Working with the Second Derivative for Inflection Points
You take the second derivative, set it equal to zero, and check where the sign changes. That part is in every textbook. The part they don't always emphasize is that setting the second derivative to zero gets you candidates, not answers. You have to verify the sign change on both sides of each candidate point, and you also have to check whether the original function is actually defined there. I've seen students lose points on exams by skipping either step. Here is how I'd actually approach it in practice. Start with your function, f(x). Compute f''(x). Find all x-values where f''(x) = 0 or where f''(x) is undefined. These are your critical values for the second derivative. Then pick test points immediately to the left and right of each critical value and evaluate the sign of f'' at those points. If the sign flips from positive to negative or negative to positive, you have an inflection point at that x-value. The y-coordinate comes from plugging back into the original f(x).
How To Use Second Derivative To Find Inflection Point in Real Problems
Let me walk through a specific example that trips people up. Take f(x) = x^4 - 4x^3. The first derivative is f'(x) = 4x^3 - 12x^2. The second derivative is f''(x) = 12x^2 - 24x. Setting that to zero gives 12x(x - 2) = 0, so your candidates are x = 0 and x = 2. Now check the sign around each one. Around x = 0: plug in x = -1, and f''(-1) = 12 + 24 = 36, which is positive. Plug in x = 0.5, and f''(0.5) = 12(0.25) - 24(0.5) = 3 - 12 = -9, negative. Sign changed from positive to negative, so there is an inflection point at x = 0. The point is (0, 0). Around x = 2: plug in x = 1, and f''(1) = 12 - 24 = -12, negative. Plug in x = 3, and f''(3) = 108 - 72 = 36, positive. Sign changed from negative to positive, so there is an inflection point at x = 2. The point is (2, -16).
I ran into a case recently where the second derivative was zero at a point but did not change sign. The function was f(x) = x^4. The second derivative is f''(x) = 12x^2, which equals zero at x = 0. But on both sides of zero, f''(x) stays positive. This is not an inflection point. It is just a point where the concavity flattens out momentarily before continuing in the same direction. I made this mistake on a preliminary exam back when I was learning this, and I have been careful about it ever since. Always verify the sign change. The zero alone means nothing. Another thing textbooks skim over is what happens when the second derivative does not exist. Consider f(x) = x^(1/3). The first derivative is (1/3)x^(-2/3), and the second derivative is (-2/9)x^(-5/3), which is undefined at x = 0. But the original function is defined there, and the concavity does switch from negative on the left to positive on the right. So x = 0 is still an inflection point even though f''(0) does not exist. Any time you see a vertical tangent or a cusp, check it anyway. Don't automatically discard points where the second derivative is undefined. Here is a practical tip that saves time: when you are working with polynomial functions, factor the second derivative completely before you do anything else. In the example above, factoring 12x^2 - 24x into 12x(x - 2) made the candidates obvious. If you try to use the quadratic formula on messier polynomials, you will waste minutes doing arithmetic that factoring would have skipped in thirty seconds. I usually factor first, then compute, and I have cut my problem set time roughly in half compared to how I used to work through it.
Get the Full Details
There are cases where this method hits a wall. If your function involves a combination of trigonometric, exponential, and polynomial terms, the second derivative can become algebraically unwieldy. I worked on a problem once where f(x) = e^x sin(x), and f''(x) expanded to 2e^x cos(x). Setting that equal to zero means cos(x) = 0, which gives x = pi/2 + n*pi for any integer n. You can solve that, but you now have infinitely many inflection points to deal with. In those situations, you typically describe the pattern rather than listing each one individually. A more annoying edge case comes up when the second derivative is a complicated rational expression. I spent about twenty minutes simplifying f''(x) for a function that was a quotient of two polynomials, only to realize I could have just looked at the sign of the numerator and denominator separately without fully simplifying. The zeros of the numerator give you where f''(x) = 0, and the zeros of the denominator give you where f''(x) is undefined. You do not need the full simplified form. Just track the signs of each factor. This shortcut usually reduces a five-minute algebra slog to about forty seconds. One more thing worth noting: an inflection point requires a sign change in the second derivative, but a sign change does not guarantee the point is on the graph of the original function. I encountered this on a homework problem where f''(x) changed sign at x = 1, but f(1) was undefined because the original function had a vertical asymptote there. There was no inflection point. The concavity changed, sure, but there is no point to call an inflection point if the function does not exist at that x-value. Always confirm f(x) is defined before you declare victory.
If you want a quick reference sheet, I keep a one-page summary on my desk that lists the steps in order, along with a checklist for the common pitfalls: did you factor completely, did you check both sides of each candidate, did you verify the original function is defined, did you consider where the second derivative is undefined rather than just where it equals zero. Following that checklist consistently has kept my error rate very low on graded assignments and exams. The method itself is straightforward. The trick is in the details, and the details are where people lose points. Find f''(x), locate where it is zero or undefined, test the sign on both sides, confirm the function exists at that point, and move on. Do not skip the verification step. It is the difference between a correct answer and a partially correct one.
