MATH 203 Calculus I • Concordia University, Montreal
Corrected exercises: Indeterminate forms and L'Hôpital's Rule (MATH 203)
This is the corrected exercise set for indeterminate forms and L'Hôpital's Rule in MATH 203, Differential and Integral Calculus I, at Concordia University, section 4.5 of Thomas' Calculus: Early Transcendentals. The course teaches it right after linearization and before extreme values; it comes back in curve sketching, for every horizontal asymptote that algebra alone cannot reach. The department-approved scientific calculator is allowed, but it computes no limit: exercise 6 shows a table of values that ends at 0 for a limit equal to 21.
The thread running through the whole set: the rule itself is one line, and the marks are lost in the ALGEBRA around it. A factor that crosses the fraction bar changes the sign of its exponent, x2/3lnx=x−2/3lnx; every round produces a fraction of fractions that must be cleared before the next form can be read; a difference goes over a common denominator; an indeterminate power goes down through lny=glnf and comes back with eL. Around that algebra, the form is written before every round, because a form that is not indeterminate is concluded, never differentiated.
The traps named in the solutions: dividing by a fraction the wrong way up, a chain factor forgotten on the inner function, a second round started where a common factor cancels, the rule applied to a form 10 or 01, the two factors of a product differentiated, a factor sent downstairs without flipping its exponent, exponents divided instead of subtracted, a table of values trusted at x=10−5, ∞−∞ answered 0, 1∞ answered 1, and the limit of lny given as the answer.
Self-checking setType your answers under each question: the page tells you right or wrong before you open the solution. With an account, every answer right on the first try earns points.
•L'Hôpital's Rule: if f(a)=g(a)=0, f and g are differentiable near a with g′(x)=0 for x=a, then limx→agf=limx→ag′f′, if the right side exists. Also for ∞∞, one-sided limits and x→±∞.
•Top and bottom are differentiated separately. The form is written before EVERY round; c0 and 0c with c=0 are not indeterminate.
•Clear the complex fraction after each round: c/da/b=ba⋅cd, and xnxm=xm−n.
•0⋅∞: fg=1/fg, with xa1=x−a; keep the logarithm upstairs. ∞−∞: common denominator or factor the dominant term.
•1∞, 00, ∞0: lny=glnf; if lny→L then y→eL. The form 0∞ gives 0.
•Growth as x→∞: lnx≪xp≪ex for every p>0. If the rule loops or worsens, use algebra.
Part A: the basics (/50)
Exercise 1: The form 0/0: check it, differentiate top and bottom, clear the fraction
L'Hôpital's Rule (Thomas 4.5). Suppose that f(a)=g(a)=0, that f and g are differentiable on an open interval I containing a, and that g′(x)=0 on I if x=a. Then limx→ag(x)f(x)=limx→ag′(x)f′(x), assuming the limit on the right side exists. The rule also holds for one-sided limits, for x→±∞ and for the form ±∞±∞.
The numerator and the denominator are differentiated SEPARATELY. The new quotient is very often a fraction of fractions: clearing it is the step where most marks are lost. The figure shows y=x2−9 and y=ln(x−2) near x=3, with their tangent lines at (3,0).
a) Find limx→3ln(x−2)x2−9, stating the form first. Explain the answer with the two tangent lines of the figure.
b) Find limx→0ex−1arctan2x.
c) Find limx→π/2cos2x1−sinx. After one round, simplify instead of starting a second one.
d) A student writes: limx→0ex+xx+sinx=limx→0ex+11+cosx=22=1. Find the error and the correct limit.
e) Find limx→02x−13x−1 as an exact number, then to four decimal places.
Show the solution
Answers
a)Form 00; the limit is 6, the quotient of the tangent slopes 6 and 1.
b)Form 00; the limit is 2.
c)Form 00; the limit is 21.
d)The form is 10, not indeterminate: the limit is 0.
e)ln2ln3=log23≈1.5850
a) At x=3: 32−9=0 and ln(3−2)=ln1=0, so the form is 00 and the rule applies. The derivative of the top is 2x; the derivative of the bottom is x−21, by the chain rule with inner function x−2. The new quotient is a complex fraction, and it is cleared before anything else: x−212x=2x⋅1x−2=2x(x−2). Dividing by a fraction is multiplying by its reciprocal; a student who writes x−22x has divided by the wrong thing and finds 06, an infinite limit for a quotient the figure shows to be tame. Now 2x(x−2)→6⋅1=6. The figure explains the number: near 3, x2−9≈6(x−3) and ln(x−2)≈1⋅(x−3), so the quotient is close to x−36(x−3)=6. When both functions vanish at a, their quotient tends to the quotient of their slopes at a: that is the whole content of the rule.
b) At 0: arctan0=0 and e0−1=0, form 00. By the rule, with the chain rule on the inner function 2x: limx→0ex1+4x22. Clear the fraction: (1+4x2)ex2, which is continuous at 0, so the limit is 1⋅12=2. The factor 2 is the one the inner function brings: forgetting it gives 1, and nothing on the page looks wrong.
