Calculus › Section D · Applications of derivatives › Topic D.4

Concavity and the second derivative test

A curve can rise while steepening, or rise while flattening. Concavity is the study of how a graph bends, and the second derivative answers it as crisply as the first derivative answered rise and fall.

Two ways to climb

Both these curves rise: a skateboard ramp that gets steeper as you go, and a hill that levels off near its crest. Both are increasing, yet they feel utterly different, and the difference matters: the ramp is accelerating its climb, the hill is decelerating. The last topic measured direction with \(f'\). This one measures bend with \(f''\), the rate of change of the slope itself. Knowing direction and bend together is most of what "the shape of the graph" means. It is what you need to sketch curves faithfully, and it gives a faster way to classify turning points than the sign chart.

Concavity, defined by the tangent slopes

Let \(f\) be differentiable on an open interval. Its graph is concave up there when \(f'\) is increasing: scanning left to right, each tangent slope is larger than the last, and the curve cups upward like a bowl that holds water. Concave down is the mirror: \(f'\) decreasing, the curve capping over like an upturned bowl that spills it. A second way to hold the two: a concave-up curve lies above each of its tangent lines, touching only at the point of contact, while a concave-down curve lies below them. Note that "slopes increasing" can mean a negative slope becoming less negative, so a falling curve can be concave up.

The second derivative decides

Concave up meant the slopes are increasing, and the last topic taught us how to test whether a function is increasing: look at its derivative's sign. Apply that test to \(f'\) itself. If \(f'' > 0\) throughout an interval, then \(f'\) is increasing there, so the graph is concave up; if \(f'' < 0\), concave down. The whole thing is the increasing-decreasing test one rung up, which is why the sign-chart machinery from D.3 transfers unchanged. For the cubic \(f(x) = x^3 - 3x\) we have been following, \(f''(x) = 6x\): negative for negative \(x\), the graph caps, and positive beyond \(0\), where the graph cups.

Inflection points

A point on the graph where \(f\) is continuous, has a tangent line, and the concavity actually changes, up to down or down to up, is an inflection point. Geometrically, the curve crosses its own tangent line there, switching from above it to below. In our examples \(f''\) is continuous wherever it exists, so a sign change requires \(f'' = 0\) or \(f''\) undefined, and those points are the suspects. But a suspect is not a culprit: for \(f(x) = x^4\), \(f''(x) = 12x^2\) is zero at \(0\) yet never changes sign. The graph stays concave up on both sides, so there is no inflection. Confirm the sign change with a test point on each side, exactly as in D.3.

Worked example: two inflection points

Find the concavity map and inflection points of \(f(x) = x^4 - 4x^3 + 2\). Differentiate twice: \(f'(x) = 4x^3 - 12x^2\), then \(f''(x) = 12x^2 - 24x = 12x(x - 2)\). The suspects are \(x = 0\) and \(x = 2\). Test the intervals: at \(x = -1\), \(f'' = 36 > 0\), concave up; at \(x = 1\), \(f'' = -12 < 0\), concave down; at \(x = 3\), \(f'' = 36 > 0\), concave up again. Both suspects flip the sign, so both are inflection points: \((0, 2)\) and \((2, -14)\), since \(f(2) = 16 - 32 + 2 = -14\). The quartic cups, caps, and cups again.

The second derivative test for extrema

Concavity gives a faster way to classify the level tangents from D.1. Suppose \(f'(c) = 0\) and \(f''\) exists on an open interval around \(c\). If \(f''(c) > 0\), the graph is cupped at \(c\), so the level point sits at the bottom of the cup: a local minimum. In sign-chart terms, \(f'\) is increasing through \(c\), so it goes from negative to positive. If \(f''(c) < 0\), capped: a local maximum. If \(f''(c) = 0\), the test gives no information, and you go back to the first derivative test. The second derivative test trades a whole sign chart for one evaluation, a bargain when \(f''\) is easy to compute.

Worked example: the test on our cubic

The critical numbers of \(f(x) = x^3 - 3x\) are \(x = \pm 1\). With \(f''(x) = 6x\): \(f''(-1) = -6 < 0\), capped, so \(x = -1\) is a local maximum; \(f''(1) = 6 > 0\), cupped, a local minimum. Both verdicts match the sign-chart conclusions from D.3, found this time with two evaluations instead of a chart. And notice the division of labour now complete: \(f'\) said where the curve turns, \(f''\) said which way, and the inflection point at \((0, 0)\) marks where the cubic switches from cap-shaped to cup-shaped halfway through its S.

The point of diminishing returns

Suppose a firm models weekly sales \(S\) (in units) against advertising spend \(x\) (in thousands of dollars) by \(S(x) = 30x^2 - x^3\) for \(0 \le x \le 20\). Then \(S'(x) = 60x - 3x^2 = 3x(20 - x)\) is positive, so more advertising always brings more sales in this range. But \(S''(x) = 60 - 6x\) is positive for \(x < 10\) and negative for \(x > 10\). Below 10, each extra thousand dollars buys more extra sales than the one before; above 10, it buys fewer. The inflection point at \(x = 10\), where \(S = 3000 - 1000 = 2000\) units, is called the point of diminishing returns. Spending beyond it still helps, but less and less.

Common mistakes to avoid

The recurring slips: concluding an inflection point from \(f''(c) = 0\) alone, when a sign change is the requirement and \(x^4\) at \(0\) shows why; mixing up the two uses of \(f''\): concavity uses its sign over intervals, while the extremum test uses its value at one point where \(f' = 0\); claiming the second derivative test "shows neither max nor min" when \(f''(c) = 0\), when it shows nothing and you must fall back (\(x^4\) at 0 is a minimum with \(f''(0) = 0\)); confusing "concave down" with "decreasing", when rising curves can cap and falling curves can cup; and dropping a factor when expanding \(f''\), so chart from the factored form where possible.

Key takeaways

Concavity is the bend of the graph, measured by the second derivative: \(f'' > 0\) cups upward, \(f'' < 0\) caps downward, because that is the increasing-decreasing test applied to the slopes. Inflection points are the genuine flips of concavity, hunted at \(f'' = 0\) or undefined and confirmed by sign change. At a level tangent, the second derivative test classifies the point in one evaluation: positive means valley, negative means summit, zero means go back to the first derivative test. Direction from \(f'\), bend from \(f''\): together with the vertical asymptotes of B.5 and the horizontal asymptotes of the next topic, that is nearly the complete anatomy of a curve, assembled in D.6.

Flashcards

Flashcard

Concave up

Slopes increasing; the graph cups

f″ > 0 on the interval.

Test your understanding

Question 1 of 5

For f(x) = x³ − 3x, where is the graph concave up?

Question 2 of 5

What is the inflection point of f(x) = x³ − 3x?

Question 3 of 5

At a critical number c with f′(c) = 0 and f″(c) < 0, the graph has what?

Question 4 of 5

What should you do when f′(c) = 0 and f″(c) = 0?

Question 5 of 5

For f(x) = x⁴, we find f″(0) = 0. What is true at x = 0?

Standards: 5

← Previous topicNext topic →