MATH 141 Calculus 2 • McGill University, Montreal

MATH 141 practice final exam with full solutions (McGill)

This is a practice final examination for MATH 141, Calculus 2, the integral calculus course of the first year at McGill University. It covers the whole course and is weighted like a real cumulative final: about a quarter on the integrals and techniques already examined at the midterm, a third on the applications of the integral (areas, centroids, volumes by two methods, work, arc length and surface area, probability densities), and a little more than a third on sequences and series, from the convergence tests to power series and Taylor polynomials with their error. Twelve questions, one hundred points, three parts.

Sit it as an exam: three hours on a timer, no calculator, no notes. Every answer is exact, ln⁡32\ln\frac{3}{2}, 32π3\frac{32\pi}{3}, 2732\frac{27}{32}, π4+19400\frac{\pi}{4} + \frac{19}{400}, and every mark goes to the method: name the substitution and its new bounds, write the limit of each improper integral, measure each radius from the axis, name each convergence test and CHECK its hypotheses, test each endpoint on its own. Not one question repeats an exercise of the twenty-four chapter sets of this site or of the practice midterm: the gestures are those of every MATH 141 final, the integrands, regions and series are new, so the paper measures what you can do and not what you remember. Each question carries its Answers box for a first quick marking, and the full reasoning underneath.

The traps named in the solutions: splitting a convergent improper integral into ∞−∞\infty - \infty, missing the second endpoint where an integrand blows up, squaring the difference in a moment instead of taking the difference of the squares, measuring a shell radius from the wrong side of the axis, keeping the wrong root when a parabola is inverted, lifting a layer of water by yy instead of 4−y4 - y, using the radius yy for a rotation about the yy-axis, reading (nn+1)n\left(\frac{n}{n + 1}\right)^n as 11, testing the first term instead of the first OMITTED one, forgetting the 22 of a second derivative, and evaluating a Maclaurin series outside its radius.

12 corrected exercises • 100 points • 180 minutes

Every MATH 141 chapter →

Part A: integrals and techniques (/24)

Exercise 1: Four integrals, and the move that opens each one

Evaluate each integral exactly. Three points each: the first line must NAME the move (the substitution with its differential and its new bounds, the choice of uu and dvdv, the identity or the algebra used), then the computation follows. No technique is announced.

  • a) ∫01xln⁡(1+x) dx\displaystyle\int_0^{1} x\ln(1 + x)\,dx
  • b) ∫0π/2cos⁡3x1+sin⁡x dx\displaystyle\int_0^{\pi/2} \frac{\cos^3 x}{1 + \sin x}\,dx
  • c) ∫0ln⁡3dxex+3e−x\displaystyle\int_0^{\ln 3} \frac{dx}{e^x + 3e^{-x}}
  • d) ∫38dxxx+1\displaystyle\int_3^{8} \frac{dx}{x\sqrt{x + 1}}
Show the solution

Answers

  • a) 14\frac{1}{4}
  • b) 12\frac{1}{2}
  • c) π63=π318\frac{\pi}{6\sqrt{3}} = \frac{\pi\sqrt{3}}{18}
  • d) ln⁡32\ln\frac{3}{2}

a) Move: integration by parts with u=ln⁡(1+x)u = \ln(1 + x), du=dx1+xdu = \frac{dx}{1 + x}, and dv=x dxdv = x\,dx. The logarithm is uu because differentiating it turns it into a rational function. For vv, ANY antiderivative of xx is allowed, and the one that factors is the smart one: v=x2−12=(x−1)(x+1)2v = \frac{x^2 - 1}{2} = \frac{(x - 1)(x + 1)}{2}. Then ∫01xln⁡(1+x) dx=[x2−12ln⁡(1+x)]01−∫01(x−1)(x+1)2(1+x) dx=0−0−12∫01(x−1) dx=−12(12−1)=14\int_0^1 x\ln(1 + x)\,dx = \Big[\frac{x^2 - 1}{2}\ln(1 + x)\Big]_0^1 - \int_0^1 \frac{(x - 1)(x + 1)}{2(1 + x)}\,dx = 0 - 0 - \frac{1}{2}\int_0^1 (x - 1)\,dx = -\frac{1}{2}\left(\frac{1}{2} - 1\right) = \frac{1}{4}. The bracket vanishes at both bounds: x2−1=0x^2 - 1 = 0 at x=1x = 1 and ln⁡1=0\ln 1 = 0 at x=0x = 0.

With the usual v=x22v = \frac{x^2}{2} the answer is the same but longer: ln⁡22−12∫01x21+x dx\frac{\ln 2}{2} - \frac{1}{2}\int_0^1 \frac{x^2}{1 + x}\,dx, and long division gives x21+x=x−1+11+x\frac{x^2}{1 + x} = x - 1 + \frac{1}{1 + x}, so the value is ln⁡22−12(12−1+ln⁡2)=14\frac{\ln 2}{2} - \frac{1}{2}\left(\frac{1}{2} - 1 + \ln 2\right) = \frac{1}{4}: the ln⁡2\ln 2 cancels. Size check: on [0,1][0, 1], the concave ln⁡(1+x)\ln(1 + x) lies between its chord xln⁡2x\ln 2 and its tangent xx, so the integral lies between ln⁡23≈0.23\frac{\ln 2}{3} \approx 0.23 and 13\frac{1}{3}, and 0.250.25 does.

b) Move: simplify BEFORE integrating. cos⁡3x=cos⁡x (1−sin⁡2x)=cos⁡x (1−sin⁡x)(1+sin⁡x)\cos^3 x = \cos x\,(1 - \sin^2 x) = \cos x\,(1 - \sin x)(1 + \sin x), and 1+sin⁡x>01 + \sin x > 0 on [0,π2][0, \frac{\pi}{2}], so the fraction reduces to cos⁡x (1−sin⁡x)=cos⁡x−sin⁡xcos⁡x\cos x\,(1 - \sin x) = \cos x - \sin x\cos x. Then ∫0π/2(cos⁡x−sin⁡xcos⁡x) dx=[sin⁡x−sin⁡2x2]0π/2=1−12=12\int_0^{\pi/2} (\cos x - \sin x\cos x)\,dx = \Big[\sin x - \frac{\sin^2 x}{2}\Big]_0^{\pi/2} = 1 - \frac{1}{2} = \frac{1}{2}. The same through u=sin⁡xu = \sin x, du=cos⁡x dxdu = \cos x\,dx, bounds 00 and 11: ∫011−u21+u du=∫01(1−u) du=12\int_0^1 \frac{1 - u^2}{1 + u}\,du = \int_0^1 (1 - u)\,du = \frac{1}{2}. Treating cos⁡3x\cos^3 x as an odd power is right, but forgetting to cancel 1+u1 + u afterwards sends people into a needless long division.

c) Move: multiply top and bottom by exe^x, so the integrand becomes exe2x+3\frac{e^x}{e^{2x} + 3}, then u=exu = e^x, du=ex dxdu = e^x\,dx, bounds x=0↦u=1x = 0 \mapsto u = 1 and x=ln⁡3↦u=3x = \ln 3 \mapsto u = 3. Now ∫13duu2+3=13[arctan⁡u3]13=13(arctan⁡3−arctan⁡13)=13(π3−π6)=π63\int_1^3 \frac{du}{u^2 + 3} = \frac{1}{\sqrt{3}}\Big[\arctan\frac{u}{\sqrt{3}}\Big]_1^3 = \frac{1}{\sqrt{3}}\left(\arctan\sqrt{3} - \arctan\frac{1}{\sqrt{3}}\right) = \frac{1}{\sqrt{3}}\left(\frac{\pi}{3} - \frac{\pi}{6}\right) = \frac{\pi}{6\sqrt{3}}. The factor 13\frac{1}{\sqrt{3}} in front and the u3\frac{u}{\sqrt{3}} inside both come from u2+3=3(1+(u3)2)u^2 + 3 = 3\left(1 + \left(\frac{u}{\sqrt{3}}\right)^2\right); forgetting either is the classic slip. Size check: ex+3e−x≥23e^x + 3e^{-x} \ge 2\sqrt{3} (sum of two positive numbers with product 33) and ≤4\le 4 on the interval, so the value lies between ln⁡34\frac{\ln 3}{4} and ln⁡323\frac{\ln 3}{2\sqrt{3}}, about 0.270.27 and 0.320.32; π318≈0.30\frac{\pi\sqrt{3}}{18} \approx 0.30.

d) Move: the root blocks everything, so rationalize with u=x+1u = \sqrt{x + 1}, x=u2−1x = u^2 - 1, dx=2u dudx = 2u\,du, bounds x=3↦u=2x = 3 \mapsto u = 2 and x=8↦u=3x = 8 \mapsto u = 3. Then ∫38dxxx+1=∫232u du(u2−1)u=∫232 duu2−1\int_3^8 \frac{dx}{x\sqrt{x + 1}} = \int_2^3 \frac{2u\,du}{(u^2 - 1)u} = \int_2^3 \frac{2\,du}{u^2 - 1}. Now partial fractions: 2(u−1)(u+1)=1u−1−1u+1\frac{2}{(u - 1)(u + 1)} = \frac{1}{u - 1} - \frac{1}{u + 1} (cover-up at u=1u = 1 and u=−1u = -1). The value is [ln⁡u−1u+1]23=ln⁡24−ln⁡13=ln⁡32\Big[\ln\frac{u - 1}{u + 1}\Big]_2^3 = \ln\frac{2}{4} - \ln\frac{1}{3} = \ln\frac{3}{2}. Two chapters in a row: the rationalizing substitution turns a root into a rational function, and the rational function is then decomposed.

