MATH 141 practice midterm with full solutions (McGill)
This is a practice midterm for MATH 141, Calculus 2, the integral calculus course of the first year at McGill University. It covers the first nine chapters of the course, the usual scope of the midterm: Riemann sums and the definite integral, the Fundamental Theorem of Calculus, the substitution rule, integration by parts, trigonometric integrals, trigonometric substitution, partial fractions, the strategy for a bare integral, and improper integrals. Eight questions, one hundred points, two parts: four short questions worth forty-two points, then four long problems worth fifty-eight.
Sit it as an exam: two hours on a timer, no calculator, no notes, because that is how the real midterm is written. Every answer is exact, ln2, 16π, arctan21, and every mark goes to the method: name the substitution with its differential and its new bounds, name u and dv, write the form of a decomposition before its constants, and state each limit of an improper integral before computing it. Not one question repeats an exercise of the nine chapter sets of this site: the gestures are the ones the examiners ask for, the integrands are new, so the paper measures what you can do and not what you remember having read. Each question has its Answers box for a first quick marking, and the full reasoning underneath for the second pass.
The traps named in the solutions: calling the left sum an overestimate without looking at monotonicity, forgetting a chain factor when both bounds move, keeping the bounds in x after a substitution, treating a leftover x3 as a constant, writing tan2θ=tanθ on the wrong branch, stopping at the first root of a quartic, evaluating across a hidden asymptote, splitting an improper integral into ∞−∞, and comparing with a divergent integral from BELOW.
Exercise 1: Right sums, the definition, and a limit read as an integral
The figure shows the region under the graph of f(x)=x2+2x for 0≤x≤2, and the four rectangles of the right Riemann sum R4. Let A be the area of the region. In parts a) and b), no antiderivative may be used.
a) Compute L4 and R4. Which one overestimates A? Justify with a property of f, and write the resulting bracket for A.
b) Use the definition of the integral, with right endpoints and the formulas for ∑i and ∑i2, to show that ∫02(x2+2x)dx=320. Check your formula for Rn at n=4.
c) Express limn→∞∑i=1nn+i1 as a definite integral, and evaluate it.
d) Evaluate ∫−22(x+4−x2)dx by geometry and the properties of the integral only.
Show the solution
Answers
a)L4=419, R4=435; f is increasing on [0,2], so R4 overestimates: 419<A<435
b)Rn=3n24(n+1)(5n+1)→320; R4=435
c)∫011+xdx=ln2
d)0+2π=2π
a) Δx=42−0=21. The left sum samples 0,21,1,23 and the right sum 21,1,23,2. The values are f(0)=0, f(21)=45, f(1)=3, f(23)=421, f(2)=8. So L4=21(0+45+3+421)=21⋅219=419 and R4=21(45+3+421+8)=21⋅235=435.
The property that decides is monotonicity, not the picture: f′(x)=2x+2>0 on [0,2], so f is increasing. On each strip the left end gives the SMALLEST value and the right end the LARGEST, so every left rectangle sits under the curve and every right rectangle sticks out above it. Hence L4≤A≤R4, that is 419≤A≤435, and R4 is the overestimate. The gap R4−L4=Δx[f(2)−f(0)]=4 is wide: four strips do not pin A down.
b) With n strips, Δx=n2 and xi=n2i, so f(xi)=n24i2+n4i. Then Rn=∑i=1nn2(n24i2+n4i)=n38∑i=1ni2+n28∑i=1ni. The sums start at i=1 and end at n, so the formulas apply: Rn=n38⋅6n(n+1)(2n+1)+n28⋅2n(n+1)=3n24(n+1)(2n+1)+n4(n+1)=3n24(n+1)(5n+1).
Check at n=4: 484⋅5⋅21=48420=435, the value of a). Now write Rn=34(1+n1)(5+n1) and let n→∞: ∫02(x2+2x)dx=34⋅5=320. It lies in the bracket of a), since 419=4.75<6.67<8.75.
c) Factor out n1: n+i1=n1⋅1+ni1. Read it in order: Δx=n1, so b−a=1; the sample point is xi=ni=0+iΔx, so a=0 and b=1; the function is f(x)=1+x1, continuous on [0,1]. The limit is ∫011+xdx=[ln(1+x)]01=ln2. Another correct reading is xi=1+ni with f(x)=x1 on [1,2], which gives ∫12xdx=ln2 as well. Sanity check: each of the n terms lies between 2n1 and n1, so the sum lies between 21 and 1, and ln2≈0.69 does.
