5=74f'(4) - 1 \cdot f'(1) = 12 - 5 = 7. The new integral is a net change (Fundamental Theorem): ∫14f′(x) dx=f(4)−f(1)=7−2=5\int_1^4 f'(x)\,dx = f(4) - f(1) = 7 - 2 = 5. So the value is 7−5=27 - 5 = 2. The derivative has moved from f′′f'' onto the factor xx, where it became 11, and what remained, f′f', has an antiderivative we can evaluate: ff.

b) With u=f′′(x)u = f''(x) we would need du=f′′′(x) dxdu = f'''(x)\,dx, and the statement only guarantees that f′′f'' is continuous: f′′′f''' may not exist, and even if it did, nothing is known about it. The new integral would be ∫14x22f′′′(x) dx\int_1^4 \frac{x^2}{2} f'''(x)\,dx, a higher power AND a higher derivative, worse on both counts. The general principle of the chapter applies to unknown functions too: put the derivative where it simplifies (xx becomes 11) and take as dvdv what you can integrate (f′′f'' integrates to f′f').

c) u=xu = x, du=dxdu = dx; dv=f′(x) dxdv = f'(x)\,dx, v=f(x)v = f(x). Then ∫14xf′(x) dx=[xf(x)]14−∫14f(x) dx=(4⋅7−1⋅2)−12=28−2−12=14\int_1^4 x f'(x)\,dx = \left[x f(x)\right]_1^4 - \int_1^4 f(x)\,dx = (4 \cdot 7 - 1 \cdot 2) - 12 = 28 - 2 - 12 = 14. This time the given value of ∫14f(x) dx\int_1^4 f(x)\,dx is indispensable: without it the question could not be answered, which is why it is in the statement. The data are consistent: a polynomial of degree 44 satisfies all five conditions, so the numbers describe a real function.

d) u=F(x)u = F(x), du=F′(x) dxdu = F'(x)\,dx; dv=dxdv = dx, v=xv = x. By the Fundamental Theorem, part 1, with the variable at the LOWER bound, F′(x)=−e−x2F'(x) = -e^{-x^2}. So ∫01F(x) dx=[xF(x)]01−∫01x⋅(−e−x2) dx\int_0^1 F(x)\,dx = \left[x F(x)\right]_0^1 - \int_0^1 x \cdot (-e^{-x^2})\,dx. The bracket vanishes at both ends: F(1)=∫11e−t2 dt=0F(1) = \int_1^1 e^{-t^2}\,dt = 0, and at x=0x = 0 the factor xx is 00. What remains is ∫01xe−x2 dx\int_0^1 x e^{-x^2}\,dx, a substitution (w=x2w = x^2, dw=2x dxdw = 2x\,dx): [−12e−x2]01=1−e−12\left[-\frac{1}{2}e^{-x^2}\right]_0^1 = \frac{1 - e^{-1}}{2}. The integrand e−t2e^{-t^2} has no elementary antiderivative, so FF has no formula; parts sidesteps that entirely, because it only uses F′F', which the Fundamental Theorem gives for free. That is the lone-factor gesture of exercise 2 again, with dv=dxdv = dx.

e) His value is the correct one with the opposite sign: he took F′(x)=e−x2F'(x) = e^{-x^2}, forgetting that the variable is the LOWER bound, which contributes a minus sign. Bounds without computing: for 0≤t≤10 \le t \le 1, e−1≤e−t2≤1e^{-1} \le e^{-t^2} \le 1, so for 0≤x≤10 \le x \le 1, e−1(1−x)≤F(x)≤1−xe^{-1}(1 - x) \le F(x) \le 1 - x. Integrating over [0,1][0, 1]: e−12≤∫01F(x) dx≤12\frac{e^{-1}}{2} \le \int_0^1 F(x)\,dx \le \frac{1}{2}. The integral is positive, about between 0.180.18 and 0.50.5, and 1−e−12≈1−0.372≈0.32\frac{1 - e^{-1}}{2} \approx \frac{1 - 0.37}{2} \approx 0.32 sits inside. A negative answer for the integral of a function that is ≥0\ge 0 is always a sign error, and here it points straight at the derivative of an integral with a variable lower bound.

