What Is the Second Derivative Test?
At a critical point c where f′(c) = 0, the second derivative test classifies the point from a single number, the concavity there:
f″(c) > 0: local minimum; f″(c) < 0: local maximum; f″(c) = 0: the test is silent
The intuition is geometric: a flat tangent on an upward-cupped (concave up) stretch must sit at the bottom of the cup; a flat tangent under a downward cap sits at the top. One derivative evaluation replaces the sign-chart legwork of the first derivative test, at the price of an honest failure mode: when f″(c) = 0, anything can happen (x⁴ has a minimum, −x⁴ a maximum, x³ neither), and a different tool must decide. This calculator finds the critical points symbolically, evaluates f″ at each, delivers the verdicts with the inconclusive cases flagged and explained, and plots f with f″ dashed so the sign being used is visible.
How to Use the Second Derivative Test Calculator
Enter a function and a search window. The steps display f′ and the critical points, then f″ and its value at each candidate with the classification: negative curvature ⇒ peak, positive ⇒ valley, zero ⇒ fall back to the first derivative test (the tool says so explicitly rather than guessing). The graph marks maxima red, minima green, and unresolved points amber, with the dashed f″ curve showing exactly which sign was read at each. The concavity calculator explores the f″ sign globally; this page uses it only at the special points where f′ vanishes.
Worked Example
Classify the critical points of f(x) = x³ − 6x² + 9x + 1. Here f′(x) = 3x² − 12x + 9 = 3(x − 1)(x − 3), so the candidates are x = 1 and x = 3. The second derivative is f″(x) = 6x − 12:
f″(1) = −6 < 0 ⇒ local maximum at (1, 5); f″(3) = 6 > 0 ⇒ local minimum at (3, 1)
Two evaluations, two verdicts, done. The x⁴ chip shows the limitation on purpose: its only critical point is x = 0, where f″(0) = 0, and the test refuses to answer even though the origin is an obvious minimum, the sign chart of f′ = 4x³ (negative then positive) settles it in seconds. Speed when decisive, silence when not: that is the test's character.
Why It Works, and Why It Sometimes Cannot
Near a critical point, Taylor expansion gives f(c + h) ≈ f(c) + ½f″(c)h², since the linear term vanishes. The local behavior is a parabola whose direction is the sign of f″(c): opening upward for positive (so c is a min), downward for negative. When f″(c) = 0 too, the parabola degenerates and the classification passes to the first nonzero higher derivative: the higher-order test says that if the first nonvanishing derivative at c is of even order, c is an extremum (min for positive, max for negative), and of odd order, an inflection-type point with no extremum. x⁴'s first nonzero derivative at 0 is the fourth (24 > 0, minimum); x³'s is the third (odd, no extremum), exactly matching the sign-chart answers. The same Taylor logic scales to several variables, where f″ becomes the Hessian matrix and its definiteness plays the role of the sign, the multivariable second derivative test.
Common Mistakes to Avoid
- Reading the inequality backwards. Positive f″ means minimum, cup holds water; negative means maximum. The mnemonic "smile/frown" prevents the classic inversion.
- Treating f″(c) = 0 as "no extremum." Zero is silence, not a verdict; x⁴ minimizes at exactly such a point. Only a first-derivative sign chart (or higher-order derivatives) can rule.
- Evaluating f″ at non-critical points. The test applies only where f′(c) = 0. The sign of f″ elsewhere describes concavity, not extrema.
- Using it at corners or cusps. Where f′ does not exist, f″ certainly does not; those candidates belong to the first derivative test exclusively.
- Concluding global extremes. The verdicts are local. On an interval, endpoints must still be compared via the closed-interval method; a classified local max can lose globally.
Real-World Applications
The second derivative test is the workhorse of unconstrained optimization wherever second-order information is cheap. Physics identifies stable equilibria as minima of potential energy: V′ = 0 finds the equilibria, and the sign of V″ separates a marble in a bowl (stable, V″ > 0) from one on a dome (unstable), with V″ = 0 marking the delicate neutral cases that require deeper analysis. Statistics uses the test on log-likelihoods: a maximum-likelihood estimate is confirmed by negative second derivative, whose magnitude (the Fisher information) simultaneously quantifies the estimate's precision, classification and error bars from one number. In economics, second-order conditions certify that a first-order optimum is genuinely a profit peak rather than a trough. Numerical optimizers embody the test's logic continuously: Newton-type methods use second derivatives both to step and to verify curvature, and detecting the "inconclusive" flat cases is exactly when they switch to safeguarded strategies.
Frequently Asked Questions
What does the second derivative test conclude, exactly?
At a point with f′(c) = 0: if f″(c) > 0 the point is a strict local minimum, if f″(c) < 0 a strict local maximum, and if f″(c) = 0 nothing is concluded, the point may be a min, a max, or neither, and further analysis is required.
Why does concavity at a flat point decide the classification?
Taylor's theorem: with the linear term gone, f(c + h) − f(c) ≈ ½f″(c)h², a parabola in h. Its sign for all small h is the sign of f″(c), so the function locally sits above its value (minimum) or below it (maximum).
How often is the test inconclusive in practice?
Rarely for generic functions, since f′ and f″ sharing a zero is a coincidence of measure zero, but reliably in symmetric and stylized textbook problems (pure powers, perfect flat spots). The calculator flags each such case and names the fallback rather than leaving it blank.
What is the higher-order derivative test?
When f″(c) = 0, look for the first nonvanishing derivative at c. Even order: extremum, minimum if that derivative is positive, maximum if negative. Odd order: no extremum (a horizontal inflection). It generalizes the parabola picture to the first surviving Taylor term.
Which test should I use, first or second derivative?
Second when f″ is easy and turns out nonzero at the candidates: it is one evaluation per point. First when f″ is messy, zero, or nonexistent (corners): it always works. They never disagree where both apply, so use whichever is cheaper and cross-check when unsure.
Can the test classify endpoints of an interval?
No. Endpoints are typically not critical points (f′ need not vanish there), and the local parabola picture requires a two-sided neighborhood. Endpoint extremes are found by direct comparison of values, as the closed-interval method prescribes.
What does the magnitude of f″(c) mean?
Sharpness. Large |f″| means a narrow, steep cup or cap; small means a broad, shallow one. Quantitatively it is the curvature of the local parabola, and in statistics this exact magnitude (of a log-likelihood) becomes the Fisher information governing how precisely a peak locates a parameter.
Is f″(c) > 0 with f′(c) = 0 enough for a global minimum?
No, only local. x⁴ − 2x² has a local max at 0 flanked by two global minima; conversely a classified local min can sit far above values elsewhere. Global claims need either interval comparison or convexity of f on the whole domain (f″ ≥ 0 everywhere), which upgrades local to global.
How does this test extend to functions of two variables?
f″ becomes the Hessian matrix of second partials. At a critical point: positive definite Hessian (D > 0, fxx > 0) gives a local min, negative definite a local max, indefinite (D < 0) a saddle, and D = 0 is the inconclusive analogue. The determinant condition is the direct heir of the sign of f″.
What does the dashed f″ curve on the graph show?
The number the test consults: at each marked critical point, the dashed curve's height is f″ there, below the axis at maxima, above at minima, on the axis exactly at the inconclusive points. Zeros of the dashed curve elsewhere are inflection candidates, the concavity story between the extrema.