Calculus3Analyzing Functions4 of 6
Analyzing Functions · 4

Concavity and Inflection Points

1

How the slope changes

Take two functions, f(x)=x22f(x) = \frac{x^2}{2} and g(x)=2x−x22g(x) = 2x - \frac{x^2}{2}, on the interval [0,2][0, 2]. Both start at (0,0)(0, 0), end at (2,2)(2, 2), and increase the whole way, yet the two curves have different shapes. To see what separates them, let's look at their tangent slopes across the interval.

22
xx
11
22
yy
0\htmlClass{lc-green}{0}
2\htmlClass{lc-green}{2}
1\htmlClass{lc-green}{1}
1\htmlClass{lc-green}{1}
2\htmlClass{lc-green}{2}
0\htmlClass{lc-green}{0}
ff
gg

The slopes of ff go 0,1,2\fpr{0, 1, 2}, so f′\fpr{f'} is increasing. When the derivative is increasing, we say the function is concave up. For gg, the slopes go 2,1,0\fpr{2, 1, 0}, so g′\fpr{g'} is decreasing. When the derivative is decreasing, we say the function is concave down.

Compare the tangent slopes of f(x)=x22f(x) = \frac{x^2}{2} and g(x)=2x−x22g(x) = 2x - \frac{x^2}{2} on [0,2][0, 2], then check ff on [−2,0][-2, 0]
11
22
xx
11
22
yy
g(x)=2x−x22g(x)=2x-\dfrac{x^2}{2}
f(x)=x22f(x)=\dfrac{x^2}{2}
Text solution
Same ends, both increasingBoth derivatives are positive on (0,2)(0, 2), but at x=1x = 1 the curve of gg is higher.
f′(x)=x,g′(x)=2−x\displaystyle \fpr{f'(x) = x}, \quad \fpr{g'(x) = 2 - x}
f(1)\displaystyle f(1) =12,g(1)\displaystyle = \frac12, \quad g(1) =32\displaystyle = \frac32
Compare the slopesThe slopes of ff go up and the slopes of gg go down.
f′(0)=0,f′(1)=1,f′(2)=2g′(0)=2,g′(1)=1,g′(2)=0\displaystyle \begin{aligned}&\fpr{f'(0) = 0}, \quad \fpr{f'(1) = 1}, \quad \fpr{f'(2) = 2} \\ &\fpr{g'(0) = 2}, \quad \fpr{g'(1) = 1}, \quad \fpr{g'(2) = 0}\end{aligned}
A decreasing curve can be concave upOn [−2,0][-2, 0], f′(x)=x\fpr{f'(x) = x} is negative, so ff decreases. But its slopes still go up, so ff is concave up there too.
f′(−2)=−2\displaystyle \fpr{f'(-2) = -2}
f′(−1)=−1\displaystyle \fpr{f'(-1) = -1}
f′(0)=0\displaystyle \fpr{f'(0) = 0}

So ff is concave up on all of [−2,2]\cvu{[-2, 2]}, even though it decreases on [−2,0][-2, 0] and increases on [0,2][0, 2]. Its slope goes up the whole way.

−2-2
22
xx
−2\htmlClass{lc-green}{-2}
−1\htmlClass{lc-green}{-1}
0\htmlClass{lc-green}{0}
1\htmlClass{lc-green}{1}
2\htmlClass{lc-green}{2}
2

Finding concavity from the second derivative

When we only have a formula for ff and no tangent lines drawn, how do we know where f′\fpr{f'} is increasing? The same way we tell where any function is increasing, from the sign of its derivative. The derivative of f′\fpr{f'} is f′′\fpp{f''}.

f′′(x)>0  ⟹  f′ increasing  ⟹  f concave up\displaystyle {\fpp{f''(x)} > 0 \implies \fpr{f'} \text{ increasing} \implies \cvu{f \text{ concave up}}}
f′′(x)>0  ⟹  f′ increasing  ⟹  f concave up\displaystyle \begin{aligned}&\fpp{f''(x)} > 0 \implies \fpr{f'} \text{ increasing} \\ &\implies \cvu{f \text{ concave up}}\end{aligned}
f′′(x)<0  ⟹  f′ decreasing  ⟹  f concave down\displaystyle {\fpp{f''(x)} < 0 \implies \fpr{f'} \text{ decreasing} \implies \cvd{f \text{ concave down}}}
f′′(x)<0  ⟹  f′ decreasing  ⟹  f concave down\displaystyle \begin{aligned}&\fpp{f''(x)} < 0 \implies \fpr{f'} \text{ decreasing} \\ &\implies \cvd{f \text{ concave down}}\end{aligned}

Our two curves agree with it. For x22\frac{x^2}{2}, f′′(x)=1\fpp{f''(x) = 1} is positive, and for 2x−x222x - \frac{x^2}{2}, f′′(x)=−1\fpp{f''(x) = -1} is negative. Now let's try a function whose second derivative changes sign. A point on the graph where the concavity changes is called an inflection point.

