Corrected exercises: Rolle's theorem and the Mean Value Theorem (MATH 140)
This is the corrected exercise set for Rolle's theorem and the Mean Value Theorem in MATH 140, Calculus 1, at McGill University, section 4.2 of Stewart. It is the most theoretical chapter of the course and the one where a correct number earns the least: an examiner who asks for c expects the hypotheses checked first, and one who asks to prove an inequality or count roots expects the theorem named and its conclusion used, not a computation. Every number is exact and chosen to be done by hand.
The thread running through the whole set: the Mean Value Theorem is an existence statement bought with hypotheses. It is licensed only once continuity on the CLOSED interval [a,b] and differentiability on the OPEN interval (a,b) are checked, with a reason. It then promises a number c without locating it, and its real power lies in what it forbids: a slope never reached, a second root, a non-constant function with zero derivative, a change of value larger than the bound on f′.
The traps named in the solutions: reading f(a)=f(b) and skipping the other two hypotheses, forgetting continuity at an endpoint, keeping a value of c outside (a,b), believing the conclusion fails whenever a hypothesis does, concluding that a function with zero derivative is constant on a domain that is not an interval, trying to compute c in a proof of an inequality, claiming exactly one root from the Intermediate Value Theorem alone, and assuming that c is unique or is the midpoint.
Self-checking setType your answers and check them question by question: the page tells you right or wrong before you open the solution. With an account, every answer right on the first try earns points, and your right answers stay filled in when you come back.
•Rolle: f continuous on [a,b], differentiable on (a,b), f(a)=f(b)⇒f′(c)=0 for some c∈(a,b).
•Mean Value Theorem: f continuous on [a,b], differentiable on (a,b)⇒f(b)−f(a)=f′(c)(b−a) for some c∈(a,b).
•f′=0 on an INTERVAL ⇒f constant there; f′=g′ on an interval ⇒f=g+C there.
•Bounds: m≤f′(x)≤M on (a,b)⇒m(b−a)≤f(b)−f(a)≤M(b−a).
•Roots: IVT for AT LEAST one; Rolle by contradiction for AT MOST one (if f′ never vanishes); both for EXACTLY one.
•If f′ has n distinct zeros, f has at most n+1 distinct roots.
Part A: the basics (/50)
Exercise 1: Rolle's theorem: three hypotheses, checked one by one
Rolle's theorem. Let f be a function such that (1) f is continuous on the CLOSED interval [a,b]; (2) f is differentiable on the OPEN interval (a,b); (3) f(a)=f(b). Then there is at least one number c in (a,b) such that f′(c)=0.
A theorem is a contract: the conclusion is owed only when every hypothesis has been checked, with a reason, on the right interval. The figure shows y=x4−2x2 on [−2,2].
a) Verify the three hypotheses of Rolle's theorem for f(x)=x4−2x2 on [−2,2], then find ALL the numbers c that satisfy the conclusion.
b) Same question for g(x)=sinx+cosx on [0,2π].
c) Let h(x)=x4/5 on [−1,1]. Show that h(−1)=h(1) but that h′(c)=0 has no solution in (−1,1). Which hypothesis fails, and where?
d) tan0=tanπ=0, yet dxdtanx=sec2x never vanishes. Does this contradict Rolle's theorem?
e) Let k(x)=x for 0≤x<1 and k(1)=0. Check each hypothesis on [0,1] and decide whether the conclusion holds.
Show the solution
Answers
a)All three hold; c=−1, c=0 and c=1.
b)All three hold (g(0)=g(2π)=1); c=4π.
c)h(±1)=1 and h′(x)=5x1/54=0; hypothesis (2) fails at x=0.
d)No: tanx is not defined at 2π∈[0,π], so hypothesis (1) fails.
e)(2) and (3) hold, (1) fails at x=1; k′(x)=1, so the conclusion is false.
a)
Hypothesis (1): f is a polynomial, so it is continuous on R, in particular on [−2,2]. Hypothesis (2): a polynomial is differentiable on R, in particular on (−2,2). Hypothesis (3): f(−2)=16−8=8 and f(2)=16−8=8. The theorem applies. By the power rule, f′(x)=4x3−4x=4x(x−1)(x+1), which vanishes at x=−1, 0 and 1, all three in (−2,2). So c∈{−1,0,1}: the theorem promised AT LEAST one number and there are three, the two low points at height −1 and the local high point at the origin that the figure shows. Writing only the hypotheses that happen to be easy, or writing them without a reason (polynomial, hence continuous), loses the method marks even with the right c.
b)
Sine and cosine are continuous and differentiable on R, so (1) and (2) hold on [0,2π] and (0,2π). For (3): g(0)=0+1=1 and g(2π)=1+0=1. Then g′(x)=cosx−sinx=0 means sinx=cosx; on (0,2π), cosx>0, so we may divide: tanx=1, and x=4π. The other solutions of tanx=1, such as 45π, lie outside the interval and are not answers. At c=4π the curve reaches g(4π)=22+22=2, its top on the interval.
c)
The fifth root is defined for every real number, so h(x)=(x1/5)4 is continuous on [−1,1], and h(−1)=(−1)4=1=h(1): hypotheses (1) and (3) hold. For x=0, the power rule gives h′(x)=54x−1/5=5x1/54, a fraction with numerator 4, never 0. At x=0, the difference quotient is th(t)−h(0)=t−1/5, which tends to +∞ as t→0+ and to −∞ as t→0−: h′(0) does not exist. Hypothesis (2) fails at ONE point, 0, inside (−1,1), and that is enough to lose the conclusion. The graph has a cusp at the origin: a lowest point without a horizontal tangent.