d) By linearity, ∫−22(x+4−x2)dx=∫−22xdx+∫−224−x2dx. The first integral is the signed area between the line y=x and the axis: a triangle of area 2 below the axis on [−2,0] and one of area 2 above it on [0,2], so it is −2+2=0. The graph of y=4−x2 is the upper half of the circle x2+y2=4, so the second integral is half the area of a disc of radius 2: 21π⋅22=2π. The value is 2π. Note that the integrand itself is negative near x=−2 (it equals −2 there): the integral is a signed area, and nothing here asks for an area.
Exercise 2: Two moving bounds, then substitution with its bounds
Parts a) to d) are independent. The function tsint has no elementary antiderivative; you may use this fact without proof.
a) Let F(x)=∫xx2tsintdt for x>0. Find F′(x), then the equation of the tangent line to the graph of F at x=1.
b) Evaluate ∫0π/21+sinxcosxdx.
c) Evaluate ∫011+x3x5dx with the substitution u=1+x3. Explain how the factor x5 is handled.
d) A classmate writes the integral of c) as 31∫01uu−1du. Find the number she obtains and explain, without redoing c), why it cannot be right. Then bracket the integral of c) between two integrals of x5 and check your answer against the bracket.
d)She gets −94<0 for a positive integrand; 621≤94−22≤61
a) Let g(t)=tsint, continuous on (0,∞). For x>0 both bounds x and x2 lie in (0,∞), so the integral can be split at the constant 1: F(x)=∫1x2g(t)dt−∫1xg(t)dt. FTC 1 and the chain rule on each piece: F′(x)=g(x2)⋅2x−g(x)⋅2x1=x2sin(x2)⋅2x−xsinx⋅2x1=x2sin(x2)−2xsinx. Looking for an antiderivative first would be a dead end: there is none in elementary terms, and FTC 1 does not need one.
At x=1 both bounds equal 1, so F(1)=∫11g(t)dt=0, and F′(1)=2sin1−21sin1=23sin1. The tangent line is y=23(sin1)(x−1). The two traps: forgetting the chain factors 2x and 2x1, and forgetting the minus sign of the lower bound.
b) The factor cosx is the derivative of 1+sinx: set u=1+sinx, du=cosxdx. Carry the bounds: x=0 gives u=1 and x=2π gives u=2. The integral becomes ∫12u−1/2du=[2u]12=22−2. Size check: the integrand lies between 0 and 1 on an interval of length 2π, and 22−2≈0.83 does.
c) With u=1+x3, du=3x2dx absorbs only x2. Split x5=x3⋅x2: the x2dx becomes 3du, and the remaining x3 is NOT a constant, it is rewritten from u=1+x3 as x3=u−1. The bounds: x=0 gives u=1, x=1 gives u=2. So ∫011+x3x5dx=31∫12uu−1du=31∫12(u1/2−u−1/2)du=31[32u3/2−2u1/2]12.
At u=2: 32⋅22−22=−322; at u=1: 32−2=−34. The difference is 34−22, and the integral is 94−22.
d) She obtains 31[32u3/2−2u1/2]01=31(32−2)=−94. The integrand 1+x3x5 is nonnegative on [0,1], so its integral cannot be negative: the sign alone condemns the answer. Her integrand is in u but her bounds are still in x, whereas u runs from 1 to 2.
Bracket: on [0,1], 1≤1+x3≤2, so 2x5≤1+x3x5≤x5, and integrating, 621≤I≤61. With 2≈1.41 that is 0.118≤I≤0.167, and 94−22≈94−2.83≈0.13 fits.
Exercise 3: Four integrals, no technique announced
Evaluate each integral exactly. Two points each: the first line must NAME the move (the substitution with its differential and its bounds, the choice of u and dv, or the identity used), then the computation follows.
a) ∫0π/3cos2xsin3xdx
b) ∫0π/4xsec2xdx
c) ∫0π/2sin2xcos2xdx
d) ∫19x+xdx
Show the solution
Answers
a)21
b)4π−21ln2
c)16π
d)2ln2
a) Move: an ODD power of sine, so detach one sinxdx and set u=cosx, du=−sinxdx, converting sin2x=1−u2. The integrand is continuous on [0,3π], where cosx≥21. Bounds: x=0 gives u=1, x=3π gives u=21. Then ∫0π/3cos2xsin2xsinxdx=∫11/2u21−u2(−du)=∫1/21(u−2−1)du=[−u1−u]1/21=(−2)−(−25)=21.
