Absolute and local extrema and the difference between them, the Extreme Value Theorem and its two hypotheses, Fermat's theorem that an interior extremum forces a zero derivative, critical points including those where the derivative fails to exist, and the closed-interval method for finding absolute extrema.
Subject: Calculus I · 67 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 4.3
Where a function is largest and smallest, and how to find out without a graph
Objectives
Chapter 3 built the derivative. This section puts it to its first big use: finding the largest and smallest values of a function, the question behind every optimization problem you will meet in Section 4.7.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 317 — learning objectives 4.3.1 to 4.3.4
You have spent a whole chapter learning to compute derivatives. This is the section where they start paying for themselves. Almost every applied question in calculus, from the cheapest can to the fastest route, comes down to finding where some function is largest or smallest, and the derivative is the tool that finds those places without a graph.
The plan has three stages. First you pin down exactly what largest and smallest mean, because there are two different notions, absolute and local, and mixing them up is the most common source of wrong answers. Then two theorems do the heavy lifting: one guarantees that the extremes exist, the other tells you where they can possibly be.
The payoff is a short, reliable procedure, the closed-interval method, that you will use again and again in Section 4.7 on optimization. By the end you should be able to run it on any continuous function over a closed interval and trust the answer.
Warm-up
Discussion prompt
Two quick recalls from Chapter 3. (1) What is the slope of the tangent line to y = 3 − (x − 1)² at its vertex? (2) Does the absolute value function have a derivative at x = 0? Say why in one sentence.
Answer both parts before revealing. These two facts from Chapter 3 are the raw material for the whole lesson, so it is worth checking they are solid.
The first is a parabola with its vertex at x equal to 1. The derivative there is zero: the tangent line at the top of a hump is horizontal. The second is the absolute value function at the origin. From the right every difference quotient is 1, from the left every one is minus 1, and a limit cannot equal two different numbers, so there is no derivative at 0.
Keep both pictures in mind. A smooth peak with a horizontal tangent, and a sharp corner with no tangent at all, are exactly the two ways a function can turn around. They will become the two kinds of critical number.
Section
Part 1
Intuition
Maximum profit, minimum material, greatest height: every one of these is the largest or smallest value of some function. A graph lets you eyeball the answer; this section gives you a way to find it exactly, from a formula, even when the interesting features are too small to see.
\[ \text{find } \max_{x \in I} f(x) \quad\text{and}\quad \min_{x \in I} f(x) \]
The answer always has two parts: the value, and where it happens. Keep them apart from the start.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 317 — introduction
Every optimization problem, however it is dressed up, asks for the largest or smallest value of a function over some set of allowed inputs. A company choosing a price, an engineer choosing a shape, a rocket reaching its peak: each is a function and an interval.
A graph is a good way to guess the answer, but it has limits. The interesting feature may be too small to see at the scale you drew, or you may not have a graph at all, only a formula. The aim of this section is a method that works from the formula alone.
Notice the two parts of the answer on the slide. The maximum is a number, the largest output. The place where it happens is a different number, an input. Most lost marks in this topic come from blurring the two, so start keeping them separate now.
Worked example
Find the largest and smallest values, if any, of the function below over the whole real line.
\[ f(x) = x^2 + 1, \quad x \in (-\infty, \infty) \]
Bound it below
Why: A square is never negative, so adding one keeps every value at least one.
\[ x^2 \ge 0 \;\Longrightarrow\; x^2 + 1 \ge 1 \]
Show the bound is reached
Why: The lower bound is only a minimum if some input attains it.
\[ f(0) = 0^2 + 1 = 1 \]
Look for a ceiling
Why: As x runs off in either direction the square grows without limit.
\[ \lim_{x\to\pm\infty}(x^2 + 1) = \infty \]
Figure (svg): The parabola y equals x squared plus 1 for x from minus 3 to 3, with its lowest point marked at (0, 1) and arrows at both ends pointing up past the top of the window.
Check with a few values
Why: Every sampled value is at least one, and they keep growing.
\[ f(\pm 1) = 2, \quad f(\pm 3) = 10, \quad f(\pm 10) = 101 \;\Longrightarrow\; \text{min } 1 \text{ at } x = 0, \text{ no max} \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 317 — Figure 4.12
This first example is simple on purpose, so you can see the logic of a minimum without any calculus getting in the way. A minimum needs two things: a lower bound that no value goes below, and an input where the bound is actually reached.
The lower bound comes from the fact that a square is never negative, so x squared plus one is always at least one. The bound is reached at x equal to 0. Both parts together make one the absolute minimum. Missing either part would not be enough: a function can have a lower bound it never reaches.
For a maximum you would need an upper bound, and there is none. The limits at plus and minus infinity are infinite, as the arrows in the figure show, so for any proposed ceiling some value goes above it. The function has an absolute minimum and no absolute maximum.
Notation
Annotate
On: \( f(c) \ge f(x) \quad \text{for all } x \in I \)
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 318 — Definition: absolute extrema
Step through the notes one at a time. The first is the most important: the absolute maximum is f of c, a value, and c is merely a place where that value occurs. When a question asks for the maximum, the answer is a height on the graph.
The inequality is not strict, which means ties are allowed. The cosine function reaches its maximum value of 1 at infinitely many inputs, and each of them is a place where the absolute maximum occurs. There is still only one maximum value.
The word absolute here means compared with every input in the interval, not absolute value; an absolute maximum can perfectly well be negative. And the interval I matters. The same formula over a different interval can have completely different extrema, so always read the interval before you start.
Concept
The book's second example flips the picture. Now the square sits in the denominator.
\[ f(x) = \frac{1}{x^2 + 1} \]
\[ x^2 + 1 \ge 1 \;\Longrightarrow\; \frac{1}{x^2+1} \le 1 = f(0) \]
So the absolute maximum is 1, at x equal to 0. There is no absolute minimum: the values are always positive and slide toward zero far out, but never reach it.
\[ f(x) > 0 \text{ for all } x, \quad \lim_{x\to\pm\infty} f(x) = 0 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 318 — the example after the definition
This is the first example turned upside down. Because x squared plus one is at least one, its reciprocal is at most one, and it equals one exactly at x equal to 0. So the absolute maximum is 1, and it occurs at 0.
The minimum is where it gets interesting. The values are all positive, and far out they get as close to zero as you like, so zero is a lower bound, and nothing larger is. But zero is never reached, because a fraction with numerator one is never zero.
So there is no smallest value. Whatever small positive value you name, the function goes lower further out. A bound that is approached but never reached is exactly the way extrema fail to exist, and you will see it again and again in this lesson.
Picture it
Figure (svg): Three panels. Left: y equals x cubed, running from far below to far above, with no highest or lowest point. Middle: y equals one over x squared plus 1, a bell shape with its peak marked at (0, 1). Right: y equals cos x, with peaks marked at 0 and plus or minus 2 pi and troughs marked at plus or minus pi.
Figure 4.13(a) to (c). On an unbounded domain, a function may have no extremum, exactly one, or both, and an extremum may be reached at infinitely many places.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 319 — Figure 4.13(a)-(c)
These are the three panels of Figure 4.13 that live on the whole real line. Look at each and say what extrema it has before reading the labels underneath.
The cube runs off to minus infinity on the left and plus infinity on the right, so it has neither extremum. The bell curve from the previous slide has a maximum and no minimum. The cosine has both, a maximum of 1 and a minimum of minus 1, and it reaches each one infinitely many times.
The lesson of the three panels is that on an unbounded domain there is no guarantee of anything. To get a guarantee, you need to restrict the domain, and the next part of the lesson says exactly what kind of restriction works.
Trap
Asked for the absolute maximum of one over x squared plus one, a student writes:
\[ \text{absolute max} = 0 \]
Wrong. Zero is the input, not the value.
The absolute maximum is the value f(c); c is the place it occurs. Say both, in that order.
\[ \text{max} = f(0) = 1 \]
\[ \text{occurs at } x = 0 \]
This mistake is so common that it is worth practising the correct sentence out loud. The maximum of the bell curve is 1. It occurs at x equal to 0. Those are two different numbers answering two different questions.
When an exam asks for the absolute maximum, it wants the value, and it usually also wants the location. A good habit is to always write both, in a full sentence: the absolute maximum is this value, and it occurs at this input. Then neither half can be mistaken for the other.
