A continuous function on a closed interval [a,b] always has a largest value and a smallest value. If one of those values happens strictly inside the interval, at a point where the derivative exists, the derivative there must be 0, so the tangent line is horizontal.
What if the function starts and ends at the same height, so f(a)=f(b)? Does the tangent line have to be horizontal somewhere between a and b?
Rolle's theorem says it does, as long as f is continuous on [a,b] and its derivative exists at every point inside (a,b).
f continuous on [a,b]f′(x) exists for every x in (a,b)f(a)=f(b)
f′(c)=0
for at least one c in (a,b)
The theorem says that at least one such number c exists, but it doesn't give the value of c. Let's see why there has to be one, on f(x)=4x−x3 over [0,2], and then find it.
Find where the tangent to f(x)=4x−x3 on [0,2] is horizontal
Text solution
The largest value is insideBoth ends are at height 0, and 4x−x3=x(4−x2) is positive between them, so the largest value can't be at 0 or at 2.
f(0)=0
f(2)=0
f(x)>0 on (0,2)
The tangent there is flatA polynomial is differentiable everywhere, so the derivative at that highest point is 0. If a function dips below its ends instead, its lowest point is inside, and if it's constant, its slope is 0 everywhere. That's why Rolle's theorem holds.
To find c, we set f′(x) equal to 0 and keep only the solution inside (0,2).
f(x)=4x−x3
f(0)=0
f(2)=8−8=0
f′(x)=4−3x2
4−3c2=0→c2=34→c=±32
4−3c2=0→c2=34
→c=±32
−32<0,not in (0,2)
c=32≈1.15,in (0,2)
2
The Mean Value Theorem
In Rolle's theorem, the endpoints were at the same height, so f(a)=f(b) forced a horizontal tangent somewhere inside the interval. What happens if the endpoints are at different heights?
The fraction b−af(b)−f(a) is the average rate of change of f on [a,b], which is the slope of the secant line connecting (a,f(a)) and (b,f(b)). The Mean Value Theorem says that as long as f is continuous on [a,b] and differentiable on (a,b), there is at least one point c in (a,b) where the instantaneous rate of change f′(c) equals that average rate of change.
f continuous on [a,b]f′(x) exists for every x in (a,b)
f′(c)=b−af(b)−f(a)
for at least one c in (a,b)
On a graph, the tangent line at c is parallel to the secant line. The conditions are the same as in Rolle's theorem, but the two endpoints no longer need the same height. Let's find this c for f(x)=x3 on [0,2], and see why it has to exist.
Find where the tangent slope to y=x3 equals the secant slope on [0,2]
1
2
8
x
y
0
Text solution
The secant slopeThe secant line joins (0,0) and (2,8).
2−0f(2)−f(0)=28−0=4
Match the tangent slopeSet the derivative 3c2 equal to 4 and keep the solution inside (0,2).
f′(c)=3c2=4
→c=±32→c=32
On a driveGoing from mile 20 at hour 1 to mile 140 at hour 3 is an average of 60 mph, so the speedometer read exactly 60 mph at some moment. With a speed limit of 55 mph, you were speeding at some point.
3−1140−20=60 mph
Notice that c=32 is about 1.15, not the midpoint x=1. It's also the exact value of c we found for 4x−x3, and that's no coincidence. The secant line is y=4x, so if we define d(x) as the height of the secant line minus the height of the curve, we get d(x)=4x−x3.
1
2
2
4
6
8
x
y
0
c=32
d(x)
y=4x
y=x3
d(x)=4x−x3
d(0)=0,d(2)=0
d′(x)=4−3x2=4−f′(x)
d′(c)=0
⟺f′(c)=4
The gap is 0 at both ends, so Rolle's theorem gives a c in (0,2) where d′(c)=0, which means f′(c)=4. The same gap works for any function that meets the two conditions, and that's why the Mean Value Theorem is true. When f(a)=f(b), the secant slope is 0, and we get back Rolle's theorem.
The theorem guarantees at least one value of c, but there can be more. A curve that intersects the secant line several times can have several parallel tangents, and a straight line matches the secant's slope at every single point. Where does the instantaneous rate equal the average rate here?