Exercise 8: Five statements to correct

Each statement below was written on a real exam paper. Say whether it is true or false; when it is false, find the error, correct it, and prove the correction with a derivative or a number.

Two marks per statement: one for locating the error, one for the corrected result with its check.

  • a) ∫xex dx=(∫x dx)(∫ex dx)=x22ex+C\int x e^x\,dx = \left(\int x\,dx\right)\left(\int e^x\,dx\right) = \frac{x^2}{2}e^x + C.
  • b) ∫xsin⁡x dx=−xcos⁡x−sin⁡x+C\int x \sin x\,dx = -x\cos x - \sin x + C.
  • c) By the tabular method, ∫x2e3x dx=e3x(x23+2x9+227)+C\int x^2 e^{3x}\,dx = e^{3x}\left(\frac{x^2}{3} + \frac{2x}{9} + \frac{2}{27}\right) + C.
  • d) ∫01xex dx=[xex]01−∫ex dx=e−ex+C\int_0^1 x e^x\,dx = \left[x e^x\right]_0^1 - \int e^x\,dx = e - e^x + C.
  • e) When vv is computed from dvdv, its constant must be taken equal to 00, otherwise integration by parts gives a wrong result.
Show the solution

Answers

  • a) False: ∫xex dx=(x−1)ex+C\int x e^x\,dx = (x - 1)e^x + C.
  • b) False: ∫xsin⁡x dx=−xcos⁡x+sin⁡x+C\int x \sin x\,dx = -x\cos x + \sin x + C.
  • c) False: e3x(x23−2x9+227)+Ce^{3x}\left(\frac{x^2}{3} - \frac{2x}{9} + \frac{2}{27}\right) + C.
  • d) False: ∫01xex dx=e−(e−1)=1\int_0^1 x e^x\,dx = e - (e - 1) = 1.
  • e) False: any constant works; v=x+1v = x + 1 gives ∫ln⁡(x+1) dx=(x+1)ln⁡(x+1)−x+C\int \ln(x + 1)\,dx = (x + 1)\ln(x + 1) - x + C in one line.

a) False. There is no product rule for integrals: the integral of a product is not the product of the integrals. The derivative of the proposed answer is xex+x22ex≠xexx e^x + \frac{x^2}{2}e^x \ne x e^x. Correct method: u=xu = x, dv=ex dxdv = e^x\,dx, v=exv = e^x, so ∫xex dx=xex−∫ex dx=(x−1)ex+C\int x e^x\,dx = x e^x - \int e^x\,dx = (x - 1)e^x + C; check, ex+(x−1)ex=xexe^x + (x - 1)e^x = x e^x. On [0,1][0, 1] the difference is visible in numbers: the correct integral is 11, while (∫01x dx)(∫01ex dx)=e−12≈0.86\left(\int_0^1 x\,dx\right)\left(\int_0^1 e^x\,dx\right) = \frac{e - 1}{2} \approx 0.86. Integration by parts EXISTS precisely because the naive rule is false.

b) False, one sign. u=xu = x, dv=sin⁡x dxdv = \sin x\,dx, v=−cos⁡xv = -\cos x: ∫xsin⁡x dx=−xcos⁡x−∫(−cos⁡x) dx=−xcos⁡x+∫cos⁡x dx=−xcos⁡x+sin⁡x+C\int x\sin x\,dx = -x\cos x - \int (-\cos x)\,dx = -x\cos x + \int \cos x\,dx = -x\cos x + \sin x + C. The student subtracted ∫v du\int v\,du without carrying the minus sign INSIDE vv. The derivative of his answer is −cos⁡x+xsin⁡x−cos⁡x=xsin⁡x−2cos⁡x-\cos x + x\sin x - \cos x = x\sin x - 2\cos x, and the stray −2cos⁡x-2\cos x is the trace of the lost sign; the corrected answer differentiates to −cos⁡x+xsin⁡x+cos⁡x=xsin⁡x-\cos x + x\sin x + \cos x = x\sin x. Writing v=−cos⁡xv = -\cos x in its own box, then −∫v du=+∫cos⁡x dx-\int v\,du = +\int \cos x\,dx on a separate line, prevents it.