Section
Part 2
Concept
Extreme Value Theorem (Theorem 4.1) — If f is continuous over the closed, bounded interval [a, b], then there is a point in [a, b] at which f has an absolute maximum, and a point in [a, b] at which f has an absolute minimum.
\[ f \text{ continuous on } [a,b] \;\Longrightarrow\; \exists\, c, d \in [a,b]: \]
\[ f(d) \le f(x) \le f(c) \quad \text{for all } x \in [a,b] \]
Two hypotheses, and both are about more than the formula: continuity everywhere on the interval, and an interval that is closed (endpoints included) and bounded (finite length).
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 319 — Theorem 4.1
This theorem is the guarantee that the previous examples lacked. It says that if you keep the function continuous and keep the interval closed and bounded, then both extrema must exist. There is no way to dodge them.
Read the hypotheses as a pair of demands, and notice what each one rules out. Continuity rules out jumps and blow-ups, where the function could escape toward a value it never takes. A closed interval includes its endpoints, so the function cannot creep toward an edge value that is missing. Bounded means finite length, so the function cannot run away to infinity along the axis.
The proof needs the careful theory of real numbers and belongs to a course in real analysis, so the book states it without proof. What you need is to check the hypotheses every time you use it.
Picture it
Figure (svg): Three panels on bounded intervals. Left: 2 minus x squared on 0 to 2 then x minus 3 on 2 to 4, with a filled dot at (0, 2), an open circle at (2, minus 2) and a filled dot at (2, minus 1). Middle: (x minus 2) squared on 1 to 4 with dots at (1, 1), (2, 0) and (4, 4). Right: x over 2 minus x on 0 to 2, rising toward a dashed vertical asymptote at x equals 2.
Figure 4.13(d) to (f). All three live on bounded intervals, but only (e) meets both hypotheses, and only (e) is guaranteed both extrema. Graph (d) is on a closed interval but jumps; graph (f) is continuous but its interval leaves out the endpoint 2.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, pp. 319-320 — Figure 4.13(d)-(f)
These are the other three panels of Figure 4.13. All three are on bounded intervals, but only the middle one satisfies both hypotheses of the theorem.
In panel d the interval is closed, from 0 to 4, but the function jumps at x equal to 2. Just before the jump the values dive toward minus 2, but the open circle shows that minus 2 itself is never taken, and at 2 the function jumps up to minus 1. So there is a maximum of 2 at the left end, but no minimum.
In panel f the function is continuous, but the interval stops short of 2, where the function blows up. It has a minimum of 0 at the left end and no maximum. Panel e is continuous on a closed interval, and sure enough it has both: 4 at the right end and 0 at x equal to 2. Each failure lines up with a missing hypothesis.
Intuition
Like the Intermediate Value Theorem of Section 2.4, this is a promise that something exists. It does not say where the maximum is, or how big it is.
\[ \text{EVT: } \exists\, c \quad\text{(but which } c\text{? what } f(c)\text{?)} \]
Its job in this section is to make the search worthwhile. Once you know an extremum must exist, the rest of the section narrows down where it can be, until only a handful of candidates are left to test.
It is easy to overestimate what this theorem gives you. It guarantees that a maximum exists somewhere in the interval. It does not tell you where, and it gives you no way to compute it.
That is the same kind of statement as the Intermediate Value Theorem from Section 2.4, which promised a root without locating it. Existence theorems are valuable because they tell you a search will succeed, so you do not waste time looking for something that is not there.
The rest of this section is the location half of the story. Fermat's theorem and Theorem 4.3 shrink the set of places the maximum could be from a whole interval to a short list of candidates.
Discrimination
Sort into buckets
Sort each function and interval by whether the theorem guarantees both an absolute maximum and an absolute minimum.
For each item, check the two hypotheses separately: is the function continuous at every point of the interval, and is the interval closed and bounded? Only if both answers are yes does it go in the guaranteed bucket.
The three failures each break something different, which is the point of the exercise. The open interval from 0 to 2 is missing its endpoints. The reciprocal function has a vertical asymptote at 0, right inside its closed interval. And the interval from 0 to infinity is closed, in the sense that it includes 0, but it is not bounded.
One subtlety: not guaranteed does not mean has no extrema. A function can fail the hypotheses and still happen to have both extrema. The theorem simply stops promising anything.
Counterexample
Discussion prompt
Give one function for each: (1) continuous on a bounded interval that is not closed, with no absolute maximum; (2) defined on a closed, bounded interval but not continuous, with no absolute maximum.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 326 — Exercises 92 and 98
Try to build your own examples before looking. Constructing a counterexample is the fastest way to understand why a hypothesis is there.
The first example keeps continuity and drops closedness. The identity function on the open interval from 0 to 2 takes values as close to 2 as you like, but never 2 itself, so there is no largest value. Its smallest value fails in the same way at the other end.
The second keeps the closed interval and drops continuity. The function climbs like x toward height 1, and then at the right endpoint it drops to 0. Every value is less than 1, and values come arbitrarily close to 1, so again nothing is largest. Put the two together and you see that each hypothesis is doing real work.
Section
Part 3
Concept
Figure (svg): A curve that rises to a rounded peak at x equals 0, dips to a valley at x equals 1, climbs to a sharp corner peak at x equals 2, then falls in a straight line. The three turning points are marked.
local maximum — f has a local maximum at c if there is an open interval I containing c, inside the domain of f, with f(c) at least f(x) for every x in I. A local minimum reverses the inequality.
\[ f(c) \ge f(x) \text{ for all } x \in I, \quad I \text{ open}, \; c \in I \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 320 — Figure 4.14 and Definition: local extrema
Look at the figure. It has two peaks with a valley between. The taller peak on the right is the highest point anywhere on the graph. The left peak is not, but it is still special: near it, nothing is higher. That is a local maximum.
The definition makes near precise. There has to be some open interval around c, however small, on which f of c is the largest value. The interval can be tiny; all that matters is that it exists and sits inside the domain.
The valley at x equal to 1 is a local minimum by the same reasoning. Notice that the function has no absolute minimum at all, since it runs off downward on the far left, yet it still has a perfectly good local minimum. Local and absolute answer different questions.
Concept
The definition asks for an open interval around c inside the domain. At an endpoint of a closed interval that is impossible: there is no room on one side.
\[ c = a \text{ in } [a, b]: \quad (a - \delta, a + \delta) \not\subset [a, b] \]
So an absolute extremum at an endpoint is called an endpoint extremum, not a local one. The reverse direction does hold: an absolute extremum at an interior point is automatically local, because it beats everything, including its neighbours.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 320 — the remark after the definition
This is a convention, but it is a useful one. At an endpoint of a closed interval there is nothing on the far side, so no open interval around the endpoint fits inside the domain. The definition of a local extremum simply cannot be satisfied there.
So when an absolute extremum happens at an endpoint, it gets its own name, an endpoint extremum. You will see why the distinction matters in Part 6: the tools for finding local extrema say nothing about endpoints, so endpoints have to be checked separately.
The other direction is worth holding onto. If an absolute maximum occurs at an interior point, it beats every value on the interval, so in particular it beats its neighbours. An interior absolute extremum is always a local one.
Concept
Look again at the three turning points of the mountain figure. At the rounded peak and in the valley the tangent is horizontal. At the sharp peak there is no tangent at all.
\[ f'(0) = 0, \quad f'(1) = 0, \quad f'(2) \text{ does not exist} \]
Those are the only two things that can happen at a local extremum. That gives a name to the points worth checking.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 320 — discussion of Figure 4.14
This slide is the bridge from pictures to formulas. Look at the three turning points of the mountain figure and ask what the tangent line does at each.
At the rounded peak and in the valley, the curve levels off, so the tangent is horizontal and the derivative is zero. At the sharp peak on the right, the curve changes direction abruptly. The slope from the left and the slope from the right disagree, just as with the absolute value function in the warm-up, so there is no derivative.
Those are the only two possibilities, and the next theorem proves it. So instead of scanning the whole graph for peaks, you only need to look where the derivative is zero or does not exist. Those points get a name on the next slide.
Notation
Annotate
On: \( c \text{ interior to } \operatorname{dom} f, \quad f'(c) = 0 \;\text{ or }\; f'(c) \text{ undefined} \)
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 321 — Definition: critical number
Step through the notes. The first condition is easy to skip and costs marks: a critical number must be an interior point of the domain. The function has to exist there, with room on both sides.
That rules out points where the function is undefined. The reciprocal function has no derivative at 0, but it has no value at 0 either, so 0 is not a critical number. You will test this idea on Exercise 117 later.
The two conditions are the two shapes from the previous slide: a horizontal tangent, or a point where the function exists but has no derivative, such as a corner, a cusp or a vertical tangent. The book uses critical number and critical point interchangeably, so do not worry about which word a question uses.