Exercise 2: Improper integrals: two values, two verdicts

For each integral, say first WHERE it is improper and write the limit or limits. Evaluate a) and b); decide c) and d) by comparison, without an antiderivative. The figure shows the integrand of b).

-0.50.511.522.533.544.50.20.40.60.811.21.41.6y = 1/√(4x − x²)area πx = 4
  • a) ∫1∞dxx2(x+1)\displaystyle\int_1^{\infty} \frac{dx}{x^2(x + 1)}. A classmate splits the integrand into partial fractions and writes one limit per fraction. What goes wrong?
  • b) ∫04dx4x−x2\displaystyle\int_0^{4} \frac{dx}{\sqrt{4x - x^2}}
  • c) ∫1∞x dxx5+1\displaystyle\int_1^{\infty} \frac{x\,dx}{\sqrt{x^5 + 1}}. Show that it converges, and prove that its value II satisfies 2≤I≤2\sqrt{2} \le I \le 2.
  • d) ∫0π/2dxsin⁡x\displaystyle\int_0^{\pi/2} \frac{dx}{\sin x}
Show the solution

Answers

  • a) Converges to 1−ln⁡21 - \ln 2; taken one fraction at a time, it is −∞+∞-\infty + \infty
  • b) Improper at 00 and at 44; converges to π2+π2=π\frac{\pi}{2} + \frac{\pi}{2} = \pi
  • c) 12x−3/2≤xx5+1≤x−3/2\frac{1}{\sqrt{2}}x^{-3/2} \le \frac{x}{\sqrt{x^5 + 1}} \le x^{-3/2} on [1,∞)[1, \infty): converges, 2≤I≤2\sqrt{2} \le I \le 2
  • d) Improper at 00; 1sin⁡x≥1x\frac{1}{\sin x} \ge \frac{1}{x} there: diverges

a) Improper at ∞\infty only: the integrand is continuous on [1,∞)[1, \infty). Decompose, with one term per power of the repeated factor: 1x2(x+1)=Ax+Bx2+Cx+1\frac{1}{x^2(x + 1)} = \frac{A}{x} + \frac{B}{x^2} + \frac{C}{x + 1}, so 1=Ax(x+1)+B(x+1)+Cx21 = Ax(x + 1) + B(x + 1) + Cx^2. At x=0x = 0: B=1B = 1. At x=−1x = -1: C=1C = 1. The x2x^2 coefficient: 0=A+C0 = A + C, so A=−1A = -1. An antiderivative is −ln⁡x−1x+ln⁡(x+1)=ln⁡x+1x−1x-\ln x - \frac{1}{x} + \ln(x + 1) = \ln\frac{x + 1}{x} - \frac{1}{x}, with the two logarithms COMBINED before any limit is taken.

Then ∫1tdxx2(x+1)=ln⁡(1+1t)−1t−(ln⁡2−1)\int_1^t \frac{dx}{x^2(x + 1)} = \ln\left(1 + \frac{1}{t}\right) - \frac{1}{t} - (\ln 2 - 1), and as t→∞t \to \infty the first two terms tend to 00: the integral converges to 1−ln⁡2≈0.311 - \ln 2 \approx 0.31. It is below ∫1∞dxx3=12\int_1^\infty \frac{dx}{x^3} = \frac{1}{2}, as it must be. The classmate writes −∫1∞dxx+∫1∞dxx2+∫1∞dxx+1-\int_1^\infty \frac{dx}{x} + \int_1^\infty \frac{dx}{x^2} + \int_1^\infty \frac{dx}{x + 1}: the first and the last are divergent, −∞-\infty and +∞+\infty, and ∞−∞\infty - \infty is not a number. Splitting a convergent improper integral is legal only into pieces that each converge.

b) Complete the square: 4x−x2=4−(x−2)24x - x^2 = 4 - (x - 2)^2, which vanishes at x=0x = 0 and at x=4x = 4. The integrand blows up at BOTH endpoints (the figure shows the two branches going up), so split at an interior point, 22 is natural, and take one limit on each side: ∫04=lim⁡s→0+∫s2+lim⁡t→4−∫2t\int_0^4 = \lim_{s \to 0^+}\int_s^2 + \lim_{t \to 4^-}\int_2^t. With x−2=2sin⁡θx - 2 = 2\sin\theta, dx=2cos⁡θ dθdx = 2\cos\theta\,d\theta and 4−4sin⁡2θ=2cos⁡θ\sqrt{4 - 4\sin^2\theta} = 2\cos\theta (positive for −π2<θ<π2-\frac{\pi}{2} < \theta < \frac{\pi}{2}), the integrand becomes dθd\theta: an antiderivative is arcsin⁡x−22\arcsin\frac{x - 2}{2}.

Left piece: arcsin⁡0−lim⁡s→0+arcsin⁡s−22=0−arcsin⁡(−1)=π2\arcsin 0 - \lim_{s \to 0^+}\arcsin\frac{s - 2}{2} = 0 - \arcsin(-1) = \frac{\pi}{2}. Right piece: lim⁡t→4−arcsin⁡t−22−0=π2\lim_{t \to 4^-}\arcsin\frac{t - 2}{2} - 0 = \frac{\pi}{2}. Both converge, so the integral converges to π\pi. The unbounded region of the figure has a finite area because the branches blow up like 12x\frac{1}{2\sqrt{x}}, an exponent 12<1\frac{1}{2} < 1 on (0,1](0, 1].

c) Improper at ∞\infty only, and the integrand is positive. For x≥1x \ge 1: x5<x5+1≤2x5x^5 < x^5 + 1 \le 2x^5, so x5<x5+1≤2 x5/2\sqrt{x^5} < \sqrt{x^5 + 1} \le \sqrt{2}\,x^{5/2}, and dividing xx by these, x−3/22≤xx5+1<x−3/2\frac{x^{-3/2}}{\sqrt{2}} \le \frac{x}{\sqrt{x^5 + 1}} < x^{-3/2}. The upper bound is a pp-integral with p=32>1p = \frac{3}{2} > 1: ∫1∞x−3/2 dx=lim⁡t→∞(2−2t)=2\int_1^\infty x^{-3/2}\,dx = \lim_{t \to \infty}\left(2 - \frac{2}{\sqrt{t}}\right) = 2. By the comparison test the integral converges, and I≤2I \le 2. Integrating the lower bound: I≥22=2I \ge \frac{2}{\sqrt{2}} = \sqrt{2}. The direction matters: the upper bound proves convergence, the lower bound only gives the bracket.

d) Improper at 00, where sin⁡x=0\sin x = 0; the integrand is positive on (0,π2](0, \frac{\pi}{2}]. For 0<x≤π20 < x \le \frac{\pi}{2}, 0<sin⁡x≤x0 < \sin x \le x (the graph of sin⁡\sin lies under its tangent at 00), so 1sin⁡x≥1x\frac{1}{\sin x} \ge \frac{1}{x}. And ∫0π/2dxx=lim⁡s→0+(ln⁡π2−ln⁡s)=∞\int_0^{\pi/2} \frac{dx}{x} = \lim_{s \to 0^+}\left(\ln\frac{\pi}{2} - \ln s\right) = \infty, the pp-integral with p=1p = 1 on (0,a](0, a]. An integral ABOVE a divergent one diverges: ∫0π/2dxsin⁡x\int_0^{\pi/2}\frac{dx}{\sin x} diverges. Confirmation: ln⁡(tan⁡x2)\ln\left(\tan\frac{x}{2}\right) is an antiderivative, and it tends to −∞-\infty as x→0+x \to 0^+.

