The inverse function theorem and the reciprocal-slope picture behind it, differentiating a general inverse, the extension of the power rule to rational exponents, and the derivatives of all six inverse trigonometric functions obtained by a right-triangle argument — every one of them algebraic.
Subject: Calculus I · 68 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 3.7
Reflect the graph, flip the slope
Objectives
You can differentiate powers, products, quotients, trigonometric functions and compositions. This lesson adds the functions that undo them: roots, and the inverse trigonometric functions. One idea covers them all, and it is the chain rule applied to the statement that a function and its inverse cancel.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, pp. 259-260 — learning objectives 3.7.1 and 3.7.2
Every derivative you have so far came from a rule for building functions: sums, products, quotients, compositions. This lesson is about the functions that run those in reverse. The square root undoes squaring, the cube root undoes cubing, and the inverse sine undoes the sine. You meet these constantly, and you have had no way to differentiate them except the limit definition.
The good news is that one idea handles them all. If g undoes f, then f of g of x is x, and differentiating that equation with the chain rule produces a formula for the derivative of g in terms of the derivative of f, which you already know.
Two payoffs follow. The power rule, which you have been trusting for fractional exponents, finally gets a proof. And the inverse trigonometric functions turn out to have derivatives that are not trigonometric at all, a fact the integration chapters lean on heavily.
Warm-up
Discussion prompt
Section 1.4 defined the inverse by the fact that f undoes it: f of f inverse of x is x, for every x in the inverse's domain. Differentiate both sides of that equation with respect to x. What comes out, and what could you solve it for?
Write your answer before revealing it. The equation says that f undoes its inverse, and the only tool you need is the chain rule, applied to the left-hand side as a composition.
The outer function is f, so its derivative is f prime, evaluated at the inside, which is the inverse of x. The inner function is the inverse itself, so its derivative is exactly the thing you do not yet know. The right-hand side, x, has derivative 1.
Now look at what you have: one equation with one unknown, the inverse's derivative, multiplied by something you can compute. Divide, and you have the main theorem of this section. Keep this derivation in mind; if you ever forget the formula, you can rebuild it in two lines from here.
Section
Part 1
Prediction
\[ f(2) = 5, \qquad f'(2) = 4 \]
Predict first
The graph of f passes through the point (2, 5) with slope 4. The graph of the inverse passes through the mirror point (5, 2). What is the inverse's slope there?
Correct: 1/4
Why: Reflecting in the diagonal swaps horizontal and vertical, so a line that rises 4 for every 1 across becomes a line that rises 1 for every 4 across. The slope turns upside down: one quarter. The next two slides show why, first with a picture and then with the chain rule.
Commit to one of the options before you reveal the answer. The trap here is to think that the inverse simply inherits the slope, or that it gets the negative of it.
Think about what the inverse's graph is: the same set of points with every pair of coordinates swapped. A tangent line that climbs 4 units for every 1 unit across, once its coordinates are swapped, climbs 1 unit for every 4 across. So the slope becomes one quarter.
Notice also where that slope lives. It belongs at the point (5, 2) on the inverse, the mirror of (2, 5) on f. The number 5 is where you evaluate the inverse's derivative, but the number 2 is where you evaluate the original's. Keeping those two numbers straight is most of the work in this lesson.
Concept
Reflecting the plane in the line y equals x swaps the two coordinates of every point. A horizontal step becomes a vertical step, and a vertical step becomes a horizontal one.
\[ (a, b) \;\longmapsto\; (b, a) \]
\[ m = \frac{\text{rise}}{\text{run}} \;\longmapsto\; \frac{\text{run}}{\text{rise}} = \frac{1}{m} \]
Figure (svg): Two lines through the origin: y equals 2x with a slope triangle of run 1 and rise 2, and its mirror image y equals x over 2 with a slope triangle of run 2 and rise 1. The dashed diagonal y equals x sits between them.
So a line of slope m reflects to a line of slope one over m. A slope of 2 becomes a slope of one half; a slope of minus 3 becomes minus one third.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 260 — the discussion before Figure 3.28
Look at the two slope triangles in the figure. On the steep blue line, a run of 1 comes with a rise of 2, so the slope is 2. Reflect that triangle in the dashed diagonal and the horizontal leg becomes vertical while the vertical leg becomes horizontal. On the orange line, a run of 2 comes with a rise of 1, so the slope is one half.