Picture it
Figure (svg): Five small graphs, each with a critical number c. (a) a smooth hump with a horizontal tangent: local maximum. (b) a sharp upward cusp: local maximum, derivative undefined. (c) a V-shaped corner: local minimum, derivative undefined. (d) a cubic flattening out with a horizontal tangent but still rising: no extremum. (e) a cube-root curve with a vertical tangent: no extremum.
Figure 4.15. Panels (a) to (c) are local extrema; (d) and (e) are critical numbers where nothing turns. Being critical makes a point a candidate, nothing more.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 322 — Figure 4.15
This is Figure 4.15, and every panel shows a critical number c. Go through them left to right and name what happens at each.
The first three turn around. Panel a is a smooth maximum with a horizontal tangent. Panel b is a cusp: the function rises steeply to a point and falls away, and the derivative does not exist. Panel c is a corner, like the absolute value, and it is a minimum.
The last two are the warning. Panel d flattens out and keeps rising: the derivative is zero, but the function does not turn. Panel e has a vertical tangent: the derivative does not exist, and again the function keeps rising. Both are critical numbers and neither is an extremum. So critical numbers are where extrema can be, not where they must be.
Matching
Match the pairs
Why: Two ways to be critical (flat or no derivative), crossed with two outcomes (turns or does not). 1 − x² and |x| turn at 0; x³ and ∛x keep increasing straight through it.
Sketch each function quickly if you need to, then match. Each graph has a critical number at 0, and the right-hand column describes both why it is critical and whether it turns.
The parabola one minus x squared has a smooth peak, derivative zero. The absolute value has a corner, no derivative, and a minimum. The cube flattens with a zero derivative but keeps increasing. The cube root has a vertical tangent at the origin, so no derivative, and it too keeps increasing.
Notice that the matches cover all four combinations of flat or broken with turns or does not turn. That is the full picture of what a critical number can be.
Section
Part 4
Concept
Fermat's theorem (Theorem 4.2) — If f has a local extremum at c and f is differentiable at c, then the derivative at c is zero.
\[ \text{local extremum at } c \;+\; f'(c) \text{ exists} \;\Longrightarrow\; f'(c) = 0 \]
Put the other case beside it, where the derivative does not exist, and every local extremum is caught: local extrema can only occur at critical numbers.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 321 — Theorem 4.2
Here is the theorem that makes critical numbers worth finding. If a function has a local extremum at c and has a derivative there, the derivative must be zero. The only escape is for the derivative not to exist.
Put those two cases together and you have the key consequence: every local extremum happens at a critical number. That turns an infinite search into a finite one, because for most functions you meet the derivative is zero or undefined at only a handful of points.
Read the direction of the arrow carefully. The theorem goes from extremum to zero derivative. It does not go back, and you already met the counterexample, the cube at the origin. The proof on the next slides shows exactly why the arrow only points one way.
Picture it
Figure (svg): A hump-shaped curve with its peak at x equals c. Two secant lines join the peak to points on its left; both slope upward. Two secant lines join the peak to points on its right; both slope downward. A dashed horizontal tangent sits on the peak.
The whole proof is in this picture. Every secant from a point on the left climbs to the peak, so its slope is at least zero; every secant to a point on the right falls, so its slope is at most zero. The derivative is the limit of both.
Before the algebra, look at the picture, because the proof is really just this picture. The green secants join points on the left to the peak. Since the peak is the highest point around, every one of them slopes upward, or at worst is flat.
The red secants join the peak to points on the right. Those points are lower, so every one of them slopes downward or is flat.
The derivative at the peak is the limit of these secant slopes from either side, and because the derivative exists, the two sides must give the same number. A number approached by slopes that are all at least zero is at least zero. A number approached by slopes that are all at most zero is at most zero. Only zero is both.
Worked example
Suppose f has a local maximum at c and is differentiable there. Show the derivative is zero.
Use the local maximum
Why: Near c no value beats f(c).
\[ f(x) - f(c) \le 0 \quad \text{for } x \text{ near } c \]
Divide by a positive step
Why: For x to the right of c the denominator is positive, so the sign stays.
\[ x > c: \quad \frac{f(x) - f(c)}{x - c} \le 0 \]
Take the right-hand limit
Why: Differentiable means the one-sided limits exist and both equal the derivative (equation 4.4).
\[ f'(c) = \lim_{x\to c^+}\frac{f(x)-f(c)}{x-c} \le 0 \]
Divide by a negative step
Why: For x to the left the denominator is negative, which flips the sign.
\[ x < c: \quad \frac{f(x) - f(c)}{x - c} \ge 0 \]
Take the left-hand limit
Why: Equation 4.5.
\[ f'(c) = \lim_{x\to c^-}\frac{f(x)-f(c)}{x-c} \ge 0 \]
Combine
Why: A number that is at most zero and at least zero is zero.
\[ 0 \le f'(c) \le 0 \;\Longrightarrow\; f'(c) = 0 \]
Check on a concrete peak
Why: For 1 minus x squared at c equal to 0 the difference quotient is minus x: positive on the left, negative on the right, and its limit is 0.
\[ \frac{(1 - x^2) - 1}{x - 0} = -x, \quad \lim_{x\to 0}(-x) = 0 = f'(0) \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 321 — Proof of Theorem 4.2
Follow the proof one line at a time, matching each line to the picture on the previous slide. The first line is just the definition of a local maximum: near c, nothing is higher than f of c.
The only delicate point is the sign of the denominator. For x to the right of c, x minus c is positive, so dividing keeps the inequality: the difference quotient is at most zero. For x to the left, x minus c is negative, so dividing flips it: the difference quotient is at least zero. The two one-sided limits carry those inequalities to the derivative.
The check shows the whole mechanism on a real function. For one minus x squared the difference quotient at 0 simplifies to minus x, which is positive on the left and negative on the right, squeezing down to zero. A local minimum works the same way with every inequality reversed.
Trap
Having found the derivative's zero, a student concludes:
\[ f(x) = x^3, \quad f'(0) = 0 \]
\[ \Longrightarrow\; \text{extremum at } 0 ? \]
Wrong. Fermat's theorem runs one way only.
The cubing function is increasing on the whole line, so it has no peak or valley at 0. Critical numbers are candidates; each one still has to be tested.
\[ x < 0: \; x^3 < 0 \]
\[ x > 0: \; x^3 > 0 \]
This is the converse of Fermat's theorem, and it is false. The cube is the standard counterexample and worth memorising: its derivative, three x squared, is zero at 0, but the function is negative to the left of 0 and positive to the right, so 0 is neither a peak nor a valley.
The practical rule is that finding the critical numbers is only the first half of the job. For local extrema you then need to test each one, which Section 4.5 will do with the sign of the derivative. For absolute extrema on a closed interval, the comparison of values in Part 6 does the testing for you.
Concept
Figure (svg): Two panels. Left: y equals x cubed, flattening to a horizontal tangent at the origin and rising on both sides. Right: its derivative y equals 3 x squared, touching zero at the origin and positive everywhere else.
\[ f'(x) = 3x^2 = 0 \iff x = 0 \]
The derivative is zero at the origin and positive on both sides, so the function is still increasing as it passes through. Figure 4.15(d) is this graph. Section 4.5 will turn the sign of the derivative on each side into a test.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 321 — the x cubed example after the proof
Look at both panels together. On the left the cube flattens at the origin, with the dashed horizontal tangent sitting right on it. On the right its derivative, three x squared, touches zero at the origin and nowhere else.
The derivative touches zero without crossing it. It is positive just to the left and positive just to the right. Positive derivative means increasing, so the cube is increasing as it approaches the origin and still increasing as it leaves. It pauses, but it never turns.
That gives you a preview of the test in Section 4.5. A critical number is a local maximum when the derivative changes from positive to negative there, a minimum when it changes from negative to positive, and neither when it does not change sign at all.
Tweak it
Parameter explorer
The curve is y = x³ − ax. Its derivative is 3x² − a. Slide a from negative to positive. When are there two critical numbers, when one, and when none? When do you see a local maximum and minimum?
\[ f(x) = x^{3} - ({a})x \]
Start with a at 3 and watch the hump and the dip. The derivative three x squared minus a is zero where x squared equals a over three, so there are two critical numbers, one a local maximum and one a local minimum.
Now slide a down toward 0. The hump and the dip move toward each other and flatten. At a equal to 0 they merge into the single flat point of the cube, which is critical but not an extremum.