c) At 2π: 1−sin2π=0 and cos22π=0, form 00. Round 1, with the chain rule on the square: 2cosx⋅(−sinx)−cosx=−2sinxcosx−cosx. Substituting now gives 00 again, and a student who starts a second round differentiates a product for nothing. The factor cosx that causes the 00 appears on both levels: for x near 2π, cosx=0, and it cancels, leaving 2sinx1→21. The same answer comes without the rule, by the identity cos2x=1−sin2x=(1−sinx)(1+sinx): the quotient is 1+sinx1→21. Cancelling a common factor is algebra from the first week of the course, and it is the shortest path in both methods.
d) At 0 the numerator is 0+sin0=0 but the denominator is e0+0=1. The form is 10, which is NOT indeterminate, and the quotient law gives the limit 10=0 at once. The hypothesis g(a)=0 of the rule is false here, so the rule may not be used, and applied anyway it produced 1, a wrong answer. The error is the missing first line: the form was never written. On an exam this costs the whole question, because the page shows no calculation error at all.
e) At 0: 30−1=0 and 20−1=0, form 00. The derivative of ax is axlna (write ax=exlna and use the chain rule). By the rule, limx→02xln23xln3=ln2ln3, which is log23 by the change of base formula. With the calculator, 0.693147…1.098612…≈1.5850. The exact answer is ln2ln3; the decimal only answers the second half of the question. Writing (3x)′=x⋅3x−1, the power rule applied to an exponential, gives 00 again at 0 and loops: the variable is in the EXPONENT.
Tick the exercises you have done or want to review: a free account, no password, keeps your ticks from one visit to the next and tells you which chapter to tackle next. Create your space, an email is enough.
Exercise 2: The form infinity over infinity: simplify between rounds
The rule applies in the same way when the top and the bottom both tend to ±∞, and when x→±∞. Each round must be licensed by its own form, written on the line before it. Between two rounds, clear the fractions and cancel what cancels: a quotient left in the form bxa/x turns the next derivative into a quotient rule, and ten lines of algebra where one was enough.
In every part, write the form before each round.
a) Find limx→∞x2(lnx)3.
b) Find limx→∞x2ex/3.
c) Find limx→(π/2)−tanxln(cosx).
d) Find limx→∞x2+4lnx3x2−lnx, first with one round of the rule, then without the rule. Do the answers agree?
Show the solution
Answers
a)0, after three rounds
b)+∞, after two rounds
c)0: the round gives −sinxcosx
d)3 both ways
a) As x→∞, (lnx)3→∞ and x2→∞: form ∞∞. Round 1, chain rule on the cube with inner function lnx: 2x3(lnx)2⋅x1. Clear it: x⋅2x3(lnx)2=2x23(lnx)2. The form is ∞∞ again. Round 2: 4x6lnx⋅x1=2x23lnx, still ∞∞. Round 3: 4x3⋅x1=4x23→0. Each round lowers the power of lnx by one, and only because the x1 was moved down next to x each time. The frequent error is to read 2xx1 as x2x, flipping the wrong level: the limit then comes out as ∞.
b) Form ∞∞. Round 1, with the factor 31 of the inner function 3x: 2x31ex/3=6xex/3, still ∞∞. Round 2: 631ex/3=18ex/3→∞. The limit is +∞: the rule also allows an infinite limit on the right side. The slower exponential ex/3 still beats every power of x, it only needs larger x to do it.
c) As x→(2π)−, cosx→0+ so ln(cosx)→−∞, and tanx→+∞: form ∞−∞. Round 1: the top gives cosx−sinx (chain rule, inner function cosx), the bottom gives sec2x=cos2x1. Clear the complex fraction: cosx−sinx⋅cos2x=−sinxcosx. This is a product of continuous functions, and it tends to −1⋅0=0. The limit is 0: the tangent blows up faster than the logarithm of the cosine goes down. Left uncleared, sec2x−tanx looks like another ∞∞, and a second round only makes it worse.
d) Form ∞∞. With the rule: 2x+x46x−x1. Multiply top and bottom by x to clear the small fractions: 2x2+46x2−1, and dividing by x2 gives 2+4/x26−1/x2→3. Without the rule: divide top and bottom by x2, 1+4x2lnx3−x2lnx, and x2lnx→0 (one round: 2x1/x=2x21→0), so the limit is 13=3. The answers agree. The second method is the chapter 5 method, dividing by the dominant term; the rule was only needed to know that lnx is negligible next to x2.
Exercise 3: Products and differences: build the quotient, watch the exponents
The rule only takes a QUOTIENT. A product f⋅g of the form 0⋅∞ is first written 1/gf or 1/fg. The algebra that costs marks is here: a factor that crosses the fraction bar changes the SIGN of its exponent, x=x−11 and x2/3=x−2/31. A difference f−g of the form ∞−∞ is first turned into a quotient by a common denominator, or by factoring out the dominant term.
In every part, write the form, then the rewritten expression and ITS form, before any derivative.
a) Find limx→0+x(lnx)2. Try both ways of writing it as a quotient and keep the one that works.
b) Find limx→1−(1−x)tan2πx exactly, then to four decimal places.
c) Find limx→0(cotx−x1).
d) Find limx→1(x2−12−x−11).
e) Find limx→∞(x1/3−lnx).