3
Why both conditions are needed
Rolle's theorem and the Mean Value Theorem both need f to be continuous on [a,b], and f′(x) to exist for every x in (a,b). The Mean Value Theorem needs the same two because we proved it with Rolle's theorem on the gap d(x), and d(x) is continuous and differentiable wherever f is. If either condition fails, the conclusion can fail too.
A smooth curve with both ends at the same height can't avoid a flat tangent. Let's see two ways a curve can avoid one, and then check the conditions on two functions whose derivative doesn't exist at a point.
Why Rolle's theorem and the Mean Value Theorem need continuity on [a,b] and differentiability on (a,b)
−1
1
x
y
(−1,1)
(1,1)
Text solution
A corner: f(x)=∣x∣Its smallest value is at x=0, but the corner means f′(0) doesn't exist, and everywhere else the slope is −1 or 1. So the derivative has to exist at every point inside (a,b).
A jump: j(x)Its slope is 1 everywhere inside (−1,1). As x approaches 1,j(x) approaches 3 without reaching it, so there's no largest value. The jump at x=1 breaks continuity at an endpoint, so f has to be continuous on the whole closed interval [a,b].
x32 on [−1,8]: a cusp insidef′(0) doesn't exist, and 0 is inside (−1,8), so the theorem doesn't apply. Solving anyway gives only the endpoint.
8−(−1)f(8)−f(−1)=31
33c2=31→c=8
x on [0,4]: only at an endpointf′(0) doesn't exist either, but 0 is an endpoint, and the theorem only needs f′(x) strictly inside (0,4). So the theorem applies.
4−0f(4)−f(0)=21
2c1=21→c=1
Here are the corner and the jump side by side, the two ways a curve with equal ends can avoid a slope of 0.
−1
1
1
3
x
y=∣x∣
f(x)=∣x∣
f(−1)=f(1)=1
f′(x)=−1(x<0)
f′(x)=1(x>0)
f′(0) does not exist
−1
1
1
3
x
y=j(x)
j(x)={x+2,1,−1≤x<1x=1
j(−1)=j(1)=1
j′(x)=1for −1<x<1
j not continuous at 1
4
Counting roots without finding them
We don't have a simple formula to write down the solutions to x3+x−1=0. But we can still prove it has exactly one real solution. "Exactly one" means there has to be at least one solution, and there can't be more than one.
Show that x3+x−1=0 has exactly one real solution
x3+x−1=0
exactly one real solution?
Text solution
At least oneLet f(x)=x3+x−1. It's negative at 0 and positive at 1, and a polynomial is continuous, so the Intermediate Value Theorem gives a root between 0 and 1.
Two roots would need a flat tangentIf f(a)=f(b)=0 with a<b, Rolle's theorem gives f′(c)=0 for some c between them.
The slope is never 0Because 3x2≥0, the derivative 3x2+1 is always at least 1. So there's no such c, and no second root.
Written out, the proof has the same two parts, each with the theorem it uses.
−1
1
x
y
f(0)=−1
f(1)=1
y=x3+x−1
f(x)=x3+x−1
f(0)=−1<0
f(1)=1>0
→f(x)=0 for some x in (0,1)
(IVT)
f(a)=f(b)=0,a<b
→f′(c)=0 for some c in (a,b)
(Rolle)
f′(x)=3x2+1≥1→f′(x)=0 for every x
f′(x)=3x2+1≥1
→f′(x)=0 for every x
5
What a table guarantees
Suppose we know a particle's position s(t) only at a few times, and we're told s is differentiable. We have no formula and no graph. What values can we still be certain about?
t(seconds)
0
3
5
8
s(t)(meters)
4
10
4
1
v(t)=2 for some t in (0,8)?
s(t)=7 for some t in (0,8)?
Remember, velocity is the derivative of position, v(t)=s′(t). Since s is differentiable, it's also continuous everywhere, so the Mean Value Theorem and the Intermediate Value Theorem apply on any interval we choose.
Decide whether v(t)=2 and whether s(t)=7 for some t in (0,8)
t
s(t)
0
3
5
8
4
10
4
1
0
1
2
3
4
5
6
7
8
5
10
t
s
Text solution
Velocity is a derivative, so use the MVTWe look for two rows whose average velocity is 2. The whole table gives −83, but t=0 to t=3 gives exactly 2.