d)
No contradiction, because the theorem does not apply. tanx=cosxsinx is not defined at x=2π, which lies in [0,π], and ∣tanx∣→∞ near it: hypothesis (1) fails (and (2) with it). The equal values at the ends are only one hypothesis out of three. Here sec2x=cos2x1≥1 wherever it is defined, so the slope is never 0, and it cannot be, since the graph is made of two separate branches. The trap is to read hypothesis (3) and stop there.
e)
Hypothesis (3): k(0)=0=k(1). Hypothesis (2): on (0,1), k(x)=x, so k is differentiable there with k′(x)=1. Hypothesis (1) fails at the right endpoint: limx→1−k(x)=1=0=k(1), so k is not continuous at 1. The conclusion is false, since k′(x)=1=0 everywhere in (0,1). The closed bracket in hypothesis (1) is not decoration: continuity AT the endpoints is what forbids the graph from jumping back down to its starting height at the last moment.
Tick the exercises you have done or want to review: a free account, no password, keeps your ticks and your right answers from one visit to the next, tells you which chapter to tackle next and lets you ask for the exercise you are missing. Create your space, an email is enough.
Exercise 2: The Mean Value Theorem: finding c, and keeping only the c in (a, b)
Mean Value Theorem. If f is continuous on [a,b] and differentiable on (a,b), then there is at least one c in (a,b) such that f′(c)=b−af(b)−f(a), or equivalently f(b)−f(a)=f′(c)(b−a).
Geometrically, somewhere strictly between a and b the tangent is parallel to the chord joining the endpoints. Rolle's theorem is the special case of a horizontal chord. The figure shows y=x3−x and its chord over [0,2].
a) Verify the hypotheses of the Mean Value Theorem for f(x)=x3−x on [0,2] and find every number c it produces.
b) Same question for f(x)=ex on [0,1]. Give c exactly, and prove without a calculator that it lies in (0,1).
c) Same question for f(x)=lnx on [1,e]. Compare with b).
d) For f(x)=x1 on [−1,1], show that no c satisfies the conclusion, and explain why.
e) For f(x)=x1/3 on [−1,1], show that a hypothesis fails, then find the numbers c anyway. What does this say about the hypotheses?
d)Slope 1 but f′(x)=−x21<0; f is not continuous on [−1,1] (undefined at 0).
e)Not differentiable at 0; yet c=±93. Hypotheses are sufficient, not necessary.
a)
f is a polynomial, hence continuous on [0,2] and differentiable on (0,2). The slope of the chord is 2−0f(2)−f(0)=26−0=3. By the power rule f′(c)=3c2−1, and 3c2−1=3 gives c2=34, so c=±32=±323. The theorem speaks about c in (0,2) only: −323 is rejected, and 323 is kept because it is positive and (323)2=34<4, so it is less than 2. Answering with both values is the most common loss on this question: solving f′(c)=m is algebra, choosing c is the theorem. The solution figure shows the tangent at c, parallel to the chord.
b)
ex is continuous and differentiable on R. The chord has slope 1−0e1−e0=e−1, and f′(c)=ec, so ec=e−1 and c=ln(e−1). That e−1>0 is needed for the logarithm to exist. Now 2<e, so e−1>1, and obviously e−1<e. Since ln is increasing, ln1<ln(e−1)<lne, that is 0<c<1. No decimal is required: the exact answer is ln(e−1), and the inequality is proved with e>2 alone.
c)
lnx is continuous and differentiable on (0,∞), which contains [1,e]. The chord has slope e−1lne−ln1=e−11, and f′(c)=c1, so c1=e−11 and c=e−1, which lies in (1,e) by the same two inequalities as in b). The two answers are linked: ln is the inverse of exp, the graph of ln on [1,e] is the mirror image of the graph of ex on [0,1] in the line y=x, the chord is reflected into a chord and the tangent into a tangent, so the point of contact (ln(e−1),e−1) becomes (e−1,ln(e−1)).
d)
The chord has slope 1−(−1)f(1)−f(−1)=21−(−1)=1, but f′(x)=−x21 is negative wherever it exists: no c can give 1. No contradiction: x1 is not defined at 0, which lies in [−1,1], so f is not continuous on [−1,1] and hypothesis (1) fails. The graph goes down on each branch and jumps from −∞ to +∞ at 0, and it is the jump that lets the endpoints be in increasing order.
e)
x1/3 is continuous on R, but at 0 the difference quotient tt1/3=t−2/3→+∞: vertical tangent, no derivative, so hypothesis (2) fails at 0∈(−1,1). Yet the chord has slope 21−(−1)=1 and, for x=0, f′(x)=3x2/31=1 gives x2/3=31, so ∣x∣=(31)3/2=331=93, and c=±93, both in (−1,1). The conclusion holds although a hypothesis fails. The hypotheses are SUFFICIENT, not necessary: when one fails, the theorem simply guarantees nothing, and the conclusion may hold, as here, or fail, as in d).
Exercise 3: Zero derivative: constant on an interval, and only on an interval
Two consequences of the Mean Value Theorem carry the rest of the course. If f′(x)=0 for every x in an INTERVAL (a,b), then f is constant on (a,b). If f′(x)=g′(x) for every x in an interval, then f−g is constant there: f(x)=g(x)+C.