c) False: the signs of the table must alternate. D column: x2x^2, 2x2x, 22, 00; I column: e3xe^{3x}, e3x3\frac{e^{3x}}{3}, e3x9\frac{e^{3x}}{9}, e3x27\frac{e^{3x}}{27}. With the signs +,−,++, -, +: x2e3x3−2xe3x9+2e3x27\frac{x^2 e^{3x}}{3} - \frac{2x e^{3x}}{9} + \frac{2e^{3x}}{27}, so ∫x2e3x dx=e3x(x23−2x9+227)+C\int x^2 e^{3x}\,dx = e^{3x}\left(\frac{x^2}{3} - \frac{2x}{9} + \frac{2}{27}\right) + C. Check: 3e3x(x23−2x9+227)+e3x(2x3−29)=e3x(x2−2x3+29+2x3−29)=x2e3x3e^{3x}\left(\frac{x^2}{3} - \frac{2x}{9} + \frac{2}{27}\right) + e^{3x}\left(\frac{2x}{3} - \frac{2}{9}\right) = e^{3x}\left(x^2 - \frac{2x}{3} + \frac{2}{9} + \frac{2x}{3} - \frac{2}{9}\right) = x^2 e^{3x}. The terms cancel in pairs only with the alternating signs; with all plus signs the derivative is e3x(x2+4x3+49)e^{3x}\left(x^2 + \frac{4x}{3} + \frac{4}{9}\right).

d) False: a definite integral is a number, and this answer is a function plus a constant. The bracket was evaluated, but the new integral was treated as indefinite. Correct: ∫01xex dx=[xex]01−∫01ex dx=(e−0)−(e−1)=1\int_0^1 x e^x\,dx = \left[x e^x\right]_0^1 - \int_0^1 e^x\,dx = (e - 0) - (e - 1) = 1. The bounds stay on BOTH halves of the formula. A quick sanity check: 0≤xex≤e0 \le x e^x \le e on [0,1][0, 1], so the value lies between 00 and ee, and 11 does.

e) False. If vv is replaced by v+Kv + K, the formula gives u(v+K)−∫(v+K) du=uv−∫v du+Ku−Ku+constu(v + K) - \int (v + K)\,du = uv - \int v\,du + Ku - Ku + \text{const}, the same family of antiderivatives. So any constant is allowed, and a clever one can shorten the work. For ∫ln⁡(x+1) dx\int \ln(x + 1)\,dx: u=ln⁡(x+1)u = \ln(x + 1), du=dxx+1du = \frac{dx}{x + 1}, dv=dxdv = dx, and choosing v=x+1v = x + 1 gives (x+1)ln⁡(x+1)−∫1 dx=(x+1)ln⁡(x+1)−x+C(x + 1)\ln(x + 1) - \int 1\,dx = (x + 1)\ln(x + 1) - x + C at once. With v=xv = x one gets xln⁡(x+1)−∫xx+1 dx=xln⁡(x+1)−x+ln⁡(x+1)+Cx\ln(x + 1) - \int \frac{x}{x + 1}\,dx = x\ln(x + 1) - x + \ln(x + 1) + C, after writing xx+1=1−1x+1\frac{x}{x + 1} = 1 - \frac{1}{x + 1}: the same function, one line longer. Check: ln⁡(x+1)+1−1=ln⁡(x+1)\ln(x + 1) + 1 - 1 = \ln(x + 1).

Exercise 9: A damped oscillation: how far does the mass really travel?

A mass on a spring, slowed by friction, moves along a horizontal line. Its velocity, in metres per second, is v(t)=e−tsin⁡(πt)v(t) = e^{-t}\sin(\pi t) for t≥0t \ge 0 in seconds. The factor sin⁡(πt)\sin(\pi t) makes it change direction every second; the factor e−te^{-t} damps the motion.

The figure shows vv on [0,2][0, 2]: the first shaded lobe is the first swing, the second lobe, below the axis, is the swing back. Displacement is the integral of vv; distance travelled is the integral of ∣v∣|v|.