Find where f(x)=x3−3xf(x) = x^3 - 3x is concave up and concave down, and find its inflection point
f(x)=x3−3xconcave up on  ?concave down on  ?inflection point  ?\displaystyle \begin{array}{l} f(x) = x^3 - 3x \\[4pt] \htmlClass{lc-blue}{\text{concave up on}}\ \ ? \\ \htmlClass{lc-pink}{\text{concave down on}}\ \ ? \\ \htmlClass{lc-violet}{\text{inflection point}}\ \ ? \end{array}
Text solution
Differentiate twiceConcavity comes from the sign of f′′\fpp{f''}.
f′(x)=3x2−3,f′′(x)=6x\displaystyle \fpr{f'(x) = 3x^2 - 3}, \qquad \fpp{f''(x) = 6x}
Read the sign of f′′\fpp{f''}6x\fpp{6x} is negative for x<0x < 0 and positive for x>0x > 0.
concave down on (−∞,0)\displaystyle \cvd{\text{concave down on } (-\infty, 0)}
concave up on (0,∞)\displaystyle \cvu{\text{concave up on } (0, \infty)}
The inflection pointThe sign of f′′\fpp{f''} changes at x=0x = 0. For the height, plug 00 into ff, not into f′′\fpp{f''}.
f(0)=0 → inflection point (0,0)\displaystyle {f(0) = 0 \ \to \ \fpp{\text{inflection point } (0, 0)}}
The smallest slopef′\fpr{f'} decreases until x=0x = 0 and increases after it, so the slope is smallest there.
f′(0)=−3\displaystyle \fpr{f'(0) = -3}

Under the curve, the sign chart of f′′\fpp{f''} shows where each piece comes from.

xx
y=x3−3xy=x^3-3x
f′′\htmlClass{lc-violet}{f''}
−\htmlClass{lc-pink}{-}
+\htmlClass{lc-blue}{+}
0\htmlClass{lc-violet}{0}
(0,0)\htmlClass{lc-violet}{(0,0)}
3

When f′′f'' is zero or undefined

At the inflection point (0,0)\fpp{(0, 0)} of x3−3xx^3 - 3x, the second derivative is 00. But does f′′=0\fpp{f'' = 0} always give an inflection point? Let's test x3x^3 and x4x^4, which both have f′(0)=0\fpr{f'(0) = 0} and f′′(0)=0\fpp{f''(0) = 0}, and then x5/3x^{5/3}, whose second derivative is undefined at 00.

y=x3y=x^3
f′′=6x\htmlClass{lc-violet}{f''=6x}
y=x4y=x^4
f′′=12x2\htmlClass{lc-violet}{f''=12x^2}
Testing x3x^3, x4x^4, and x5/3x^{5/3} for inflection points
f′′=0 ⇒ inflection point ?\htmlClass{lc-violet}{f''=0}\ \Rightarrow\ \htmlClass{lc-violet}{\text{inflection point}}\,?
f(x)=x3f(x)=x^3
f(x)=x4f(x)=x^4
Text solution
x3x^3: the sign changes6x\fpp{6x} is negative left of 00 and positive right of it, so (0,0)\fpp{(0, 0)} is an inflection point.
x3: f′′(x)=6x,−∣+ at 0 → (0,0)\displaystyle \begin{aligned}&x^3: \ \fpp{f''(x) = 6x}, \\ &\quad \cvd{-} \mid \cvu{+} \text{ at } 0 \ \to \ \fpp{(0, 0)}\end{aligned}
x4x^4: the sign stays the same12x2\fpp{12x^2} is 00 at 00 but positive on both sides, so x4x^4 is concave up on both sides and has no inflection point.
x4: f′′(x)=12x2,+∣+ at 0 → none\displaystyle \begin{aligned}&x^4: \ \fpp{f''(x) = 12x^2}, \\ &\quad \cvu{+} \mid \cvu{+} \text{ at } \mut{0} \ \to \ \text{none}\end{aligned}
The candidatesLike f′\fpr{f'}, f′′\fpp{f''} can only change sign where it's 00 or undefined. Those points are candidates, and we check the sign of f′′\fpp{f''} on each side.
x53x^{\frac53}: undefined at 00The numerator 1010 is never 00, but the cube root makes f′′\fpp{f''} undefined at 00 and gives it the sign of xx. So (0,0)\fpp{(0, 0)} is an inflection point.
f′′(x)=109x3\displaystyle \fpp{f''(x) = \frac{10}{9\sqrt[3]{x}}}
−∣+ at 0 → (0,0)\displaystyle \cvd{-} \mid \cvu{+} \text{ at } 0 \ \to \ \fpp{(0, 0)}

Is every zero of the second derivative an inflection point?

4

Reading concavity from the graph of f′f'

Sometimes we're only given a graph of f′\fpr{f'}, with no formula for ff. Whether the graph is above or below the x-axis still tells us where ff increases and decreases. For concavity, we look at something else, whether f′\fpr{f'} itself is increasing or decreasing.

f increasing: f′>0,f decreasing: f′<0\displaystyle {f \text{ increasing: } \fpr{f'} > 0, \qquad f \text{ decreasing: } \fpr{f'} < 0}
f increasing: f′>0,f decreasing: f′<0\displaystyle \begin{aligned}&f \text{ increasing: } \fpr{f'} > 0, \\ &f \text{ decreasing: } \fpr{f'} < 0\end{aligned}
f concave up: f′ increasing,f concave down: f′ decreasing\displaystyle {\cvu{f \text{ concave up}}\text{: } \fpr{f'} \text{ increasing}, \qquad \cvd{f \text{ concave down}}\text{: } \fpr{f'} \text{ decreasing}}
f concave up: f′ increasing,f concave down: f′ decreasing\displaystyle \begin{aligned}&\cvu{f \text{ concave up}}\text{: } \fpr{f'} \text{ increasing}, \\ &\cvd{f \text{ concave down}}\text{: } \fpr{f'} \text{ decreasing}\end{aligned}

Let's read both from the graph of f′(x)=3x2−3\fpr{f'(x) = 3x^2 - 3} on [−2,2][-2, 2].