The word interval is not a detail: the proof applies the Mean Value Theorem between any two points of the set, which requires the whole segment between them to be inside it.
a) Prove the first consequence from the Mean Value Theorem.
b) Prove that arcsinx+arccosx=2π for every x in [−1,1].
c) A function f is differentiable on R, with f′(x)=3x2 for every x and f(2)=5. Prove that f(x)=x3−3 for every x.
d) Prove that if f′′(x)=0 for every real x, then f(x)=mx+b for some constants m and b.
e) Let D(x)=arctan(1−x1+x)−arctanx for x=1. Show that D′(x)=0 at every point of its domain. A student concludes that D(x)=D(0)=4π for all x=1. Compute D(3) and explain.
Show the solution
Answers
a)For x1<x2 in (a,b): f(x2)−f(x1)=f′(c)(x2−x1)=0.
b)F′=1−x21−1−x21=0 on (−1,1), F(0)=2π, and F(±1)=2π directly.
c)g=f−x3 has g′=0 on R, so g=C=f(2)−8=−3.
d)f′=m constant, then f−mx has zero derivative: f=mx+b.
e)D(3)=−43π: D=4π on (−∞,1) and −43π on (1,∞); the domain is not an interval.
a)
Take any two numbers x1<x2 in (a,b). The segment [x1,x2] lies inside (a,b), because (a,b) is an interval; there f is differentiable, hence continuous on [x1,x2] and differentiable on (x1,x2). The Mean Value Theorem gives c in (x1,x2) with f(x2)−f(x1)=f′(c)(x2−x1). Since c is in (a,b), f′(c)=0, so f(x2)=f(x1). Any two values of f are equal: f is constant. The unknown c is never computed; all we need is that it lands where f′=0. For the second consequence, apply this to h=f−g, whose derivative is f′−g′=0.
b)
Let F(x)=arcsinx+arccosx. On (−1,1), F′(x)=1−x21+(−1−x21)=0, and (−1,1) is an interval, so F is constant there, equal to F(0)=0+2π=2π. The endpoints need their own line, since the derivatives do not exist at ±1: F(1)=2π+0 and F(−1)=−2π+π=2π. The identity holds on all of [−1,1]. The shape of the proof is the one to learn: derivative zero, constant on the interval, the constant read at one convenient point.
c)
Let g(x)=f(x)−x3. By the difference and power rules, g′(x)=3x2−3x2=0 for every x in R, which is an interval, so g is a constant C. At x=2: C=g(2)=f(2)−8=5−8=−3. Hence f(x)=x3−3 for every x, and no other function fits. This is the uniqueness behind every initial value problem: two functions with the same derivative on an interval differ by a constant, and one value fixes it.
d)
f′′(x)=(f′)′(x)=0 for every real x, so by a) applied to f′ on R, f′ is a constant, call it m. Then k(x)=f(x)−mx has k′(x)=m−m=0 on R, so k is a constant b, and f(x)=mx+b. The consequence is used twice, once per derivative: a function whose second derivative vanishes has a graph that is a straight line, and the proof says why, instead of saying that it is obvious.
e)
Let u=1−x1+x. The quotient rule gives u′=(1−x)2(1)(1−x)−(1+x)(−1)=(1−x)22, and 1+u2=(1−x)2(1−x)2+(1+x)2=(1−x)22+2x2. By the chain rule with inner function u, dxdarctanu=1+u2u′=2+2x22=1+x21, so D′(x)=1+x21−1+x21=0 for x=1. At x=3: 1−31+3=1−3(1+3)2=−(2+3), and tan125π=tan(4π+6π)=1−311+31=3−13+1=2+3, so arctan(−(2+3))=−125π. Hence D(3)=−125π−3π=−43π=4π. The student's error: the domain (−∞,1)∪(1,∞) is NOT an interval, and the Mean Value Theorem cannot be applied on [0,3], which contains the point 1 where D is not even defined. D is constant on EACH piece: 4π on the left, −43π on the right, as the solution figure shows.
Exercise 4: Inequalities from the Mean Value Theorem: bound f', not c
The Mean Value Theorem turns a bound on the derivative into a bound on the function: if m≤f′(x)≤M for every x in (a,b), and f satisfies the hypotheses on [a,b], then m(b−a)≤f(b)−f(a)≤M(b−a). The number c stays unknown; we only use the interval it lives in.
The figure shows y=x, y=arctanx and y=1+x2x for x≥0.
a) Prove that ∣sina−sinb∣≤∣a−b∣ for all real numbers a and b. Deduce that ∣sinx∣≤∣x∣ for every x.
b) Prove that 1+x2x<arctanx<x for every x>0.
c) Deduce from b), with x=31, that 233<π<23.
d) A function f is differentiable on R, with f(2)=−3 and f′(x)≤5 for every x. How large can f(7) possibly be? Show that this value is reached.
e) Prove that 1+x<1+2x for every x>0.
Show the solution
Answers
a)sinb−sina=cosc(b−a) and ∣cosc∣≤1; with b=0, ∣sinx∣≤∣x∣.
b)arctanx=1+c2x with 0<c<x, and 1+x21<1+c21<1.
c)43<6π<33, so 233<π<23.
d)f(7)≤22, reached by f(x)=5x−13.
e)1+x−1=21+cx<2x with 0<c<x.
a)
If a=b both sides are 0. Otherwise, by symmetry of the inequality we may assume a<b. Sine is continuous and differentiable on R, so the Mean Value Theorem on [a,b] gives c in (a,b) with sinb−sina=cosc(b−a). Whatever c is, ∣cosc∣≤1, hence ∣sinb−sina∣=∣cosc∣∣b−a∣≤∣b−a∣. With b=0: ∣sina∣≤∣a∣ for every a. The student who tries to find c is stuck: it depends on a and b and has no formula. The whole method is to bound f′ on the interval and let c be anywhere in it.