0.511.522.5-0.4-0.20.20.40.60.8swing 1swing 2v(t) = e^(-t) sin(πt)t (s)v (m/s)
  • a) Find an antiderivative PP of vv by cyclic integration by parts.
  • b) Compute the displacement during the first second, ∫01v(t) dt\int_0^1 v(t)\,dt.
  • c) Compute the displacement during the next second, and the ratio of the two displacements.
  • d) Give the net displacement and the total distance travelled over [0,2][0, 2].
  • e) Show that the displacement during [k,k+1][k, k + 1] is (−1)ke−k(-1)^k e^{-k} times the first one. With π≈3.14\pi \approx 3.14 and e≈2.72e \approx 2.72, estimate the first swing, and find the first swing shorter than 11 cm.
Show the solution

Answers

  • a) P(t)=−e−t(sin⁡(πt)+πcos⁡(πt))1+π2P(t) = -\frac{e^{-t}\left(\sin(\pi t) + \pi\cos(\pi t)\right)}{1 + \pi^2}
  • b) ∫01v dt=π1+π2(1+e−1)\int_0^1 v\,dt = \frac{\pi}{1 + \pi^2}(1 + e^{-1}) m
  • c) ∫12v dt=−π1+π2e−1(1+e−1)\int_1^2 v\,dt = -\frac{\pi}{1 + \pi^2}e^{-1}(1 + e^{-1}) m; ratio −e−1-e^{-1}.
  • d) Net π1+π2(1−e−2)\frac{\pi}{1 + \pi^2}(1 - e^{-2}) m; distance π1+π2(1+e−1)2\frac{\pi}{1 + \pi^2}(1 + e^{-1})^2 m.
  • e) First swing about 0.400.40 m; the swing during [4,5][4, 5] is the first under 11 cm (about 0.70.7 cm).

a) Let K=∫e−tsin⁡(πt) dtK = \int e^{-t}\sin(\pi t)\,dt. First trade: u=sin⁡(πt)u = \sin(\pi t), du=πcos⁡(πt) dtdu = \pi\cos(\pi t)\,dt; dv=e−t dtdv = e^{-t}\,dt, v=−e−tv = -e^{-t}: K=−e−tsin⁡(πt)+π∫e−tcos⁡(πt) dtK = -e^{-t}\sin(\pi t) + \pi\int e^{-t}\cos(\pi t)\,dt. Second trade, SAME type as uu: u=cos⁡(πt)u = \cos(\pi t), du=−πsin⁡(πt) dtdu = -\pi\sin(\pi t)\,dt, v=−e−tv = -e^{-t}: ∫e−tcos⁡(πt) dt=−e−tcos⁡(πt)−πK\int e^{-t}\cos(\pi t)\,dt = -e^{-t}\cos(\pi t) - \pi K. So K=−e−tsin⁡(πt)−πe−tcos⁡(πt)−π2KK = -e^{-t}\sin(\pi t) - \pi e^{-t}\cos(\pi t) - \pi^2 K, that is (1+π2)K=−e−t(sin⁡(πt)+πcos⁡(πt))(1 + \pi^2)K = -e^{-t}\left(\sin(\pi t) + \pi\cos(\pi t)\right), and P(t)=−e−t(sin⁡(πt)+πcos⁡(πt))1+π2P(t) = -\frac{e^{-t}\left(\sin(\pi t) + \pi\cos(\pi t)\right)}{1 + \pi^2}. Check: P′(t)=e−t(sin⁡πt+πcos⁡πt)−e−t(πcos⁡πt−π2sin⁡πt)1+π2=e−t(1+π2)sin⁡πt1+π2=v(t)P'(t) = \frac{e^{-t}(\sin \pi t + \pi\cos \pi t) - e^{-t}(\pi\cos \pi t - \pi^2 \sin \pi t)}{1 + \pi^2} = \frac{e^{-t}(1 + \pi^2)\sin \pi t}{1 + \pi^2} = v(t). The number 1+π21 + \pi^2 is the a2+b2a^2 + b^2 of exercise 4, with a=−1a = -1 and b=πb = \pi.