Push a below 0 and the derivative becomes positive everywhere, so the function is always increasing and there are no critical numbers at all. You have just watched two extrema collide and annihilate, which is a good way to remember that a critical number with a zero derivative can sit right on the boundary between turning and not turning.
Section
Part 5
Worked example
Find all critical points and decide from the graph whether each is a local extremum.
\[ f(x) = \tfrac13 x^3 - \tfrac52 x^2 + 4x \]
Differentiate
Why: Power rule term by term.
\[ f'(x) = x^2 - 5x + 4 \]
Where is it undefined?
Why: A polynomial is differentiable everywhere, so only zeros can give critical numbers.
\[ f'(x) \text{ exists for all } x \]
Factor and solve
Why: Two numbers with product 4 and sum minus 5.
\[ x^2 - 5x + 4 = (x - 4)(x - 1) = 0 \;\Longrightarrow\; x = 1, \; x = 4 \]
Evaluate at each
Why: The heights of the two flat points.
\[ f(1) = \tfrac13 - \tfrac52 + 4 = \tfrac{11}{6}, \quad f(4) = \tfrac{64}{3} - 40 + 16 = -\tfrac83 \]
Figure (svg): The cubic one third x cubed minus five halves x squared plus 4x for x from minus 1 to 6, with a local maximum marked at (1, 11/6) and a local minimum at (4, minus 8/3).
Check the derivative's signs
Why: Positive before 1, negative between, positive after 4: rise, fall, rise, so a local maximum at 1 and a local minimum at 4, as the graph shows.
\[ f'(0) = 4 > 0, \quad f'(2) = -2 < 0, \quad f'(5) = 4 > 0 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 322 — Example 4.12a and Figure 4.16
This is the routine case, and the routine is worth learning exactly. Differentiate, then ask the two questions: where is the derivative zero, and where does it fail to exist?
A polynomial has a derivative everywhere, so the second question has no answers, and the first comes down to factoring a quadratic. The critical numbers are 1 and 4. Evaluating the function there gives the heights of the two flat points, eleven sixths and minus eight thirds.
The book decides the type of each point from the graph. The check does it from the derivative instead, by sampling one value in each stretch: positive, then negative, then positive. The function rises into 1 and falls out of it, a maximum; it falls into 4 and rises out of it, a minimum. The picture and the signs agree.
Worked example
\[ f(x) = \left(x^2 - 1\right)^3 \]
Differentiate with the chain rule
Why: Outer power three, inner function x squared minus 1.
\[ f'(x) = 3\left(x^2 - 1\right)^2 (2x) \]
Tidy the product
Why: Collect the constants.
\[ f'(x) = 6x\left(x^2 - 1\right)^2 \]
Set each factor to zero
Why: The derivative is a polynomial, defined everywhere.
\[ 6x = 0 \;\text{ or }\; x^2 - 1 = 0 \;\Longrightarrow\; x = 0, \; x = \pm 1 \]
Figure (svg): The curve y equals (x squared minus 1) cubed for x from minus 1.6 to 1.6: a single valley at (0, minus 1), and flat shoulders at x equals minus 1 and 1 where the curve levels off momentarily and keeps going.
Read the graph
Why: Only 0 is a turning point, and f(0) is minus 1, the lowest value anywhere.
\[ f(0) = -1: \text{ local and absolute min}; \quad x = \pm 1: \text{ no extremum} \]
Check the sign of the derivative
Why: The squared factor is never negative, so the sign of the derivative is the sign of x alone: it changes only at 0.
\[ \operatorname{sign} f'(x) = \operatorname{sign}(x) \quad (x \ne \pm 1) \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 323 — Example 4.12b and Figure 4.17
The chain rule gives a product of simple factors, which is exactly what you want when you are about to set something equal to zero. Each factor gives its own critical numbers: 0 from the six x, and plus and minus 1 from the squared bracket.
The graph shows that only 0 is a turning point. At plus and minus 1 the curve flattens into a shoulder and carries on in the same direction. These are two more examples of the cube's behaviour, hiding inside a bigger function.
The check explains why. The squared factor is never negative, so it cannot change the sign of the derivative. The sign is decided by x alone, which changes only at 0. A squared factor in the derivative is a strong hint that its zero is a pause, not a turn.
Worked example
\[ f(x) = \frac{4x}{1 + x^2} \]
Apply the quotient rule
Why: Bottom times derivative of top, minus top times derivative of bottom, over bottom squared.
\[ f'(x) = \frac{(1 + x^2)(4) - 4x(2x)}{(1 + x^2)^2} \]
Simplify the numerator
Why: Four plus four x squared minus eight x squared.
\[ f'(x) = \frac{4 - 4x^2}{(1 + x^2)^2} \]
Where is it undefined?
Why: The denominator is at least 1, never zero.
\[ (1 + x^2)^2 \ge 1 > 0 \]
Set the numerator to zero
Why: Divide by 4 and take square roots.
\[ 4 - 4x^2 = 0 \;\Longrightarrow\; x^2 = 1 \;\Longrightarrow\; x = \pm 1 \]
Figure (svg): The curve y equals 4x over 1 plus x squared for x from minus 5 to 5: it falls from near 0 on the far left to a lowest point at (minus 1, minus 2), rises through the origin to a highest point at (1, 2), and then decays back toward 0.
Check the values against the bound
Why: f at 1 is 2 and f at minus 1 is minus 2; and the absolute value of 4x is at most 2 times one plus x squared, because the difference is twice a square.
\[ 2(1 + x^2) - 4|x| = 2(|x| - 1)^2 \ge 0 \;\Longrightarrow\; -2 \le f(x) \le 2 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 323 — Example 4.12c and Figure 4.18
The quotient rule is the only real work here. Be careful with the order of the two products in the numerator and with the sign between them; after that, the numerator simplifies to four minus four x squared.
The denominator is a square of something at least one, so it is never zero and the derivative exists everywhere. The critical numbers come from the numerator alone: plus and minus 1. The function's values there are 2 and minus 2.
The graph suggests these are not just local but absolute extrema, and the check proves it without the graph. The difference between two times one plus x squared and four times the absolute value of x is twice a perfect square, so it is never negative. Rearranged, that says the function never goes above 2 or below minus 2.
Concept
Example 4.12(c) found absolute extrema at 1 and minus 1, points with room on both sides. The book notes the general fact:
\[ \text{absolute extremum at interior } c \;\Longrightarrow\; \text{local extremum at } c \]
\[ \Longrightarrow\; c \text{ is a critical number (Fermat)} \]
This chain is exactly why absolute extrema on an interval can only be at endpoints or at critical numbers, which is Theorem 4.3 in Part 6.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 323 — the note at the end of Example 4.12
This observation looks small but it carries the whole method of Part 6. An absolute maximum at an interior point beats every value on the interval, so it certainly beats the values nearby, which makes it a local maximum.
Now apply Fermat. A local extremum happens only at a critical number. So an absolute extremum that is not at an endpoint must be at a critical number.
That leaves only two places to look for absolute extrema on a closed interval: the endpoints, and the critical numbers. Everything else can be ignored. Keep this two-step chain in mind, because it is exactly the proof of Theorem 4.3.
Fill the middle
\[ f(x) = x^3 - \tfrac12 x^2 - 2x + 1 \quad\Longrightarrow\quad f'(x) = 3x^2 - x - 2 \]
Fill in the blanks
3x^2 - x - 2 = (3x + 2)(x - 1)
Why: Expanding (3x + 2)(x − 1) gives 3x² − 3x + 2x − 2 = 3x² − x − 2. So the critical numbers are x = −2/3 and x = 1; the derivative is a polynomial, so there are no others.
Fill in both blanks before checking. The derivative of the cubic is a quadratic with leading coefficient 3, so one factor must start with three x and the other with x.
You need the constants to multiply to minus 2 and to produce a middle term of minus x. Try plus 2 and minus 1: the cross terms are minus three x and plus two x, which add to minus x. That works.
Always expand your factorisation to confirm it, as the explanation does. A sign slip at this stage sends you to the wrong critical numbers, and nothing later in the problem will catch it.