Part B: applications of the integral (/38)

Exercise 3: One region: its area, its centroid, and the volume Pappus gives for free

Let RR be the region enclosed by y=xy = \sqrt{x} and y=x3y = x^3, shown in the figure with its centroid.

0.10.20.30.40.50.60.70.80.911.11.21.30.10.20.30.40.50.60.70.80.911.11.2y = √xy = x³centroid
  • a) Find the points where the curves meet and the area AA of RR.
  • b) Find the centroid (xˉ,yˉ)(\bar{x}, \bar{y}) of RR. Check that the red point of the figure is plausible.
  • c) RR is rotated about the line y=−1y = -1. Use the theorem of Pappus to find the volume of the solid, then check it with washers.
Show the solution

Answers

  • a) x=0x = 0 and x=1x = 1; A=23−14=512A = \frac{2}{3} - \frac{1}{4} = \frac{5}{12}
  • b) xˉ=1225\bar{x} = \frac{12}{25}, yˉ=37\bar{y} = \frac{3}{7}
  • c) V=2π(37+1)512=25π21V = 2\pi\left(\frac{3}{7} + 1\right)\frac{5}{12} = \frac{25\pi}{21}, confirmed by washers

a) x=x3\sqrt{x} = x^3 gives x=x6x = x^6 for x≥0x \ge 0, so x(x5−1)=0x(x^5 - 1) = 0: x=0x = 0 or x=1x = 1, at the points (0,0)(0, 0) and (1,1)(1, 1). On (0,1)(0, 1), x>x3\sqrt{x} > x^3 (at x=14x = \frac{1}{4}: 12\frac{1}{2} against 164\frac{1}{64}), so the top is x\sqrt{x}. A=∫01(x1/2−x3)dx=23−14=512A = \int_0^1 \left(x^{1/2} - x^3\right)dx = \frac{2}{3} - \frac{1}{4} = \frac{5}{12}.

b) A vertical strip at xx has area (x−x3) dx(\sqrt{x} - x^3)\,dx, its arm about the yy-axis is xx, and its midpoint sits at height x+x32\frac{\sqrt{x} + x^3}{2}. So My=∫01x(x1/2−x3)dx=25−15=15M_y = \int_0^1 x\left(x^{1/2} - x^3\right)dx = \frac{2}{5} - \frac{1}{5} = \frac{1}{5} and Mx=∫0112((x)2−(x3)2)dx=12∫01(x−x6) dx=12(12−17)=528M_x = \int_0^1 \frac{1}{2}\left((\sqrt{x})^2 - (x^3)^2\right)dx = \frac{1}{2}\int_0^1 (x - x^6)\,dx = \frac{1}{2}\left(\frac{1}{2} - \frac{1}{7}\right) = \frac{5}{28}. Divide by the area: xˉ=1/55/12=1225=0.48\bar{x} = \frac{1/5}{5/12} = \frac{12}{25} = 0.48 and yˉ=5/285/12=1228=37≈0.43\bar{y} = \frac{5/28}{5/12} = \frac{12}{28} = \frac{3}{7} \approx 0.43.

The trap in MxM_x is to square the difference, 12(x−x3)2\frac{1}{2}(\sqrt{x} - x^3)^2, instead of taking the difference of the squares: the strip is a washer-like band from x3x^3 to x\sqrt{x}, and its moment is 12(f2−g2)\frac{1}{2}(f^2 - g^2). Plausibility: the red point (0.48,0.43)(0.48, 0.43) lies inside RR, as it must, since RR is convex (it lies under the concave curve x\sqrt{x} and over the convex curve x3x^3) and the centroid of a convex region is inside it. Its abscissa is also close to where the strip height x−x3\sqrt{x} - x^3 is largest: the derivative 12x−3x2\frac{1}{2\sqrt{x}} - 3x^2 vanishes when x5/2=16x^{5/2} = \frac{1}{6}, a little below 12\frac{1}{2}.

c) Pappus: the volume is the area times the distance travelled by the centroid. The centroid is at height 37\frac{3}{7}, so its distance to the axis y=−1y = -1 is 37+1=107\frac{3}{7} + 1 = \frac{10}{7}, and it travels 2π⋅1072\pi \cdot \frac{10}{7}. V=2π⋅107⋅512=100π84=25π21V = 2\pi \cdot \frac{10}{7} \cdot \frac{5}{12} = \frac{100\pi}{84} = \frac{25\pi}{21}.

Check with washers perpendicular to the xx-axis: the radii are measured FROM the axis y=−1y = -1, so Rout=x+1R_{\text{out}} = \sqrt{x} + 1 and Rin=x3+1R_{\text{in}} = x^3 + 1. V=π∫01[(x+1)2−(x3+1)2]dx=π∫01(x+2x−x6−2x3)dx=π(12+43−17−12)=25π21V = \pi\int_0^1 \left[(\sqrt{x} + 1)^2 - (x^3 + 1)^2\right]dx = \pi\int_0^1 \left(x + 2\sqrt{x} - x^6 - 2x^3\right)dx = \pi\left(\frac{1}{2} + \frac{4}{3} - \frac{1}{7} - \frac{1}{2}\right) = \frac{25\pi}{21}. The two routes agree; with the radii x\sqrt{x} and x3x^3, forgetting the shift, one would get the volume about the xx-axis, 5π14\frac{5\pi}{14}.

Exercise 4: One solid, two methods: shells, then washers that force an inverse

Let RR be the region between the parabolas y=x2y = x^2 and y=4x−x2y = 4x - x^2. The figure shows RR, one vertical strip, and the axis x=3x = 3. The solid SS is obtained by rotating RR about the line x=3x = 3.

-0.50.511.522.533.5-0.50.511.522.533.544.5y = 4x − x²y = x²x = 33 − x
  • a) Find the volume of SS by cylindrical shells.
  • b) Set up and evaluate the volume of SS with washers. Explain why the inner and outer radii require solving y=4x−x2y = 4x - x^2 for xx, and which root to keep.
  • c) Find the volume of the solid obtained by rotating RR about the line x=−1x = -1 instead. Explain the result with a symmetry of RR.
Show the solution

Answers

  • a) V=2π∫02(3−x)(4x−2x2) dx=32π3V = 2\pi\int_0^2 (3 - x)(4x - 2x^2)\,dx = \frac{32\pi}{3}
  • b) x=2−4−yx = 2 - \sqrt{4 - y} on the left, x=yx = \sqrt{y} on the right; V=π∫04[(1+4−y)2−(3−y)2]dy=32π3V = \pi\int_0^4 \left[(1 + \sqrt{4 - y})^2 - (3 - \sqrt{y})^2\right]dy = \frac{32\pi}{3}
  • c) 32π3\frac{32\pi}{3} again: the height 4x−2x24x - 2x^2 is symmetric about x=1x = 1, which is at distance 22 from both axes

a) Intersections: x2=4x−x2x^2 = 4x - x^2 gives 2x(x−2)=02x(x - 2) = 0, so x=0x = 0 and x=2x = 2, at (0,0)(0, 0) and (2,4)(2, 4); between them 4x−x2≥x24x - x^2 \ge x^2 (at x=1x = 1: 33 against 11). A vertical strip at xx, parallel to the axis, sweeps a shell of height h(x)=(4x−x2)−x2=4x−2x2h(x) = (4x - x^2) - x^2 = 4x - 2x^2 and radius equal to its distance to the axis. The region lies to the LEFT of x=3x = 3, so the radius is 3−x3 - x, not xx and not x−3x - 3. V=2π∫02(3−x)(4x−2x2) dx=2π∫02(12x−10x2+2x3)dx=2π(24−803+8)=2π⋅163=32π3V = 2\pi\int_0^2 (3 - x)(4x - 2x^2)\,dx = 2\pi\int_0^2 \left(12x - 10x^2 + 2x^3\right)dx = 2\pi\left(24 - \frac{80}{3} + 8\right) = 2\pi \cdot \frac{16}{3} = \frac{32\pi}{3}.

b) Washers need slices PERPENDICULAR to the axis, so horizontal slices at height yy, 0≤y≤40 \le y \le 4. A horizontal line at height yy crosses RR between the right branch of y=4x−x2y = 4x - x^2 on the left and y=x2y = x^2 on the right, so both curves must be solved for xx. From y=x2y = x^2 with x≥0x \ge 0: x=yx = \sqrt{y}. From x2−4x+y=0x^2 - 4x + y = 0: x=2±4−yx = 2 \pm \sqrt{4 - y}, and on RR the xx runs over [0,2][0, 2], so the root to keep is x=2−4−yx = 2 - \sqrt{4 - y} (the ++ root is the other half of the parabola, for 2≤x≤42 \le x \le 4, outside RR). Check at y=3y = 3: the slice runs from x=1x = 1 to x=3x = \sqrt{3}, and indeed 4⋅1−1=34 \cdot 1 - 1 = 3 and (3)2=3(\sqrt{3})^2 = 3.