The colours track the legs: the yellow run of 1 reappears as a yellow rise of 1, and the pink rise of 2 reappears as a pink run of 2. Nothing about the lengths changed, only which one is horizontal. That is why the new slope is the reciprocal of the old one.
The same thing happens for a negative slope, since both the rise and the run keep their signs when they swap roles. A line of slope minus 3 reflects to a line of slope minus one third, not plus one third.
Picture it
Figure (svg): The curve y equals x squared for x from 0 to 2.3 and its reflection y equals the square root of x, with the dashed diagonal between them. The point (1.5, 2.25) on the parabola carries a tangent of slope 3; the mirror point (2.25, 1.5) on the root curve carries a tangent of slope one third.
The graph of the inverse is the mirror image of the graph of f, and the tangent at a point reflects to the tangent at the mirror point. The parabola's slope at (1.5, 2.25) is 3, so the root curve's slope at (2.25, 1.5) must be one third.
\[ \big(f^{-1}\big)'(2.25) = \frac{1}{f'(1.5)} = \frac{1}{3} \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 260 — Figure 3.28
This is the book's Figure 3.28 drawn with real functions: the parabola y equals x squared, restricted to x at least zero, and its inverse, the square root. The yellow tangent touches the parabola at (1.5, 2.25), where the slope is twice 1.5, which is 3.
Reflect the whole picture in the diagonal. The parabola lands on the root curve, the point (1.5, 2.25) lands on (2.25, 1.5), and the yellow tangent lands on the pink one. A line that touches a curve still touches it after both are reflected, so the pink line really is the tangent to the root curve, and its slope must be one third.
You can confirm this with what you already know: the derivative of the square root is one over twice the root, and at 2.25 the root is 1.5, giving one over 3. The picture predicted it without any differentiation of the root at all.
Concept
The picture says the answer is a reciprocal. The chain rule says exactly which reciprocal. Start from the cancellation equation.
\[ x = f\big(f^{-1}(x)\big) \]
\[ 1 = f'\big(f^{-1}(x)\big)\cdot\big(f^{-1}\big)'(x) \]
\[ \big(f^{-1}\big)'(x) = \frac{1}{f'\big(f^{-1}(x)\big)} \]
The first line differentiates both sides; the second divides. Dividing is allowed only when the original's derivative is not zero at the point in question, a condition that will matter in a few slides.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 260 — equation 3.19
The picture tells you the answer is a reciprocal, but not precisely which slope to take the reciprocal of. The chain rule settles that. This is the same computation you did in the warm-up, now written as a derivation.
Read the middle line carefully. The original's derivative is evaluated at the inverse of x, not at x. That is the algebraic version of the mirror point: the inverse's slope at x matches the original's slope at the point whose output is x.
The last line divides by the original's derivative. You can only divide by something that is not zero, so the formula comes with a condition built in. The picture will show you in a moment what happens when that condition fails.
Concept
Inverse function theorem (Theorem 3.11) — Let f be invertible and differentiable, with inverse g. At every x where the original's derivative at g of x is not zero, the inverse is differentiable, and its derivative is one over that derivative.
\[ \frac{dy}{dx} = \big(f^{-1}\big)'(x) = \frac{1}{f'\big(f^{-1}(x)\big)} \]
\[ g = f^{-1}: \qquad g'(x) = \frac{1}{f'\big(g(x)\big)} \]
The second form is the one you will use most. Name the inverse g, and the theorem reads: the inverse's slope is one over the original's slope, taken at the inverse's own output.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 260 — Theorem 3.11
This is the statement to learn. It has two forms, and they say the same thing. The first uses the inverse notation throughout. The second gives the inverse its own name, g, which is usually easier to read: the derivative of g at x is one over f prime of g of x.
In words, the inverse's slope is one over the original's slope, where the original's slope is taken at the inverse's own output. That last phrase is the part people drop, and dropping it gives wrong answers that look perfectly reasonable.
The hypothesis is modest: f must be invertible and differentiable, and its derivative must not be zero at the relevant point. Almost every function you will invert in this course satisfies that except at a few isolated points, and those points are exactly the ones worth looking at.
Notation
Annotate
On: \( \big(f^{-1}\big)'(a) = \frac{1}{f'\big(f^{-1}(a)\big)} \)
Step through the notes in order, and notice that they describe an order of work as well as a formula. You start at the left with the question, the inverse's slope at a.