Worked example
\[ f(x) = x^3 - \tfrac12 x^2 - 2x + 1 \]
Differentiate
Why: Power rule term by term.
\[ f'(x) = 3x^2 - x - 2 \]
Factor
Why: From the previous slide.
\[ f'(x) = (3x + 2)(x - 1) \]
Solve
Why: Each factor gives one critical number; there are no points where the derivative is undefined.
\[ x = -\tfrac23, \quad x = 1 \]
Find the critical points
Why: Evaluate f at each critical number to get the points on the graph.
\[ f\left(-\tfrac23\right) = \tfrac{49}{27} \approx 1.815, \quad f(1) = -\tfrac12 \]
Check by substituting back
Why: The derivative really is zero at both numbers.
\[ f'\left(-\tfrac23\right) = \tfrac43 + \tfrac23 - 2 = 0, \quad f'(1) = 3 - 1 - 2 = 0 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 323 — Checkpoint 4.12
With the factorisation in hand, the critical numbers are immediate: minus two thirds and 1. The derivative is a polynomial, so there are no points where it fails to exist, and these two are all.
The question asks for critical points, so evaluate the function too. The book uses critical point and critical number interchangeably, but giving the full points, with their heights, is more useful and costs one line.
The check substitutes each number back into the derivative and gets zero. It is fast and it catches arithmetic slips. The two heights also tell you the shape: the function is higher at minus two thirds than at 1, so for a cubic with a positive leading coefficient the first is a local maximum and the second a local minimum.
Fill the middle
\[ y = 4x^3 - 3x \]
Fill in the blanks
y' = 12x^2 - 3 = 0 \iff x^2 = 1/4 \iff x = \pm 1/2
Why: Add 3 and divide by 12: x² = 3/12 = 1/4, so x = ±1/2. The derivative is a polynomial, defined everywhere, so these two are the only critical numbers.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 327 — Exercise 108
This is a quick one, so do it without writing anything down if you can. The derivative is twelve x squared minus three, and it exists everywhere.
Setting it to zero gives x squared equal to a quarter, so x is plus or minus one half. Type the answers as fractions or decimals; both are accepted.
Notice that the plus-or-minus comes from the square root. Forgetting the negative root is one of the commonest ways to lose a critical number, and with it possibly the answer to the whole problem.
Worked example
Find every critical number of this function, which is defined for all real x.
\[ f(x) = x^{2/3}(x - 4) \]
Expand before differentiating
Why: Two powers of x are easier than a product.
\[ f(x) = x^{5/3} - 4x^{2/3} \]
Differentiate
Why: Power rule on each term.
\[ f'(x) = \tfrac53 x^{2/3} - \tfrac83 x^{-1/3} \]
Combine over one denominator
Why: Multiply the first term by x to the one third over itself.
\[ f'(x) = \frac{5x - 8}{3x^{1/3}} \]
Zero of the numerator
Why: That gives the smooth critical number.
\[ 5x - 8 = 0 \;\Longrightarrow\; x = \tfrac85 \]
Zero of the denominator
Why: At 0 the derivative is undefined, yet f(0) is 0, so 0 is in the domain: it is critical too.
\[ 3x^{1/3} = 0 \;\Longrightarrow\; x = 0, \quad f(0) = 0 \]
Figure (svg): The curve x to the two thirds times x minus 4 for x from minus 1.5 to 5: it rises to a sharp cusp at the origin, falls to a smooth valley near (1.6, minus 3.28), then climbs.
Check the formula at a sample point
Why: At x equal to 1 the two forms of the derivative agree.
\[ \tfrac53 - \tfrac83 = -1 = \frac{5 - 8}{3} \]
This example exists to show you the kind of critical number that careless solutions miss. Expanding first makes the differentiation clean: two powers of x instead of a product.
Writing the derivative over a single denominator is the key move. The numerator vanishes at eight fifths, the ordinary kind of critical number. The denominator vanishes at 0, where the derivative does not exist. But the function itself is fine at 0, with value 0, so 0 is in the domain and it is a critical number too.
Look at the figure. The critical number at 0 is a cusp, and it is a local maximum. If you had only solved the derivative equal to zero, you would have found the valley and completely missed the peak. Whenever the derivative has a fractional power or an absolute value in it, look for points where it breaks.
Prediction
\[ y = x + \frac1x, \quad y' = 1 - \frac{1}{x^2} \]
Predict first
The derivative is undefined at x = 0. Is 0 a critical number?
Correct: No: 0 is not in the domain of the function
Why: A critical number must be a point of the domain; x + 1/x is not defined at 0 at all. The only critical numbers are x = ±1, where 1 − 1/x² = 0. An undefined derivative counts only where the function itself exists.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 327 — Exercise 117
Commit to an answer before revealing. This question is designed to test the first word of the definition, interior point of the domain.
The derivative is undefined at 0, which tempts you to call 0 critical. But the function is undefined there as well. A critical number has to be a place where the function exists; otherwise there is no point on the graph to be a peak or a valley.
The real critical numbers come from one minus one over x squared equal to zero, which gives plus and minus 1. At 1 the function has a local minimum of 2, and at minus 1 a local maximum of minus 2. Oddly, the local maximum is lower than the local minimum, which is possible only because the graph is broken at 0.
Section
Part 6
Concept
Location of Absolute Extrema (Theorem 4.3) — If f is continuous over a closed, bounded interval I, its absolute maximum and absolute minimum over I occur at endpoints of I or at critical numbers of f in I.
Figure (svg): A number line from minus 1.5 to 2.5 with the closed interval from 0 to 2 highlighted. Four points are marked: minus 1, crossed out because it lies outside the interval; 0, an endpoint where the derivative is also undefined; 1, where the derivative is zero; and 2, the other endpoint.
The argument is two lines long. An extremum not at an endpoint is interior, hence local, hence critical by Fermat.
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 324 — Theorem 4.3
This theorem is the payoff of everything so far. For a continuous function on a closed, bounded interval, the absolute extremes can only be at endpoints or at critical numbers inside the interval.
The figure shows what that means for Example 4.13(b), coming up shortly. The derivative is zero at minus 1 and at 1 and undefined at 0. Minus 1 is outside the interval, so it is crossed out. What remains is the two endpoints and the critical number 1: three candidates, three values to compare.
The reasoning is the two-step chain from before. If an extremum is not at an endpoint, it is interior, so it is local, so Fermat puts it at a critical number. And the Extreme Value Theorem guarantees there is something to find in the first place.
Notation
Annotate
On: \( \max_{[a,b]} f, \;\min_{[a,b]} f \;\in\; \{ f(a), \; f(b), \; f(c_1), \dots, f(c_k) \} \)
Go through the notes one at a time. The first reminds you that the theorem sits on top of the Extreme Value Theorem: the hypotheses are the same, continuous on a closed, bounded interval, and they guarantee that both extrema exist.
The endpoints are always candidates, and you never need to know anything about the derivative there. The interior critical numbers are candidates only if they lie inside the interval; one that lands outside is irrelevant, however natural it looked in the algebra.
The last note is the reason the method is so pleasant. The candidates form a short list of numbers, and you just pick the largest and the smallest. You do not need to decide whether each candidate is a peak or a valley; the comparison decides for you.
Concept
\[ \text{candidates} = \{a, b\} \cup \{c \in (a,b) : f'(c) = 0 \text{ or undefined}\} \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 324 — Problem-Solving Strategy: Locating Absolute Extrema over a Closed Interval
This is the book's problem-solving strategy, and you should be able to recite it. Endpoints, critical numbers inside, compare.
The middle step has two parts, and the second part is the one people forget. Solve for where the derivative is zero, and also look for points in the interval where the function is defined but the derivative is not. Then throw away any candidate outside the open interval.
The formula on the slide says the same thing as a set: the candidates are the two endpoints together with every interior point where the derivative is zero or undefined. In practice that set is small, often three or four numbers, and a table of their values finishes the problem.
Worked example
\[ f(x) = -x^2 + 3x - 2 \quad \text{over } [1, 3] \]
Step 1: evaluate at the endpoints
Why: Substitute 1 and 3.
\[ f(1) = -1 + 3 - 2 = 0, \quad f(3) = -9 + 9 - 2 = -2 \]
Step 2: differentiate
Why: A polynomial, so the derivative exists everywhere.
\[ f'(x) = -2x + 3 \]
Solve for the critical number
Why: It lies inside the interval, so it is a candidate.
\[ -2x + 3 = 0 \;\Longrightarrow\; x = \tfrac32 \in (1, 3) \]
Evaluate there
Why: Nine quarters subtracted from nine halves, minus two.
\[ f\left(\tfrac32\right) = -\tfrac94 + \tfrac92 - 2 = \tfrac14 \]
Step 3: compare
Why: Three candidates, three values.
\[ \begin{array}{c|ccc} x & 1 & \tfrac32 & 3 \\ \hline f(x) & 0 & \tfrac14 & -2 \end{array} \]
Figure (svg): The parabola minus x squared plus 3x minus 2 drawn solid over the interval from 1 to 3 and faint outside it, with dots at (1, 0), at the vertex (3/2, 1/4) and at (3, minus 2).