Second route, as a check: cos2xsin3x=cos2xsinx(1−cos2x)=secxtanx−sinx, whose antiderivative is secx+cosx; from 0 to 3π it gives (2+21)−(1+1)=21.
b) Move: integration by parts with u=x, du=dx, dv=sec2xdx, v=tanx. The polynomial is u because differentiating it removes it, and sec2x is dv because it integrates at sight; the opposite choice would need v=2x2 and raise the power. Then ∫0π/4xsec2xdx=[xtanx]0π/4−∫0π/4tanxdx=4π−[ln∣secx∣]0π/4=4π−ln2=4π−21ln2. Size check: about 0.79−0.35=0.44, positive as it must be.
c) Move: only EVEN powers, so no factor can be saved; lower the degree. First sinxcosx=21sin2x, so sin2xcos2x=41sin22x, then the half-angle formula sin22x=21−cos4x: the integrand is 81(1−cos4x). Then ∫0π/281(1−cos4x)dx=81[x−4sin4x]0π/2=81⋅2π=16π, since sin2π=0.
d) Move: the root blocks everything, so rationalize with u=x, x=u2, dx=2udu; the WHOLE of dx is converted. Bounds: x=1 gives u=1, x=9 gives u=3. Then ∫19x+xdx=∫13u2+u2udu=∫13u+12du, because u2+u=u(u+1) and the factor u cancels: no partial fractions are needed once the fraction is simplified. The value is 2[ln(u+1)]13=2(ln4−ln2)=2ln2.
Exercise 4: Converge or diverge: four improper integrals
For each integral, say first WHERE it is improper, then write the limit or limits, and evaluate the integral or show that it diverges.
a) ∫0∞x2+4x+5dx
b) ∫01xlnxdx
c) ∫1e2x3(lnx−1)2dx
d) ∫01x2−4x+3dx
Show the solution
Answers
a)Converges: 2π−arctan2=arctan21
b)Converges: −4
c)Improper at x=e, inside the interval; converges: 3+3=6
d)Improper at x=1; diverges
a) Improper at ∞ only: x2+4x+5=(x+2)2+1>0, so the integrand is continuous on [0,∞). Completing the square gives the arctangent: ∫0t(x+2)2+1dx=arctan(t+2)−arctan2, and as t→∞, arctan(t+2)→2π. The integral converges to 2π−arctan2, which is also arctan21 since arctana+arctana1=2π for a>0. It is an exact answer: no calculator is needed, nor wanted.
b) Improper at 0: xlnx→−∞ as x→0+. By parts with u=lnx, dv=x−1/2dx, v=2x: ∫xlnxdx=2xlnx−∫x2xdx=2xlnx−4x+C. So ∫t1xlnxdx=−4−2tlnt+4t. The limit limt→0+tlnt is of the form 0⋅(−∞): write it t−1/2lnt, of the form ∞−∞, and L'Hôpital's rule gives −21t−3/21/t=−2t→0. The integral converges to −4, negative as it must be, since lnx<0 on (0,1).
c) Both endpoints are harmless (lnx−1=−1 at x=1 and 1 at x=e2), but lnx−1=0 at x=e, which lies INSIDE [1,e2]: the integrand blows up there. Split at e, each piece a limit: ∫1e2=lims→e−∫1s+limr→e+∫re2. With u=lnx−1, du=xdx, an antiderivative is 33u=33lnx−1, valid on each side of e (the cube root of a negative number is real). Left piece: lims→e−(33lns−1)−33−1=0+3=3. Right piece: 331−limr→e+33lnr−1=3. Both converge, so the integral converges to 6.
Evaluating 33lnx−1 from 1 to e2 in one line also prints 6, by luck: that line uses FTC 2 across a point where the integrand is not continuous. The marks go to the split and the two limits, because the same shortcut prints a finite number for integrals that diverge.
d) Look at the endpoints: x2−4x+3=(x−1)(x−3) vanishes at x=1, an endpoint, so the integral is improper at 1. Partial fractions: (x−1)(x−3)1=21(x−31−x−11). Then ∫0t(x−1)(x−3)dx=21[ln∣x−3∣−ln∣x−1∣]0t=21(ln(3−t)−ln(1−t)−ln3). As t→1−, −ln(1−t)→+∞: the integral diverges. Evaluating from 0 to 1 directly would ask for ln0, which is exactly the warning that a limit was needed.