Show the solution
Answers
a)x−1(lnx)2 works, in two rounds: the limit is 0.
b)π2≈0.6366
c)0
d)−21, by the common denominator alone
e)+∞: x1/3(1−x1/3lnx) with x1/3lnx→0
a) Form 0⋅∞, since x→0+ and (lnx)2→+∞. First way: 1/(lnx)2x, form 00; the derivative of (lnx)−2 is −2(lnx)−3⋅x1, and the new quotient is −2(lnx)−3/x1=−2x(lnx)3, a product of the same kind with a HIGHER power of the logarithm. Abandon it. Second way: x=x−11, so x(lnx)2=x−1(lnx)2, form ∞∞. Round 1: −x−22lnx⋅x−1=−2lnx⋅x−1⋅x2=−2xlnx, again 0⋅(−∞). Rewrite the same way: −2xlnx=x−1−2lnx, form ∞−∞, and round 2 gives −x−2−2x−1=2x→0. The limit is 0. The rule of thumb: keep the logarithm upstairs, where differentiating simplifies it, and send the power down with its exponent flipped.
b) As x→1−, 1−x→0+ and 2πx→(2π)−, so tan2πx→+∞: form 0⋅∞. Since tanu=cotu1, write cot(πx/2)1−x, form 00. The derivative of cotu is −csc2u, times the factor 2π of the inner function u=2πx. Round 1: −2πcsc2(πx/2)−1=π2sin22πx→π2⋅1=π2, about 0.6366. The other choice, (1−x)−1tan(πx/2), gives after one round (1−x)−22πsec2(πx/2), which is worse. The exact answer is π2; forgetting the chain factor 2π gives 1.
c) As x→0+, cotx→+∞ and x1→+∞; as x→0−, both tend to −∞: form ∞−∞ on each side. Common denominator: cotx−x1=sinxcosx−x1=xsinxxcosx−sinx, form 00. Round 1: the top gives cosx−xsinx−cosx=−xsinx (product rule), the bottom gives sinx+xcosx. The quotient sinx+xcosx−xsinx is still 00; divide top and bottom by x instead of differentiating: xsinx+cosx−sinx→1+10=0, using xsinx→1 (Thomas 2.4). The limit is 0. Answering 0 at the start because both terms are infinite would be the right number for a wrong reason, and part d) shows that the reason fails.
d) As x→1+ both terms tend to +∞, and as x→1− both tend to −∞ (since x2−1→0− and x−1→0−): form ∞−∞. Common denominator, with x2−1=(x−1)(x+1): (x−1)(x+1)2−(x−1)(x+1)x+1=(x−1)(x+1)1−x=x+1−1 for x=1. The sign is the algebra trap: 1−x=−(x−1), and dropping that minus gives +21. So the limit is −21, with no derivative at all. The rule would also work on x2−11−x, giving 2x−1→−21, but the factorization is shorter. The two infinities differ by a quantity that tends to −21, not 0.
e) Both terms tend to ∞: form ∞−∞. There is no fraction to combine, so factor out the term expected to dominate: x1/3−lnx=x1/3(1−x1/3lnx). The quotient x1/3lnx is ∞∞; one round gives 31x−2/3x−1=3x−1+2/3=3x−1/3=x1/33→0. Subtracting the exponents, −1−(−32)=−31, is the step to write out. So the bracket tends to 1, the factor x1/3 to ∞, and the limit is +∞: even a cube root beats the logarithm. Not everywhere, though: at x=27 the root is 3 while ln27≈3.30, and the logarithm is ahead between about x=6 and x=93; at x=1000 the root is 10 against ln1000≈6.91, and it leads for good. A limit at infinity only speaks of large x.
Exercise 4: Repeated rounds, and the round too many
When a round gives 00 again, the rule may be used again, and the form is written again. Two habits keep the page short and right: simplify between rounds (cancel a common factor, use the conjugate, split off a factor with a known limit), and STOP as soon as the form is no longer indeterminate.
Every limit here is at x=0.
a) Find limx→0x2ln(1+x)−sinx.
b) Find limx→0xsin2x1−cos4x.
c) A student writes: limx→0x2+x3sin2x−2x=limx→02x+3x22cos2x−2=limx→02+6x−4sin2x=limx→06−8cos2x=−34. Find the error and the correct limit.
d) Find limx→0x3arcsinx−x, finishing with the conjugate instead of a second round.