Both radii are measured from x=3x = 3: the far edge is the left one, Rout=3−(2−4−y)=1+4−yR_{\text{out}} = 3 - (2 - \sqrt{4 - y}) = 1 + \sqrt{4 - y}, and Rin=3−yR_{\text{in}} = 3 - \sqrt{y}. Expanding, Rout2−Rin2=(5−y+24−y)−(9−6y+y)=−4−2y+24−y+6yR_{\text{out}}^2 - R_{\text{in}}^2 = (5 - y + 2\sqrt{4 - y}) - (9 - 6\sqrt{y} + y) = -4 - 2y + 2\sqrt{4 - y} + 6\sqrt{y}. Then V=π∫04(−4−2y+2(4−y)1/2+6y1/2)dy=π(−16−16+43⋅8+4⋅8)=π(323)V = \pi\int_0^4 \left(-4 - 2y + 2(4 - y)^{1/2} + 6y^{1/2}\right)dy = \pi\left(-16 - 16 + \frac{4}{3} \cdot 8 + 4 \cdot 8\right) = \pi\left(\frac{32}{3}\right), since ∫04(4−y)1/2dy=∫04y1/2dy=163\int_0^4 (4 - y)^{1/2}dy = \int_0^4 y^{1/2}dy = \frac{16}{3}. Same volume, twice the work: this is why shells are the method of choice when the axis is vertical and the curves are given as y=f(x)y = f(x).

c) Shells again, with radius x−(−1)=x+1x - (-1) = x + 1 since RR is to the RIGHT of x=−1x = -1: V=2π∫02(x+1)(4x−2x2) dx=2π∫02(4x+2x2−2x3)dx=2π(8+163−8)=32π3V = 2\pi\int_0^2 (x + 1)(4x - 2x^2)\,dx = 2\pi\int_0^2 \left(4x + 2x^2 - 2x^3\right)dx = 2\pi\left(8 + \frac{16}{3} - 8\right) = \frac{32\pi}{3}. The same value is no accident: the height h(x)=2x(2−x)h(x) = 2x(2 - x) is symmetric about x=1x = 1, so the centroid of RR has xˉ=1\bar{x} = 1, at distance 22 from both x=3x = 3 and x=−1x = -1. With A=∫02(4x−2x2) dx=83A = \int_0^2 (4x - 2x^2)\,dx = \frac{8}{3}, Pappus gives 2π⋅2⋅83=32π32\pi \cdot 2 \cdot \frac{8}{3} = \frac{32\pi}{3} for both axes, although the two solids have different shapes.

Exercise 5: Pumping out a bowl shaped like a paraboloid

A tank is the surface obtained by rotating the parabola y=x2y = x^2, 0≤x≤20 \le x \le 2, about the yy-axis (lengths in metres). It is full of water, of density 10001000 kg/m3^3; take g=9.8g = 9.8 m/s2^2. The figure shows a cross-section and one horizontal layer of water. Give exact answers, in terms of π\pi.

-2.5-2-1.5-1-0.50.511.522.50.511.522.533.544.5rim, y = 4layerheight yy = x²
  • a) Find the work needed to pump all the water out over the rim.
  • b) Find the work needed to pump out only enough water to lower the level from 44 m to 22 m. Compare the fraction of the work and the fraction of the water.
  • c) The water must in fact be pumped to a spout 11 m above the rim. Find the new total work.
Show the solution

Answers

  • a) W=9800π∫04y(4−y) dy=313600π3W = 9800\pi\int_0^4 y(4 - y)\,dy = \frac{313600\pi}{3} J
  • b) 9800π⋅163=156800π39800\pi \cdot \frac{16}{3} = \frac{156800\pi}{3} J: half the work for three quarters of the water
  • c) 9800π(323+8)=548800π39800\pi\left(\frac{32}{3} + 8\right) = \frac{548800\pi}{3} J

a) Measure yy upward from the bottom of the bowl. The layer at height yy, of thickness dydy, is a disc of radius x=yx = \sqrt{y} (the tank wall is y=x2y = x^2), so its volume is π(y)2 dy=πy dy\pi(\sqrt{y})^2\,dy = \pi y\,dy, its mass 1000πy dy1000\pi y\,dy kg and its weight 9800πy dy9800\pi y\,dy N. It must rise from height yy to the rim at height 44: a distance 4−y4 - y, not yy. So W=∫049800π y(4−y) dy=9800π[2y2−y33]04=9800π(32−643)=9800π⋅323=313600π3W = \int_0^4 9800\pi\,y(4 - y)\,dy = 9800\pi\Big[2y^2 - \frac{y^3}{3}\Big]_0^4 = 9800\pi\left(32 - \frac{64}{3}\right) = 9800\pi \cdot \frac{32}{3} = \frac{313600\pi}{3} J, about 3.3×1053.3 \times 10^5 J.

b) Only the layers between y=2y = 2 and y=4y = 4 leave, each still lifted to the rim: Wtop=9800π∫24y(4−y) dy=9800π(323−(8−83))=9800π⋅163=156800π3W_{\text{top}} = 9800\pi\int_2^4 y(4 - y)\,dy = 9800\pi\left(\frac{32}{3} - \left(8 - \frac{8}{3}\right)\right) = 9800\pi \cdot \frac{16}{3} = \frac{156800\pi}{3} J, exactly HALF of a). The water removed is π∫24y dy=6π\pi\int_2^4 y\,dy = 6\pi m3^3 out of π∫04y dy=8π\pi\int_0^4 y\,dy = 8\pi m3^3: three quarters of the water for half of the work. The bowl is wide at the top, so most of the water is near the rim, where it has little distance to travel; the last quarter, at the narrow bottom, costs the other half. Treating the bowl as a cylinder, with a constant layer area, would miss exactly this.

c) Every layer now travels 5−y5 - y instead of 4−y4 - y: one extra metre for each. W=9800π∫04y(5−y) dy=9800π(40−643)=9800π⋅563=548800π3W = 9800\pi\int_0^4 y(5 - y)\,dy = 9800\pi\left(40 - \frac{64}{3}\right) = 9800\pi \cdot \frac{56}{3} = \frac{548800\pi}{3} J. The difference with a) is 9800π⋅89800\pi \cdot 8 J, the total weight of the water, 9800⋅8π9800 \cdot 8\pi N, times the extra metre: the extra lift is the same for every layer, so it simply multiplies the total weight.

Exercise 6: A curve whose length element opens, and the surface it sweeps

Let CC be the curve y=x48+14x2y = \frac{x^4}{8} + \frac{1}{4x^2} for 1≤x≤21 \le x \le 2.