b) P(1)=−e−1(0+π⋅(−1))1+π2=πe−11+π2P(1) = -\frac{e^{-1}(0 + \pi \cdot (-1))}{1 + \pi^2} = \frac{\pi e^{-1}}{1 + \pi^2} and P(0)=−0+π1+π2=−π1+π2P(0) = -\frac{0 + \pi}{1 + \pi^2} = -\frac{\pi}{1 + \pi^2}. So ∫01v dt=P(1)−P(0)=π1+π2(1+e−1)\int_0^1 v\,dt = P(1) - P(0) = \frac{\pi}{1 + \pi^2}(1 + e^{-1}) metres. The sine vanishes at every integer, so only the cosine term survives at the bounds, and cos⁡(πk)=(−1)k\cos(\pi k) = (-1)^k does the rest: evaluating the bracket is quick once that is seen.

c) P(2)=−e−2(0+π)1+π2P(2) = -\frac{e^{-2}(0 + \pi)}{1 + \pi^2}, so ∫12v dt=P(2)−P(1)=−π(e−2+e−1)1+π2=−π1+π2e−1(1+e−1)\int_1^2 v\,dt = P(2) - P(1) = -\frac{\pi(e^{-2} + e^{-1})}{1 + \pi^2} = -\frac{\pi}{1 + \pi^2}e^{-1}(1 + e^{-1}) metres. It is negative: the mass comes back, as the lobe below the axis shows. The ratio of the second displacement to the first is −e−1-e^{-1}: the swing back is the first swing reversed and multiplied by e−1≈0.37e^{-1} \approx 0.37.

d) Net displacement: P(2)−P(0)=π−πe−21+π2=π1+π2(1−e−2)P(2) - P(0) = \frac{\pi - \pi e^{-2}}{1 + \pi^2} = \frac{\pi}{1 + \pi^2}(1 - e^{-2}) metres, which is also the sum of b) and c), since (1+e−1)−e−1(1+e−1)=(1+e−1)(1−e−1)=1−e−2(1 + e^{-1}) - e^{-1}(1 + e^{-1}) = (1 + e^{-1})(1 - e^{-1}) = 1 - e^{-2}. The distance travelled adds the two swings in absolute value, because v≥0v \ge 0 on [0,1][0, 1] and v≤0v \le 0 on [1,2][1, 2]: π1+π2(1+e−1)+π1+π2e−1(1+e−1)=π1+π2(1+e−1)2\frac{\pi}{1 + \pi^2}(1 + e^{-1}) + \frac{\pi}{1 + \pi^2}e^{-1}(1 + e^{-1}) = \frac{\pi}{1 + \pi^2}(1 + e^{-1})^2 metres. Integrating vv straight over [0,2][0, 2] gives the displacement, never the distance: the distance requires splitting at the zero of vv, t=1t = 1.

e) Since sin⁡(πk)=0\sin(\pi k) = 0 and cos⁡(πk)=(−1)k\cos(\pi k) = (-1)^k, P(k)=−π(−1)ke−k1+π2P(k) = -\frac{\pi(-1)^k e^{-k}}{1 + \pi^2}, so P(k+1)−P(k)=π(−1)ke−k(1+e−1)1+π2P(k + 1) - P(k) = \frac{\pi(-1)^k e^{-k}(1 + e^{-1})}{1 + \pi^2}, that is (−1)ke−k(-1)^k e^{-k} times the first swing. Estimates: π2≈9.87\pi^2 \approx 9.87, so π1+π2≈3.1410.87≈0.29\frac{\pi}{1 + \pi^2} \approx \frac{3.14}{10.87} \approx 0.29, and 1+e−1≈1.371 + e^{-1} \approx 1.37: the first swing is about 0.400.40 m. On the figure the first lobe rises to about 0.640.64 over a base of 11 s; a triangle would give 0.320.32, and the lobe is fuller than a triangle, so 0.400.40 is plausible. The swings shrink by the factor ee each second: with e3≈20e^3 \approx 20 and e4≈55e^4 \approx 55, the swing during [3,4][3, 4] is about 0.4020=0.02\frac{0.40}{20} = 0.02 m, still 22 cm, and the one during [4,5][4, 5] about 0.4055≈0.007\frac{0.40}{55} \approx 0.007 m. The first swing under 11 cm is the fifth, during [4,5][4, 5].