Part B: long problems (/58)
Exercise 5: Integration by parts, and a reduction formula for powers of sine
For n≥0, let Wn=∫0π/2sinnxdx. X Part a) is independent of the rest.
a) Evaluate ∫1e2xlnxdx. Name u and dv, and say why the opposite choice is worse.
b) Prove that, for n≥2, ∫sinnxdx=−n1sinn−1xcosx+nn−1∫sinn−2xdx.
c) Deduce that Wn=nn−1Wn−2 for n≥2. Compute W0 and W1, then W4 and W5 exactly.
d) Explain with the figure why W5≤W4≤W3, and check your values. Then confirm W4 by a half-angle computation.
e) Evaluate ∫0π/2cos5xdx with no new integration by parts.
Show the solution
Answers
a)u=lnx, dv=xdx: 98e3+4
b)u=sinn−1x, dv=sinxdx, then cos2x=1−sin2x: the original integral reappears and is solved for
c)W0=2π, W1=1, W4=163π, W5=158
d)0≤sinx≤1 gives sinn+1x≤sinnx; 0.53≤0.59≤0.67
e)x=2π−t turns it into W5=158
a) Take u=lnx, du=xdx, and dv=x1/2dx, v=32x3/2. The logarithm is u because its derivative is algebraic and kills the logarithm; the opposite choice, dv=lnxdx, would need v=xlnx−x and leave ∫2xxlnx−xdx, no simpler than the start. Then ∫1e2xlnxdx=[32x3/2lnx]1e2−32∫1e2x1/2dx. At x=e2, x3/2=e3 and lnx=2, so the bracket is 34e3−0. The new integral is 32⋅32[x3/2]1e2=94(e3−1). The value is 34e3−94e3+94=98e3+4.
Size check: on [1,e2] the integrand grows from 0 to 2e, over an interval of length e2−1; so the value lies between 0 and 2e(e2−1)≈5.4×6.4≈35, and 98e3+4≈98×20+4≈18 does (with e3≈20).
b) Write sinnx=sinn−1x⋅sinx and take u=sinn−1x, du=(n−1)sinn−2xcosxdx, dv=sinxdx, v=−cosx. Then ∫sinnxdx=−sinn−1xcosx+(n−1)∫sinn−2xcos2xdx. Convert cos2x=1−sin2x: ∫sinnxdx=−sinn−1xcosx+(n−1)∫sinn−2xdx−(n−1)∫sinnxdx.
The integral we started with reappears on the right with the coefficient −(n−1). Treat it as an unknown and bring it to the left: n∫sinnxdx=−sinn−1xcosx+(n−1)∫sinn−2xdx. Dividing by n gives the formula. The hypothesis n≥2 makes sinn−2x an honest power, so the right-hand side is a genuine integral of lower degree.
c) On [0,2π] the bracket vanishes at both ends: sinn−1xcosx is 0 at x=2π because cos2π=0, and 0 at x=0 because sin0=0 and n−1≥1. So Wn=nn−1Wn−2. The two starting values are W0=∫0π/2dx=2π and W1=[−cosx]0π/2=1. Even indices come from W0: W2=21⋅2π=4π, W4=43⋅4π=163π. Odd indices come from W1: W3=32, W5=54⋅32=158. Note that an even n always ends on π and an odd n never does.
d) On [0,2π], 0≤sinx≤1, so multiplying sinnx by sinx can only lower it: sinn+1x≤sinnx, which is the order of the curves on the figure. Integrating, Wn+1≤Wn. The values agree: W5=158≈0.53, W4=163π≈0.59, W3=32≈0.67 (with π≈3.14).
Half-angle check: sin4x=(21−cos2x)2=41−21cos2x+41cos22x=83−21cos2x+81cos4x. On [0,2π] the two cosines integrate to 0 (sinπ=sin2π=0), so W4=83⋅2π=163π. Same value, by a longer road.
e) Substitute x=2π−t, dx=−dt: x=0 gives t=2π, x=2π gives t=0, and cos(2π−t)=sint. So ∫0π/2cos5xdx=∫π/20sin5t(−dt)=W5=158. The odd-power method agrees: with u=sinx, ∫01(1−u2)2du=1−32+51=158.