  • a) Show that 1+(dydx)21 + \left(\frac{dy}{dx}\right)^2 is a perfect square, and find the length of CC exactly. Compare with the length of the chord joining its endpoints.
  • b) CC is rotated about the yy-axis. Find the area of the surface generated.
Show the solution

Answers

  • a) 1+(y′)2=(x32+12x3)21 + (y')^2 = \left(\frac{x^3}{2} + \frac{1}{2x^3}\right)^2; L=3316L = \frac{33}{16}, longer than the chord 98516\frac{\sqrt{985}}{16}
  • b) S=2π∫12x(x32+12x3)dx=67π10S = 2\pi\int_1^2 x\left(\frac{x^3}{2} + \frac{1}{2x^3}\right)dx = \frac{67\pi}{10}

a) dydx=x32−12x3\frac{dy}{dx} = \frac{x^3}{2} - \frac{1}{2x^3}. Write it a−ba - b with a=x32a = \frac{x^3}{2} and b=12x3b = \frac{1}{2x^3}, whose product is ab=14ab = \frac{1}{4}. Then 1+(a−b)2=a2−2ab+b2+1=a2+12+b2=(a+b)21 + (a - b)^2 = a^2 - 2ab + b^2 + 1 = a^2 + \frac{1}{2} + b^2 = (a + b)^2, because 1−2ab=12=2ab1 - 2ab = \frac{1}{2} = 2ab. So 1+(y′)2=x32+12x3\sqrt{1 + (y')^2} = \frac{x^3}{2} + \frac{1}{2x^3}, positive on [1,2][1, 2], and L=∫12(x32+x−32)dx=[x48−14x2]12=(2−116)−(18−14)=3116+18=3316L = \int_1^2 \left(\frac{x^3}{2} + \frac{x^{-3}}{2}\right)dx = \Big[\frac{x^4}{8} - \frac{1}{4x^2}\Big]_1^2 = \left(2 - \frac{1}{16}\right) - \left(\frac{1}{8} - \frac{1}{4}\right) = \frac{31}{16} + \frac{1}{8} = \frac{33}{16}.

The endpoints are (1,38)\left(1, \frac{3}{8}\right) and (2,3316)\left(2, \frac{33}{16}\right), a rise of 2716\frac{27}{16} over a run of 11, so the chord is 1+729256=98516\sqrt{1 + \frac{729}{256}} = \frac{\sqrt{985}}{16}. Since 985<332=1089985 < 33^2 = 1089, the chord is shorter than 3316\frac{33}{16}, as a chord must be. Notice the antiderivative: x48−14x2\frac{x^4}{8} - \frac{1}{4x^2} is yy with one sign changed, the signature of these designed curves.

b) About the yy-axis the radius of the circle described by a point of CC is its distance to that axis, xx, not yy. So S=∫122πx ds=2π∫12x(x32+12x3)dx=2π∫12(x42+12x2)dx=2π[x510−12x]12=2π(3210−14−110+12)=2π⋅6720=67π10S = \int_1^2 2\pi x\,ds = 2\pi\int_1^2 x\left(\frac{x^3}{2} + \frac{1}{2x^3}\right)dx = 2\pi\int_1^2 \left(\frac{x^4}{2} + \frac{1}{2x^2}\right)dx = 2\pi\Big[\frac{x^5}{10} - \frac{1}{2x}\Big]_1^2 = 2\pi\left(\frac{32}{10} - \frac{1}{4} - \frac{1}{10} + \frac{1}{2}\right) = 2\pi \cdot \frac{67}{20} = \frac{67\pi}{10}. Two errors cost the mark here: writing 2πx dx2\pi x\,dx instead of 2πx ds2\pi x\,ds, which gives the area of a flat ring, and taking the radius yy, which rotates about the other axis.

Exercise 7: How far from the target does the drone land?

A delivery drone aims at a target on the ground. In a model of its precision, the distance XX, in metres, between the target and the landing point has the density f(x)=kx(1+x2)2f(x) = \frac{kx}{(1 + x^2)^2} for x≥0x \ge 0, and f(x)=0f(x) = 0 for x<0x < 0. The figure shows ff for k=2k = 2.

0.511.522.533.544.550.10.20.30.40.50.60.70.8y = 2x/(1 + x²)²medianP(X > 2)
  • a) Find kk.
  • b) Find the cumulative distribution function FF, the probability that the drone lands more than 22 m from the target, and the median distance.
  • c) Find the mean distance, and explain why it is larger than the median.
Show the solution

Answers

  • a) ∫0∞x dx(1+x2)2=12\int_0^\infty \frac{x\,dx}{(1 + x^2)^2} = \frac{1}{2}, so k=2k = 2
  • b) F(x)=1−11+x2F(x) = 1 - \frac{1}{1 + x^2} for x≥0x \ge 0; P(X>2)=15P(X > 2) = \frac{1}{5}; median 11 m
  • c) μ=∫0∞2x2(1+x2)2 dx=π2\mu = \int_0^\infty \frac{2x^2}{(1 + x^2)^2}\,dx = \frac{\pi}{2} m, about 1.571.57 m

a) A density must be nonnegative, which holds for k>0k > 0 and x≥0x \ge 0, and have total area 11. With u=1+x2u = 1 + x^2, du=2x dxdu = 2x\,dx: ∫0tx dx(1+x2)2=12[−1u]11+t2=12(1−11+t2)→12\int_0^t \frac{x\,dx}{(1 + x^2)^2} = \frac{1}{2}\Big[-\frac{1}{u}\Big]_1^{1 + t^2} = \frac{1}{2}\left(1 - \frac{1}{1 + t^2}\right) \to \frac{1}{2} as t→∞t \to \infty. So k⋅12=1k \cdot \frac{1}{2} = 1 and k=2k = 2.

b) For x≥0x \ge 0, F(x)=∫0x2t dt(1+t2)2=1−11+x2F(x) = \int_0^x \frac{2t\,dt}{(1 + t^2)^2} = 1 - \frac{1}{1 + x^2}, and F(x)=0F(x) = 0 for x<0x < 0. Then P(X>2)=1−F(2)=11+4=15P(X > 2) = 1 - F(2) = \frac{1}{1 + 4} = \frac{1}{5}: one landing in five misses by more than 22 m, the shaded tail of the figure. The median mm solves F(m)=12F(m) = \frac{1}{2}: 11+m2=12\frac{1}{1 + m^2} = \frac{1}{2}, so m2=1m^2 = 1 and m=1m = 1 m, the positive root since X≥0X \ge 0. The median is NOT the peak of the density, which is at x=13x = \frac{1}{\sqrt{3}}: it is the point that splits the AREA in two.

c) μ=∫0∞xf(x) dx=∫0∞2x2(1+x2)2 dx\mu = \int_0^\infty x f(x)\,dx = \int_0^\infty \frac{2x^2}{(1 + x^2)^2}\,dx, an improper integral; it converges because the integrand behaves like 2x2\frac{2}{x^2} at infinity. By parts with u=xu = x and dv=2x dx(1+x2)2dv = \frac{2x\,dx}{(1 + x^2)^2}, v=−11+x2v = -\frac{1}{1 + x^2}: ∫0t2x2 dx(1+x2)2=[−x1+x2]0t+∫0tdx1+x2=−t1+t2+arctan⁡t→0+π2\int_0^t \frac{2x^2\,dx}{(1 + x^2)^2} = \Big[-\frac{x}{1 + x^2}\Big]_0^t + \int_0^t \frac{dx}{1 + x^2} = -\frac{t}{1 + t^2} + \arctan t \to 0 + \frac{\pi}{2}. So μ=π2≈1.57\mu = \frac{\pi}{2} \approx 1.57 m. (With x=tan⁡θx = \tan\theta the integrand becomes 2sin⁡2θ dθ2\sin^2\theta\,d\theta on [0,π2][0, \frac{\pi}{2}], the same π2\frac{\pi}{2}.) The mean exceeds the median 11 because the density has a long right tail: the rare landings far from the target pull the average up, but not the middle value.

Part C: sequences and series (/38)

Exercise 8: Five series, five verdicts, five tests

Decide whether each series converges or diverges. Two points each: name the test, CHECK its hypotheses, and state the conclusion. A verdict without its test earns nothing.