Show the solution
Answers
a)−21, after two rounds
b)4, after two rounds
c)The third quotient is 20, not indeterminate: the limit is 0, not −34.
d)61
a) At 0: ln1−sin0=0 over 0, form 00. Round 1: 2x1+x1−cosx, and at 0 the top is 1−1=0: form 00 again. Round 2: the derivative of (1+x)−1 is −(1+x)−2, and that of −cosx is +sinx, so the quotient is 2−(1+x)−2+sinx→2−1+0=−21. Writing 1+x1 as (1+x)−1 before differentiating is what keeps the minus sign: a student who differentiates 1+x1 as ln(1+x), or forgets the minus of the power rule, answers +21. The sign is plausible: near 0, ln(1+x) lies below x by about 2x2, while sinx stays much closer to x.
b) At 0: 1−cos0=0 and 0⋅sin0=0, form 00. Round 1: the top gives 4sin4x (chain rule, inner function 4x), the bottom sin2x+2xcos2x (product rule, with the chain factor 2). At 0: 00 again. Round 2: the top gives 16cos4x; the bottom gives 2cos2x+2cos2x−4xsin2x=4cos2x−4xsin2x. At 0: 416, not indeterminate: stop. The limit is 4. The product rule on xsin2x is where the page goes wrong: (xsin2x)′=sin2x+2xcos2x, two terms, each with its chain factor. A student who writes (xsin2x)′=2cos2x finds 20 after round 1 and answers 0.
c) Rounds 1 and 2 are legal: at 0 the starting quotient is 00, and so is 2x+3x22cos2x−2, since 2cos0−2=0. But the third quotient, 2+6x−4sin2x, has a denominator that tends to 2: its form is 20, NOT indeterminate, and its limit is 0. The third round had no permit, and it produced −34. The correct limit is 0. A check confirms it: for small x, sin2x−2x is about −34x3, so the quotient behaves like x2−34x3=−34x→0. The −34 the student found is the limit of x3sin2x−2x, another question with another denominator. One round too many costs the whole question, and it only shows if the form is written before EVERY round.
d) At 0: arcsin0−0=0 over 0, form 00. Round 1, with (arcsinx)′=1−x21: 3x21−x21−1, still 00. A second round would differentiate (1−x2)−1/2 and bring a new fraction. Algebra instead. Put the top over one denominator: 3x21−x21−1−x2. Multiply top and bottom by the conjugate 1+1−x2: the top becomes 1−(1−x2)=x2, which cancels with the x2 below, for x=0: 31−x2(1+1−x2)1→3⋅1⋅21=61. The conjugate is the chapter 3 gesture, and it is exactly what makes the factor x2 appear so that it can cancel.
Exercise 5: Indeterminate powers: down with the logarithm, back with the exponential
When both the base and the exponent move, y=f(x)g(x) with f(x)>0, the forms 1∞, 00 and ∞0 are indeterminate. Thomas's method: take the logarithm, lny=g(x)lnf(x), a form 0⋅∞; rewrite it as a quotient and find limlny=L; then, since ex is continuous, limy=eL. The exponent comes DOWN as a factor, ln(fg)=glnf, and it never becomes (lnf)g. The form 0∞ is not indeterminate: it gives 0.
The figure shows y=(tanx)tan2x on (0,2π); the function is not defined at x=4π.
a) Find limx→0+(sinx)tanx.
b) Find limx→1x3/(x−1) exactly, then to two decimal places. Check with the calculator at x=1.001.
c) Find limx→∞(lnx)1/x.
d) Find limx→π/4(tanx)tan2x, and give the height of the hole of the figure to four decimal places.
e) Compare limx→(π/2)−(cosx)tanx and limx→(π/2)−(cosx)cosx: which one is an indeterminate form?
Show the solution
Answers
a)Form 00; lny→0, so the limit is 1.
b)Form 1±∞; lny→3, so the limit is e3≈20.09.
c)Form ∞0; lny→0, so the limit is 1.
d)Form 1±∞; lny→−1, so the limit is e−1≈0.3679, the height of the hole.
e)(cosx)tanx is 0∞, not indeterminate: limit 0. (cosx)cosx is 00: limit 1.
a) As x→0+, the base sinx→0+ and the exponent tanx→0: form 00. Let y=(sinx)tanx; then lny=tanxln(sinx), of the form 0⋅(−∞). Send tanx down as its reciprocal: lny=cotxln(sinx), form ∞−∞. Round 1: the top gives sinxcosx (chain rule, inner function sinx), the bottom −csc2x=−sin2x1. Clear the fraction of fractions: sinxcosx⋅(−sin2x)=−sinxcosx→0. So lny→0 and y=elny→e0=1. Flipping the wrong level, −sin3xcosx, would give −∞ and the answer 0. The answer is 1 because the logarithm tends to 0, not because 00 is 1: part e) shows a base tending to 0 with another ending.
b) As x→1 the base tends to 1, and the exponent x−13 tends to +∞ from the right and −∞ from the left: form 1±∞, NOT 1. lny=x−13lnx=x−13lnx, form 00 on both sides. Round 1: 13/x→3. So lny→3 and y→e3≈20.09, from both sides. Calculator check: 1.0013000=e3000ln1.001≈20.06, close to 20.09 and far from 1. The check is done at a MODERATE value: at x=1+10−12 a calculator rounds the base to 1 and displays 1, which proves nothing.
c) As x→∞ the base lnx→∞ and the exponent x1→0: form ∞0. lny=xln(lnx), form ∞∞. Round 1: the derivative of ln(lnx) is lnx1⋅x1 (chain rule, inner function lnx, named), so the quotient is xlnx1→0. Hence lny→0 and y→e0=1. The logarithm grows, but taking its x-th root crushes it.