Check with the vertex formula
Why: A downward parabola peaks at minus b over 2a, and the far endpoint is lower than the near one.
\[ -\frac{3}{2(-1)} = \tfrac32 \;\Longrightarrow\; \max = \tfrac14 \text{ at } x = \tfrac32, \quad \min = -2 \text{ at } x = 3 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, pp. 324-325 — Example 4.13a and Figure 4.19
Follow the three steps exactly as the strategy lists them. Step 1 is just substitution: the function is 0 at the left end and minus 2 at the right end.
Step 2 finds one critical number, three halves, and it is inside the interval, so it joins the list. Its value is one quarter. Step 3 lays the three values out in a table, and reading off the largest and smallest finishes the problem.
Look at where the answers landed in the figure. The maximum is at an interior critical number, the vertex. The minimum is at the right endpoint, where the derivative is minus 3, not zero. If you had only looked at critical numbers you would have found the maximum and missed the minimum completely. The vertex formula confirms the maximum location.
Step zero
\[ f(x) = x^2 - 3x^{2/3} \quad \text{over } [0, 2] \]
Discussion prompt
Example 4.13(b) is next. Before computing anything, list what your candidate list must contain, and say which kind of critical number a power like x to the two thirds tends to produce.
Write your list before revealing. The point of this slide is to set up your search before the algebra distracts you.
The endpoints are always on the list. For the interior, a fractional power is a warning sign: differentiating x to the two thirds brings down a power of minus one third, which means dividing by a cube root, and that fails at 0.
In this example 0 happens to be an endpoint, so it is on the list anyway. But in the cusp example earlier, 0 was interior and was the local maximum. Getting into the habit of asking where the derivative breaks, every time, is what keeps those points from slipping away.
Worked example
\[ f(x) = x^2 - 3x^{2/3} \quad \text{over } [0, 2] \]
Step 1: evaluate at the endpoints
Why: The cube root of 4 is about 1.587.
\[ f(0) = 0, \quad f(2) = 4 - 3\sqrt[3]{4} \approx -0.762 \]
Step 2: differentiate
Why: Power rule, for x not zero.
\[ f'(x) = 2x - 2x^{-1/3} \]
Write over one denominator
Why: Multiply the first term by x to the one third over itself.
\[ f'(x) = \frac{2x^{4/3} - 2}{x^{1/3}} \]
Find all critical numbers
Why: The numerator vanishes when x to the four thirds is 1; the denominator vanishes at 0.
\[ 2x^{4/3} = 2 \;\Longrightarrow\; x = \pm 1; \quad f'(0) \text{ undefined} \]
Keep only candidates in the interval
Why: Minus 1 is outside; 0 is an endpoint, already evaluated.
\[ x = 1: \quad f(1) = 1 - 3 = -2 \]
Step 3: compare
Why: Three values.
\[ \begin{array}{c|ccc} x & 0 & 1 & 2 \\ \hline f(x) & 0 & -2 & -0.762 \end{array} \]
Figure (svg): The curve x squared minus 3 times x to the two thirds drawn solid on the interval from 0 to 2 and faint for x from minus 1.3 to 0. Dots mark (0, 0), where the curve has a cusp, the lowest point (1, minus 2), the right endpoint (2, minus 0.762), and, faintly, the excluded point (minus 1, minus 2).
Check the minimum with the derivative
Why: Just left of 1 the derivative is negative, just right it is positive: the graph falls then rises, as a minimum should.
\[ f'(0.5) \approx -1.52 < 0, \quad f'(1.5) \approx 1.25 > 0 \;\Longrightarrow\; \max = 0 \text{ at } 0, \; \min = -2 \text{ at } 1 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 325 — Example 4.13b and Figure 4.20
The endpoint values need the cube root of 4, about 1.587, so the right endpoint gives about minus 0.762. The derivative is two x minus two over the cube root of x, which over a common denominator becomes a fraction whose numerator and denominator can each be set to zero.
That produces three critical numbers: plus and minus 1 from the numerator, and 0 from the denominator. Now apply the filter. Minus 1 is outside the interval. Zero is an endpoint already on the list. Only 1 is new, and its value is minus 2.
The table gives a maximum of 0 at the left endpoint and a minimum of minus 2 at 1. The check samples the derivative on either side of 1 and finds it negative then positive, so the graph really does fall and then rise there, as the figure shows.
Trap
In Example 4.13(b) a student keeps every solution of the derivative equation:
\[ x = -1, \; 0, \; 1, \; 2 \]
\[ \min = f(-1) = -2 \text{ at } x = -1 \]
Wrong place, even though the value happens to match.
Minus 1 is not in [0, 2], so f(minus 1) is not a value the function takes on the interval. The minimum occurs at x = 1.
\[ -1 \notin [0, 2] \]
\[ \min = f(1) = -2 \text{ at } x = 1 \]
This trap is sneaky because the wrong answer has the right value. The function is even, so f at minus 1 equals f at 1, both minus 2. Keep minus 1 and you get the minimum value right and the location wrong.
The fix is a single line in the method: after solving for critical numbers, cross out every one outside the interval before evaluating anything. On most problems the values will not conveniently match, and an outside candidate will give you a wrong maximum or minimum outright. The faint part of the curve in the Example 4.13(b) figure shows the region the question never asked about.
Comparison
\[ f(x) = x^2 - 4x + 3 \text{ over } [1, 4], \quad f'(x) = 2x - 4 \]
Comparison matrix
| x | why it is a candidate | f(x) | conclusion |
|---|---|---|---|
| 1 | left endpoint | 0 | neither |
| 2 | f′(x) = 0 | −1 | absolute minimum |
| 4 | right endpoint | 3 | absolute maximum |
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 326 — Checkpoint 4.13
Fill in every blank before checking. The derivative is two x minus four, so the only critical number is 2, and it is inside the interval from 1 to 4.
The values are 0 at the left end, minus 1 at the critical number, and 3 at the right end. The largest, 3, is the absolute maximum, at x equal to 4. The smallest, minus 1, is the absolute minimum, at x equal to 2. The left endpoint is neither.
Laying the candidates out as a table like this is a good habit even when the arithmetic is easy. It makes your reasoning visible, and it makes it very hard to forget an endpoint.
Worked example
\[ f(x) = x^3 - 6x^2 + 9x \quad \text{over } [0, 5] \]
Evaluate at the endpoints
Why: Left and right ends.
\[ f(0) = 0, \quad f(5) = 125 - 150 + 45 = 20 \]
Differentiate
Why: A polynomial; no undefined points.
\[ f'(x) = 3x^2 - 12x + 9 \]
Factor
Why: Take out the 3 first.
\[ f'(x) = 3(x^2 - 4x + 3) = 3(x - 1)(x - 3) \]
Evaluate at the critical numbers
Why: Both 1 and 3 lie inside the interval.
\[ f(1) = 1 - 6 + 9 = 4, \quad f(3) = 27 - 54 + 27 = 0 \]
Compare
Why: Four candidates; the smallest value is attained twice.
\[ \begin{array}{c|cccc} x & 0 & 1 & 3 & 5 \\ \hline f(x) & 0 & 4 & 0 & 20 \end{array} \]
Figure (svg): The cubic x cubed minus 6 x squared plus 9x on the interval from 0 to 5, with dots at (0, 0), (1, 4), (3, 0) and (5, 20). The curve rises to a small hump at x equals 1, dips back to 0 at x equals 3, then shoots up to 20.
Check by factoring the function
Why: f is x times (x minus 3) squared, which is never negative on the interval, so 0 really is the floor, reached exactly at 0 and 3.
\[ f(x) = x(x - 3)^2 \ge 0 \text{ on } [0, 5] \;\Longrightarrow\; \max = 20 \text{ at } 5, \; \min = 0 \text{ at } 0 \text{ and } 3 \]
Run the method in order. The endpoint values are 0 and 20. Factoring the derivative gives critical numbers 1 and 3, both inside the interval, with values 4 and 0.
The table holds two surprises. The hump at x equal to 1, the only local maximum, is not the absolute maximum: the right endpoint is five times higher. And the minimum value 0 is reached twice, at the left endpoint and at the valley. Both locations should be reported.
The check finds a cleaner reason for the minimum. Factoring the function itself gives x times the square of x minus 3. On the interval both factors are never negative, so the function is never below 0, and it equals 0 exactly where one of the factors does: at 0 and at 3.
Sorting
For the cubic on the interval from 0 to 5 from the previous slide.