Find the intervals of concavity and the inflection points of ff from the graph of f′(x)=3x2−3f'(x) = 3x^2 - 3
−2-2
−1-1
00
11
22
f(x)f(x)
=?{}={?}
f′(x)\htmlClass{lc-green}{f'(x)}
=3x2−3\htmlClass{lc-green}{{}=3x^2-3}
Text solution
Where f′\fpr{f'} crosses the axisf′\fpr{f'} is positive, then negative, then positive, so ff has a local maximum at −1-1 and a local minimum at 11.
f′(−1)=0, f′ ⁣:+→−→ local max\displaystyle \begin{aligned}&\fpr{f'(-1) = 0, \ f' \colon + \to -} \\ &\quad \fpr{\to \ \text{local max}}\end{aligned}
f′(1)=0, f′ ⁣:−→+→ local min\displaystyle \begin{aligned}&\fpr{f'(1) = 0, \ f' \colon - \to +} \\ &\quad \fpr{\to \ \text{local min}}\end{aligned}
Where f′\fpr{f'} falls and risesf′\fpr{f'} decreases until x=0x = 0 and increases after it. Its lowest point is where the concavity of ff changes.
concave down on (−2,0)\displaystyle \cvd{\text{concave down on } (-2, 0)}
concave up on (0,2)\displaystyle \cvu{\text{concave up on } (0, 2)}
inflection point at x=0\displaystyle \fpp{\text{inflection point at } x = 0}
Both at once on (−2,−1)(-2, -1)There f′\fpr{f'} is positive but decreasing, so ff is increasing and concave down\cvd{\text{concave down}}.
From a graph of f′′\fpp{f''}Read concavity from its sign. f′′(x)=6x\fpp{f''(x) = 6x} crosses the x-axis at 00, so there's an inflection point there. A graph like 12x2\fpp{12x^2} that only touches the axis gives none.

On the graph of f(x)=x3−3xf(x) = x^3 - 3x itself, the local maximum (−1,2)\fpr{(-1, 2)} and local minimum (1,−2)\fpr{(1, -2)} sit above the zeros of f′\fpr{f'}, and the inflection point (0,0)\fpp{(0, 0)} sits above its lowest point.

xx
−2-2
22
xx
ff
f′\htmlClass{lc-green}{f'}
(−1,2)\htmlClass{lc-green}{(-1,2)}
(1,−2)\htmlClass{lc-green}{(1,-2)}
(0,0)\htmlClass{lc-violet}{(0,0)}
−1\htmlClass{lc-green}{-1}
1\htmlClass{lc-green}{1}
(0,−3)\htmlClass{lc-violet}{(0,-3)}

Can a function be decreasing and concave up at the same time?

5

Concavity and tangent lines

When a function is concave up, does the curve lie above its tangent lines or below them? The answer tells us whether a tangent-line estimate comes out too big or too small. Remember, the tangent line to y=f(x)y = f(x) at x=ax = a is

y=f(a)+f′(a)(x−a)\displaystyle {\tng{y = f(a) + f'(a)(x - a)}}

Let's compare f(x)=x22f(x) = \frac{x^2}{2} with its tangent line at x=1x = 1, and then use what we find to judge an estimate of 4.2\sqrt{4.2}.

Tangent lines to concave-up and concave-down curves, and estimating 4.2\sqrt{4.2}
11
xx
f(x)=x22f(x)=\dfrac{x^2}{2}
Text solution
Right of 11, the curve is steeperThe tangent line has slope 1\fpr{1}, and for x>1x > 1 the curve's slope is larger. They start at the same point (1,12)\left(1, \frac12\right), so the curve stays above the line.
y=x−12,f′(x)=x>1\displaystyle \tng{y = x - \frac12}, \qquad \fpr{f'(x) = x} > \fpr{1}
Left of 11, the curve rises lessFor x<1x < 1, the curve's slope is less than 1\fpr{1}, so on the way to 11 the curve rises less than the line. They meet at 11, so the curve started higher. At x=0x = 0, the curve is at 00 and the line is at −12\tng{-\frac12}.
Every tangent lineBecause f′\fpr{f'} is increasing, the same holds at any point aa, so a concave-up curve lies above every tangent line. For a concave-down curve, f′\fpr{f'} is decreasing, and the curve lies below every tangent line.
Estimate 4.2\sqrt{4.2}The tangent to x\sqrt{x} at 44 gives 2.05\tng{2.05}. Since f′′<0\fpp{f''} < 0, x\sqrt{x} is concave down and lies below the tangent, so 2.05\tng{2.05} is too big.
y=2+14(x−4)\displaystyle \tng{y = 2 + \frac14(x - 4)}
f′′(x)=−14x32<0\displaystyle \fpp{f''(x) = -\frac{1}{4x^{\frac32}}} < 0
2.052=4.2025>4.2\displaystyle {\tng{2.05}^2 = 4.2025 > 4.2}

So the concavity tells us right away which way a tangent-line estimate is off. At an inflection point the concavity switches, so there the curve crosses from one side of its tangent line to the other.

11
xx
0\htmlClass{lc-green}{0}
2\htmlClass{lc-green}{2}
y=x−12\htmlClass{lc-orange}{y=x-\tfrac12}
f concave up: curve above its tangent lines→ tangent estimate too small\displaystyle \begin{aligned}&\cvu{f \text{ concave up}}\text{: curve above its }\tng{\text{tangent lines}} \\ &\quad \to \ \tng{\text{tangent estimate}}\text{ too small}\end{aligned}
f concave up: curve above its tangent lines→ tangent estimate too small\displaystyle \begin{aligned}&\cvu{f \text{ concave up}}\text{: } \\ &\quad \text{curve above its }\tng{\text{tangent lines}} \\ &\quad \to \ \tng{\text{tangent estimate}}\text{ too small}\end{aligned}
f concave down: curve below its tangent lines→ tangent estimate too big\displaystyle \begin{aligned}&\cvd{f \text{ concave down}}\text{: curve below its }\tng{\text{tangent lines}} \\ &\quad \to \ \tng{\text{tangent estimate}}\text{ too big}\end{aligned}
f concave down: curve below its tangent lines→ tangent estimate too big\displaystyle \begin{aligned}&\cvd{f \text{ concave down}}\text{: } \\ &\quad \text{curve below its }\tng{\text{tangent lines}} \\ &\quad \to \ \tng{\text{tangent estimate}}\text{ too big}\end{aligned}

Does concavity change at a local maximum or minimum?