b)
Fix x>0. arctan is continuous and differentiable on R, with derivative 1+t21. The Mean Value Theorem on [0,x] gives c in (0,x) with arctanx−arctan0=1+c2x. From 0<c<x: 1<1+c2<1+x2, so 1+x21<1+c21<1, and multiplying by x>0 keeps the directions: 1+x2x<arctanx<x. The inequalities are strict because c is STRICTLY between 0 and x. The figure shows the three curves in that order for every x>0, touching only at the origin.
c)
With x=31: arctan31=6π, and 1+x2x=4/31/3=433=43. So 43<6π<31=33, and multiplying by 6: 233<π<23. In decimals, about 2.60<π<3.46: a crude bracket, but a proved one, obtained by hand from a derivative bound. The same method on a smaller angle gives a sharper bracket, since c is then trapped in a shorter interval.
d)
f is differentiable on R, hence continuous on [2,7] and differentiable on (2,7). By the Mean Value Theorem, f(7)−f(2)=f′(c)(7−2)=5f′(c) for some c in (2,7), and f′(c)≤5, so f(7)−f(2)≤25 and f(7)≤−3+25=22. The bound is reached: f(x)=5x−13 has f(2)=−3, f′(x)=5≤5 and f(7)=22. So the largest possible value is 22. A bound is only the answer to how large can it be when it is shown to be attained: that second line earns its own mark.
e)
Fix x>0 and let f(t)=1+t, continuous on [0,x] and differentiable on (0,x) since 1+t>0 there, with f′(t)=21+t1 by the chain rule. The Mean Value Theorem gives c in (0,x) with 1+x−1=21+cx. Since c>0, 1+c>1, so 21+cx<2x, and 1+x<1+2x. Check at x=3: 4=2<2.5. The same four lines prove every inequality of the form f(x)−f(0) against x times a bound: choose f, name the interval, apply the theorem, bound f′(c) using only where c lies.
Exercise 5: At most one root, exactly one root: Rolle by contradiction
The Intermediate Value Theorem proves that a root EXISTS. Rolle's theorem proves that there cannot be TWO: if f(r1)=f(r2)=0 with r1<r2, and f is continuous on [r1,r2] and differentiable on (r1,r2), then f′(c)=0 for some c between them. So if f′ never vanishes, f has at most one root; existence plus this gives exactly one.
The figure, for part e), shows y=x3−4x on [−3,3].
a) Prove that x5+3x+1=0 has exactly one real root, and locate it between two consecutive integers.
b) Prove that 3x+cosx=2 has exactly one real solution, and that it lies in (0,1).
c) Prove that for every constant k, the equation x4+4x+k=0 has at most two real roots.
d) Prove that for every constant k, the equation x3−15x+k=0 has at most one root in [−2,2].
e) For f(x)=x3−4x, f(−3)=−15<0<15=f(3). A student writes: by the Intermediate Value Theorem, f has exactly one root in (−3,3). Correct this using the figure, and say which argument is missing.
Show the solution
Answers
a)One root, in (−1,0): f(−1)=−3, f(0)=1, and f′(x)=5x4+3>0.
b)One solution, in (0,1): f(0)=−1, f(1)=1+cos1>0, f′(x)=3−sinx≥2.
c)Three roots would give two zeros of f′(x)=4(x3+1), which has only x=−1.
d)Two roots would give c∈(−2,2) with 3c2=15, but ∣c∣=5>2.
e)Three roots, −2, 0, 2; the IVT gives AT LEAST one, and f′ vanishes at ±323.
a)
Let f(x)=x5+3x+1, a polynomial, so continuous and differentiable on R. Existence: f(−1)=−1−3+1=−3<0 and f(0)=1>0, so by the Intermediate Value Theorem on [−1,0] there is a root in (−1,0). Uniqueness, by contradiction: suppose f had two roots r1<r2. Then f is continuous on [r1,r2], differentiable on (r1,r2), and f(r1)=f(r2)=0, so Rolle's theorem gives c with f′(c)=0. But f′(x)=5x4+3≥3>0 for every x: contradiction. The equation has exactly one real root, and it lies in (−1,0). The three hypotheses of Rolle must be named in the contradiction as in any other use of the theorem.
b)
Let f(x)=3x+cosx−2, continuous and differentiable on R. f(0)=0+1−2=−1<0 and f(1)=3+cos1−2=1+cos1>0, because 0<1<2π gives cos1>0. The Intermediate Value Theorem gives a solution in (0,1). If there were two, Rolle's theorem would give c with f′(c)=0; but f′(x)=3−sinx≥3−1=2>0. Exactly one solution, in (0,1). No value of cos1 is needed, only its sign: the no-calculator rule is met by an inequality, not by a decimal.
c)
Let f(x)=x4+4x+k. Its derivative f′(x)=4x3+4=4(x3+1) vanishes only at x=−1, the only real cube root of −1. Suppose f had three roots r1<r2<r3. Rolle's theorem on [r1,r2] and on [r2,r3] (a polynomial satisfies the first two hypotheses on any interval, and the values at the ends are 0) gives c1 in (r1,r2) and c2 in (r2,r3) with f′(c1)=f′(c2)=0. The two intervals do not overlap, so c1<r2<c2 and c1=c2: two distinct zeros of f′, which has only one. Contradiction: at most two real roots, whatever k is. Saying why c1=c2 is the step most papers forget.