The most important piece is the inside of the denominator. The inverse of a is a number, call it b, and it is the input on the original that produces a. You must find b before anything else, and in many problems finding b is the only real work: once you know it, the rest is ordinary differentiation of a function you know.
Then evaluate the original's derivative at b, and take the reciprocal. The last note is a reminder that this last step can fail. If the original is flat at b, the reciprocal does not exist, and neither does the inverse's derivative.
Counterexample
Discussion prompt
Someone claims: if f is differentiable everywhere and has an inverse, then the inverse is differentiable everywhere too. Find a counterexample, and say at which point the inverse fails.
Try to find an example yourself before revealing. A good strategy with any theorem that has a hypothesis is to ask what goes wrong when the hypothesis fails. Here that means looking for an invertible function with a flat spot.
The cube is the standard example. It is smooth everywhere and one-to-one, since it always rises, but at the origin its slope is zero: the curve flattens for an instant and then keeps rising. Its inverse, the cube root, rises everywhere too, but at the origin it goes straight up.
So differentiability of f is not enough to make the inverse differentiable everywhere. What is needed is a nonzero slope at the matching point, which is exactly the condition Theorem 3.11 includes. The next slide shows the picture.
Concept
Figure (svg): The curves y equals x cubed and y equals the cube root of x crossing at the origin. A short horizontal segment marks the cube's flat tangent at the origin; a short vertical segment marks the cube root's vertical tangent there.
\[ \frac{d}{dx}\sqrt[3]{x} = \frac{1}{f'\big(\sqrt[3]{x}\big)} = \frac{1}{3\big(\sqrt[3]{x}\big)^2} = \frac{1}{3x^{2/3}} \]
At x equal to 0 the denominator is zero, so the formula gives no value, exactly where the picture shows a vertical tangent. Everywhere else the formula works.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 260 — the hypothesis of Theorem 3.11
Look at the origin in the figure. The yellow segment lies along the cube's tangent there, which is flat. Reflect it in the diagonal and it becomes the pink segment, which is vertical, and that pink segment is the cube root's tangent at the origin.
A vertical line has no slope, and the formula agrees: at x equal to zero the derivative of the cube root is one over three times zero, which is undefined. The theorem is not failing here; it is correctly reporting that there is nothing to report.
You will see this pattern again with the inverse sine. Sine is flat at the top and bottom of its restricted piece, at minus pi over 2 and pi over 2, and the inverse sine goes vertical at the matching points, 1 and minus 1. Whenever an inverse has a vertical tangent, look for a flat spot on the original.
Worked example
Use the inverse function theorem to differentiate g, then compare with differentiating g directly.
\[ g(x) = \frac{x+2}{x} \]
Solve for x to find the function g undoes
Why: Set y equal to g of x, clear the denominator, and isolate x.
\[ y = \frac{x+2}{x} \;\Longrightarrow\; xy - x = 2 \;\Longrightarrow\; x = \frac{2}{y-1} \]
Name that function f
Why: Swap the letters. Then g is the inverse of f.
\[ f(x) = \frac{2}{x-1} \]
Differentiate f
Why: Write it as 2 times x minus 1 to the minus one and use the chain rule.
\[ f'(x) = -\frac{2}{(x-1)^2} \]
Evaluate f prime at g of x
Why: Put the whole of g of x in place of x.
\[ f'\big(g(x)\big) = -\frac{2}{\left(\frac{x+2}{x} - 1\right)^2} \]
Simplify the bracket
Why: x plus 2 over x, minus x over x, leaves 2 over x.
\[ \frac{x+2}{x} - 1 = \frac{2}{x} \]
Square and divide
Why: Two divided by four over x squared.
\[ f'\big(g(x)\big) = -\frac{2}{4/x^2} = -\frac{x^2}{2} \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = \frac{1}{f'\big(g(x)\big)} = -\frac{2}{x^2} \]
Figure (svg): The right-hand branches of g of x equals x plus 2 over x and of its inverse 2 over x minus 1, with their asymptotes dashed. At (1, 3) the tangent to g has slope minus 2; at the mirror point (3, 1) the tangent to the inverse has slope minus one half.