d) As x→4π, tanx→1 and 2x→2π, so tan2x→+∞ from the left and −∞ from the right: form 1±∞. lny=tan2xln(tanx)=cot2xln(tanx), form 00. Round 1: the top gives tanxsec2x, the bottom −2csc22x. Clear the top first with the identities: tanxsec2x=cos2x1⋅sinxcosx=sinxcosx1=sin2x2, by the double angle formula sin2x=2sinxcosx. Then −2/sin22x2/sin2x=−sin2x→−sin2π=−1. So lny→−1 and y→e−1=e1≈0.3679 from both sides: the hole of the figure is at (4π,e1). The double angle identity turns a page of quotient rules into one line: this is the chapter 1 algebra paying off in chapter 16.
e) As x→(2π)−, cosx→0+ and tanx→+∞. So (cosx)tanx is of the form 0∞: a small positive base raised to a large power, which is NOT indeterminate. The limit is 0 directly; the logarithm confirms it, lny=tanxln(cosx)→(+∞)(−∞)=−∞ and e−∞=0. On the other hand (cosx)cosx is of the form 00. With u=cosx→0+: lny=ulnu=u−1lnu, form ∞−∞, and one round gives −u−2u−1=−u→0, so y→1. Same base, two limits: only the exponent decides, and only after the logarithm.
Part B: problems and reasoning (/50)
Exercise 6: When the rule goes in circles, and when the calculator lies
The rule is a tool, not an obligation. It can give back the quotient you started from, it can make the quotient worse, and a table of values can suggest a limit that is false. In each case, algebra decides: divide by the dominant term, use a law of exponents, or factor.
For part c), a calculator that keeps 10 significant digits for tanx and sinx gives this table for q(x)=x3tanx−sinx: xq(x)10−10.5012610−20.5000210−30.4997010−40.4700010−50. The figure shows the graph of q drawn by a computer.
a) Apply the rule twice to ex−e−xex+e−x as x→∞ and describe what happens. Find the limit at ∞, then at −∞.
b) Same question for limx→(π/2)−tanxsecx.
c) What does the table suggest? Find limx→0q(x), and explain the last two lines of the table.
d) Find limx→∞ex2ex. Show that the rule makes the quotient worse, and conclude with a law of exponents.
Show the solution
Answers
a)The rule gives the reciprocal, then the starting quotient: a loop. Limits 1 at ∞ and −1 at −∞.
b)Loop again; tanxsecx=sinx1→1.
c)limx→0q(x)=21; the 0.47 and the 0 are roundoff in tanx−sinx.
d)ex−x2→0
a) Form ∞∞, since e−x→0. Round 1: ex+e−xex−e−x (the chain rule turns e−x into −e−x), form ∞∞. Round 2: ex−e−xex+e−x, the starting quotient. The rule turns in a circle and will never finish. Algebra instead: multiply top and bottom by e−x, which divides by the dominant term ex: 1−e−2x1+e−2x→1−01+0=1. As x→−∞ the dominant term is e−x; multiply by ex: e2x−1e2x+1→0−10+1=−1. The law e−x⋅ex=1 and e−x⋅e−x=e−2x is the whole computation.
b) As x→(2π)−, secx→+∞ and tanx→+∞: form ∞∞. Round 1: sec2xsecxtanx=secxtanx, form ∞∞. Round 2: secxtanxsec2x=tanxsecx: back to the start. Rewrite in sines and cosines instead, and clear the complex fraction: tanxsecx=cosx1⋅sinxcosx=sinx1, for cosx=0. So the limit is sin(π/2)1=1. When a trigonometric quotient loops, go back to sin and cos: the identity usually cancels what the rule cannot.
c) The first three lines suggest 21, the last two suggest something falling towards 0. The algebra decides. Factor sinx: tanx−sinx=cosxsinx−sinx=cosxsinx(1−cosx), so q(x)=xsinx⋅x21−cosx⋅cosx1. The first factor tends to 1 (Thomas 2.4) and the last to 1. The middle one is 00: one round gives 2xsinx→21. So limx→0q(x)=1⋅21⋅1=21, and the graph, which never dips, agrees. The last two lines are ROUNDOFF: for x=10−5, tanx and sinx agree in their first 10 digits, both are stored as 1.000000000×10−5, and the calculator subtracts two equal numbers. The difference it should find, about 2x3=5×10−16, is far below its last digit. The line for 10−4 is already damaged. A table of values supports a limit only at MODERATE values of x; it never replaces the computation, and a scientific calculator cannot compute a limit.
d) Form ∞∞. Round 1, with the chain rule on the inner function x2: 2xex2ex. The denominator has gained a factor 2x, the next round would bring a product rule and a factor 4x2+2: every round makes it worse. Law of exponents instead: ex2ex=ex−x2, and x−x2=x(1−x)→−∞ (a positive factor times one that tends to −∞). Since eu→0 as u→−∞, the limit is 0. Writing ex2ex=ex/x2, dividing the exponents instead of subtracting them, gives e1/x→1: the algebra slip costs the whole question.