d)
Let f(x)=x3−15x+k. Suppose f had two roots r1<r2 in [−2,2]. A polynomial satisfies Rolle's hypotheses on [r1,r2], so f′(c)=3c2−15=0 for some c in (r1,r2), which is inside (−2,2). But 3c2=15 gives c=±5, and 5>2 since 5>4: neither value is in (−2,2). Contradiction: at most one root in [−2,2]. Whether there is one depends on k: f(−2)=22+k and f(2)=k−22, so the Intermediate Value Theorem gives one as soon as −22≤k≤22. The restriction to [−2,2] is essential; on R, k=0 gives the three roots 0 and ±15.
e)
x3−4x=x(x−2)(x+2) has three roots in (−3,3): −2, 0 and 2, the three red points of the figure. The Intermediate Value Theorem only says that there is AT LEAST one root between two values of opposite signs; it never says how many. Uniqueness needs the second argument, Rolle by contradiction, and here it cannot even start: f′(x)=3x2−4 vanishes at ±32=±323, both inside (−3,3), so nothing forbids several roots. Correct statement: f has at least one root in (−3,3). The word exactly costs the whole question when only the IVT has been used.
Part B: problems and reasoning (/50)
Exercise 6: A function known by three values: the slopes it must take
A function f is differentiable on R, and all we know about it is the table f(0)=1, f(2)=7, f(5)=4. The theorems of the chapter still force some of its slopes, whatever f is.
The figure shows two different functions consistent with the table: the parabola q(x)=−54x2+523x+1 in blue, and a wavier function w in orange.
a) Prove that there are numbers c1 in (0,2) and c2 in (2,5) such that f′(c1)=3 and f′(c2)=−1.
b) Prove that f′(c3)=53 for some c3 in (0,5).
c) Prove that f′(c)=0 for some c in (0,5), although f(0)=f(5).
d) Can one guarantee that f′(x)=5 for some x in (0,5)?
e) On the figure, count the horizontal tangents of q and of w on (0,5). Which of the conclusions a) to d) depend on which function f is?
Show the solution
Answers
a)MVT on [0,2]: slope 27−1=3; on [2,5]: slope 34−7=−1.
b)MVT on [0,5]: slope 54−1=53.
c)IVT: f(d)=4 for some d in (0,2); Rolle on [d,5] gives f′(c)=0.
d)No: q′(x)=523−8x<523<5 on (0,5).
e)q: one (x=823); w: three. The existence in a) to c) holds for both; the number and position of the c depend on f.
a)
f is differentiable on R, so on any closed interval it is continuous and on the open interval differentiable: the Mean Value Theorem applies everywhere, and this sentence must still be written. On [0,2]: f′(c1)=2−0f(2)−f(0)=27−1=3 for some c1 in (0,2). On [2,5]: f′(c2)=5−2f(5)−f(2)=34−7=−1 for some c2 in (2,5). Each pair of values in the table gives one guaranteed slope, and nothing more: we know neither c1 nor c2.
b)
The third pair of values: on [0,5], f′(c3)=5−0f(5)−f(0)=54−1=53 for some c3 in (0,5). Three data points give three chords and three guaranteed slopes, 3, −1 and 53. On the parabola q, where q′(x)=523−8x, these are reached at x=1, x=27 and x=25, the midpoints of the three intervals: for a quadratic, and only for a quadratic, c is always the midpoint.
c)
Rolle's theorem cannot be applied on [0,5] directly, since f(0)=1=4=f(5). First create two equal values. f is continuous on [0,2] and 4 lies between f(0)=1 and f(2)=7, so by the Intermediate Value Theorem there is d in (0,2) with f(d)=4. Now f(d)=f(5)=4, f is continuous on [d,5] and differentiable on (d,5), and Rolle's theorem gives c in (d,5)⊂(0,5) with f′(c)=0. The two theorems work in sequence: the IVT manufactures the equal values that Rolle needs. Another route, the Extreme Value Theorem on [0,5] and Fermat's theorem, works too, provided one proves that the maximum is not at an endpoint, which f(2)=7>f(0) and f(2)>f(5) do.
d)
No. A guarantee must hold for EVERY function matching the table, so one counterexample is enough. The parabola q matches: q(0)=1, q(2)=−516+546+1=7, q(5)=−20+23+1=4. Its derivative q′(x)=523−8x decreases from 523 at 0 to −517 at 5, so on (0,5) it stays strictly below 523<5: q′ never equals 5 there. The theorem gives only the slopes of chords; a slope steeper than every chord may or may not occur.
e)
On the figure, q has one horizontal tangent on (0,5), at its vertex x=823, where q′(x)=0. The orange function w has three: a low point near x=0.5, a high point near x=2.8, and a low point near x=4.7. Both satisfy a), b) and c), because those conclusions were proved from the table alone. What differs is how many c there are and where they sit; the theorems are statements of existence and are silent on both. The answer to d) is the one that depends on f: for q the slope 5 never occurs on (0,5), while w, which dips before climbing to 7, is steeper than 5 on part of its climb and so takes the slope 5, near x=0.9 for instance. This is why only a proof, never a drawing, answers a question of the form must there be.