Exercise 6: Three trigonometric substitutions, one of them after completing the square
In each part, state the substitution with the interval of θ, and justify the sign of every root that comes out of it. The figure belongs to part d): it shows y=x2−4x+81, the region under it for 2≤x≤3, and dashed segments at the heights 21 and 51 over that interval.
a) Evaluate ∫22x3x2−1dx, changing the bounds with the variable.
b) Deduce ∫−2−2x3x2−1dx without any substitution, and explain the sign.
c) Find ∫(x2+4)3/2x2dx, returning to x with a right triangle. Check your answer by differentiating it.
d) Evaluate ∫23x2−4x+8dx, then use the figure to check that the value is plausible, without a calculator.
Show the solution
Answers
a)x=secθ: ∫π/4π/3cos2θdθ=24π+83−41
b)The integrand is odd: −24π−83+41, negative since x3<0
c)ln(x+x2+4)−x2+4x+C
d)ln21+5, between 51 and 21
a) The radicand has the form x2−a2 with a=1 and x≥1: set x=secθ with 0≤θ<2π. Then dx=secθtanθdθ and x2−1=tan2θ=∣tanθ∣=tanθ, since tanθ≥0 on that interval. Bounds: secθ=2 gives θ=4π, and secθ=2 gives θ=3π. The integrand times dx becomes sec3θtanθsecθtanθdθ=cos2θdθ.
With the half-angle formula, ∫π/4π/3cos2θdθ=[2θ+4sin2θ]π/4π/3=21(3π−4π)+41(23−1)=24π+83−41. Size check: about 0.131+0.216−0.25≈0.10. The integrand decreases from 221≈0.35 to 831≈0.07 on an interval of length 2−2≈0.59, so the value lies between 0.04 and 0.21: it does.
b) Let g(x)=x3x2−11. Then g(−x)=−g(x): the root is even and x3 is odd. With u=−x, ∫−2−2g(x)dx=∫22g(−u)(−du)=∫22g(u)du=−∫22g(u)du=−24π−83+41. The sign is right: on [−2,−2], x3<0, so the integrand is negative. A student who substitutes x=secθ on this branch and writes x2−1=tanθ where tanθ<0 finds a POSITIVE value, and the sign check catches him at once.
c) A sum of squares: set x=2tanθ with −2π<θ<2π, so dx=2sec2θdθ and x2+4=4sec2θ. Since secθ>0 on this interval, (x2+4)3/2=8sec3θ with no absolute value. The integral becomes ∫8sec3θ4tan2θ⋅2sec2θdθ=∫secθtan2θdθ=∫secθsec2θ−1dθ=∫(secθ−cosθ)dθ=ln∣secθ+tanθ∣−sinθ+C.
Triangle: tanθ=2x, so the opposite side is x, the adjacent side 2 and the hypotenuse x2+4. Then secθ=2x2+4 and sinθ=x2+4x, and the antiderivative is ln2x+x2+4−x2+4x+C=ln(x+x2+4)−x2+4x+C′: the −ln2 goes into the constant, and no absolute value is needed because x2+4>∣x∣.
Check: dxdln(x+x2+4)=x+x2+41+x2+4x=x2+41, and by the quotient rule dxdx2+4x=(x2+4)3/2(x2+4)−x2=(x2+4)3/24. The difference is (x2+4)3/2x2+4−4=(x2+4)3/2x2, the integrand.
d) Complete the square BEFORE substituting: x2−4x+8=(x−2)2+4. With u=x−2 the bounds become 0 and 1, and the integral is ∫01u2+4du. Now u=2tanθ, du=2sec2θdθ, u2+4=2secθ, and the integral is ∫secθdθ=ln∣secθ+tanθ∣, taken from θ=0 to the angle where tanθ=21, at which secθ=1+41=25. The value is ln(25+21)−ln1=ln21+5, the logarithm of the golden ratio φ=21+5.
Plausibility: on [2,3] the integrand decreases from 21 to 51, so the shaded area lies between the two dashed rectangles, 51≈0.447 and 21. The upper bound can be checked exactly: lnφ<21 means φ2<e, and φ2=φ+1≈2.62 (since φ solves t2=t+1), which is less than e≈2.72. The value is about 0.48, inside the bracket.