Exercise 10: Final exam problem: x arctan x, the constant hidden in v, and an exact area

A typical final exam question, in four moves. The figure shows the region AA under y=xarctan⁡xy = x\arctan x on [0,1][0, 1], and the dashed parabola y=x2y = x^2.

Recall that arctan⁡1=π4\arctan 1 = \frac{\pi}{4} and ddxarctan⁡x=11+x2\frac{d}{dx}\arctan x = \frac{1}{1 + x^2}.

0.20.40.60.811.21.40.20.40.60.811.2Ay = x²y = x arctan xx
  • a) Compute ∫xarctan⁡x dx\int x\arctan x\,dx with u=arctan⁡xu = \arctan x and v=x22v = \frac{x^2}{2}.
  • b) Redo a) with v=x2+12v = \frac{x^2 + 1}{2}. Why is it allowed, and what does it save?
  • c) Compute the area of AA exactly.
  • d) Prove that π4x≤arctan⁡x≤x\frac{\pi}{4}x \le \arctan x \le x on [0,1][0, 1], deduce bounds for the area of AA, and check c) against them.
  • e) Compute ∫01x2arctan⁡x dx\int_0^1 x^2\arctan x\,dx exactly. Does a clever constant in vv help this time?
Show the solution

Answers

  • a) ∫xarctan⁡x dx=x2+12arctan⁡x−x2+C\int x\arctan x\,dx = \frac{x^2 + 1}{2}\arctan x - \frac{x}{2} + C
  • b) v du=12 dxv\,du = \frac{1}{2}\,dx: same answer in one line; any antiderivative of xx is a valid vv.
  • c) A=π4−12A = \frac{\pi}{4} - \frac{1}{2}
  • d) π12≤A≤13\frac{\pi}{12} \le A \le \frac{1}{3}, and π4−12≈0.285\frac{\pi}{4} - \frac{1}{2} \approx 0.285 lies between 0.2620.262 and 0.3330.333.
  • e) ∫01x2arctan⁡x dx=π12−16+ln⁡26\int_0^1 x^2\arctan x\,dx = \frac{\pi}{12} - \frac{1}{6} + \frac{\ln 2}{6}; no constant helps.

a) u=arctan⁡xu = \arctan x, du=dx1+x2du = \frac{dx}{1 + x^2}; dv=x dxdv = x\,dx, v=x22v = \frac{x^2}{2}. The inverse trigonometric factor is uu for the reason of exercise 1: its derivative is algebraic. Then ∫xarctan⁡x dx=x22arctan⁡x−12∫x21+x2 dx\int x\arctan x\,dx = \frac{x^2}{2}\arctan x - \frac{1}{2}\int \frac{x^2}{1 + x^2}\,dx. Add and subtract 11 in the numerator: x21+x2=(1+x2)−11+x2=1−11+x2\frac{x^2}{1 + x^2} = \frac{(1 + x^2) - 1}{1 + x^2} = 1 - \frac{1}{1 + x^2}. So the new integral is x−arctan⁡xx - \arctan x, and ∫xarctan⁡x dx=x22arctan⁡x−x2+12arctan⁡x+C=x2+12arctan⁡x−x2+C\int x\arctan x\,dx = \frac{x^2}{2}\arctan x - \frac{x}{2} + \frac{1}{2}\arctan x + C = \frac{x^2 + 1}{2}\arctan x - \frac{x}{2} + C.

b) Any function whose derivative is xx is a valid vv, since only dv=x dxdv = x\,dx is prescribed (exercise 8, statement e). With v=x2+12v = \frac{x^2 + 1}{2}: ∫xarctan⁡x dx=x2+12arctan⁡x−∫x2+12⋅11+x2 dx=x2+12arctan⁡x−∫12 dx=x2+12arctan⁡x−x2+C\int x\arctan x\,dx = \frac{x^2 + 1}{2}\arctan x - \int \frac{x^2 + 1}{2} \cdot \frac{1}{1 + x^2}\,dx = \frac{x^2 + 1}{2}\arctan x - \int \frac{1}{2}\,dx = \frac{x^2 + 1}{2}\arctan x - \frac{x}{2} + C. The constant 12\frac{1}{2} was chosen so that vv contains the factor 1+x21 + x^2 of dudu: the new integral collapses to a constant, and the algebra of a) disappears. The trade was already good; the constant made it free.