Position is the function itself, so use the IVT7 is between s(0)=4 and s(3)=10, and s is continuous, so the particle passes through position 7 between those times.
Two equal rows, so use Rolle's theoremThe position is 4 at both t=0 and t=5, so the average velocity between them is 0, and the particle stops at some time in between.
Each answer is written with the theorem that guarantees it.
8−0s(8)−s(0)=81−4=−83
3−0s(3)−s(0)=310−4=2
→v(t)=2 for some t in (0,3)
(MVT)
s(0)=4<7<10=s(3)
→s(t)=7 for some t in (0,3)
(IVT)
s(0)=s(5)=4
→v(t)=0 for some t in (0,5)
(Rolle)
6
Bounding a function with its derivative
Suppose we know that f(0)=1, and that the derivative f′(x) is never greater than 2. How large could f(4) possibly be?
If f(0)=1 and f′(x)≤2 for every x, how large can f(4) be?
4
5
10
x
y
(0,1)
f(4)=?
Text solution
Could f(4) be 11?The average rate of change would be 25, so the Mean Value Theorem would need f′(c)=25 somewhere in (0,4). But f′(x) is never more than 2.
4−011−1=25
Any value of f(4)Since f′ exists everywhere, f is continuous, so the Mean Value Theorem applies on [0,4]. Multiply by 4 and use f′(c)≤2.
4−0f(4)−f(0)=f′(c)
→f(4)=1+4f′(c)
f′(c)≤2→f(4)≤1+4⋅2=9
A lower bound works the same wayIf the derivative is at least −1, the smallest possible value is −3.
f′(x)≥−1→f(4)≥1+4(−1)=−3
The line y=1+2x has slope 2 everywhere and reaches y=9 at x=4, so the largest possible value of f(4) is 9. No allowed f can get above that line.
1
2
3
4
1
5
9
x
y
(4,9)
y=1+2x
For any b>a, the Mean Value Theorem gives the same equation, so a bound on the derivative becomes a bound on f(b).
f(b)=f(a)+f′(c)(b−a)
Summary
The Mean Value Theorem and Rolle's theorem
x
a
b
c
The Mean Value Theorem
f′(c)=b−af(b)−f(a)
Rolle's theorem
f(a)=f(b)→f′(c)=0
If f is continuous on [a,b] and differentiable on (a,b), the tangent line at some point c inside the interval has the same slope as the secant line between the endpoints.
When f(a)=f(b), the secant line is horizontal, so f′(c)=0. That special case is Rolle's theorem.
Finding c
4x−x3 on [0,2]
4−3c2=0→c=±32
c=32,in (0,2)
To find c, set f′(c) equal to the secant slope b−af(b)−f(a), solve for c, and keep only the solutions inside the open interval (a,b). There can be more than one valid c.
For 4x−x3 on [0,2], the equation 4−3c2=0 gives two solutions, but only c=32 lies inside (0,2).
Checking the conditions
Before using either theorem, check both conditions. The function must be continuous on the closed interval [a,b], ends included, and its derivative must exist at every point strictly inside (a,b).
If f has a jump, a hole, or a vertical asymptote anywhere on [a,b], or a sharp corner, cusp, or vertical tangent inside (a,b), neither theorem guarantees a value c. But if f′ is undefined only at an endpoint, like x at x=0 on [0,4], that's fine because the theorems only need the derivative to exist inside the interval.
Showing an equation has exactly one solution
To show an equation has exactly one real root, prove there is at least one, and then prove there cannot be two.
Use the Intermediate Value Theorem to show at least one root exists. If f is continuous and changes sign between two numbers, say f(0)=−1 and f(1)=1, then f(x)=0 somewhere between them.
Then use Rolle's theorem to rule out a second root. If there were two roots, f′(c) would have to be 0 somewhere between them. If f′(x)=0 everywhere, having two roots is impossible.
Reading values from a table
To guarantee a value of the derivative, use the Mean Value Theorem. Find two rows in the table whose average rate of change b−af(b)−f(a) equals that value. If two rows have the exact same function value, Rolle's theorem guarantees that the derivative is 0 at some point between them.