Exercise 7: Proving the Mean Value Theorem from Rolle's theorem
Rolle's theorem is the special case of the Mean Value Theorem where the chord is horizontal, and the general case reduces to it by one idea: subtract the chord. Assume f is continuous on [a,b] and differentiable on (a,b), let m=b−af(b)−f(a), and define
h(x)=f(x)−[f(a)+m(x−a)].
a) Interpret h(x) geometrically, and compute h(a) and h(b).
b) Prove that h satisfies the three hypotheses of Rolle's theorem on [a,b].
c) Complete the proof of the Mean Value Theorem.
d) Illustrate with f(x)=x3 on [0,3]: write h, find c, and show that at c the vertical distance between the curve and the chord is the largest on [0,3].
e) Rolle's theorem itself is proved with the Extreme Value Theorem and Fermat's theorem. Say at which step each hypothesis of the Mean Value Theorem is used, and why the proof gives no formula for c.
Show the solution
Answers
a)h(x) is the signed vertical gap between the curve and the chord; h(a)=h(b)=0.
b)h = f minus a linear function: continuous on [a,b], differentiable on (a,b), equal ends.
c)Rolle: h′(c)=f′(c)−m=0, so f′(c)=b−af(b)−f(a).
d)h(x)=x3−9x, c=3, largest gap ∣h(3)∣=63.
e)Continuity on [a,b] feeds the EVT, differentiability at c feeds Fermat; the EVT only asserts existence.
a)
The chord through (a,f(a)) and (b,f(b)) is the line L(x)=f(a)+m(x−a), so h(x)=f(x)−L(x) is the signed vertical distance from the chord to the curve: positive where the curve is above the chord. At the ends, h(a)=f(a)−f(a)−m⋅0=0 and h(b)=f(b)−f(a)−m(b−a)=f(b)−f(a)−(f(b)−f(a))=0, by the definition of m. The whole proof rests on this choice: subtracting the chord makes the endpoint values equal, which is the one hypothesis Rolle has and the Mean Value Theorem lacks.
b)
L is a polynomial of degree at most 1, continuous and differentiable on R. Hypothesis (1): f and L are continuous on [a,b], so their difference h is. Hypothesis (2): f and L are differentiable on (a,b), so h is, with h′(x)=f′(x)−m. Hypothesis (3): h(a)=h(b)=0 by a). Each hypothesis of Rolle is inherited from a hypothesis of the Mean Value Theorem, and the write-up must say from which one, since this is the entire content of the step.
c)
By Rolle's theorem there is c in (a,b) with h′(c)=0, that is f′(c)−m=0, so f′(c)=m=b−af(b)−f(a). This is the conclusion of the Mean Value Theorem. The proof takes four lines once h is chosen, which is why an exam may ask for it: the marks are in the definition of h, the check h(a)=h(b), and the derivative h′=f′−m.
d)
With f(x)=x3, a=0, b=3: m=327−0=9, the chord is L(x)=9x, and h(x)=x3−9x, with h(0)=0 and h(3)=27−27=0. Then h′(x)=3x2−9=0 gives x=±3, and only c=3 lies in (0,3); indeed f′(3)=3⋅3=9=m. On the closed interval [0,3], the continuous function h has its extreme values at the critical number or at the ends: h(0)=0, h(3)=0, h(3)=33−93=−63. So the curve is below the chord, and the gap ∣h∣ is largest, 63, exactly at c=3. The tangent parallel to the chord touches the curve where the curve is farthest from the chord. Note c=3 is not the midpoint 23: 3>23 since 3>49.
e)
Proof of Rolle: if h is constant, h′=0 everywhere and any c works. Otherwise h takes a value different from h(a)=h(b), say larger. Because h is continuous on the CLOSED interval [a,b], the Extreme Value Theorem gives a point where h reaches its maximum, and that maximum exceeds h(a)=h(b), so the point c is inside (a,b). Because h is differentiable at that interior point, Fermat's theorem gives h′(c)=0. So continuity on [a,b] is consumed by the Extreme Value Theorem, differentiability on (a,b) by Fermat's theorem, and the equal values by the placement of c away from the ends. The Extreme Value Theorem asserts that a maximum exists without saying where it is: that is why the Mean Value Theorem, built on it, never locates c. To find c one must solve f′(c)=m, which is possible for x3 and impossible in closed form for most functions.
Exercise 8: Five statements to correct
Each statement below was written by a student on a MATH 140 assignment, and each is false. Say what is wrong, give a counterexample, and write a correct statement.
a) If f(a)=f(b), then f′(c)=0 for some c in (a,b).
b) The Mean Value Theorem says that the slope of the chord equals the average 2f′(a)+f′(b).
c) If ∣f′(x)∣≤1 at every point where f is differentiable, then ∣f(b)−f(a)∣≤∣b−a∣.
d) Between two roots of f there is always a root of f′.
e) The number c of the Mean Value Theorem is unique.
Show the solution
Answers
a)False: x21 on [−1,1]. True with continuity on [a,b] and differentiability on (a,b).
b)False: x3 on [0,1], chord slope 1, average 23. True only for quadratics.
c)False: ⌊x⌋ on [21,1]. True if f is continuous on [a,b] and differentiable on (a,b).
d)False: 1−x2/3, roots ±1, f′ never 0. True if f is differentiable between the roots.
e)False: x3 on [−2,2] gives c=±323. The theorem says at least one.
a)
FALSE: the hypotheses have been dropped. f(x)=x21 has f(−1)=f(1)=1, yet f′(x)=−x32 is never 0. The function is not defined at 0, so it is not continuous on [−1,1]; its graph climbs to infinity on both sides of 0 and never turns. Correct statement (Rolle): if f is continuous on [a,b], differentiable on (a,b) and f(a)=f(b), then f′(c)=0 for some c in (a,b). Equal values at the ends are the least important of the three hypotheses.
