Tutorials › Calculus I › Higher Derivatives Of Inverse Functions

Implicit and Inverse Differentiation · Tutorial 596 of 1495

Higher Derivatives Of Inverse Functions

Learn how the derivatives of an inverse function depend on the derivatives of the original function, and how to evaluate and check these formulas.

Advanced 10 min read

What You'll Learn

  • State the local differentiability conditions needed for inverse derivative formulas
  • Derive the first, second, and third derivatives from the inverse identity
  • Evaluate inverse derivatives using values of the original function and its derivatives
  • Check computed values by differentiating the defining composition
  • Identify common errors involving signs, powers, and evaluation points

Introduction

An inverse function reverses the input-output relationship of a function. If \(f(a)=b\), then \(f^{-1}(b)=a\), provided the inverse is defined on the relevant range. Differentiating this reversal once gives a reciprocal-slope relationship. Differentiating again and a third time reveals how the curvature and higher-order behavior of \(f\) contribute to the derivatives of its inverse.

The key is to keep track of which variable is being differentiated. Write \(g=f^{-1}\), and set \(y=g(x)\). The defining identity is then \(f(y)=x\). Derivatives of \(f\) in the formulas below are evaluated at \(y=g(x)\), not at \(x\), unless the two happen to be equal.

Conditions and notation

Definition: Let \(f\) be three times continuously differentiable on an open interval containing \(a\), and suppose \(f'(a)\ne 0\). Locally around \(a\), \(f\) has an inverse \(g=f^{-1}\) around \(b=f(a)\). The inverse is three times differentiable there, and its derivatives can be expressed using derivatives of \(f\) at \(g(x)\).

The nonzero derivative condition matters: it ensures that \(f\) is locally invertible and that the reciprocal in the first-derivative formula is defined. For the second-derivative formula, \(f''\) must exist; for the third-derivative formula, \(f'''\) must exist. We use \(y=g(x)\), so \(f(y)=x\).

Theorem: Under the stated conditions, at points \(x\) near \(b=f(a)\), $$ g'(x)=\frac{1}{f'(y)},\qquad g''(x)=-\frac{f''(y)}{(f'(y))^3},\qquad g'''(x)=\frac{3(f''(y))^2-f'(y)f'''(y)}{(f'(y))^5}, \quad y=g(x). $$

These expressions are useful when the inverse is difficult to write explicitly. To find its derivatives at a particular \(x\), first find \(y\) satisfying \(f(y)=x\), then evaluate the derivatives of \(f\) at that \(y\).

Deriving the inverse derivative formulas

Start with \(f(g(x))=x\). The chain rule gives the first derivative. To obtain the next formulas, differentiate the resulting identities with respect to \(x\), remembering that \(y=g(x)\) depends on \(x\).

1
Differentiate once.
From \(f(y)=x\), the chain rule gives \(f'(y)g'(x)=1\), so \(g'(x)=1/f'(y)\). This requires \(f'(y)\ne 0\).
2
Differentiate twice.
Differentiate \(f'(y)g'(x)=1\) by the product and chain rules, then solve for \(g''(x)\).
3
Differentiate a third time.
Differentiate the second-derivative expression, using \(dy/dx=g'(x)=1/f'(y)\), and simplify.

Here is the full calculation for the second derivative. In the product rule, differentiating \(f'(y)\) with respect to \(x\) gives \(f''(y)g'(x)\):

$$ \frac{d}{dx}\bigl(f'(y)g'(x)\bigr) =f''(y)(g'(x))^2+f'(y)g''(x)=0. $$

Substitute \(g'(x)=1/f'(y)\), then isolate \(g''(x)\):

$$ f''(y)\frac{1}{(f'(y))^2}+f'(y)g''(x)=0, \qquad g''(x)=-\frac{f''(y)}{(f'(y))^3}. $$

For the third derivative, differentiate \(g''(x)=-f''(y)(f'(y))^{-3}\). Applying the product and chain rules gives

$$ \begin{aligned} g'''(x) &=-f'''(y)g'(x)(f'(y))^{-3} +3(f''(y))^2g'(x)(f'(y))^{-4}\\ &=-\frac{f'''(y)}{(f'(y))^4} +\frac{3(f''(y))^2}{(f'(y))^5}\\ &=\frac{3(f''(y))^2-f'(y)f'''(y)}{(f'(y))^5}. \end{aligned} $$

The second line uses \(g'(x)=1/f'(y)\) in each term. This derivation also explains why the denominator in the third-derivative formula has the fifth power: differentiating with respect to \(x\) introduces another factor \(1/f'(y)\), in addition to the factors already present.