To guarantee a value of the function itself, use the Intermediate Value Theorem. If f is continuous, it must take every value between two given heights. If a function is differentiable, it is also continuous.
Bounding a function with its derivative
f(b)=f(a)+f′(c)(b−a)
When we know a starting value f(a) and bounds on the derivative, the Mean Value Theorem gives bounds on f(b).
For b>a, the largest allowed derivative gives the largest possible value for f(b), and the smallest allowed derivative gives the smallest.
7
Practice
Enter exact values, separated by commas if there is more than one.
1
Find every value of c in (0,π) that satisfies the conclusion of Rolle's theorem for f(x)=sin2x on [0,π].
The derivative is f′(x)=2cos2x. If x is in (0,π), then 2x is in (0,2π). Where is cosine 0 on (0,2π)?
If you got 4π only Check the interval for 2x. When x is between 0 and π,2x goes from 0 to 2π, so cos2x=0 at two angles in that interval.
Both endpoints give 0, and sin2x is continuous and differentiable everywhere, so Rolle's theorem applies.
f(0)=sin0=0,f(π)=sin2π=0
Set the derivative equal to 0. Because x is in (0,π), the angle 2x is in (0,2π).
f′(x)=2cos2x=0→cos2x=0,0<x<π→0<2x<2π
On (0,2π), cosine is 0 at 2π and 23π, and dividing each by 2 gives two values inside (0,π).
2c=2π,23π→c=4π,43π
2
At what value of x between 1 and 4 is the tangent line to y=x1 parallel to the secant line through (1,1) and (4,41)?
The secant line has rise 41−1=−43. What is the run from x=1 to x=4?
If you got 32 Check the secant slope. The rise is −43, but remember to divide by the run 4−1=3.If you got 25 For a quadratic function, the midpoint of [1,4] gives the parallel tangent, but that doesn't work for x1.
Parallel lines have the same slope, so we start with the slope of the secant line.
4−141−1=3−43=−41
The derivative gives the slope of the tangent line, so we set y′ equal to −41.
y′=−x21:−c21=−41→c2=4→c=±2
We only keep values strictly between 1 and 4, so we discard −2.
−2 not in (1,4)→x=2
3
Let f(x)=tanx. Then f(0)=f(π)=0, but f′(x)=sec2x is never 0. Explain why this does not contradict Rolle's theorem.
Is there a value of x in (0,π) where tanx is undefined?
If you got tanx is not differentiable at 0 and π The derivative sec2x exists at both 0 and π. Check whether tanx is defined everywhere inside the interval.
The denominator cosx is 0 at 2π, so tanx has a vertical asymptote inside [0,π].
tanx=cosxsinx,cos2π=0→tanx undefined at 2π,in [0,π]
Because tanx is not continuous on [0,π], the conditions of Rolle's theorem are not met, so it makes no promise that f′(c)=0.
tanx not continuous on [0,π]→Rolle’s theorem does not apply
4
Let f(x)=x2 for x≤1 and f(x)=2x−1 for x>1. Find every value of c in (0,2) that satisfies the conclusion of the Mean Value Theorem for f on [0,2].
Check the boundary at x=1. Do the two pieces have the same value and the same derivative there?
If you got the theorem does not apply Check the boundary at x=1. Both pieces give a value of 1 and a slope of 2, so the function is continuous and differentiable there.
If you got 1 Check the formula used for f(2). Because 2>1,f(2) comes from 2x−1, not x2.
Check the boundaryAt x=1, both pieces give a value of 1 and a derivative of 2, so f is continuous on [0,2] and differentiable on (0,2).
x=1:12=1,2⋅1−1=1;(x2)′=2x=2,(2x−1)′=2
The Mean Value Theorem applies, so we find the average rate of change on [0,2].
f(0)=0,f(2)=3→2−03−0=23
On (1,2), the derivative is always 2, so f′(c)=23 can only happen on (0,1).
0<c<1:2c=23→c=43;1<c<2:f′(c)=2=23
5
A particle's velocity v(t) is differentiable, with v(0)=2,v(4)=10,v(6)=10 and v(10)=4. (a) Must the particle's acceleration be 0 at some time between t=0 and t=10? Explain. (b) Must its velocity be 7 at some time between t=0 and t=10? Explain.