6

The second derivative test

To decide whether a critical point cc is a local maximum or a local minimum, we've checked the sign of f′\fpr{f'} on both sides of cc. At cc the tangent line is horizontal, so the concavity there tells us which side of that line the curve is on. Let's use that on both critical points of f(x)=x3−3xf(x) = x^3 - 3x.

Classify the critical points of f(x)=x3−3xf(x) = x^3 - 3x, and test x4x^4 and x3x^3 where f′′=0f'' = 0
−1-1
11
xx
f(x)=x3−3xf(x)=x^3-3x
Text solution
The critical points
f′(x)=3x2−3=0 → x=±1\displaystyle \fpr{f'(x) = 3x^2 - 3 = 0} \ \to \ \fpr{x = \pm 1}
f′′(x)=6x\displaystyle \fpp{f''(x) = 6x}
Concave up at 11, concave down at −1-1Near 11 the curve lies above its tangent y=−2\tng{y = -2}, and near −1-1 it lies below y=2\tng{y = 2}.
f′′(1)=6>0  ⟹  local min (1,−2)\displaystyle \fpp{f''(1) = 6} > 0 \implies \fpr{\text{local min } (1, -2)}
f′′(−1)=−6<0  ⟹  local max (−1,2)\displaystyle \fpp{f''(-1) = -6} < 0 \implies \fpr{\text{local max } (-1, 2)}
When f′′(c)=0\fpp{f''(c) = 0}x4x^4, −x4-x^4 and x3x^3 all have f′(0)=0\fpr{f'(0) = 0} and f′′(0)=0\fpp{f''(0) = 0}, but at 00 they have a local minimum, a local maximum and neither. So f′′(c)=0\fpp{f''(c) = 0} tells us nothing, and we check the sign of f′\fpr{f'} on each side.
−1-1
11
xx
y=2\htmlClass{lc-orange}{y=2}
y=−2\htmlClass{lc-orange}{y=-2}

Plugging a critical point into f′′\fpp{f''} this way is the second derivative test. When f′′(c)=0\fpp{f''(c) = 0}, we go back to the first derivative test.

f′(c)=0, f′′(c)>0  ⟹  local min at c\displaystyle {\fpr{f'(c) = 0}, \ \fpp{f''(c)} > 0 \implies \fpr{\text{local min at } c}}
f′(c)=0, f′′(c)>0  ⟹  local min at c\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c)} > 0 \\ &\implies \fpr{\text{local min at } c}\end{aligned}
f′(c)=0, f′′(c)<0  ⟹  local max at c\displaystyle {\fpr{f'(c) = 0}, \ \fpp{f''(c)} < 0 \implies \fpr{\text{local max at } c}}
f′(c)=0, f′′(c)<0  ⟹  local max at c\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c)} < 0 \\ &\implies \fpr{\text{local max at } c}\end{aligned}
f′(c)=0, f′′(c)=0  ⟹  no conclusion: use the first derivative test\displaystyle {\fpr{f'(c) = 0}, \ \fpp{f''(c) = 0} \implies \text{no conclusion: use the first derivative test}}
f′(c)=0, f′′(c)=0  ⟹  no conclusion: use the first derivative test\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c) = 0} \\ &\implies \text{no conclusion: } \\ &\quad \text{use the first derivative test}\end{aligned}
7

Sketching with both derivatives

To sketch the graph of f(x)=x4−4x3f(x) = x^4 - 4x^3, we put the sign charts of f′\fpr{f'} and f′′\fpp{f''} together on a single axis. The sign of f′\fpr{f'} tells us where the curve is increasing or decreasing, and the sign of f′′\fpp{f''} tells us where it's concave up or concave down.

Sketch the graph of f(x)=x4−4x3f(x) = x^4 - 4x^3 using the sign charts of f′f' and f′′f''
1010
−10-10
−20-20
f(x)=x4−4x3f(x)=x^4-4x^3
f′\htmlClass{lc-green}{f'}
f′′\htmlClass{lc-violet}{f''}
Text solution
Factor both derivatives
f′(x)=4x3−12x2=4x2(x−3)f′′(x)=12x2−24x=12x(x−2)\displaystyle \begin{aligned}&\fpr{f'(x) = 4x^3 - 12x^2 = 4x^2(x - 3)} \\ &\fpp{f''(x) = 12x^2 - 24x = 12x(x - 2)}\end{aligned}
The sign of f′\fpr{f'}Since 4x2≥0\fpr{4x^2} \ge 0, the sign comes from x−3\fpr{x - 3}. So f′\fpr{f'} is negative on both sides of 00, and the only extremum is at x=3\fpr{x = 3}.
f(3)\displaystyle f(3) =81−108\displaystyle = 81 - 108 =−27 → local min (3,−27)\displaystyle = -27 \ \to \ \fpr{\text{local min } (3, -27)}
The sign of f′′\fpp{f''}f′′\fpp{f''} goes from positive to negative at 00 and back to positive at 22, so both are inflection points.
(0,0),(2,−16)\displaystyle \fpp{(0, 0)}, \quad \fpp{(2, -16)}
At the originBoth f′(0)=0\fpr{f'(0) = 0} and f′′(0)=0\fpp{f''(0) = 0}, so the second derivative test gives no conclusion. The charts show ff keeps decreasing while it switches from concave up to concave down, so the curve crosses its horizontal tangent y=0\tng{y = 0} from above to below.

Read together, interval by interval, the two charts give the shape of the curve.