Exercise 7: Who grows faster: the rule against the table
As x→∞, f grows faster than g if limg(x)f(x)=∞, and f and g grow at the same rate if limg(x)f(x) is a finite number other than 0. L'Hôpital's Rule decides these limits. A calculator table, on the other hand, only sees the values you type, and for slowly growing functions the interesting values are out of reach.
Let q(x)=x(lnx)10. The figure shows q as a function of t=lnx, that is q=ett10, in units of 100000; the orange dots are x=10, 100 and 1000.
a) With the calculator, compute q(10) to the nearest unit, q(100) to the nearest hundred and q(1000) to the nearest thousand. What does this table suggest?
b) Substitute x=et and find limx→∞q(x) with the rule.
c) Compute q(e20) to the nearest unit and q(e40) to four decimal places. What do they say about the table of a)?
d) Decide which grows faster: x0.1 or (lnx)10; then ex or x5.
e) Do log2x and lnx grow at the same rate? Find limx→∞lnxlog2x.
Show the solution
Answers
a)q(10)≈419, q(100)≈42900, q(1000)≈247000: the table suggests q→∞.
b)q=ett10; ten rounds give et10!→0.
c)q(e20)≈21106, q(e40)≈0.0445: the table was on the rising side of the hump.
d)x0.1 grows faster than (lnx)10; ex grows faster than x5.
e)Same rate: lnxlog2x=ln21≈1.4427.
a) q(10)=10(2.302585…)10≈418.9, so 419 to the nearest unit. q(100)=100(4.605170…)10≈42900 and q(1000)=1000(6.907755…)10≈247383, so 247000 to the nearest thousand. Each line is about a hundred or six times the previous one: the table suggests that q(x)→∞, that the tenth power of the logarithm beats x. Keep the logarithm unrounded in the calculator: rounding ln10 to 2.30 before raising it to the tenth power already moves q(10) by several units.
b) As x→∞, t=lnx→∞ and x=et, so q=ett10, form ∞∞. Round 1: et10t9, still ∞∞; each round lowers the power by one and leaves et unchanged. After ten rounds, each on a form ∞∞: et10!=et3628800→0. So limx→∞q(x)=0: x grows faster than (lnx)10. Working in t is the algebra that makes the computation possible: directly in x, each round would bring a x1 to clear, ten times.
c) q(e20)=e202010≈4.852×1081.024×1013≈21106 and q(e40)=e404010≈0.0445. Between x=1000 and x=e20≈4.9×108, q went up and then came down: the figure shows the hump, whose top is near t=10, that is x=e10≈22026. The three values of a) all sit on the rising side, which is why the table misled. The quotient only drops below 1 near t=36, that is for x around 3×1015, far beyond any table you would type; the limit is 0 all the same.
d) First pair: with x=et, x0.1(lnx)10=e0.1tt10, form ∞∞; each round brings a factor 0.1 downstairs, and after ten rounds the quotient is 0.110e0.1t10!→0. So x0.1 grows faster than (lnx)10, although the crossing happens at astronomically large x. Second pair: with u=x, x5=u10 and x5ex=u10eu→∞ (ten rounds, or the reciprocal of b). So ex grows faster than x5. In both cases a substitution turns the question into the ranking ln≪ power ≪ exponential, which the rule proves.
e) By the change of base formula from chapter 2, log2x=ln2lnx, so lnxlog2x=ln21 for every x>1, and the limit is ln21≈1.4427. The rule gives the same, 1/x1/(xln2)=ln21, but it is not needed. A finite nonzero limit: the two logarithms grow at the SAME rate. Changing the base of a logarithm, like changing the base of an exponential from 2 to 3, is not the same thing: 2x3x=(23)x→∞.
Exercise 8: Five statements to correct
Each line below was written by a student on a MATH 203 assignment, and each one is false. Say what is wrong, give the correct limit, and write the correct statement.
a) By L'Hôpital's Rule, limx→0+xlnx=limx→0+(x)′⋅(lnx)′=limx→0+1⋅x1=+∞.
b) ∞∞=1, so limx→∞ln(x2)lnx=1.
c) ∞−∞=0, so limx→0+(x1−x21)=0.
d) limx→0x2ex−1=limx→02xex=limx→02ex=21.
e) The base tends to 1 and 1 to any power is 1, so limx→∞(1+x1)x2=1.