  • a) ∑n=1∞n+1nn+3\displaystyle\sum_{n=1}^{\infty} \frac{n + 1}{n\sqrt{n + 3}}
  • b) ∑n=1∞(2n)!n! nn\displaystyle\sum_{n=1}^{\infty} \frac{(2n)!}{n!\,n^n}
  • c) ∑n=1∞(nn+1)n2\displaystyle\sum_{n=1}^{\infty} \left(\frac{n}{n + 1}\right)^{n^2}
  • d) ∑n=2∞1nln⁡n\displaystyle\sum_{n=2}^{\infty} \frac{1}{n\sqrt{\ln n}}
  • e) ∑n=1∞(1n−sin⁡1n)\displaystyle\sum_{n=1}^{\infty} \left(\frac{1}{n} - \sin\frac{1}{n}\right)
Show the solution

Answers

  • a) Limit comparison with 1n\frac{1}{\sqrt{n}}, limit 11: diverges
  • b) Ratio test, L=4e>1L = \frac{4}{e} > 1: diverges
  • c) Root test, L=1e<1L = \frac{1}{e} < 1: converges
  • d) Integral test, ∫2∞dxxln⁡x=∞\int_2^\infty \frac{dx}{x\sqrt{\ln x}} = \infty: diverges
  • e) Terms positive; limit comparison with 1n3\frac{1}{n^3}, limit 16\frac{1}{6}: converges

a) Positive terms. For large nn the numerator behaves like nn and the denominator like n⋅n1/2n \cdot n^{1/2}, so compare with bn=1nb_n = \frac{1}{\sqrt{n}}: anbn=(n+1)nnn+3=1+1n1+3n→1\frac{a_n}{b_n} = \frac{(n + 1)\sqrt{n}}{n\sqrt{n + 3}} = \frac{1 + \frac{1}{n}}{\sqrt{1 + \frac{3}{n}}} \to 1, a finite positive limit. ∑1n\sum \frac{1}{\sqrt{n}} is a pp-series with p=12≤1p = \frac{1}{2} \le 1, divergent, so the series diverges by the limit comparison test. The terms do tend to 00: the divergence test says nothing here.

b) Factorials: the ratio test. an+1an=(2n+2)!(2n)!⋅n!(n+1)!⋅nn(n+1)n+1=(2n+2)(2n+1)n+1⋅nn(n+1)n+1=2(2n+1)n+1⋅(nn+1)n\frac{a_{n+1}}{a_n} = \frac{(2n + 2)!}{(2n)!} \cdot \frac{n!}{(n + 1)!} \cdot \frac{n^n}{(n + 1)^{n+1}} = \frac{(2n + 2)(2n + 1)}{n + 1} \cdot \frac{n^n}{(n + 1)^{n+1}} = \frac{2(2n + 1)}{n + 1} \cdot \left(\frac{n}{n + 1}\right)^n. The first factor tends to 44; the second is 1(1+1/n)n→1e\frac{1}{(1 + 1/n)^n} \to \frac{1}{e}. So L=4eL = \frac{4}{e}, and L>1L > 1 because e<4e < 4: the series diverges, its terms even tend to infinity. The two traps: writing (2n+2)!(2n)!=2n+2\frac{(2n + 2)!}{(2n)!} = 2n + 2, forgetting the factor 2n+12n + 1, and reading (nn+1)n\left(\frac{n}{n + 1}\right)^n as 11 because its base tends to 11.

c) The whole term is an nn-th power, an=[(nn+1)n]na_n = \left[\left(\frac{n}{n + 1}\right)^n\right]^n: the root test. ann=(nn+1)n=1(1+1n)n→1e<1\sqrt[n]{a_n} = \left(\frac{n}{n + 1}\right)^n = \frac{1}{\left(1 + \frac{1}{n}\right)^n} \to \frac{1}{e} < 1, so the series converges. Same trap as in b), one level up: the base nn+1\frac{n}{n + 1} tends to 11, but 1∞1^\infty is an indeterminate form, and here the exponent n2n^2 wins.

d) Let f(x)=1xln⁡xf(x) = \frac{1}{x\sqrt{\ln x}} on [2,∞)[2, \infty). Hypotheses: ff is continuous and positive there (ln⁡x≥ln⁡2>0\ln x \ge \ln 2 > 0), and decreasing, since xx and ln⁡x\sqrt{\ln x} are positive and increasing, so their product increases. With u=ln⁡xu = \ln x, du=dxxdu = \frac{dx}{x}: ∫2tdxxln⁡x=[2ln⁡x]2t=2ln⁡t−2ln⁡2→∞\int_2^t \frac{dx}{x\sqrt{\ln x}} = \Big[2\sqrt{\ln x}\Big]_2^t = 2\sqrt{\ln t} - 2\sqrt{\ln 2} \to \infty. The integral diverges, so the series diverges by the integral test, however slowly. Comparing with 1n\frac{1}{n} is useless here: 1nln⁡n<1n\frac{1}{n\sqrt{\ln n}} < \frac{1}{n} for n≥3n \ge 3, and being smaller than a divergent series proves nothing.

e) Sign first: for 0<t≤10 < t \le 1, sin⁡t<t\sin t < t, so every term 1n−sin⁡1n\frac{1}{n} - \sin\frac{1}{n} is positive. Size: with t=1nt = \frac{1}{n} and the Maclaurin series sin⁡t=t−t36+t5120−⋯\sin t = t - \frac{t^3}{6} + \frac{t^5}{120} - \cdots, t−sin⁡t=t36−t5120+⋯t - \sin t = \frac{t^3}{6} - \frac{t^5}{120} + \cdots, so compare with bn=1n3b_n = \frac{1}{n^3}: anbn=t−sin⁡tt3=16−t2120+⋯→16\frac{a_n}{b_n} = \frac{t - \sin t}{t^3} = \frac{1}{6} - \frac{t^2}{120} + \cdots \to \frac{1}{6} as n→∞n \to \infty. The limit is finite and positive and ∑1n3\sum \frac{1}{n^3} converges (p=3p = 3), so the series converges. Comparing with 1n\frac{1}{n}, the natural first guess, gives the limit 00, which decides nothing for a divergent benchmark.

Exercise 9: Two alternating series: a late decrease, then an error to control

Parts a) and b) are independent. The figure shows the partial sums S1,…,S8S_1, \dots, S_8 of the series of b), with its sum dashed.

1234567890.270.280.290.30.310.320.330.34ln(4/3)S₁S₂
  • a) Show that ∑n=1∞(−1)nn2n3+4\displaystyle\sum_{n=1}^{\infty} (-1)^n\frac{n^2}{n^3 + 4} converges, checking that its terms decrease in absolute value from some rank on (give the rank). Does it converge absolutely?
  • b) Let S=∑n=1∞(−1)n−1n 3nS = \displaystyle\sum_{n=1}^{\infty} \frac{(-1)^{n-1}}{n\,3^n}. How many terms guarantee an error of at most 10−310^{-3}? Give that partial sum as a fraction, say whether it is too large or too small, and give an interval that contains SS.
  • c) Find SS exactly, and check it against b).
Show the solution

Answers

  • a) f′(x)=x(8−x3)(x3+4)2<0f'(x) = \frac{x(8 - x^3)}{(x^3 + 4)^2} < 0 for x>2x > 2: decreasing from n=2n = 2; converges, conditionally (n2n3+4∼1n\frac{n^2}{n^3 + 4} \sim \frac{1}{n})
  • b) b5=11215<10−3<b4=1324b_5 = \frac{1}{1215} < 10^{-3} < b_4 = \frac{1}{324}: four terms; S4=31108S_4 = \frac{31}{108}, too small; 31108<S<31108+11215\frac{31}{108} < S < \frac{31}{108} + \frac{1}{1215}
  • c) S=ln⁡(1+13)=ln⁡43≈0.2877S = \ln\left(1 + \frac{1}{3}\right) = \ln\frac{4}{3} \approx 0.2877

a) Here bn=n2n3+4>0b_n = \frac{n^2}{n^3 + 4} > 0 and bn=1/n1+4/n3→0b_n = \frac{1/n}{1 + 4/n^3} \to 0. The decrease is not obvious, numerator and denominator both grow, so study f(x)=x2x3+4f(x) = \frac{x^2}{x^3 + 4}: f′(x)=2x(x3+4)−x2⋅3x2(x3+4)2=8x−x4(x3+4)2=x(8−x3)(x3+4)2f'(x) = \frac{2x(x^3 + 4) - x^2 \cdot 3x^2}{(x^3 + 4)^2} = \frac{8x - x^4}{(x^3 + 4)^2} = \frac{x(8 - x^3)}{(x^3 + 4)^2}, negative exactly when x3>8x^3 > 8, that is x>2x > 2. So (bn)(b_n) decreases from n=2n = 2 on: b1=15<b2=13b_1 = \frac{1}{5} < b_2 = \frac{1}{3}, then b3=931<13b_3 = \frac{9}{31} < \frac{1}{3}, and so on. The alternating series test only needs the decrease EVENTUALLY, since the first term does not affect convergence: the series converges.

