Units and Interpretation of the Second Derivative

What a Second Derivative Measures

A first derivative describes the instantaneous rate of change of a quantity with respect to an input. A second derivative describes how that first rate is changing. Keeping these two interpretations separate is essential: a positive second derivative does not necessarily mean that the modeled quantity is increasing, and a negative second derivative does not necessarily mean that it is decreasing.

Units help keep the distinction clear. If a model’s output is measured in one unit and its input in another, the first derivative has output-units per input-unit. The second derivative has output-units per input-unit squared. Those units describe the change in the rate, not the rate of the original quantity.

For example, if \(f\) measures temperature in degrees Celsius and \(x\) measures time in minutes, then \(f'(x)\) is measured in degrees Celsius per minute, and \(f''(x)\) is measured in degrees Celsius per minute squared. The second derivative tells us how quickly the temperature’s rate of change is itself changing.

Why the Units Have This Form

The derivative is defined by a limit of ratios. In a model, the numerator represents a change in output, while the denominator represents a change in input. Thus the units of \(f'\) are output-units divided by input-units. Applying the same reasoning to the function \(f'\), whose units are output-units per input-unit, gives the units of \(f''\).

This unit calculation follows directly from the derivative definition. If \(f\) has units \(U\) and \(x\) has units \(X\), then a difference quotient for \(f'\) has units \(U/X\). A difference quotient for \(f'\) has units \((U/X)/X=U/X^2\), as required for \(f''\).

The limits describe instantaneous rates. For small changes in the input, the second derivative also gives a useful local interpretation: a change of about \(\Delta x\) in the input corresponds to a change of about \(f''(a)\Delta x\) in the first derivative, when \(\Delta x\) is sufficiently small and the model behaves smoothly near \(a\). The units agree: \((U/X^2)X=U/X\), the units of the first derivative.

Worked Example: Temperature Changing Over Time

A Second Derivative Is Not the Rate Itself

A common error is to use the sign of \(f''\) to decide whether \(f\) is increasing or decreasing. The first derivative answers that question: \(f'(a)>0\) means the function has a positive instantaneous rate of change at \(a\), and \(f'(a)<0\) means it has a negative instantaneous rate of change there. The second derivative describes the rate of change of \(f'\).

What a Sign at One Point Does—and Does Not—Establish

At a point \(a\) where \(f''(a)>0\), the precise statement is that \(f'\) has a positive instantaneous rate of change at \(a\). That single value does not, by itself, show that \(f'\) is increasing throughout an interval around \(a\). To establish monotonicity on an interval, we need information about the second derivative throughout the interval.

This follows from the Mean Value Theorem applied to \(f'\). For any \(x_1<x_2\) in \([u,v]\), the theorem gives a point \(c\) between them such that the change in \(f'\) equals its derivative at \(c\) multiplied by the interval length. The sign of \(f''(c)\), together with \(x_2-x_1>0\), determines the sign of that change.

Thus, if \(f''>0\) throughout the interior, \(f'(x_2)-f'(x_1)>0\) whenever \(x_1<x_2\), so \(f'\) is strictly increasing. If \(f''<0\) throughout, the difference is negative, so \(f'\) is strictly decreasing. The condition concerns the whole interval between the two points, not just one endpoint.

Worked Example: Volume Rising While Its Rate Falls

A Common Interpretation Error

For the tank model, the incorrect calculation “\(V''(2)=-2\), so the volume is falling” treats the second derivative as if it were the first. It is not: the volume’s instantaneous rate is \(V'(2)=4\ \mathrm{L}/\mathrm{h}\), which is positive. The value \(V''(2)=-2\ \mathrm{L}/\mathrm{h}^2\) says that this rate has a negative instantaneous rate of change.

There is a second, subtler mistake: “\(f''(a)>0\), so \(f'\) must be increasing throughout a neighborhood of \(a\).” A derivative’s sign at one point alone does not justify that conclusion. For instance, define \(g(x)=x+x^2\sin(1/x^2)\) for \(x\neq 0\), and \(g(0)=0\). Let \(f\) be an antiderivative of \(g\), so \(f'=g\). Then \(f''(0)=g'(0)=1>0\), since \(\lim_{h\to 0}g(h)/h=1\). But for \(x_n=1/\sqrt{2\pi n}\), differentiation gives \(g'(x_n)=1-2/x_n<0\) for sufficiently large \(n\). Thus \(f'\) is not increasing on any neighborhood of zero. To conclude that \(f'\) is increasing on an interval, verify that \(f''>0\) throughout that interval.

Why Units and Interpretation Matter

Units are more than labels attached to a final answer. They help identify which quantity a derivative describes and can reveal an interpretation error before it spreads through a calculation. A rate measured in litres per hour cannot be confused dimensionally with a rate of that rate measured in litres per hour squared.

When interpreting a second derivative in a model, keep three questions separate:

• What quantity does the model measure, and in what units?
• What is the first derivative at the point, including its sign and units?
• What is the second derivative at the point, and does its sign apply only there or throughout an interval?

Answering all three makes the interpretation precise. The first derivative describes how the modeled quantity is changing. The second describes how that rate is changing. An interval-wide claim about the behavior of the rate requires an interval-wide condition, not merely a value at one point.