Check with the quotient rule, and at a point
Why: The direct route gives the same formula. At x equal to 1 the slope is minus 2, and at the mirror point (3, 1) the inverse's slope is minus one half, the reciprocal.
\[ g'(x) = \frac{x\cdot 1 - (x+2)\cdot 1}{x^2} = -\frac{2}{x^2}, \qquad g'(1) = -2, \quad f'(3) = -\frac12 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — Example 3.60
This example is chosen because you can check the answer by the quotient rule, so it shows the theorem working on familiar ground. The strategy is to find the function f that g undoes, differentiate f, and then evaluate that derivative at g of x.
Finding f takes one line of algebra: set y equal to g of x, multiply through by x, gather the x terms and divide. The result is 2 over y minus 1, and renaming the variable gives f. Its derivative follows from the chain rule.
The step that needs care is substituting g of x into f prime. Simplify the bracket first: x plus 2 over x, minus 1, is just 2 over x. Then the reciprocal comes out as minus 2 over x squared.
Look at the figure for the geometric check. At (1, 3) the slope of g is minus 2, and at the mirror point (3, 1) the slope of f is minus one half. The two tangent lines are reflections of each other.
Worked example
\[ g(x) = \frac{1}{x+2} \]
Find the function g undoes
Why: Flip both sides, then subtract 2.
\[ y = \frac{1}{x+2} \;\Longrightarrow\; x + 2 = \frac1y \;\Longrightarrow\; f(x) = \frac1x - 2 \]
Differentiate f
Why: Power rule on x to the minus one; the constant disappears.
\[ f'(x) = -\frac{1}{x^2} \]
Evaluate at g of x
Why: One over one over x plus 2, squared, is x plus 2 squared.
\[ f'\big(g(x)\big) = -\frac{1}{\left(\frac{1}{x+2}\right)^2} = -(x+2)^2 \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = -\frac{1}{(x+2)^2} \]
Check by differentiating g directly
Why: Chain rule on x plus 2 to the minus one gives the same answer; at x equal to 1 both give minus one ninth.
\[ \frac{d}{dx}(x+2)^{-1} = -(x+2)^{-2}, \qquad g'(1) = -\frac19 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — Checkpoint 3.42
Same method, simpler function. Flipping both sides of y equals one over x plus 2 gives x plus 2 equals one over y, so the function g undoes is one over x, minus 2.
Its derivative is minus one over x squared. Evaluated at g of x, the x squared becomes one over x plus 2, squared, and dividing by that is the same as multiplying by x plus 2 squared. Take the reciprocal of minus x plus 2 squared and you have the answer.
The direct check is the chain rule on x plus 2 to the minus one power, and it gives the identical formula. In practice you would always differentiate this function directly; the point of doing it the long way is to build confidence that the theorem gives right answers, so you can trust it for functions where there is no direct route.
Error analysis
\[ f(x) = x^3 + 2x, \qquad f(1) = 3 \]
Annotate
On: \( \big(f^{-1}\big)'(3) = \frac{1}{f'(3)} = \frac{1}{29} \)
\[ \big(f^{-1}\big)'(3) = \frac{1}{f'(1)} = \frac{1}{3 + 2} = \frac15 \]
Read the line and find the problem before revealing the notes. Every symbol in it looks like the theorem, which is exactly why this error survives on so many exams.
The mistake is the number inside f prime. The question asks about the inverse at 3, and 3 is an output of f. The theorem needs the original's slope at the input that produces 3, and since f of 1 is 3, that input is 1. The derivative of f is 3 x squared plus 2, which is 5 at 1 but 29 at 3.
A useful habit: before writing f prime of anything, say out loud which point on the graph of f you are standing on. Here it is (1, 3), and the inverse's point is (3, 1). The slope at (1, 3) is 5, so the inverse's slope is one fifth.
Step zero
\[ f(x) = x^3 + 2x + 3, \qquad \text{find } \big(f^{-1}\big)'(0) \]
Discussion prompt
There is no algebraic formula for this inverse that you would want to use. What single number must you find before Theorem 3.11 can do anything, and how do you find it?
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 266 — Exercise 269
Write down your plan before revealing. The temptation with a cubic like this is to try to find a formula for its inverse, which is a dead end: there is one, but it is enormous and you would never differentiate it by hand.
The theorem does not need the formula. It needs only one number, the input b that f sends to 0, and for a textbook exercise that is almost always a small integer. Try 0, 1, minus 1, and so on. Here minus 1 works: minus 1, minus 2, plus 3 is zero.