Acceleration is the derivative of velocity. Which two times have the same velocity?
Which two velocities in the table does 7 lie between?
If you got (a) no Check the times t=4 and t=6. The average acceleration across the whole interval is not 0, but Rolle's theorem applies between these two times.
If you got (b) yes, by the Mean Value Theorem7 is a value of the velocity v(t) itself, not of its derivative, so this is the Intermediate Value Theorem.
AccelerationAcceleration is a(t)=v′(t). Because v(4)=v(6)=10 and v is differentiable, Rolle's theorem guarantees v′(t)=0 for some t in (4,6), so the answer is yes.
v(4)=v(6)=10→a(t)=v′(t)=0 for some t in (4,6)(Rolle)
VelocityBecause v is differentiable, it is continuous. Since 7 lies between v(0)=2 and v(4)=10, the Intermediate Value Theorem guarantees v(t)=7 for some t in (0,4), so the answer is yes.
v(0)=2<7<10=v(4)→v(t)=7 for some t in (0,4)(IVT)
6
Suppose f(1)=3 and −1≤f′(x)≤4 for all x. What are the smallest and largest possible values of f(5)?
By the Mean Value Theorem, f(5)=f(1)+f′(c)(5−1). What are the smallest and largest values that f′(c) can take?
If you got −2 and 23 Check the width of the interval. From x=1 to x=5, the width is 5−1=4, not 5.If you got 2 and 7 Remember to multiply the slope f′(c) by the interval width 5−1=4.
By the Mean Value Theorem on [1,5], the change in f equals the width times the derivative at some point c.
5−1f(5)−f(1)=f′(c) for some c in (1,5)→f(5)=3+4f′(c)
Use the smallest derivative, −1, to find the smallest possible value, and the largest derivative, 4, to find the largest.
−1≤f′(c)≤4→3+4(−1)≤f(5)≤3+4⋅4→−1≤f(5)≤19
Lines through (1,3) with slopes −1 and 4 reach −1 and 19 at x=5, so the smallest possible value is −1 and the largest is 19.
y=3−(x−1):y=−1 at x=5;y=3+4(x−1):y=19 at x=5
7
Show that 2x+cosx=0 has exactly one real solution.
To show there is at least one solution, find an x where 2x+cosx is negative and an x where it is positive.
If there were two solutions, what would Rolle's theorem say about the derivative between them?
If you got only that there is at least one solution That shows there is a solution, but not that it is the only one. Check the derivative to show there cannot be a second solution.
At least one solutionAt x=0,f(0)=1>0, and at x=−1,f(−1)=−2+cos1<0 because cos1≤1. By the Intermediate Value Theorem, f(x)=0 for some x in (−1,0).
f(x)=2x+cosx:f(−1)=−2+cos1<0,f(0)=1>0→f(x)=0 for some x in (−1,0)(IVT)
At most one solutionBecause sinx≤1, the derivative 2−sinx is always at least 1, so it is never 0.
f′(x)=2−sinx≥2−1=1→f′(x)=0 for every x
If there were two solutions a and b, Rolle's theorem would require f′(c)=0 between them. Since f′ is never 0, there cannot be a second solution.
f(a)=f(b)=0,a<b→f′(c)=0 for some c in (a,b)(Rolle)→no second solution
8
Suppose f is twice differentiable, with f(0)=0,f(1)=1 and f(2)=0. Show that f′′(c)<0 for some c between 0 and 2. (Challenge)
Apply the Mean Value Theorem to f on [0,1], and then on [1,2]. What values do you get for f′?
The second derivative is the derivative of f′. What does the Mean Value Theorem give when applied to f′ between those two points?
If you got f′(c)=0 for some c in (0,2), and stopped That gives a point where the first derivative is 0. To reach the second derivative, apply the Mean Value Theorem to f on two separate intervals first.
The Mean Value Theorem on [0,1] gives a point p where the derivative is 1.
[0,1]:f′(p)=1−01−0=1,0<p<1
The Mean Value Theorem on [1,2] gives a point q where the derivative is −1.
[1,2]:f′(q)=2−10−1=−1,1<q<2
Since p<1<q, we have p<q. Because f is twice differentiable, we can apply the Mean Value Theorem to f′ on [p,q].