The first derivative describes slope. The second derivative describes how that slope changes. Concavity is therefore about increasing or decreasing slopes, not about whether the graph itself is rising or falling.
Concavity is a story about slopes
If on an interval, is concave up there: tangent slopes increase as increases. They might progress from to to 0, so a decreasing graph can still be concave up. If tangent slopes decrease and the graph is concave down.
Think of a car's position: velocity is the first derivative and acceleration is the second. Positive acceleration means velocity increases, even when velocity is initially negative. The same distinction separates increasing function values from increasing slopes.
An inflection point is a point on a continuous graph where concavity changes. Candidates occur where or the second derivative does not exist, but a candidate must be checked on both sides.
Example: locate a genuine inflection point
For we have and When the second derivative is negative; when it is positive. Concavity changes from down to up at zero, and so is an inflection point.
The slope there is not zero. An inflection point need not have a horizontal tangent. It marks a change in bending, not necessarily a turning point of the function.
For equals zero at zero but is positive on both sides. There is no concavity change, so zero is not an inflection point. This is why solving is only the beginning.
At the origin the cubic changes from concave down to concave up, even though its slope is negative.
Describe this figure
Function values for the solid curve:
| x | f(x) |
|---|---|
| -2 | -2 |
| -1 | 2 |
| 0 | 0 |
| 1 | -2 |
| 2 | 2 |
The second derivative test classifies stationary points
Suppose and the second derivative exists with the usual local smoothness. If then is a local minimum; if then is a local maximum. Near a stationary point, increasing slopes pass from negative toward positive, while decreasing slopes pass from positive toward negative.
If the test is inconclusive. It does not establish a minimum, maximum, or inflection point. Use a first derivative sign chart or another argument.
Example: a decisive test and an inconclusive one
Let Then so the stationary points occur at 1 and 3. Since we have and Thus there is a local maximum at and a local minimum at
The second derivative changes sign at 2, so the inflection point is Its location lies between the stationary points, but that placement is a feature of this example, not a universal rule.
For at zero, both and are zero. The second derivative test says nothing. Yet proves a minimum. For the same two zero derivatives occur with no extremum. An inconclusive test can accompany different outcomes.
Use concavity to assess tangent approximations
Where a differentiable function is concave up, its graph lies above its tangent lines; tangent-line estimates are low. Where it is concave down, the estimates are high. Check concavity on the relevant interval, not just at a distant point.
On a graph of look for rising or falling portions to determine concavity of . Looking at whether lies above zero instead determines whether increases, a different question.
Try it yourself
For find the inflection points.
Answer: It is positive outside and negative inside, so concavity changes at both and 1. Since the inflection points are and
What to remember
Concavity tracks changes in slope. Test for an actual sign change at inflection candidates, and treat a zero second derivative in the extremum test as inconclusive.