Once b is known, the rest is routine. The derivative of f is 3 x squared plus 2, which is 5 at minus 1, so the inverse's slope at 0 is one fifth. This is the most common exam format for the theorem, and it always goes the same way: find b, then differentiate the function you know.
Fill the middle
\[ f(x) = x + \sin x \]
Fill in the blanks
f(0) = 0 \;\Longrightarrow\; \big(f^0\big)'(0) = \frac1/2___})} = ___
Why: The sum of x and sine of x is zero at x equal to 0, so the inverse sends 0 to 0. The derivative of f is 1 plus cosine of x, which is 2 at 0, and the inverse's slope is its reciprocal, one half.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 266 — Exercise 272
Fill in each blank before checking. The first question is which input f sends to 0, and the answer is almost too easy: x plus sine of x is zero when x is zero.
The derivative of f is 1 plus cosine of x, and at 0 that is 1 plus 1, which is 2. The inverse's slope at 0 is therefore one half. Both blanks for the matching input hold 0, because at the origin the input and the output happen to coincide, which is also why the point sits on the diagonal and is its own mirror image.
Nobody can write down the inverse of x plus sine x with ordinary functions, yet its slope at 0 took two lines. That is the real power of the theorem.
Worked example
Find the slope of the inverse at P, and the equation of the inverse's tangent line there. Restrict f to x at least 0 so it is one-to-one.
\[ f(x) = \frac{4}{1+x^2}, \quad x \ge 0, \qquad P(2, 1) \text{ on } f^{-1} \]
Find the mirror point on f
Why: P lies on the inverse, so (1, 2) lies on f.
\[ f(1) = \frac{4}{1+1} = 2 \]
Differentiate f
Why: Chain rule on 4 times 1 plus x squared to the minus one.
\[ f'(x) = -\frac{8x}{(1+x^2)^2} \]
Evaluate at the matching input
Why: At 1, not at 2.
\[ f'(1) = -\frac{8}{4} = -2 \]
Take the reciprocal
Why: Theorem 3.11.
\[ \big(f^{-1}\big)'(2) = \frac{1}{-2} = -\frac12 \]
Write the tangent line
Why: Point-slope form through (2, 1).
\[ y - 1 = -\frac12(x - 2) \;\Longrightarrow\; y = -\frac{x}{2} + 2 \]
Figure (svg): The curve y equals 4 over 1 plus x squared for x at least 0, and its reflection in the diagonal. The point (1, 2) on f has a tangent of slope minus 2; the mirror point (2, 1) on the inverse has the tangent line y equals minus x over 2 plus 2.
Check with an explicit inverse
Why: Here the inverse can be found: solve y equals 4 over 1 plus x squared for x. Its derivative at 2 is also minus one half.
\[ f^{-1}(x) = \sqrt{\frac4x - 1}, \quad \big(f^{-1}\big)'(2) = \frac{1}{2\sqrt{1}}\cdot\left(-\frac{4}{4}\right) = -\frac12 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 266 — Exercise 274
Here the question gives you a point on the inverse's graph, P at (2, 1), and asks for the tangent line there. Swap the coordinates to get the point on f, which is (1, 2), and confirm it: 4 over 1 plus 1 is 2.
The derivative of f comes from the chain rule, and at 1 it equals minus 2. The inverse's slope at P is the reciprocal, minus one half, and point-slope form gives the line. Notice that the exercise does not state a domain; the function is only one-to-one once you restrict it, and x at least 0 is the natural choice because the point (1, 2) lies on that side.
The figure shows both tangent lines and how they mirror each other. The final step checks the answer a second way, by actually finding the inverse, which is possible here. You would rarely bother, but it confirms minus one half exactly.
Section
Part 2
Concept
The nth root undoes the nth power, so the inverse function theorem hands you the derivative of every root. For an even n, restrict the power to x at least 0 so that it is one-to-one.
\[ g(x) = x^{1/n} \quad \text{is the inverse of} \quad f(x) = x^{n} \]
\[ g'(x) = \frac{1}{f'\big(g(x)\big)} = \frac{1}{n\big(x^{1/n}\big)^{n-1}} \]
Up to now the power rule was proved only for integer exponents. Roots are where it gets extended, and the extension to every fraction follows with one more use of the chain rule.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — the remark before Theorem 3.12
When the power rule was introduced, it was proved for whole-number exponents and then used for fractional ones on trust. This part of the lesson pays off that debt. The key is that the nth root is the inverse of the nth power, so the inverse function theorem applies directly.