Sort into buckets
Sort each marked point of the graph.
Use the table from the previous slide and the figure to place each point. There are four buckets, so think about each point carefully.
The hump at 1 is a local maximum and nothing more, since the endpoint value 20 is larger. The valley at 3 is a local minimum whose value is also the smallest on the interval, so it is both local and absolute. The two endpoints are extrema, a tied absolute minimum on the left and the absolute maximum on the right, but by convention not local ones.
The point at x equal to 2 is included to keep you honest. The graph is falling there, with derivative minus 3, so it is not any kind of extremum; it is just a point on the way down.
Worked example
\[ f(x) = x + \sin x \quad \text{over } [0, 2\pi] \]
Evaluate at the endpoints
Why: Sine is zero at both.
\[ f(0) = 0, \quad f(2\pi) = 2\pi \]
Differentiate
Why: Defined everywhere.
\[ f'(x) = 1 + \cos x \]
Solve for critical numbers
Why: Cosine equals minus one only at pi in this interval.
\[ \cos x = -1 \;\Longrightarrow\; x = \pi \]
Evaluate there
Why: Sine of pi is zero.
\[ f(\pi) = \pi + 0 = \pi \]
Compare
Why: The critical value sits between the endpoint values.
\[ 0 < \pi < 2\pi \;\Longrightarrow\; \min = 0 \text{ at } x = 0, \quad \max = 2\pi \text{ at } x = 2\pi \]
Figure (svg): The curve y equals x plus sin x on the interval from 0 to 2 pi. It rises the whole way, flattening for a moment at x equals pi, where a dashed horizontal tangent is drawn. The endpoints (0, 0) and (2 pi, 2 pi) are marked.
Check why pi is not an extremum
Why: Cosine is at least minus one, so the derivative is never negative: f never decreases, and pi is a pause like the flat point of x cubed.
\[ f'(x) = 1 + \cos x \ge 0 \text{ for all } x \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 328 — Exercise 123
The derivative, one plus cosine x, is zero only where cosine is minus one, which on this interval is only at pi. So the candidates are 0, pi and two pi, with values 0, pi and two pi.
The comparison is immediate: the minimum 0 is at the left end and the maximum two pi is at the right end. The critical number at pi does not win anything.
The check shows why. Cosine is never less than minus one, so the derivative is never negative. The function never decreases; at pi it pauses with a horizontal tangent and then carries on climbing, exactly like the cube at the origin. The figure shows the dashed tangent at pi with the curve rising on both sides. The closed-interval method handled this correctly without you needing to notice.
Worked example
\[ f(x) = |x + 1| + |x - 1| \quad \text{over } [-3, 2] \]
Evaluate at the endpoints
Why: Add the two distances.
\[ f(-3) = 2 + 4 = 6, \quad f(2) = 3 + 1 = 4 \]
Rewrite without absolute values
Why: Split at the corners minus 1 and 1.
\[ f(x) = \begin{cases} -2x & x < -1 \\ 2 & -1 \le x \le 1 \\ 2x & x > 1 \end{cases} \]
Read off the derivative
Why: Slope minus 2, then 0, then 2; at the two corners the one-sided slopes disagree.
\[ f'(x) = -2, \; 0, \; 2; \quad f'(\pm 1) \text{ undefined} \]
List the critical numbers
Why: Both corners, and every point of the open middle piece, where the derivative is zero.
\[ \text{critical: every } x \in [-1, 1], \quad f(x) = 2 \]
Figure (svg): The graph of the absolute value of x plus 1 plus the absolute value of x minus 1 on the interval from minus 3 to 2: a straight line falling from (minus 3, 6) to (minus 1, 2), a flat floor at height 2 from minus 1 to 1, then a line rising to (2, 4).
Check by comparing all values
Why: The only values that occur at candidates are 6, 4 and 2.
\[ \max = 6 \text{ at } x = -3, \quad \min = 2 \text{ at every } x \in [-1, 1] \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 328 — Exercise 125
Absolute values are easiest to handle by splitting the line at the points where the insides change sign, here minus 1 and 1. On each piece the function is a simple line: slope minus 2 on the left, a constant 2 in the middle, slope 2 on the right.
The critical numbers are unusual. At minus 1 and 1 the slope jumps, so the derivative does not exist. And on the whole middle stretch the derivative is zero, so every point there is critical too. This function has infinitely many critical numbers, but they all share the value 2.
So the comparison still involves only three values: 6 at the left end, 4 at the right end, and 2 on the floor. The maximum is 6 at x equal to minus 3, and the minimum 2 occurs at every point of the floor. The figure shows that flat floor as a thick green segment.
Error analysis
Exercise 126 asks for the absolute extrema of the square root of x minus the square root of x cubed, over the interval from 0 to 4. Here is a submitted answer.
Annotate
On: \( f'(x) = \frac{1}{2\sqrt x} - \tfrac32\sqrt x = 0 \Rightarrow x = \tfrac13, \; \max = f\left(\tfrac13\right) \approx 0.385 \)
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 328 — Exercise 126
Read the submitted work and decide what is wrong before revealing the notes. The derivative and the critical number are both correct, which makes the mistake easy to overlook.
The error is an omission. The endpoints were never evaluated. At the left end the function is 0; at the right end it is the square root of 4 minus the square root of 64, which is 2 minus 8, or minus 6.
Once all three candidates are in a table, the answer changes. The maximum of about 0.385 at one third survives, but only by luck. The minimum, minus 6 at the right endpoint, was not found at all, and the submitted answer wrongly concluded there was none. On a closed interval the Extreme Value Theorem says there must be a minimum, which should have been the clue.
Estimation
\[ h(t) = -4.9t^2 + 60t + 5 \quad \text{(metres, } t \text{ in seconds)} \]
Predict first
A ball is thrown upward at 60 m/s from 5 m. Estimate its greatest height before you compute it.
Correct: About 190 m
Why: The velocity 60 − 9.8t reaches zero near t ≈ 6.1 s; in that time the average upward speed is about 30 m/s, so the ball climbs about 30 × 6.1 ≈ 184 m above its 5 m start: about 189 m.
Make a guess before you calculate. A quick physical estimate is a good habit, because it tells you whether the exact answer you compute later is reasonable.
Gravity takes away 9.8 metres per second of speed every second, so a ball launched at 60 metres per second stops rising after a little over 6 seconds. Its speed falls steadily from 60 to 0, so its average upward speed is about 30, and 30 metres per second for about 6 seconds is roughly 180 metres of climb.
Add the 5 metre starting height and you land near 190 metres. The next slide gets the exact value with a derivative.
Worked example
Find the height at which the ball stops ascending, and when that happens.
\[ h(t) = -4.9t^2 + 60t + 5 \]
Differentiate
Why: The derivative of height is the velocity.
\[ h'(t) = -9.8t + 60 \]
Set the velocity to zero
Why: The ball stops ascending when it stops moving up.
\[ -9.8t + 60 = 0 \;\Longrightarrow\; t = \frac{60}{9.8} \approx 6.12 \text{ s} \]
Evaluate the height
Why: Substitute that time.
\[ h(6.12) \approx -4.9(6.122)^2 + 60(6.122) + 5 \approx 188.7 \text{ m} \]
Figure (svg): The height of the ball, minus 4.9 t squared plus 60t plus 5 metres, for t from 0 to about 12.3 seconds: a downward parabola starting at 5 metres, peaking at about 188.7 metres at t about 6.12 seconds, and returning to the ground.
Check with the vertex formula
Why: The peak of a downward parabola is at minus b over 2a, and the height there is 5 plus b squared over minus 4a.
\[ -\frac{60}{2(-4.9)} \approx 6.122, \quad 5 + \frac{60^2}{4(4.9)} \approx 188.7 \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 328 — Exercise 141
The derivative of height with respect to time is the velocity. The ball stops ascending at the moment its velocity is zero: before that it is moving up, after that it is moving down. That is a critical number of the height function.
Solving gives about 6.12 seconds, and substituting gives about 188.7 metres, very close to the estimate. The figure shows the velocity changing sign at that moment: rising before, falling after.
The check uses the vertex formula for a parabola, which you will prove on the last worked example of this part. Two different routes giving the same numbers is a good sign that no arithmetic went wrong.
Real world
\[ C(v) = 0.005v^2 - 0.6v + 25 \quad \text{litres per 100 km, } 40 \le v \le 120 \]
Discussion prompt
A car's fuel use depends on its speed v (km/h) as above, over the speeds it actually drives. At what speed is fuel use lowest, and at what speed highest? Then: on a road where you must drive at least 70 km/h, what is the best speed?