(−∞,0):f′<0, f′′>0decreasing, concave up\displaystyle {(-\infty, 0): \quad \fpr{f' < 0}, \ \fpp{f'' > 0} \qquad \text{decreasing, }\cvu{\text{concave up}}}
(−∞,0):f′<0, f′′>0decreasing, concave up\displaystyle \begin{aligned}&(-\infty, 0): \quad \fpr{f' < 0}, \ \fpp{f'' > 0} \\ &\quad \text{decreasing, }\cvu{\text{concave up}}\end{aligned}
(0,2):f′<0, f′′<0decreasing, concave down\displaystyle {(0, 2): \quad \fpr{f' < 0}, \ \fpp{f'' < 0} \qquad \text{decreasing, }\cvd{\text{concave down}}}
(0,2):f′<0, f′′<0decreasing, concave down\displaystyle \begin{aligned}&(0, 2): \quad \fpr{f' < 0}, \ \fpp{f'' < 0} \\ &\quad \text{decreasing, }\cvd{\text{concave down}}\end{aligned}
(2,3):f′<0, f′′>0decreasing, concave up\displaystyle {(2, 3): \quad \fpr{f' < 0}, \ \fpp{f'' > 0} \qquad \text{decreasing, }\cvu{\text{concave up}}}
(2,3):f′<0, f′′>0decreasing, concave up\displaystyle \begin{aligned}&(2, 3): \quad \fpr{f' < 0}, \ \fpp{f'' > 0} \\ &\quad \text{decreasing, }\cvu{\text{concave up}}\end{aligned}
(3,∞):f′>0, f′′>0increasing, concave up\displaystyle {(3, \infty): \quad \fpr{f' > 0}, \ \fpp{f'' > 0} \qquad \text{increasing, }\cvu{\text{concave up}}}
(3,∞):f′>0, f′′>0increasing, concave up\displaystyle \begin{aligned}&(3, \infty): \quad \fpr{f' > 0}, \ \fpp{f'' > 0} \\ &\quad \text{increasing, }\cvu{\text{concave up}}\end{aligned}
(0,0),(2,−16),(3,−27),(4,0)\displaystyle {\fpp{(0, 0)}, \quad \fpp{(2, -16)}, \quad \fpr{(3, -27)}, \quad (4, 0)}
xx
1010
−10-10
−20-20
yy
y=0\htmlClass{lc-orange}{y=0}
(0,0)\htmlClass{lc-violet}{(0,0)}
(2,−16)\htmlClass{lc-violet}{(2,-16)}
(3,−27)\htmlClass{lc-green}{(3,-27)}
(4,0)(4,0)
y=x4−4x3y=x^4-4x^3

Across the whole interval (−∞,3)(-\infty, 3), ff keeps decreasing while its concavity switches from up to down to up at the inflection points x=0\fpp{x = 0} and x=2\fpp{x = 2}.

Summary

Concavity and the second derivative
xx
concave up\htmlClass{lc-blue}{\text{concave up}}
decreasing\text{decreasing}
increasing\text{increasing}
f′′(x)>0  ⟹  f′ increasing  ⟹  f concave up\displaystyle \begin{aligned}&\fpp{f''(x)} > 0 \implies \fpr{f'} \text{ increasing} \\ &\implies \cvu{f \text{ concave up}}\end{aligned}
f′′(x)<0  ⟹  f′ decreasing  ⟹  f concave down\displaystyle \begin{aligned}&\fpp{f''(x)} < 0 \implies \fpr{f'} \text{ decreasing} \\ &\implies \cvd{f \text{ concave down}}\end{aligned}

A function is concave up where its derivative f′\fpr{f'} is increasing, and concave down where f′\fpr{f'} is decreasing. We find where f′\fpr{f'} is increasing from the sign of its derivative, f′′\fpp{f''}. Where f′′\fpp{f''} is positive, ff is concave up. Where f′′\fpp{f''} is negative, ff is concave down.

A function doesn't have to be increasing to be concave up. For example, x22\frac{x^2}{2} decreases on [−2,0][-2, 0] and increases on [0,2][0, 2], but f′′(x)=1\fpp{f''(x) = 1} is positive everywhere, so the curve is concave up on all of [−2,2]\cvu{[-2, 2]}.

Finding inflection points
x3: f′′(x)=6x,−∣+ at 0 → (0,0)\displaystyle \begin{aligned}&x^3: \ \fpp{f''(x) = 6x}, \\ &\quad \cvd{-} \mid \cvu{+} \text{ at } 0 \ \to \ \fpp{(0, 0)}\end{aligned}
x4: f′′(x)=12x2,+∣+ at 0 → none\displaystyle \begin{aligned}&x^4: \ \fpp{f''(x) = 12x^2}, \\ &\quad \cvu{+} \mid \cvu{+} \text{ at } \mut{0} \ \to \ \text{none}\end{aligned}

An inflection point is a point (c,f(c))\fpp{(c, f(c))} on the graph where the concavity changes. Points where f′′(x)=0\fpp{f''(x) = 0} or where f′′(x)\fpp{f''(x)} is undefined are candidates, but a candidate is an inflection point only if f′′\fpp{f''} changes sign across it.

For x3x^3, f′′(x)=6x\fpp{f''(x) = 6x} changes sign at 00, so (0,0)\fpp{(0, 0)} is an inflection point. For x4x^4, f′′(0)=0\fpp{f''(0) = 0}, but f′′(x)=12x2\fpp{f''(x) = 12x^2} is positive on both sides of 00, so there's no inflection point.

Reading graphs of f′\fpr{f'} and f′′\fpp{f''}
ff
f′\htmlClass{lc-green}{f'}

Where the graph of f′\fpr{f'} is above the x-axis, ff is increasing, and where it is below, ff is decreasing. A zero where f′\fpr{f'} crosses the x-axis gives a local maximum or minimum of ff.

Where the graph of f′\fpr{f'} is increasing, ff is concave up, and where it is decreasing, ff is concave down. A local maximum or minimum of f′\fpr{f'} gives an inflection point of ff. If you're given the graph of f′′\fpp{f''} instead, look for a change of sign, including across a point where it is undefined. If that graph only touches the x-axis without changing sign, there's no inflection point.

