Calculus › Section F · Logarithmic, exponential, and other transcendental functions › Topic F.3

Inverse functions and their derivatives

Many famous functions exist only as answers to a reversal question: the square root, the logarithm, the arcsine. Inverse functions have their own theory, and a remarkably clean derivative rule.

Reversing the machine

A function moves inputs to outputs; its inverse is the machine that walks back, turning each output into the input it came from. Reversal only makes sense when the journey never loses information: two different inputs must never collide into one output, or there is no way back. This is why so many "named" functions in mathematics are inverses: \(\sqrt{x}\) answers the square, \(\ln x\) answers the exponential, and the inverse trig functions of F.7 answer the sine. This topic builds the general theory: when inverses exist, what their graphs look like, and the crisp rule that hands you their derivatives.

The definition and one-to-one

A function \(g\) is the inverse of \(f\), written \(f^{-1}\), when composing the two in either order returns the input: \(f(g(x)) = x\) and \(g(f(x)) = x\), on their domains. The existence condition is that \(f\) be one-to-one: different inputs always give different outputs, so every value in the range came from exactly one input. Note also what swaps: the domain of \(f^{-1}\) is the range of \(f\), and vice versa. Any candidate inverse is proven by composing it both ways and watching \(x\) come back.

Tests: the horizontal line and monotonicity

The graph test mirrors A.3's vertical line test. A function is one-to-one exactly when no horizontal line meets its graph twice, because a repeated height is precisely two inputs sharing an output. On the algebra side, monotonicity settles it: a function strictly increasing or strictly decreasing on an interval is automatically one-to-one there, and the sign of the derivative, which D.3 taught you to chart, is the practical test. So \(x^3 - 2\) and \(\ln x\) are one-to-one everywhere they live, while \(x^2\) fails the horizontal line test on the whole real line, since \(2^2 = (-2)^2\).

Worked example: finding and verifying an inverse

Find the inverse of \(f(x) = 3x + 2\). The working recipe is swap and solve: write \(y = 3x + 2\), exchange the roles to \(x = 3y + 2\), and solve for \(y\): \(y = \frac{x - 2}{3}\), so \(f^{-1}(x) = \frac{x - 2}{3}\). Proof by composition, both directions: \(f(f^{-1}(x)) = 3 \cdot \frac{x-2}{3} + 2 = x\) and \(f^{-1}(f(x)) = \frac{3x + 2 - 2}{3} = x\). Both return \(x\), so the inverse is confirmed. The verification is not ceremony: it is the two-second check that catches every swap-and-solve slip.

Worked example: roots as inverses

The cube root is the inverse of the cube, and restricted powers follow the same pattern. Take \(f(x) = x^3 - 2\), strictly increasing because its derivative \(3x^2 \ge 0\) vanishes only at the isolated point \(0\), so it never stalls. Swap and solve: \(y = x^3 - 2\) becomes \(x = y^3 - 2\), so \(y = \sqrt[3]{x + 2}\), and \(f^{-1}(x) = \sqrt[3]{x + 2}\). Verify: \(f(f^{-1}(x)) = (\sqrt[3]{x+2})^3 - 2 = x\). Correct. For \(f(x) = x^2\), which fails one-to-one globally, the fix is a restricted domain: on \([0, \infty)\) it is increasing, and its inverse there is the square root, exactly where the square root came from all along.

The picture: reflection in y = x

Graphically the inverse is obtained by swapping coordinates, so the graph of \(f^{-1}\) is the graph of \(f\) reflected in the line \(y = x\). Every point \((a, b)\) of \(f\) becomes \((b, a)\) of the inverse. Two consequences worth keeping. First, since reflection interchanges input and output, the steepness also interchanges: a tangent of slope \(m\) on \(f\) becomes a tangent of slope \(\frac{1}{m}\) on the inverse. That sentence, made precise, is the derivative rule of the next cards. Second, the reflection shows at a glance when an inverse exists: a graph that passes the horizontal line test reflects into a graph that passes the vertical one.