Exercise 7: A quartic denominator: factor, decompose, integrate
Let Q(x)=x4+2x3+2x2+2x+1. The figure shows y=Q(x)x2+2x+3 for 0≤x≤2.5 and the region under it for 0≤x≤1, whose area is I=∫01Q(x)x2+2x+3dx. Part a) is independent of the rest.
a) Write the FORM of the partial fraction decomposition of x3−2x2x3+4 and of (x−1)(x2+2x+2)25x. Do not solve for the constants.
b) Show that −1 is a root of Q, and factor Q completely over the real numbers, justifying that the factorization is complete.
c) Find the partial fraction decomposition of Q(x)x2+2x+3, and check it at one value of x.
d) Evaluate I exactly, and check its size against the figure.
a) First function: numerator and denominator both have degree 3, so the degree check forces a long division first: x3−2x2x3+4=1+x3−2x22x2+4. The denominator factors as x2(x−2), a repeated linear factor and a distinct one, so the form is 1+xA+x2B+x−2C, one term for EACH power of x. Second function: it is proper (1<5), and x2+2x+2 has discriminant 4−8=−4<0, so it is irreducible, and squared. The form is x−1A+x2+2x+2Bx+C+(x2+2x+2)2Dx+E: five constants for a denominator of degree five, as it must be.
b) By the rational root test, a rational root of Q divides 1, so only ±1 can be one. Q(1)=8 and Q(−1)=1−2+2−2+1=0, so x+1 is a factor. Dividing (synthetic division with −1 on the coefficients 1,2,2,2,1 gives 1,1,1,1 and remainder 0): Q(x)=(x+1)(x3+x2+x+1). The cubic factors by grouping: x3+x2+x+1=x2(x+1)+(x+1)=(x+1)(x2+1). So Q(x)=(x+1)2(x2+1), and the factorization is complete because x2+1 has no real root (its discriminant is −4). A student who stops after the first root misses that −1 is a DOUBLE root, and his form has no term in (x+1)21: it cannot work.
c) The fraction is proper (2<4). Form: (x+1)2(x2+1)x2+2x+3=x+1A+(x+1)2B+x2+1Cx+D, that is x2+2x+3=A(x+1)(x2+1)+B(x2+1)+(Cx+D)(x+1)2. At x=−1: 2=2B, so B=1. The x3 coefficients: 0=A+C. At x=0: 3=A+B+D, so A+D=2. At x=1: 6=4A+2B+4(C+D)=4(A+C)+2+4D, so 4D=4 and D=1, then A=1 and C=−1.
The decomposition is x+11+(x+1)21+x2+11−x. Check at x=2: the left side is 9⋅54+4+3=4511, the right side 31+91−51=4515+5−9=4511.
d) Split the last term as x2+11−x2+1x. An antiderivative is ln∣x+1∣−x+11+arctanx−21ln(x2+1): the squared factor gives a POWER, not a logarithm, and the part of the numerator proportional to 2x gives the logarithm, the rest the arctangent. At x=1: ln2−21+4π−21ln2; at x=0: 0−1+0−0=−1. So I=21ln2+21+4π≈0.35+0.5+0.79=1.63.
Size check: each of the three terms of c) decreases on [0,1] (for the last, the derivative of 1+x21−x has numerator x2−2x−1=(x−1)2−2<0 there), so the integrand decreases from 3 at x=0 to 86=43 at x=1, as on the figure. Hence 43≤I≤3, and 1.63 fits. Forgetting the constant −1 at x=0, the usual slip here, would give 0.63, below the floor of 43.
Exercise 8: Decide before computing: comparison, and the infinity minus infinity trap
For each improper integral, the verdict must be justified by the comparison test with its hypotheses checked (positive integrands, and the inequality in the direction that concludes), or by an exact limit. You may use without proof that xarctanx has no elementary antiderivative, and that on [0,1] the convex function ex lies below its chord, ex≤1+(e−1)x.
The figure shows y=ex−11 (solid) and y=x1 (dashed) for 0<x≤4.
a) Show by comparison that ∫1∞x(1+x2)dx converges, and give an upper bound for its value.
b) Compute it exactly. A classmate writes ∫1∞x(1+x2)dx=∫1∞xdx−∫1∞1+x2xdx: explain why this line is illegal, and how the correct computation gets around it.
c) Show that ∫1∞xarctanxdx diverges. Which comparison with x1 proves nothing?
d) Show that ∫1∞ex−1dx converges, with a bound, then compute it exactly with the substitution u=ex.
e) The figure shows ex−11 BELOW x1. Explain why that proves nothing about ∫01ex−1dx, then decide its convergence. What can you conclude about ∫0∞ex−1dx?