Absolutely? ∑bn\sum b_n against 1n\frac{1}{n}: bn1/n=n3n3+4→1\frac{b_n}{1/n} = \frac{n^3}{n^3 + 4} \to 1, and the harmonic series diverges, so ∑bn\sum b_n diverges by limit comparison. The series converges conditionally, not absolutely.

b) bn=1n 3nb_n = \frac{1}{n\,3^n} is positive, decreasing (both nn and 3n3^n increase) and tends to 00, so the alternating series estimate applies: ∣S−SN∣≤bN+1|S - S_N| \le b_{N+1}. The first omitted term is the one to test: b4=14⋅81=1324>10−3b_4 = \frac{1}{4 \cdot 81} = \frac{1}{324} > 10^{-3}, while b5=15⋅243=11215<10−3b_5 = \frac{1}{5 \cdot 243} = \frac{1}{1215} < 10^{-3}. So N=4N = 4 terms suffice (and three do not guarantee it). S4=13−118+181−1324=108−18+4−1324=93324=31108S_4 = \frac{1}{3} - \frac{1}{18} + \frac{1}{81} - \frac{1}{324} = \frac{108 - 18 + 4 - 1}{324} = \frac{93}{324} = \frac{31}{108}.

The error S−S4S - S_4 has the sign of the first omitted term, +11215+\frac{1}{1215}: S4S_4 is too SMALL. Hence 31108<S<31108+11215\frac{31}{108} < S < \frac{31}{108} + \frac{1}{1215}. The figure shows it: the odd partial sums sit above the dashed line, the even ones below, and S4S_4 is an even one.

c) For −1<x≤1-1 < x \le 1, ln⁡(1+x)=∑n=1∞(−1)n−1xnn\ln(1 + x) = \sum_{n=1}^{\infty} \frac{(-1)^{n-1}x^n}{n}. At x=13x = \frac{1}{3}, inside the interval, this is exactly our series: S=ln⁡43S = \ln\frac{4}{3}. Check: ln⁡43=2ln⁡2−ln⁡3≈1.3863−1.0986=0.2877\ln\frac{4}{3} = 2\ln 2 - \ln 3 \approx 1.3863 - 1.0986 = 0.2877, while 31108≈0.2870\frac{31}{108} \approx 0.2870: the difference, about 0.00060.0006, is positive and below 11215≈0.0008\frac{1}{1215} \approx 0.0008, as b) promised.

Exercise 10: An interval with its two endpoints, and a series found by differentiating

Parts a) and b) are independent.

  • a) Find the centre, the radius and the interval of convergence of ∑n=1∞(x+1)n3nn2+1\displaystyle\sum_{n=1}^{\infty} \frac{(x + 1)^n}{3^n\sqrt{n^2 + 1}}. Test each endpoint with its own test.
  • b) Starting from the geometric series of 11+x\frac{1}{1 + x}, find a power series for 1(1+x)3\frac{1}{(1 + x)^3} centred at 00, with its radius. Deduce the exact sum of ∑n=0∞(−1)n(n+1)(n+2)3n\displaystyle\sum_{n=0}^{\infty} (-1)^n\frac{(n + 1)(n + 2)}{3^n}.
Show the solution

Answers

  • a) Centre −1-1, radius 33; diverges at x=2x = 2, converges conditionally at x=−4x = -4: interval [−4,2)[-4, 2)
  • b) 1(1+x)3=∑n=0∞(−1)n(n+1)(n+2)2xn\frac{1}{(1 + x)^3} = \sum_{n=0}^{\infty} (-1)^n\frac{(n + 1)(n + 2)}{2}x^n, R=1R = 1; the sum is 2(4/3)3=2732\frac{2}{(4/3)^3} = \frac{27}{32}

a) Centre a=−1a = -1, read in the power (x−(−1))n(x - (-1))^n. Ratio test on the absolute values, x≠−1x \ne -1 fixed: ∣an+1an∣=∣x+1∣3⋅n2+1(n+1)2+1→∣x+1∣3\left|\frac{a_{n+1}}{a_n}\right| = \frac{|x + 1|}{3} \cdot \frac{\sqrt{n^2 + 1}}{\sqrt{(n + 1)^2 + 1}} \to \frac{|x + 1|}{3}. The series converges absolutely when ∣x+1∣<3|x + 1| < 3 and diverges when ∣x+1∣>3|x + 1| > 3: radius R=3R = 3, and the open interval is (−4,2)(-4, 2). The ratio test is silent at the two endpoints, where the limit is exactly 11, so each is tested separately.

At x=2x = 2: x+1=3x + 1 = 3 and the series is ∑1n2+1\sum \frac{1}{\sqrt{n^2 + 1}}, positive terms. Limit comparison with 1n\frac{1}{n}: nn2+1=11+1/n2→1\frac{n}{\sqrt{n^2 + 1}} = \frac{1}{\sqrt{1 + 1/n^2}} \to 1, and the harmonic series diverges, so the series diverges at x=2x = 2. At x=−4x = -4: x+1=−3x + 1 = -3 and the series is ∑(−1)nn2+1\sum \frac{(-1)^n}{\sqrt{n^2 + 1}}. Alternating series test: bn=1n2+1b_n = \frac{1}{\sqrt{n^2 + 1}} is positive, decreasing because n2+1n^2 + 1 increases, and tends to 00: it converges, conditionally by the computation at x=2x = 2. The interval of convergence is [−4,2)[-4, 2).

b) For ∣x∣<1|x| < 1, 11+x=∑n=0∞(−1)nxn\frac{1}{1 + x} = \sum_{n=0}^{\infty} (-1)^n x^n. Differentiate twice, term by term, which keeps the radius 11: −1(1+x)2=∑n=1∞(−1)nnxn−1\frac{-1}{(1 + x)^2} = \sum_{n=1}^{\infty} (-1)^n n x^{n-1} and 2(1+x)3=∑n=2∞(−1)nn(n−1)xn−2\frac{2}{(1 + x)^3} = \sum_{n=2}^{\infty} (-1)^n n(n - 1)x^{n-2}. Shift the index with m=n−2m = n - 2, so (−1)n=(−1)m(-1)^n = (-1)^m: 2(1+x)3=∑m=0∞(−1)m(m+2)(m+1)xm\frac{2}{(1 + x)^3} = \sum_{m=0}^{\infty} (-1)^m (m + 2)(m + 1)x^m, hence 1(1+x)3=∑m=0∞(−1)m(m+1)(m+2)2xm\frac{1}{(1 + x)^3} = \sum_{m=0}^{\infty} (-1)^m\frac{(m + 1)(m + 2)}{2}x^m for ∣x∣<1|x| < 1. Check the first terms against the binomial series: 1−3x+6x2−⋯1 - 3x + 6x^2 - \cdots.

The numerical series is ∑(−1)n(n+1)(n+2)(13)n=2⋅1(1+1/3)3\sum (-1)^n (n + 1)(n + 2)\left(\frac{1}{3}\right)^n = 2 \cdot \frac{1}{(1 + 1/3)^3}, evaluated at x=13x = \frac{1}{3}, which is inside the interval ∣x∣<1|x| < 1. So the sum is 2⋅2764=27322 \cdot \frac{27}{64} = \frac{27}{32}. Forgetting the factor 22 of the second derivative, or evaluating at x=−13x = -\frac{1}{3} because of the sign (−1)n(-1)^n, are the two ways to lose this mark.

Exercise 11: A Maclaurin series from the table, read three ways

Let f(x)=xln⁡(1+x2)f(x) = x\ln(1 + x^2). Use the Maclaurin series of the table; no derivative of ff is to be computed by hand.

  • a) Find the Maclaurin series of ff, its general term, and its interval of convergence.
  • b) Find f(7)(0)f^{(7)}(0) and f(8)(0)f^{(8)}(0).
  • c) Compute lim⁡x→0xln⁡(1+x2)−xsin⁡2xx5\displaystyle\lim_{x \to 0} \frac{x\ln(1 + x^2) - x\sin^2 x}{x^5}.
Show the solution