Worked examples

Worked Example: A cubic inverse at a specified point

Let \(f(t)=t+t^3\), and let \(g=f^{-1}\) near \(t=1\). Find \(g'(2)\), \(g''(2)\), and \(g'''(2)\).

First locate the corresponding input to \(f\). Since \(f(1)=1+1=2\), we have \(g(2)=1\). Also, $$ f'(t)=1+3t^2,\qquad f''(t)=6t,\qquad f'''(t)=6. $$ At \(t=1\), these values are \(f'(1)=4\), \(f''(1)=6\), and \(f'''(1)=6\). In particular \(f'(1)\ne0\), as required.

Apply the formulas, evaluating each derivative of \(f\) at \(g(2)=1\): $$ g'(2)=\frac{1}{4},\qquad g''(2)=-\frac{6}{4^3}=-\frac{6}{64}=-\frac{3}{32}, $$ $$ g'''(2)=\frac{3(6)^2-(4)(6)}{4^5} =\frac{108-24}{1024} =\frac{84}{1024} =\frac{21}{256}. $$

Check the values by differentiating the composition identity. At \(x=2\), the first-derivative identity gives \(f'(1)g'(2)=4(1/4)=1\). The second-derivative identity gives $$ f''(1)(g'(2))^2+f'(1)g''(2) =6\left(\frac14\right)^2+4\left(-\frac{3}{32}\right) =\frac{3}{8}-\frac{3}{8}=0. $$ For the third derivative, differentiating the second-derivative identity yields \(f'''(y)(g')^3+3f''(y)g'g''+f'(y)g'''=0\). At the point in question, $$ 6\left(\frac14\right)^3 +3(6)\left(\frac14\right)\left(-\frac{3}{32}\right) +4\left(\frac{21}{256}\right) =\frac{3}{32}-\frac{27}{64}+\frac{21}{64}=0. $$ All three values satisfy the differentiated composition identities.

Worked Example: An inverse with zero second derivative

Let \(f(t)=t+\sin t\), and consider its local inverse \(g\) near \(x=0\). Find the first three derivatives of \(g\) at \(0\).

We have \(f(0)=0\), so \(g(0)=0\). The derivatives are $$ f'(t)=1+\cos t,\qquad f''(t)=-\sin t,\qquad f'''(t)=-\cos t. $$ Thus \(f'(0)=2\), \(f''(0)=0\), and \(f'''(0)=-1\). Because \(f'(0)=2\ne0\), the inverse exists locally and the formulas apply.

Substitution gives $$ g'(0)=\frac12,\qquad g''(0)=-\frac{0}{2^3}=0, $$ $$ g'''(0)=\frac{3(0)^2-(2)(-1)}{2^5} =\frac{2}{32} =\frac{1}{16}. $$

To check, the first differentiated identity gives \(2(1/2)=1\). The second gives \(0(1/2)^2+2(0)=0\). The third gives $$ (-1)\left(\frac12\right)^3 +3(0)\left(\frac12\right)(0) +2\left(\frac{1}{16}\right) =-\frac18+\frac18=0. $$ Each composition identity is satisfied.

Worked Example: The inverse of the natural logarithm

Let \(f(t)=\ln t\) for \(t>0\). Its inverse is \(g(x)=f^{-1}(x)\). Find \(g'(0)\), \(g''(0)\), and \(g'''(0)\) using the inverse formulas.

Because \(f(1)=\ln 1=0\), the corresponding input is \(g(0)=1\). The derivatives of \(f\) are $$ f'(t)=\frac1t,\qquad f''(t)=-\frac1{t^2},\qquad f'''(t)=\frac2{t^3}. $$ At \(t=1\), these are \(1\), \(-1\), and \(2\), respectively, and \(f'(1)\ne0\).