b)
FALSE. For f(x)=x3 on [0,1]: the chord has slope 1−01−0=1, while 2f′(0)+f′(1)=20+3=23. The theorem says the chord slope is the derivative at SOME interior point, here 3c2=1, c=31, not the average of the derivatives at the ends. The student probably checked on x2: for a quadratic, f′ is linear, and the average of f′(a) and f′(b) is f′ at the midpoint, which is the chord slope. A rule verified on one family is not a theorem.
c)
FALSE. f(x)=⌊x⌋, the greatest integer ≤x, is differentiable with f′(x)=0 at every non-integer x, so ∣f′∣≤1 wherever f′ exists. Yet on [21,1]: ∣f(1)−f(21)∣=∣1−0∣=1>21. The jump at 1 is where the function changes, and no derivative sees it. Correct statement: if f is continuous on [a,b], differentiable on (a,b), and ∣f′(x)∣≤1 there, then ∣f(b)−f(a)∣≤∣b−a∣, by the Mean Value Theorem as in Exercise 4.
d)
FALSE. f(x)=1−x2/3 has roots −1 and 1, since (±1)2/3=1. For x=0, f′(x)=−3x1/32, never 0, and f′(0) does not exist: the graph has a cusp at the top. Rolle's hypothesis (2) fails at one point and the conclusion goes with it. Correct statement: if f is continuous on [r1,r2] and differentiable on (r1,r2), where r1<r2 are roots, then f′ has a root in (r1,r2). For a polynomial the hypotheses always hold, which is why the statement feels true.
e)
FALSE. f(x)=x3 on [−2,2]: the chord slope is 48−(−8)=4, and 3c2=4 gives c=±32=±323, both in (−2,2), since 34<4. Two values. Correct statement: there is AT LEAST one c. In a question that asks for all the numbers c, stopping at the first one found is a lost mark; in a proof, assuming uniqueness is a lost argument.
Exercise 9: Average speed cameras: the theorem that writes the ticket
On some highways, speed is not measured at one point: two gantries A and B, 30 km apart, photograph each vehicle and record the times. The speed limit is 100 km/h. A car passes gantry A at 14:00 and gantry B at 14:15. Let s(t) be its position along the road, in km from A, at time t; the motion is continuous and s is differentiable, its derivative being the speed shown on the speedometer.
The figure shows ONE possible graph of s (time in minutes after 14:00), for part e).
a) Compute the average speed between the gantries and prove, naming the theorem and its hypotheses, that at some instant the car was driving at exactly 120 km/h.
b) The driver shows that the cameras measured 100 km/h at each gantry. Does this invalidate a)?
c) A truck whose speed is limited by its governor to 90 km/h passes gantry A at 9:00. Prove that it cannot pass gantry B before 9:20.
d) The driver of a) now claims to have been stopped by an accident from 14:05 to 14:08. Assuming this is true, prove that at some instant the car was driving at 150 km/h or more.
e) On the figure, the car covers 11 km before the stop and 19 km after it. On which stretch does the Mean Value Theorem guarantee a speed of 150 km/h or more for this graph? Can the police say at what instant, and for how long, the limit was exceeded?
Show the solution
Answers
a)1/430=120 km/h; MVT on [0,41] gives s′(c)=120.
b)No: the theorem gives an instant strictly between the gantries, whatever the speeds at the ends.
c)30=s′(c)T≤90T, so T≥31 h =20 min.
d)d1+d2=30; if both averages were below 2.5 km/min, d1<12.5 and d2<17.5: contradiction.
e)Second stretch: 719 km/min =71140 km/h >150. Neither the instant nor the duration.
a)
The car covers 30 km in 15 minutes, that is 41 hour: average speed 1/430=120 km/h. Measure t in hours from 14:00. The position s is continuous on [0,41] and differentiable on (0,41), as the statement says of any motion. By the Mean Value Theorem there is c in (0,41) with s′(c)=1/4−0s(1/4)−s(0)=1/430−0=120. At that instant the speedometer showed exactly 120 km/h, above the limit of 100: the offence is proved, although no camera saw the car speeding. Average speed cameras rest on exactly this theorem.
b)
No. The theorem guarantees an instant c STRICTLY between the gantries where s′(c)=120; the speeds at the two endpoints, s′(0) and s′(41), play no role in it and may be anything. On the figure the car passes each gantry at 100 km/h (the slope at both ends is 35 km per minute) and still has to drive faster in between to cover 30 km in 15 minutes. Slowing down in front of the gantries changes the photographs, not the average.
c)
Let T be the time in hours taken by the truck, with s(0)=0 and s(T)=30. By the Mean Value Theorem on [0,T], 30=s(T)−s(0)=s′(c)T for some c in (0,T), and s′(c)≤90, so 30≤90T and T≥9030=31 hour, that is 20 minutes. The truck cannot pass gantry B before 9:20. This is the inequality form of the theorem: a bound on the speed at every instant bounds the distance covered.
d)
Work in km and minutes, with t=0 at 14:00: the car moves on [0,5] and [8,15] and stands still on [5,8]. Let d1=s(5)−s(0) and d2=s(15)−s(8); since s(8)=s(5), d1+d2=30. The Mean Value Theorem on [0,5] gives an instant where the speed is 5d1 km per minute, and on [8,15] an instant where it is 7d2. Suppose both were below 2.5 km per minute, that is 150 km/h: then d1<12.5 and d2<17.5, so d1+d2<30, a contradiction. Hence one of the two instants has a speed of at least 2.5 km per minute, 150 km/h. The alibi makes the case worse: 30 km in the 12 minutes of actual driving is an average of 150 km/h.
e)
For the graph shown, the first stretch has average speed 511=2.2 km per minute, 132 km/h, and the second 719 km per minute, that is 719×60=71140 km/h, more than 162 km/h since 7×162=1134<1140. So for THIS graph the Mean Value Theorem places a speed above 150 km/h on the second stretch, between 14:08 and 14:15. For another car with d1=15, the first stretch would carry it, at 3 km per minute. As for when and how long: the theorem gives neither. It asserts that an instant exists; it does not say at what time, and a speed above the limit may have lasted two seconds or ten minutes. The ticket is for exceeding the limit, and existence is all it needs.