Answers

  • a) f(x)=∑n=1∞(−1)n−1x2n+1n=x3−x52+x73−⋯f(x) = \sum_{n=1}^{\infty} \frac{(-1)^{n-1}x^{2n+1}}{n} = x^3 - \frac{x^5}{2} + \frac{x^7}{3} - \cdots on [−1,1][-1, 1]
  • b) f(7)(0)=7!3=1680f^{(7)}(0) = \frac{7!}{3} = 1680; f(8)(0)=0f^{(8)}(0) = 0
  • c) −12+13=−16-\frac{1}{2} + \frac{1}{3} = -\frac{1}{6}

a) The table gives ln⁡(1+u)=∑n=1∞(−1)n−1unn\ln(1 + u) = \sum_{n=1}^{\infty} \frac{(-1)^{n-1}u^n}{n} for −1<u≤1-1 < u \le 1. Substitute u=x2u = x^2: ln⁡(1+x2)=∑n=1∞(−1)n−1x2nn\ln(1 + x^2) = \sum_{n=1}^{\infty} \frac{(-1)^{n-1}x^{2n}}{n}, then multiply by xx: f(x)=∑n=1∞(−1)n−1x2n+1n=x3−x52+x73−⋯f(x) = \sum_{n=1}^{\infty} \frac{(-1)^{n-1}x^{2n+1}}{n} = x^3 - \frac{x^5}{2} + \frac{x^7}{3} - \cdots. The interval moves with the variable: the condition is −1<x2≤1-1 < x^2 \le 1, and x2>−1x^2 > -1 always holds, so it is ∣x∣≤1|x| \le 1. At x=±1x = \pm 1 the series is ±∑(−1)n−1n\pm\sum \frac{(-1)^{n-1}}{n}, the alternating harmonic series, convergent, which confirms both endpoints. For ∣x∣>1|x| > 1 the terms ∣x∣2n+1n\frac{|x|^{2n+1}}{n} tend to infinity. Interval: [−1,1][-1, 1].

b) The series of ff at 00 IS its Taylor series, so the coefficient of xkx^k is f(k)(0)k!\frac{f^{(k)}(0)}{k!}. The coefficient of x7x^7 is the term n=3n = 3, (−1)23=13\frac{(-1)^2}{3} = \frac{1}{3}, so f(7)(0)=7!3=50403=1680f^{(7)}(0) = \frac{7!}{3} = \frac{5040}{3} = 1680. There is no x8x^8: every power is odd, as it must be for the odd function ff, so f(8)(0)=0f^{(8)}(0) = 0. Seven differentiations of xln⁡(1+x2)x\ln(1 + x^2) by the product and chain rules would take a page.

c) Expand both terms up to x5x^5. From a): xln⁡(1+x2)=x3−x52+O(x7)x\ln(1 + x^2) = x^3 - \frac{x^5}{2} + O(x^7). For sin⁡2x\sin^2 x, use sin⁡2x=1−cos⁡2x2\sin^2 x = \frac{1 - \cos 2x}{2} and the series of cosine: cos⁡2x=1−2x2+2x43−⋯\cos 2x = 1 - 2x^2 + \frac{2x^4}{3} - \cdots, so sin⁡2x=x2−x43+O(x6)\sin^2 x = x^2 - \frac{x^4}{3} + O(x^6) and xsin⁡2x=x3−x53+O(x7)x\sin^2 x = x^3 - \frac{x^5}{3} + O(x^7). The x3x^3 terms cancel, which is why the expansion must go one step further: the numerator is −x52+x53+O(x7)=−x56+O(x7)-\frac{x^5}{2} + \frac{x^5}{3} + O(x^7) = -\frac{x^5}{6} + O(x^7), and the limit is −16-\frac{1}{6}. Stopping at x3x^3 gives 0x5\frac{0}{x^5} and no answer; squaring only the first term of sin⁡x\sin x loses the −x43-\frac{x^4}{3}.

Exercise 12: Taylor's inequality away from zero: arctan 1.1 by hand

The figure shows y=arctan⁡xy = \arctan x and its Taylor polynomial T2T_2 of degree 22 centred at a=1a = 1 (dashed).

-0.50.511.522.53-0.4-0.20.20.40.60.811.21.4y = arctan xy = T₂(x)(1, π/4)
  • a) Compute T2(x)T_2(x) for f(x)=arctan⁡xf(x) = \arctan x at a=1a = 1.
  • b) Use it to approximate arctan⁡1.1\arctan 1.1, then bound the error with Taylor's inequality, justifying your MM without a calculator.
  • c) Is the approximation too large or too small? Give an interval that contains arctan⁡1.1\arctan 1.1.
  • d) Why not simply use the Maclaurin series x−x33+x55−⋯x - \frac{x^3}{3} + \frac{x^5}{5} - \cdots at x=1.1x = 1.1?
Show the solution

Answers

  • a) T2(x)=π4+x−12−(x−1)24T_2(x) = \frac{\pi}{4} + \frac{x - 1}{2} - \frac{(x - 1)^2}{4}
  • b) arctan⁡1.1≈π4+19400\arctan 1.1 \approx \frac{\pi}{4} + \frac{19}{400}; f′′′≤1f''' \le 1 on [1,1.1][1, 1.1], ∣R2∣≤(0.1)36=16000|R_2| \le \frac{(0.1)^3}{6} = \frac{1}{6000}
  • c) f′′′>0f''' > 0: too small; π4+19400<arctan⁡1.1<π4+19400+16000\frac{\pi}{4} + \frac{19}{400} < \arctan 1.1 < \frac{\pi}{4} + \frac{19}{400} + \frac{1}{6000}
  • d) That series has radius 11 and diverges at 1.11.1

a) f(x)=arctan⁡xf(x) = \arctan x, f′(x)=11+x2f'(x) = \frac{1}{1 + x^2}, f′′(x)=−2x(1+x2)2f''(x) = -\frac{2x}{(1 + x^2)^2}. At a=1a = 1: f(1)=π4f(1) = \frac{\pi}{4}, f′(1)=12f'(1) = \frac{1}{2}, f′′(1)=−24=−12f''(1) = -\frac{2}{4} = -\frac{1}{2}. So T2(x)=f(1)+f′(1)(x−1)+f′′(1)2!(x−1)2=π4+x−12−(x−1)24T_2(x) = f(1) + f'(1)(x - 1) + \frac{f''(1)}{2!}(x - 1)^2 = \frac{\pi}{4} + \frac{x - 1}{2} - \frac{(x - 1)^2}{4}. The division by 2!2! is the usual slip: without it, the last coefficient would be −12-\frac{1}{2}.

b) With x=1.1x = 1.1, x−1=0.1x - 1 = 0.1: T2(1.1)=π4+0.05−0.0025=π4+19400T_2(1.1) = \frac{\pi}{4} + 0.05 - 0.0025 = \frac{\pi}{4} + \frac{19}{400}. The answer keeps π\pi: it is exact up to the Taylor error, and no calculator is needed. For the bound, f′′′(x)=6x2−2(1+x2)3f'''(x) = \frac{6x^2 - 2}{(1 + x^2)^3}. On the WHOLE interval [1,1.1][1, 1.1]: the numerator is at most 6(1.21)−2=5.266(1.21) - 2 = 5.26 and the denominator at least 23=82^3 = 8, so 0<f′′′(x)≤5.268<10 < f'''(x) \le \frac{5.26}{8} < 1. Take M=1M = 1: ∣R2(1.1)∣≤M3!∣1.1−1∣3=16⋅11000=16000|R_2(1.1)| \le \frac{M}{3!}|1.1 - 1|^3 = \frac{1}{6} \cdot \frac{1}{1000} = \frac{1}{6000}.

c) Taylor's formula with the Lagrange remainder writes R2(1.1)=f′′′(c)6(0.1)3R_2(1.1) = \frac{f'''(c)}{6}(0.1)^3 for some cc between 11 and 1.11.1, and f′′′(c)>0f'''(c) > 0 there since 6c2−2≥46c^2 - 2 \ge 4. So the error is positive: T2(1.1)T_2(1.1) is too SMALL, and π4+19400<arctan⁡1.1<π4+19400+16000\frac{\pi}{4} + \frac{19}{400} < \arctan 1.1 < \frac{\pi}{4} + \frac{19}{400} + \frac{1}{6000}. With π≈3.1416\pi \approx 3.1416 that is between 0.83280.8328 and 0.83310.8331. The figure agrees: just to the right of 11 the dashed parabola passes under the curve. No alternating series is in sight here, so Taylor's inequality, with the sign of f′′′f''', is the tool that gives both the bound and the side.

d) The series ∑(−1)nx2n+12n+1\sum \frac{(-1)^n x^{2n+1}}{2n + 1} has radius 11: at x=1.1x = 1.1 its terms (1.1)2n+12n+1\frac{(1.1)^{2n+1}}{2n + 1} tend to infinity, so it DIVERGES, although arctan⁡1.1\arctan 1.1 is a perfectly good number. A series centred at 00 says nothing beyond its radius. This is why the polynomial is centred at a=1a = 1, the nearest point where arctan⁡\arctan and its derivatives are known exactly, and why the error bound shrinks with ∣x−a∣3=10−3|x - a|^3 = 10^{-3}.

See also

Preparing for the MATH 141 final?

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