Show the solution
Answers
a)False: the rule takes a quotient. x−1lnx→lim(−x)=0.
b)False: ln(x2)=2lnx, so the limit is 21.
c)False: x2x−1→−∞.
d)False: after one round the form is 01; the limit does not exist (+∞ on the right, −∞ on the left).
e)False: lny=x⋅xln(1+x1)→∞, so the limit is +∞.
a) FALSE. L'Hôpital's Rule applies to a QUOTIENT, and it never differentiates the two factors of a product. The form here is 0⋅(−∞): first rewrite, sending x down with its exponent flipped, xlnx=x−1lnx, form ∞−∞. One round: −x−2x−1=−x−1⋅x2=−x→0. So limx→0+xlnx=0, and the curve y=xlnx comes back to the origin. Correct statement: a product of the form 0⋅∞ is first written as a quotient 1/gf, and only then does the rule apply, to that quotient.
b) FALSE. The form ∞∞ says nothing about the value: two quantities that both grow can grow at different rates. Here no rule is needed, only the law of logarithms: for x>0, ln(x2)=2lnx, so ln(x2)lnx=2lnxlnx=21 for every x>1. The limit is 21. Correct statement: ∞∞ is an indeterminate form; the limit is decided by algebra or by the rule, never by the symbols.
c) FALSE. Common denominator: x1−x21=x2x−1. As x→0+ the top tends to −1 and the bottom to 0+: the form 0+−1 is not indeterminate and gives −∞. The term x21 wins the race. Correct statement: ∞−∞ is an indeterminate form; the common denominator turns it into a quotient whose form decides.
d) FALSE. The first round is legal: at 0 the form is 00. But 2xex has a top that tends to 1 and a bottom that tends to 0: its form is 01, NOT indeterminate, and the second round had no permit. The signs decide: as x→0+, 2xex→+∞; as x→0−, 2x<0 and 2xex→−∞. The one-sided limits differ, so limx→0x2ex−1 does not exist. The original quotient behaves like x2x=x1 near 0, which confirms it. Correct statement: before every round the form is written, and a form 0c with c=0 is concluded with the signs of each side.
e) FALSE. The form is 1∞, which is indeterminate: the base moves towards 1 while the exponent grows. With y=(1+x1)x2, lny=x2ln(1+x1)=x⋅[xln(1+x1)]. The bracket is x−1ln(1+1/x), form 00; one round gives −x−21+1/x−x−2=1+1/x1→1. So lny behaves like x⋅1→∞, and y→+∞. Correct statement: for 1∞, compute limlny=L first; the answer is eL, here e∞=∞, and it is 1 only when L=0.
Exercise 9: Compounding more and more often: the effective annual rate
A savings account pays a nominal annual rate r, compounded x times a year: each period adds xr of the balance. After one year every dollar has become (1+xr)x dollars, so the EFFECTIVE annual rate, the one printed in small print as the annual percentage yield, is E(x)=(1+xr)x−1. Banks compound yearly, monthly, daily; here x is treated as a positive real variable so that the rule can be used.
The figure shows, for r=5%, the gain E(x)−5% in basis points (one basis point is 0.01%), for 1≤x≤52, with the dots x=1, 2, 4, 12, 52 and a dashed ceiling.
a) For r=5%, compute E(4), E(12) and E(365) as percentages, to four decimal places.
b) Name the form of (1+xr)x as x→∞. Substitute h=x1 and find its limit with the rule. Deduce limx→∞E(x), and its value in percent for r=5%: this is the dashed ceiling.
c) Bank A offers 4.95% compounded daily, bank B 5% compounded annually, bank C 4.9% with the frequency increased without bound (the limit of b). Rank the three offers by effective annual rate.
d) With ANNUAL compounding at rate r, the balance doubles after T(r)=ln(1+r)ln2 years. Name the form of rT(r) as r→0+, and prove that its limit is ln2. Deduce the rule of 70: the doubling time is about p70 years at p percent.
e) With the calculator, compute T(0.05) and T(0.08) to two decimal places, and compare with 570 and with the rule of 72, 872.
Show the solution
Answers
a)E(4)≈5.0945%, E(12)≈5.1162%, E(365)≈5.1267%
b)Form 1∞; hln(1+rh)→r, so (1+xr)x→er and E(x)→er−1≈5.1271%.
c)A ≈5.0742%, then C ≈5.0220%, then B =5%.
d)Form 0⋅∞; ln2⋅ln(1+r)r→ln2≈0.693, so T≈p69.3.
e)T(0.05)≈14.21 years against 14; T(0.08)≈9.01 years against 9.
a) E(4)=1.01254−1≈0.050945, that is 5.0945%. E(12)=(1+120.05)12−1≈0.051162, that is 5.1162%. E(365)=(1+3650.05)365−1≈0.051267, that is 5.1267%. Type 120.05 as a division inside the brackets rather than a rounded 0.0042: rounding the rate per period before raising it to the twelfth power changes the fourth decimal. The gains shrink: from yearly to quarterly, 9.45 basis points; from monthly to daily, barely one more.
b) As x→∞ the base 1+xr→1 and the exponent x→∞: form 1∞, NOT 1. Put h=x1, so h→0+ and (1+xr)x=(1+rh)1/h. Take the logarithm: lny=hln(1+rh), form 00. Round 1, with the chain rule on the inner function 1+rh: 11+rhr→r. Since the exponential is continuous, y→er, and E(x)→er−1. For r=0.05: e0.05−1≈0.051271, that is 5.1271%, the dashed ceiling, 12.71 basis points above 5%. The substitution h=x1 is the algebra that avoids a quotient of −x2r over −x21; both routes give r.