Therefore $$ g'(0)=\frac11=1,\qquad g''(0)=-\frac{-1}{1^3}=1, $$ $$ g'''(0)=\frac{3(-1)^2-(1)(2)}{1^5} =\frac{3-2}{1}=1. $$

For a check, the inverse here is \(g(x)=e^x\), whose first, second, and third derivatives at \(0\) are all \(e^0=1\). We can also check the composition identities: the first gives \(1\cdot1=1\), the second gives \((-1)(1)^2+(1)(1)=0\), and the third gives \(2(1)^3+3(-1)(1)(1)+(1)(1)=2-3+1=0\). The inverse formulas agree with both checks.

Common mistakes and how to avoid them

A frequent error is to evaluate derivatives of \(f\) at the output \(x\), rather than at the corresponding input \(y=g(x)\). For instance, in the cubic example, the requested inverse derivatives were at \(x=2\), but the derivatives of \(f\) had to be evaluated at \(y=1\), because \(f(1)=2\).

Another common error is losing a power or a minus sign. For the cubic example, a wrong second-derivative calculation might use $$ g''(2)=-\frac{f''(1)}{(f'(1))^2} =-\frac{6}{16} =-\frac38. $$

The denominator must be \((f'(y))^3\), not its square. The correct result is $$ g''(2)=-\frac{6}{4^3}=-\frac{3}{32}. $$ A useful check is the twice-differentiated identity: with the incorrect value, \(6(1/4)^2+4(-3/8)=3/8-3/2\ne0\), so the composition identity fails.

For the third derivative, preserve the entire numerator \(3(f'')^2-f'f'''\). It is not enough to differentiate only the numerator informally or to omit the factor introduced by \(dy/dx\). Keeping \(y=g(x)\) visible throughout the calculation helps prevent both errors.

Why higher inverse derivatives matter

The first derivative describes the slope of the inverse; it is the reciprocal of the original slope at the corresponding point. The second derivative describes how that inverse slope changes, and the third derivative records how the second derivative changes. The formulas show that inverse curvature depends not only on \(f''\), but also on the size and sign of \(f'\). The third derivative combines \(f''\) and \(f'''\), with a fifth power of the original slope in the denominator.

These relationships are especially useful when \(f^{-1}\) has no convenient explicit formula. One can still compute derivatives at a point by solving \(f(y)=x\) and evaluating \(f'(y)\), \(f''(y)\), and \(f'''(y)\). The local condition \(f'(y)\ne0\) remains essential; if it fails, these formulas cannot be used at that point.

Check Your Understanding

For each question, keep the inverse input \(y=g(x)\) distinct from the point \(x\) where the inverse derivative is requested.

  1. State the condition on \(f'\) needed to use the inverse derivative formulas locally.
  2. Write the formula for \(g''(x)\) in terms of derivatives of \(f\) at \(y=g(x)\).
  3. If \(f(2)=5\), which input to \(f\) is used to evaluate the derivatives when finding \(g'(5)\)?
  4. For \(f(t)=t^3+t\), use the formula to find \(g'(2)\), where \(g=f^{-1}\) near \(1\).
  5. Explain why differentiating with respect to \(x\) introduces a factor of \(g'(x)\) when differentiating an expression involving \(y=g(x)\).

Test yourself: Calculus I final exam practice

Ready to check what you know? Work a full-length practice final, then open the complete worked solutions.

Want all 1,114 practice problems?

This tutorial covers one topic. Calc Pass covers the full Calculus AB & BC curriculum (the core of Calculus I and II): full review, every practice question, and step-by-step solutions to each one.

QR code to download the Calc Pass app

Get the Calc Pass App

Scan to download and practice on the go, with instant scoring.

Calc Pass Volume 1: 1,114 Practice Questions book cover

Vol. 1: 1,114 Questions

The complete practice question bank, organized by AB & BC topic.

View on Amazon
Calc Pass Volume 2: Solutions Manual book cover

Vol. 2: Solutions Manual

Fully worked solutions to every question in Volume 1.

View on Amazon