Show the solution
Answers
a)0<x(1+x2)1≤x31 and p=3>1: converges, value at most 21
b)21ln2≈0.35; both pieces diverge, so combine the logarithms on [1,t] before the limit
c)xarctanx≥4xπ for x≥1: diverges; xarctanx≤2xπ proves nothing
d)ex−11≤e−1ee−x: converges, at most e−11; exact value lne−1e=1−ln(e−1)
e)Smaller than a divergent integral proves nothing; ex−11≥(e−1)x1 on (0,1]: diverges, so ∫0∞ diverges
a) For x≥1 the integrand is continuous and positive, and 1+x2≥x2, so 0<x(1+x2)1≤x⋅x21=x31. The integral ∫1∞x3dx converges (p=3>1) and equals 21. Smaller than a convergent integral: by comparison, ∫1∞x(1+x2)dx converges, and its value is at most 21.
b) Partial fractions: x(1+x2)1=xA+x2+1Bx+C gives 1=A(x2+1)+(Bx+C)x; x=0 gives A=1, the x2 coefficients give B=−1, the x coefficients C=0. So ∫1t(x1−x2+1x)dx=[lnx−21ln(x2+1)]1t=lnt2+1t−ln21. As t→∞, t2+1t=1+1/t21→1, so the first logarithm tends to 0, and the integral equals ln2=21ln2≈0.35, below the bound 21 of a).
The classmate's line splits the improper integral into two improper integrals, and BOTH diverge: ∫1txdx=lnt and ∫1t1+x2xdx=21ln21+t2 each tend to ∞. The rule ∫(f−g)=∫f−∫g holds for improper integrals only when both converge; otherwise the right side is ∞−∞, which is not a number. The correct computation subtracts on the FINITE interval [1,t], combines the two logarithms into one, and only then lets t→∞.
c) For x≥1, arctanx is increasing, so arctanx≥arctan1=4π, and xarctanx≥4π⋅x1>0. The integral ∫1∞xdx diverges (p=1), so ∫1∞4xπdx diverges too, and LARGER than a divergent integral: ∫1∞xarctanxdx diverges. No antiderivative was needed, which is fortunate, since there is none in elementary terms. The other natural bound, arctanx<2π, gives xarctanx≤2xπ: smaller than a divergent integral, which proves nothing.
d) For x≥1, ex≥e, so 1≤eex and ex−1≥ex−eex=ee−1ex. Hence 0<ex−11≤e−1ee−x, and ∫1∞e−xdx=limt→∞(e−1−e−t)=e1 converges. By comparison the integral converges, and its value is at most e−1e⋅e1=e−11≈0.58.
Exact value: with u=ex, du=exdx, so dx=udu, and ∫ex−1dx=∫u(u−1)du=∫(u−11−u1)du=lnuu−1+C=ln(1−e−x)+C (here u>1, so no absolute values are needed). Then ∫1tex−1dx=ln(1−e−t)−ln(1−e−1)→0−lnee−1, so the integral equals lne−1e=1−ln(e−1). It respects the bound: lny≤y−1 gives lne−1e≤e−1e−1=e−11.
e) Since ex≥1+x for all x, ex−1≥x and ex−11≤x1 for x>0: that is the figure. But ∫01xdx DIVERGES, and being smaller than a divergent integral proves nothing. The comparison that concludes goes the other way: by the chord inequality, ex−1≤(e−1)x on (0,1], so ex−11≥e−11⋅x1>0, and ∫01xdx diverges (p=1 on (0,1]). Larger than a divergent integral: ∫01ex−1dx diverges. The antiderivative of d) confirms it: ∫t1ex−1dx=ln(1−e−1)−ln(1−e−t)→+∞ as t→0+.
So ∫0∞ex−1dx diverges, although its tail ∫1∞ converges: an integral improper at both ends converges only if EVERY piece converges, and one divergent piece is enough to sink it. The figure tells the same story: near 0 the two curves are almost the same, so they must share the verdict of x1 there; far away ex−11 dies like e−x.