c) xarctan⁡x≥0x\arctan x \ge 0 on [0,1][0, 1], so A=[x2+12arctan⁡x−x2]01=(22⋅π4−12)−(0−0)=π4−12A = \left[\frac{x^2 + 1}{2}\arctan x - \frac{x}{2}\right]_0^1 = \left(\frac{2}{2} \cdot \frac{\pi}{4} - \frac{1}{2}\right) - (0 - 0) = \frac{\pi}{4} - \frac{1}{2}. With π≈3.14\pi \approx 3.14, A≈0.285A \approx 0.285.

d) Upper bound: g(x)=x−arctan⁡xg(x) = x - \arctan x has g′(x)=1−11+x2=x21+x2≥0g'(x) = 1 - \frac{1}{1 + x^2} = \frac{x^2}{1 + x^2} \ge 0 and g(0)=0g(0) = 0, so g≥0g \ge 0 on [0,1][0, 1], that is arctan⁡x≤x\arctan x \le x. Lower bound: (arctan⁡x)′′=−2x(1+x2)2≤0(\arctan x)'' = -\frac{2x}{(1 + x^2)^2} \le 0 on [0,1][0, 1], so arctan⁡\arctan is concave there and lies above its chord from (0,0)(0, 0) to (1,π4)(1, \frac{\pi}{4}), that is arctan⁡x≥π4x\arctan x \ge \frac{\pi}{4}x. Multiplying by x≥0x \ge 0: π4x2≤xarctan⁡x≤x2\frac{\pi}{4}x^2 \le x\arctan x \le x^2, and integrating over [0,1][0, 1]: π12≤A≤13\frac{\pi}{12} \le A \le \frac{1}{3}. Numerically 0.262≤0.285≤0.3330.262 \le 0.285 \le 0.333: c) passes. The figure shows the upper bound: the dashed parabola stays above the curve. A value like π4+12\frac{\pi}{4} + \frac{1}{2}, from a sign slip, would be outside at once.

e) u=arctan⁡xu = \arctan x, dv=x2 dxdv = x^2\,dx, v=x33v = \frac{x^3}{3}: ∫01x2arctan⁡x dx=[x33arctan⁡x]01−13∫01x31+x2 dx\int_0^1 x^2\arctan x\,dx = \left[\frac{x^3}{3}\arctan x\right]_0^1 - \frac{1}{3}\int_0^1 \frac{x^3}{1 + x^2}\,dx. The bracket is π12\frac{\pi}{12}. For the new integral, x3=x(1+x2)−xx^3 = x(1 + x^2) - x, so x31+x2=x−x1+x2\frac{x^3}{1 + x^2} = x - \frac{x}{1 + x^2}, whose antiderivative is x22−12ln⁡(1+x2)\frac{x^2}{2} - \frac{1}{2}\ln(1 + x^2) (substitution w=1+x2w = 1 + x^2); on [0,1][0, 1] it gives 12−ln⁡22\frac{1}{2} - \frac{\ln 2}{2}. Hence the value π12−16+ln⁡26\frac{\pi}{12} - \frac{1}{6} + \frac{\ln 2}{6}, about 0.262−0.167+0.115=0.210.262 - 0.167 + 0.115 = 0.21, below AA as it must be since x2≤xx^2 \le x on [0,1][0, 1]. No constant helps here: v=x3+K3v = \frac{x^3 + K}{3} would need x3+Kx^3 + K to be a multiple of 1+x21 + x^2, and no constant KK makes it so. The trick of b) works only when the antiderivative of dvdv can be made to contain 1+x21 + x^2.

See also

Struggling with MATH 141?

I tutor first-year calculus at McGill and Concordia, in English or in French, in Montreal or online. Get in touch for a first session.

Site by Studio Squalli