Exercise 10: A final exam question: exactly three roots, and one of them pinned down
A graphing program draws y=x5−5x+1 as in the figure: three crossings of the axis. A picture is not a proof, and on the final exam you have no graphing program. This question is the shape of a long final exam problem: existence, then an upper bound on the number of roots, then a guaranteed location, all without a calculator.
Let fk(x)=x5−5x+k, where k is a constant, and f=f1.
a) Prove that f(x)=x5−5x+1 has at least three real roots, and locate each between two consecutive integers.
b) Prove that f has at most three real roots.
c) Let r be the root in (0,1). Show that 0<r<41, then apply the Mean Value Theorem on [0,r] to prove that 51<r<1275256.
d) Prove that for every k with −4<k<4, the equation x5−5x+k=0 has exactly three real roots.
e) For k=4, check that f4(x)=(x−1)2(x3+2x2+3x+4) and find how many distinct real roots f4 has. Why does the argument of d) break down?
Show the solution
Answers
a)f(−2)=−21, f(−1)=5, f(0)=1, f(1)=−3, f(2)=23: roots in (−2,−1), (0,1), (1,2).
b)Four roots would give three zeros of f′(x)=5(x2+1)(x−1)(x+1), which has two.
c)f(41)=−1024255<0; r=5−5c41 with 0<c<41, so 51<r<1275256.
d)fk(−2)<0<fk(−1), fk(1)<0<fk(2), and the bound of b) holds for every k.
e)Two distinct roots: 1 (double) and one in (−2,−1). The sign change at 1 disappears.
a)
f is a polynomial, so continuous on R. Its values: f(−2)=−32+10+1=−21, f(−1)=−1+5+1=5, f(0)=1, f(1)=1−5+1=−3, f(2)=32−10+1=23. The sign changes on [−2,−1], on [0,1] and on [1,2], so by the Intermediate Value Theorem on each of these three intervals there is a root in (−2,−1), one in (0,1) and one in (1,2). The open intervals are disjoint, so these are three DIFFERENT roots. Note that [−1,0] shows no sign change, and the IVT says nothing there, neither yes nor no.
b)
f′(x)=5x4−5=5(x2−1)(x2+1)=5(x−1)(x+1)(x2+1), and since x2+1>0, f′ has exactly two real zeros, −1 and 1. Suppose f had four roots r1<r2<r3<r4. Rolle's theorem on [r1,r2], [r2,r3] and [r3,r4], whose hypotheses a polynomial meets, gives three zeros of f′ in three disjoint open intervals, hence three distinct zeros. Contradiction: f has at most three real roots, and with a), exactly three.
c)
f(0)=1>0 and f(41)=10241−45+1=10241−1280+1024=−1024255<0, so the IVT gives a root in (0,41); since f has only one root in (0,1), by b), it is r. Now apply the Mean Value Theorem to f on [0,r]: f(r)−f(0)=f′(c)r for some c in (0,r), that is −1=(5c4−5)r, so r=5−5c41. From 0<c<r<41: 0<c4<2561, hence 5−2565<5−5c4<5, that is 2561275<5−5c4<5, and taking reciprocals reverses the inequalities: 51<r<1275256. Since 1275256−51=12751, the root is pinned down to within 12751, less than 0.001, by hand. The unknown c did its job without ever being found.
d)
fk is a polynomial for every k. fk(−2)=−22+k<0 and fk(2)=22+k>0 because ∣k∣<4<22; fk(−1)=4+k>0 because k>−4; fk(1)=k−4<0 because k<4. So fk changes sign on [−2,−1], [−1,1] and [1,2], and the IVT gives three roots in three disjoint open intervals. The derivative fk′=f′ does not depend on k, so the argument of b) gives at most three roots for every k. Exactly three. The values fk(−1)=4+k and fk(1)=k−4 are the heights at the two points where the tangent is horizontal, and the condition −4<k<4 says that they have opposite signs.
e)
Expanding, (x2−2x+1)(x3+2x2+3x+4)=x5+2x4+3x3+4x2−2x4−4x3−6x2−8x+x3+2x2+3x+4=x5−5x+4. So 1 is a root, and a double one. For g(x)=x3+2x2+3x+4: g(−2)=−2<0<2=g(−1) gives a root in (−2,−1), and g′(x)=3x2+4x+3 has discriminant 16−36<0, so g′>0 never vanishes and, by Rolle, g has no second root. Also g(1)=10=0. So f4 has exactly two distinct real roots: 1 and one number in (−2,−1). The argument of d) needed fk(1)<0; at k=4, f4(1)=0 and the graph touches the axis at x=1 without crossing it, so two of the three sign changes merge into a single tangency. The upper bound of b) still holds, three at most; it is the existence part that loses a root.
Something missing from this set?A type of exercise your teacher gives, a problem you are stuck on, a mistake you spotted: tell me, I read every request.
Sending goes through your free space: that is how I can answer you and tell you when the exercise is added. An email is enough, no password. Create your space.
Rather not create an account? Write to me directly.