Concavity and tangent lines

Where a curve is concave up, it lies above its tangent lines, so a tangent-line estimate is too small. Where a curve is concave down, it lies below its tangent lines, so a tangent-line estimate is too big. For example, f(x)=xf(x) = \sqrt{x} has f′′(x)<0\fpp{f''(x)} < 0 for positive xx, so the curve lies below its tangent line at x=4x = 4, and the tangent estimate of 2.05\tng{2.05} is greater than 4.2\sqrt{4.2}.

The second derivative test
f′(c)=0, f′′(c)>0  ⟹  local min\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c)} > 0 \\ &\implies \fpr{\text{local min}}\end{aligned}
f′(c)=0, f′′(c)<0  ⟹  local max\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c)} < 0 \\ &\implies \fpr{\text{local max}}\end{aligned}
f′(c)=0, f′′(c)=0  ⟹  first derivative test\displaystyle \begin{aligned}&\fpr{f'(c) = 0}, \ \fpp{f''(c) = 0} \\ &\implies \text{first derivative test}\end{aligned}

At a critical point where f′(c)=0\fpr{f'(c) = 0}, the tangent line is horizontal. If f′′(c)>0\fpp{f''(c)} > 0, the curve is concave up and lies above the tangent line, so cc is a local minimum. If f′′(c)<0\fpp{f''(c)} < 0, the curve is concave down and lies below the tangent line, so cc is a local maximum.

If f′′(c)=0\fpp{f''(c) = 0}, the test gives no conclusion. The point can be a local minimum (like x4x^4 at 00), a local maximum (like −x4-x^4 at 00), or neither (like x3x^3 at 00). Use the first derivative test to check the signs of f′\fpr{f'} on both sides of cc.

8

Practice

Give intervals as open intervals, points as (x,y)(x, y), and values in exact form.

1
For f(x)=x3−6x2+5f(x) = x^3 - 6x^2 + 5, find the intervals on which ff is concave up and concave down, and find the inflection point.
concave up\text{concave up}
concave down\text{concave down}

Does concavity come from the sign of f′f' or the sign of f′′f''?

Where is f′′(x)=6(x−2)f''(x) = 6(x - 2) negative, and where is it positive?

If you got (0,5)(0, 5) and (4,−27)(4, -27) Check which derivative you set to 00. Inflection points come from where f′′f'' changes sign, not from the zeros of f′f'.

Concavity comes from the sign of the second derivative, so we differentiate twice.
f′(x)=3x2−12x,f′′(x)=6x−12=6(x−2)\displaystyle \fpr{f'(x) = 3x^2 - 12x}, \qquad \fpp{f''(x) = 6x - 12 = 6(x - 2)}
Check the sign of f′′\fpp{f''} on either side of x=2\fpp{x = 2}.
x<2: f′′(x)<0 → concave down on (−∞,2);x>2: f′′(x)>0 → concave up on (2,∞)\displaystyle x < 2: \ \fpp{f''(x)} < 0 \ \to \ \cvd{\text{concave down on } (-\infty, 2)}; \qquad x > 2: \ \fpp{f''(x)} > 0 \ \to \ \cvu{\text{concave up on } (2, \infty)}
The second derivative changes sign at x=2\fpp{x = 2}, so plug 22 into f(x)f(x) to get the y-coordinate of the inflection point.
f(2)\displaystyle f(2) =8−24+5\displaystyle = 8 - 24 + 5 =−11 → (2,−11)\displaystyle = -11 \ \to \ \fpp{(2, -11)}
2
Find all inflection points of f(x)=3x5−5x4f(x) = 3x^5 - 5x^4.
inflection points\text{inflection points}

Where is f′′(x)f''(x) equal to 00?

For f′′(x)=60x2(x−1)f''(x) = 60x^2(x - 1), what is the sign of f′′f'' just to the left of 00, and just to the right of 00?

If you got (0,0)(0, 0) and (1,−2)(1, -2) Check the sign of f′′f'' on both sides of x=0x = 0. An inflection point requires f′′f'' to change sign, not just equal 00.

Find the candidatesTake the derivative twice and factor f′′\fpp{f''} to find where it equals 00.
f′(x)=15x4−20x3,f′′(x)=60x3−60x2=60x2(x−1)\displaystyle \fpr{f'(x) = 15x^4 - 20x^3}, \qquad \fpp{f''(x) = 60x^3 - 60x^2 = 60x^2(x - 1)}
Test the sign of f′′\fpp{f''}The factor 60x260x^2 is positive for every non-zero xx, so the sign of f′′\fpp{f''} depends entirely on x−1\fpp{x - 1}.
x<0: f′′<0;0<x<1: f′′<0;x>1: f′′>0\displaystyle x < 0: \ \fpp{f''} < 0; \qquad 0 < x < 1: \ \fpp{f''} < 0; \qquad x > 1: \ \fpp{f''} > 0
The sign changes only at x=1\fpp{x = 1}, so plug 11 into ff to get the y-coordinate.
f(1)\displaystyle f(1) =3−5\displaystyle = 3 - 5 =−2 → (1,−2)\displaystyle = -2 \ \to \ \fpp{(1, -2)}
3
The graph of f′f', the derivative of a function ff, is shown for −3≤x≤3-3 \le x \le 3. (a) On what intervals is ff concave up? (b) At what values of xx does ff have an inflection point? (c) At what value of xx does ff have a relative maximum?
(a) concave up\text{(a) concave up}

On (−3,−2)(-3, -2), is the graph of f′f' increasing or decreasing?

Where does the graph of f′f' have a local maximum or a local minimum?

Where does f′f' change from positive to negative?

If you got (−3,0)(-3, 0) in (a) Check whether you looked at the sign of f′f' or where f′f' is increasing. The function ff is concave up where f′f' is increasing, not where f′f' is positive.