Work through the question before revealing. This is the closed-interval method in a real setting: the interval is the range of speeds the car actually drives, from 40 to 120 kilometres per hour.
The derivative is zero at 60, which is inside the interval. Comparing the three candidates gives 9 at 40, 7 at 60 and 25 at 120. Fuel use is lowest at 60 and highest at the top speed, an endpoint.
The last part shows why the endpoints matter in practice. On a road with a minimum speed of 70, the interval is from 70 to 120, and the critical number 60 is outside it. The best you can do is the endpoint 70, with 7.5 litres per 100 kilometres. Real constraints push optimal answers to boundaries all the time, and only the endpoint check finds them.
Worked example
Prove that the parabola below, with a not zero, has its maximum or minimum at x equal to minus b over 2a.
\[ f(x) = ax^2 + bx + c, \quad a \ne 0 \]
Differentiate
Why: The derivative exists everywhere.
\[ f'(x) = 2ax + b \]
Solve for the only critical number
Why: Divide by 2a, which is not zero.
\[ 2ax + b = 0 \;\Longrightarrow\; x = -\frac{b}{2a} \]
Explain why it is an extremum
Why: The derivative is a line with slope 2a, so it changes sign at that point: from negative to positive if a is positive (a minimum), the reverse if a is negative (a maximum).
\[ f'(x) = 2a\left(x + \frac{b}{2a}\right) \]
Check on Example 4.13(a)
Why: There a is minus 1 and b is 3, and the maximum was found at three halves.
\[ -\frac{3}{2(-1)} = \tfrac32 \;\checkmark \]
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 326 — Exercise 90
You memorised this formula in precalculus, probably by completing the square. Calculus gets it in two lines, and it is worth seeing how.
The derivative of a quadratic is a straight line, two a x plus b, and it has exactly one zero, at minus b over two a. That is the only critical number, so it is the only place the parabola can turn.
Writing the derivative as two a times x minus the critical number shows the sign change directly. If a is positive the derivative goes from negative to positive, so the vertex is a minimum; if a is negative it is a maximum. The check confirms the formula on Example 4.13(a), where it gives three halves, exactly the maximum you found with the table.
Section
Part 7
Pattern
Figure (svg): A flow diagram. Two boxes on the left, endpoints f(a) and f(b), and critical numbers inside (a, b) where f prime is 0 or undefined, both feed arrows into a middle box, a table of values. From the table, one arrow leads up to the largest value, the absolute maximum, and one down to the smallest, the absolute minimum.
This is the procedure to follow every time you are asked for absolute extrema on a closed interval. The diagram shows its shape: two sources of candidates flowing into one table.
Start by checking the hypotheses, because without continuity on a closed, bounded interval the extrema may not exist, and the method could report an answer that is not real. Then gather the candidates: always the two endpoints, and every interior point where the derivative is zero or undefined. Discard anything outside the interval.
Finally, evaluate, tabulate, compare. Report each extremum as a value and a location, and do not be alarmed by ties. For local extrema the story is not finished: critical numbers are only candidates, and Section 4.5 supplies the test that decides them.
Ranking
Put in order
Order the moves of a complete solution.
Why: The hypothesis check comes first because without it there may be no extremum to find. Endpoints and critical numbers can be swapped, but every candidate must be evaluated before the comparison.
Drag the steps into order before checking. Most of the order is forced: you cannot compare values you have not computed.
The step that tends to be dropped is the first one, confirming continuity on a closed interval. It feels like a formality, but it is what guarantees the answer exists. The other step that gets skipped is discarding critical numbers outside the interval, which you saw can give a wrong location.
Check
Check your understanding
What is the absolute minimum value of f(x) = x³ − 3x over [0, 3]?
Answer: A
Why: f′(x) = 3x² − 3 = 0 at x = ±1; only x = 1 is in the interval. The candidates give f(0) = 0, f(1) = −2 and f(3) = 18, so the minimum value is −2, occurring at x = 1.
Run the method yourself before choosing. The derivative is zero at plus and minus 1, but only 1 is in the interval.
The values at 0, 1 and 3 are 0, minus 2 and 18. The smallest value is minus 2. Each distractor is one of the traps from this lesson: reporting the location instead of the value, stopping at an endpoint, or using a critical number outside the interval.
Check
Check your understanding
f is differentiable everywhere and f′(c) = 0. What can you conclude?
Answer: B
Why: Fermat's theorem runs from extremum to zero derivative, never back. x³ has f′(0) = 0 with no extremum at 0, so a zero derivative only makes c a candidate.
Think about which way the implication in Fermat's theorem runs before answering. It starts from a local extremum and concludes a zero derivative.
Here you are given the zero derivative, so you are being asked about the converse, which is false. The cube at the origin is the counterexample. All you can say is that c is a critical number, a candidate that still needs testing.
Check
Check your understanding
For which function does the Extreme Value Theorem guarantee both an absolute maximum and an absolute minimum?
Answer: C
Why: √x is continuous on the closed, bounded interval [0, 4] (it is defined and continuous for x ≥ 0), so both hypotheses hold: minimum 0 at x = 0 and maximum 2 at x = 4.
Check both hypotheses for each option: continuity at every point of the interval, and an interval that is closed and bounded.
Only the square root on the interval from 0 to 4 passes both. It is continuous for all non-negative x, including the endpoint 0. The other three each fail one hypothesis: two have a discontinuity inside a closed interval, and one has an open interval.
Explain it to yourself
Discussion prompt
Fermat's theorem is the reason we look at critical numbers. In two or three sentences, explain why the same theorem gives no information at an endpoint, and so why endpoints must be evaluated separately. (Exercise 91.)
OpenStax Calculus Volume 1, §4.3 Maxima and Minima §4.3, p. 326 — Exercise 91
Write your explanation before revealing. This is the idea that turns following the method into understanding it.
The proof of Fermat's theorem needs points on both sides of c: the right-hand quotients give one inequality, the left-hand quotients the other, and together they force zero. At an endpoint only one side exists, so you only get one inequality, and the derivative can be anything.
Example 4.13(a) is the proof in practice. Its minimum is at the right endpoint 3, where the derivative is minus 3, not zero. No search for critical numbers would ever find it. That is why the endpoints go on the candidate list unconditionally.
Exit ticket
\[ f(x) = 2x^3 - 3x^2 - 12x \quad \text{over } [-2, 3] \]
Discussion prompt
Find the absolute maximum and minimum values and where they occur. Show your candidate table.
This problem uses the whole method in one go. Check the hypotheses, gather the candidates, and build the table.
The derivative factors as six times x minus 2 times x plus 1, so the critical numbers are minus 1 and 2, both inside the interval. With the endpoints minus 2 and 3, there are four candidates, with values minus 4, 7, minus 20 and minus 9.
The maximum is 7 at minus 1 and the minimum is minus 20 at 2. Here both happen to be interior, so the endpoints lose, but you only know that because you checked them.
Recap
| idea | what it says | watch out for |
|---|---|---|
| absolute extremum | f(c) is the largest or smallest value on the whole interval | report the value f(c) and where it occurs |
| local extremum | f(c) beats its neighbours on some open interval | endpoints are not local extrema |
| Extreme Value Theorem | continuous on closed, bounded [a, b] gives both extrema | drop either hypothesis and they can vanish |
| critical number | interior c with f′(c) = 0 or f′(c) undefined | only a candidate: x³ at 0 |
| Fermat's theorem | local extremum and differentiable gives f′(c) = 0 | one-way implication |
| closed-interval method | compare f at endpoints and interior critical numbers | discard critical numbers outside the interval |
Next, Section 4.4 proves the Mean Value Theorem, the tool that lets Section 4.5 decide which critical numbers really are maxima or minima.
Stewart, Calculus: Early Transcendentals 8e, §4.1 Maximum and Minimum Values §4.1, pp. 276-286 — the same material in Stewart
Absolute extrema are the largest and smallest values on the whole interval; local extrema only beat their neighbours. The Extreme Value Theorem guarantees absolute extrema for a continuous function on a closed, bounded interval, and each hypothesis matters.
Fermat's theorem says a local extremum with a derivative has a zero derivative, so local extrema live at critical numbers: interior points where the derivative is zero or undefined. The converse is false, as the cube shows.
Put together, those give the closed-interval method: evaluate at the endpoints and the interior critical numbers, and compare. The one question left open is how to tell, without a graph, whether a critical number is a maximum, a minimum or neither. Section 4.4's Mean Value Theorem is the tool that answers it in Section 4.5.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.