The derivative of an inverse

Here is the rule, and its derivation is one chain-rule line. Starting from \(f(f^{-1}(y)) = y\), differentiate with respect to \(y\): \(f'\!\left(f^{-1}(y)\right) \cdot (f^{-1})'(y) = 1\), so

(f^{-1})'(y) = \frac{1}{f'\!\left(f^{-1}(y)\right)}

provided the denominator is not zero. In words: the inverse's slope at \(b\) is the reciprocal of the original's slope at the matching point \(a\), where \(f(a) = b\). Exactly what the reflection picture promised. The condition \(f'(a) \ne 0\) is essential: where the original has a level tangent, the inverse's tangent is vertical and its derivative does not exist.

Worked example: the rule, without a formula

Suppose all you know about a one-to-one \(f\) is that \(f(2) = 7\) and \(f'(2) = 5\). What is \((f^{-1})'(7)\)? You hold no formula for \(f\) or its inverse, and none is needed. The rule says evaluate the reciprocal at the matching point: \((f^{-1})'(7) = \frac{1}{f'(f^{-1}(7))}\), and since \(f(2) = 7\) means \(f^{-1}(7) = 2\), the answer is \(\frac{1}{5}\). This style of question, compute an inverse derivative from a table of values of \(f\) and \(f'\), is a favourite of the AP exam, precisely because it tests whether you understand the rule rather than merely quote it.

Worked example: the cube root, two independent ways

Differentiate \(g(x) = \sqrt[3]{x}\) using the inverse rule, with \(f(x) = x^3\). The matching point for an output \(y\) is \(x = y^{1/3}\), so \(g'(y) = \frac{1}{f'(x)} = \frac{1}{3x^2} = \frac{1}{3 y^{2/3}}\). Now check against the ordinary power rule: \(\frac{d}{dx} x^{1/3} = \frac{1}{3} x^{-2/3} = \frac{1}{3 x^{2/3}}\). Identical answers from two different routes (both valid for \(x \ne 0\), where the cube's slope is not zero). The same manoeuvre of inverting, differentiating, and re-expressing will generate every inverse trig derivative in F.7.

Common mistakes to avoid

The classics: reading \(f^{-1}(x)\) as \(\frac{1}{f(x)}\), where the superscript means inverse, never reciprocal; verifying an inverse with one composition only; forgetting that one-to-one is required before an inverse can exist, and that restricted domains repair it; differentiating with the rule at a point where \(f'\) is zero, where the inverse has no finite slope; and mixing the matching points, the reciprocal is evaluated at \(f^{-1}(b)\), not at \(b\) itself. Also note the graph mirror: the inverse reflection is across \(y = x\), not across an axis.

Key takeaways

An inverse exists exactly when \(f\) is one-to-one, certified by the horizontal line test or by strict monotonicity via the derivative's sign. Find it by swap and solve, and prove it by composing both ways. The graph is the reflection in \(y = x\), and the slopes obey the reciprocal rule \((f^{-1})'(b) = \frac{1}{f'(f^{-1}(b))}\), valid while the original slope is nonzero. Computing from a table, as with \((f^{-1})'(7) = \frac{1}{5}\), is the signature skill. These are the mechanics F.4 and F.7 will apply to exponentials and inverse trig functions respectively.

Flashcards

Flashcard

Inverse function

f(f⁻¹(x)) = x and f⁻¹(f(x)) = x

Compose both ways and recover x.

Test your understanding

Question 1 of 5

What is the inverse of f(x) = 3x + 2?

Question 2 of 5

Is f(x) = x² one-to-one on all real numbers?

Question 3 of 5

Given f(2) = 7 and f′(2) = 5 for a one-to-one f, what is (f⁻¹)′(7)?

Question 4 of 5

For any invertible f, what is f(f⁻¹(9))?

Question 5 of 5

When does the inverse-derivative formula fail to give a real number?

Standards: 3

← Previous topicNext topic →