If you got x=0x = 0 in (b) Check what f′(0)=0f'(0) = 0 means for ff. A zero of f′f' gives a critical point of ff, while inflection points happen where f′f' has a local maximum or minimum.

Find where ff is concave upThe second derivative is the slope of f′\fpr{f'}, so ff is concave up where f′\fpr{f'} is increasing.
f′ increasing on (−3,−2) and (2,3), decreasing on (−2,2) → f concave up on (−3,−2) and (2,3)\displaystyle \fpr{f'} \text{ increasing on } (-3, -2) \text{ and } (2, 3), \ \text{decreasing on } (-2, 2) \ \to \ \cvu{f \text{ concave up on } (-3, -2) \text{ and } (2, 3)}
Find the inflection pointsThe concavity changes where f′\fpr{f'} changes from increasing to decreasing, or decreasing to increasing.
f′: local max at x=−2, local min at x=2 → inflection points at x=−2, x=2\displaystyle \fpr{f'} \text{: }\fpp{\text{local max at } x = -2, \ \text{local min at } x = 2} \ \to \ \fpp{\text{inflection points at } x = -2, \ x = 2}
Find the relative maximumBy the first derivative test, ff has a relative maximum where f′\fpr{f'} changes from positive to negative.
f′>0 on (−3,0), f′<0 on (0,3) → relative max at x=0\displaystyle \fpr{f'} > 0 \text{ on } (-3, 0), \ \fpr{f'} < 0 \text{ on } (0, 3) \ \to \ \fpr{\text{relative max at } x = 0}
4
The function ff is twice differentiable. (a) f′(3)=0f'(3) = 0 and f′′(3)=−2f''(3) = -2. Does ff have a relative maximum, a relative minimum, or neither at x=3x = 3? Give a reason. (b) f′(5)=0f'(5) = 0 and f′′(5)=0f''(5) = 0. What can you conclude about ff at x=5x = 5?
(a) at x=3\text{(a) at }x=3
(b) at x=5\text{(b) at }x=5

Because f′(3)=0f'(3) = 0, the tangent line is horizontal. Does f′′(3)<0f''(3) < 0 put the curve above or below that horizontal tangent near 33?

When f′′(5)=0f''(5) = 0, what does the second derivative test tell you?

If you got relative minimum in (a) Check what a negative second derivative means for concavity. When f′′(3)<0f''(3) < 0, the curve is concave down, so it lies below its horizontal tangent line.

If you got neither in (b) Check whether f′′(5)=0f''(5) = 0 rules out a maximum or a minimum. For x4x^4, both derivatives are 00 at x=0x = 0, but the function still has a relative minimum there.

Classify x=3x = 3The tangent line at x=3x = 3 is horizontal, and f′′(3)<0\fpp{f''(3)} < 0 means ff is concave down, so the curve lies below the tangent line near 33.
f′(3)=0 → tangent y=f(3);f′′(3)=−2<0 → concave down → relative max at x=3\displaystyle \fpr{f'(3) = 0} \ \to \ \tng{\text{tangent } y = f(3)}; \qquad \fpp{f''(3) = -2} < 0 \ \to \ \cvd{\text{concave down}} \ \to \ \fpr{\text{relative max at } x = 3}
Test x=5x = 5When both f′(5)=0\fpr{f'(5) = 0} and f′′(5)=0\fpp{f''(5) = 0}, the second derivative test gives no conclusion. For x4x^4, −x4-x^4, and x3x^3, both derivatives are 00 at 00, but one has a minimum, one has a maximum, and one has neither, so you cannot conclude anything about x=5x = 5 from these two values alone.
f′(0)=0, f′′(0)=0:x4 min,−x4 max,x3 neither\displaystyle \fpr{f'(0) = 0}, \ \fpp{f''(0) = 0}: \quad x^4 \fpr{\text{ min}}, \quad -x^4 \fpr{\text{ max}}, \quad x^3 \text{ neither}
5
The graph of f(x)=ax3+bx2f(x) = ax^3 + bx^2 has an inflection point at (1,2)(1, 2). Find aa and bb.

If the graph has an inflection point at x=1x = 1, which derivative must be 00 there?

The point (1,2)(1, 2) lies on the curve, so what is f(1)f(1)?

If you got a=−4a = -4, b=6b = 6 Check which derivative is 00 at an inflection point. Setting f′(1)=0f'(1) = 0 gives a horizontal tangent, while an inflection point requires f′′(1)=0f''(1) = 0.

Find the first two derivatives in terms of aa and bb.
f′(x)=3ax2+2bx,f′′(x)=6ax+2b\displaystyle \fpr{f'(x) = 3ax^2 + 2bx}, \qquad \fpp{f''(x) = 6ax + 2b}
An inflection point at x=1x = 1 requires f′′(1)=0\fpp{f''(1) = 0}, and since (1,2)\fpp{(1, 2)} is on the graph, f(1)=2f(1) = 2.
f′′(1)=6a+2b=0,f(1)\displaystyle \fpp{f''(1) = 6a + 2b = 0}, \qquad f(1) =a+b\displaystyle = a + b =2\displaystyle = 2
b\displaystyle b =−3a → −2a\displaystyle = -3a \ \to \ -2a =2 → a\displaystyle = 2 \ \to \ a =−1, b\displaystyle = -1, \ b =3\displaystyle = 3
Check that f′′\fpp{f''} actually changes sign across x=1x = 1.
f′′(x)=−6x+6:f′′>0 for x<1, f′′<0 for x>1\displaystyle \fpp{f''(x) = -6x + 6}: \quad \fpp{f''} > 0 \text{ for } x < 1, \ \fpp{f''} < 0 \text{ for } x > 1
6
For f(x)=xe−xf(x) = x e^{-x}, find the intervals on which ff is increasing and decreasing, the relative extrema, the intervals on which ff is concave up and concave down, and the inflection points.
increasing\text{increasing}
decreasing\text{decreasing}
relative minimum\text{relative minimum}
concave up\text{concave up}
concave down\text{concave down}

Using the product rule gives f′(x)=(1−x)e−xf'(x) = (1 - x)e^{-x}. Because e−xe^{-x} is always positive, what determines the sign of f′f'?