c) A: (1+3650.0495)365−1≈0.050742, that is 5.0742%. B: 5% exactly. C: by b), e0.049−1≈0.050220, that is 5.0220%. Ranking: A first, then C, then B. The daily compounding of A does not reach the ceiling of b), but its higher nominal rate more than makes up for it: at a nominal 5%, the whole gain from frequency is at most e0.05−1.05, about 13 basis points, and daily compounding already collects nearly all of it, so the 5 basis points by which A's nominal rate exceeds C's decide the race. Comparing nominal rates is the error; the effective rates decide.
d) As r→0+, r→0 and T(r)=ln(1+r)ln2→+∞, since ln(1+r)→0+: form 0⋅∞. Rewrite as a quotient: rT(r)=ln2⋅ln(1+r)r, and ln(1+r)r is of the form 00. Round 1: 1+r11=1+r→1, clearing the complex fraction. So rT(r)→ln2≈0.693. With r=100p, T≈r0.693=p69.3 years, rounded up to p70 because 70 has more divisors: the rule of 70 is a limit.
e) T(0.05)=ln1.05ln2≈14.21 years, against 570=14. T(0.08)=ln1.08ln2≈9.01 years, against 872=9. The rule of 70 is exact only in the limit r→0; at 8% the product rT(r) is about 0.7205 (since ln1.080.08≈1.039), which is why bankers quote the rule of 72 for rates near 8%. The limit gives the number, and the calculator shows how far from the limit a real rate is.
Exercise 10: A final exam question: one function, three kinds of limit
Let f(x)=2lnxx2−1 for x>0, x=1. This is the shape of a long question on the final: a continuous extension, a derivative computed from its definition, the behaviour at both ends of the domain, with a form to write at every step, and one limit that is not indeterminate at all.
Every limit must be justified: name the form, then the tool.
a) Find limx→1f(x), and define the continuous extension F of f at x=1.
b) Find limx→0+f(x), stating the form.
c) Compute F′(1) from the definition of the derivative.
d) Find limx→∞f(x), limx→∞x2f(x) and limx→∞x1.9f(x).
e) With the calculator, compute F(1.1) to four decimal places and compare it with the value given by the tangent line of c). Is the graph above or below its tangent there?
Show the solution
Answers
a)Form 00; the limit is 1, so F(1)=1.
b)Form −∞−1, not indeterminate: the limit is 0.
c)F′(1)=limx→12(x−1)lnxx2−1−2lnx=1, after two rounds.
d)+∞; 0 (form ∞1); +∞
e)F(1.1)≈1.1017, the tangent gives 1.1: the graph is above its tangent.
a) As x→1, x2−1→0 and 2lnx→0: form 00. By the rule, 2/x2x; clear the complex fraction, 2x⋅2x=x2→1. So limx→1f(x)=1, from both sides. The continuous extension is F(x)=f(x) for x=1 and F(1)=1: F is defined at 1, its limit exists there, and the two agree, the three conditions of continuity. The rule is what turns a hole into a point of the graph.
b) As x→0+, x2−1→−1 and 2lnx→−∞. The form is −∞−1, which is NOT indeterminate: a bounded quantity divided by one that grows without bound tends to 0, here through positive values since top and bottom are both negative. So the limit is 0, and the graph ends at an open point at the origin. The rule must not be used: its hypothesis, a form 00 or ∞∞, is false. Applied anyway, it gives 2/x2x=x2→0, the right number by luck and a method error on the page, which the marker sees.
c) By definition, F′(1)=limx→1x−1F(x)−F(1)=limx→1x−12lnxx2−1−1. First the algebra: put the top over one denominator, then clear, 2lnxx2−1−2lnx⋅x−11=2(x−1)lnxx2−1−2lnx. Form 00. Round 1: the top gives 2x−x2, the bottom gives 2lnx+2(x−1)⋅x1 (product rule). Multiply top and bottom by x to clear the small fractions: 2xlnx+2x−22x2−2, still 00 at 1. Round 2: 2lnx+2+24x→0+44=1. So F′(1)=1. The quotient rule cannot be used at 1, where F is not given by the formula: the definition is the only way, and the rule makes it computable. Clearing before each round is what keeps round 2 to one line.
d) As x→∞, form ∞∞; one round gives x2→∞ as in a), so f(x)→+∞. Next, x2f(x)=2lnx1−x−2: the top tends to 1 and the bottom to ∞, form ∞1, not indeterminate, limit 0. Last, x1.9f(x)=2lnxx0.1−x−1.9, form ∞∞; one round gives 2x−10.1x−0.9+1.9x−2.9=20.1x0.1+1.9x−1.9→∞, multiplying top and bottom by x, that is ADDING 1 to each exponent. So f grows slower than x2 but faster than x1.9: the logarithm in the denominator costs less than any power.
e) F(1.1)=2ln1.11.21−1=0.190620…0.21≈1.1017. The tangent line at (1,1) has slope F′(1)=1: y=1+1⋅(x−1)=x, which gives 1.1 at x=1.1. Since 1.1017>1.1, the graph is above its tangent there, as the figure of the solution shows on the whole domain. The linearization of chapter 15 needed F′(1), and only L'Hôpital's Rule could provide it.