Differentiate (1−x)e−x(1 - x)e^{-x} with the product rule to find f′′(x)f''(x).

If you got concave up on (−∞,2)(-\infty, 2) Check the signs when differentiating (1−x)e−x(1 - x)e^{-x}. Distributing the minus sign gives f′′(x)=(x−2)e−xf''(x) = (x - 2)e^{-x}, not (2−x)e−x(2 - x)e^{-x}.

If you got an inflection point at (1,1e)\left(1, \frac1e\right) Check whether x=1x = 1 is a zero of f′f' or f′′f''. The point (1,1e)\left(1, \frac1e\right) is a relative maximum where f′(1)=0f'(1) = 0, while inflection points come from where f′′f'' changes sign.

Increasing, decreasing, and extremaUse the product rule on xe−xx e^{-x}, then use the sign of 1−x\fpr{1 - x} to find where ff increases and decreases.
f′(x)=e−x−xe−x=(1−x)e−x\displaystyle \fpr{f'(x) = e^{-x} - x e^{-x} = (1 - x)e^{-x}}
e−x>0:x<1: f′>0;x>1: f′<0 → increasing on (−∞,1), decreasing on (1,∞)\displaystyle e^{-x} > 0: \quad x < 1: \ \fpr{f'} > 0; \quad x > 1: \ \fpr{f'} < 0 \ \to \ \text{increasing on } (-\infty, 1), \ \text{decreasing on } (1, \infty)
f(1)\displaystyle f(1) =1e → relative max (1,1e)\displaystyle = \frac1e \ \to \ \fpr{\text{relative max } \left(1, \frac1e\right)}
Concavity and inflection pointDifferentiate f′\fpr{f'} with the product rule, then use the sign of x−2\fpp{x - 2} to find concavity and the inflection point.
f′′(x)=−e−x−(1−x)e−x=(x−2)e−x\displaystyle \fpp{f''(x) = -e^{-x} - (1 - x)e^{-x} = (x - 2)e^{-x}}
x<2: f′′<0;x>2: f′′>0 → concave down on (−∞,2), concave up on (2,∞)\displaystyle x < 2: \ \fpp{f''} < 0; \quad x > 2: \ \fpp{f''} > 0 \ \to \ \cvd{\text{concave down on } (-\infty, 2)}, \ \cvu{\text{concave up on } (2, \infty)}
f(2)\displaystyle f(2) =2e2 → inflection point (2,2e2)\displaystyle = \frac{2}{e^2} \ \to \ \fpp{\text{inflection point } \left(2, \frac{2}{e^2}\right)}
7
Find all inflection points of f(x)=x5/3−5x2/3f(x) = x^{5/3} - 5x^{2/3}. (Challenge)
inflection points\text{inflection points}

For f′′(x)=109x−4/3(x+1)f''(x) = \frac{10}{9}x^{-4/3}(x + 1), where does f′′f'' equal 00, and where is it undefined?

What is the sign of x−4/3=1(x3)4x^{-4/3} = \frac{1}{(\sqrt[3]{x})^4} for all x≠0x \neq 0?

If you got (−1,−6)(-1, -6) and (0,0)(0, 0) Check whether the sign of f′′f'' changes at x=0x = 0. Because f′′(0)f''(0) is undefined, x=0x = 0 is only a candidate. The second derivative is positive on both sides of 00, so concavity does not change across 00 and (0,0)(0, 0) is not an inflection point.

Use the power rule twice, then factor out the smaller power, x−4/3x^{-4/3}.
f′(x)=53x23−103x−13,f′′(x)=109x−13+109x−43=109x−43(x+1)\displaystyle \fpr{f'(x) = \frac53 x^{\frac23} - \frac{10}{3} x^{-\frac13}}, \qquad \fpp{f''(x) = \frac{10}{9} x^{-\frac13} + \frac{10}{9} x^{-\frac43} = \frac{10}{9} x^{-\frac43}(x + 1)}
Find the candidatesThe second derivative can change sign where it is 00 or undefined.
f′′(−1)=0,f′′(0) undefined\displaystyle \fpp{f''(-1) = 0}, \qquad \fpp{f''(0) \text{ undefined}}
Test the sign of f′′\fpp{f''}Because the power 44 is even, x−4/3x^{-4/3} is positive for every non-zero xx, so the sign of f′′\fpp{f''} matches the sign of x+1\fpp{x + 1}.
x−43\displaystyle x^{-\frac43} =1(x3)4>0 for x≠0 → sign of f′′\displaystyle = \frac{1}{\left(\sqrt[3]{x}\right)^4} > 0 \text{ for } x \neq 0 \ \to \ \text{sign of } \fpp{f''} =sign of x+1\displaystyle = \text{sign of } \fpp{x + 1}
x=−1: −∣+;x=0: +∣+\displaystyle \fpp{x = -1}: \ \cvd{-} \mid \cvu{+}; \qquad \mut{x = 0}: \ \cvu{+} \mid \cvu{+}
The only sign change is at x=−1\fpp{x = -1}, so plug −1-1 back into f(x)f(x) for the y-coordinate.
f(−1)\displaystyle f(-1) =(−1)53−5(−1)23\displaystyle = (-1)^{\frac53} - 5(-1)^{\frac23} =−1−5\displaystyle = -1 - 5 =−6 → (−1,−6)\displaystyle = -6 \ \to \ \fpp{(-1, -6)}