For odd n, the nth power is one-to-one on the whole line. For even n it is not, since a number and its negative have the same square, so you restrict the power to x at least 0, which is also where the even root is defined.
The second line of the slide is the theorem written out for the root. The next two slides simplify it, first for the cube root and then in general, and you will see the familiar power-rule answer appear.
Worked example
\[ g(x) = \sqrt[3]{x} \]
Name the function it undoes
Why: The cube root is the inverse of the cube.
\[ f(x) = x^3 \]
Differentiate f
Why: Power rule.
\[ f'(x) = 3x^2 \]
Evaluate f prime at g of x
Why: Square the cube root.
\[ f'\big(g(x)\big) = 3\big(\sqrt[3]{x}\big)^2 = 3x^{2/3} \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = \frac{1}{3x^{2/3}} \]
Rewrite as a power
Why: Move the power to the numerator.
\[ g'(x) = \frac13 x^{-2/3} \]
Check at x equal to 8
Why: The cube root has slope one twelfth at (8, 2); the cube has slope 12 at the mirror point (2, 8). The power rule with exponent one third predicts the same formula.
\[ g'(8) = \frac{1}{3\cdot 4} = \frac{1}{12}, \qquad f'(2) = 12, \qquad \frac13 x^{\frac13 - 1} = \frac13 x^{-2/3} \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — Example 3.61
Follow the same four moves as before: name the function the cube root undoes, which is the cube; differentiate it, giving 3 x squared; evaluate that at the cube root of x; take the reciprocal.
The evaluation step is where the exponents meet. The square of the cube root of x is x to the two thirds, so f prime of g of x is 3 x to the two thirds, and the reciprocal is one third of x to the minus two thirds.
That is exactly what the power rule would have given with exponent one third: bring down one third and subtract 1 from the exponent. The difference is that now it has been proved rather than assumed.
The check at 8 is the picture again. The cube root has slope one twelfth at (8, 2), and the cube has slope 12 at the mirror point (2, 8).
Fill the middle
\[ g(x) = \sqrt[5]{x} \quad \text{undoes} \quad f(x) = x^5 \]
Fill in the blanks
f'\big(g(x)\big) = 5\big(\sqrt[5]4/5\big)^4 = 5x^-4/5} \;\Longrightarrow\; g'(x) = \frac______x^___}
Why: The derivative of x to the fifth is 5 x to the fourth. At the fifth root of x, the fourth power becomes x to the four fifths. Taking the reciprocal moves the power to the numerator with its sign changed: one fifth times x to the minus four fifths, exactly what the power rule with exponent one fifth predicts.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — Checkpoint 3.43
Fill in both exponents before checking. This is the cube-root example with 5 in place of 3, and working it yourself is the fastest way to see the general pattern.
The derivative of x to the fifth is 5 x to the fourth. Put the fifth root of x in place of x, and the fourth power of the fifth root is x to the four fifths. The reciprocal then moves that power to the numerator with its sign flipped, giving one fifth times x to the minus four fifths.
Compare with the power rule for exponent one fifth: bring down one fifth, and one fifth minus 1 is minus four fifths. They match, and the next slide proves that they always will.
Worked example
Prove the power rule for every rational exponent, in two stages: first one over n, then m over n.
Differentiate the nth power
Why: g of x equals x to the one over n undoes f of x equals x to the n.
\[ f'(x) = nx^{n-1} \]
Evaluate at g of x
Why: Multiply the exponents.
\[ f'\big(g(x)\big) = n\big(x^{1/n}\big)^{n-1} = nx^{(n-1)/n} \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = \frac{1}{nx^{(n-1)/n}} = \frac1n x^{(1-n)/n} \]
Tidy the exponent
Why: One minus n, over n, is one over n minus 1.
\[ \frac{d}{dx}x^{1/n} = \frac1n x^{(1/n) - 1} \]
Write x to the m over n as a power of a root
Why: Now any integer m.
\[ x^{m/n} = \big(x^{1/n}\big)^{m} \]
Apply the chain rule
Why: Integer power rule outside, the first stage inside.
\[ \frac{d}{dx}\big(x^{1/n}\big)^{m} = m\big(x^{1/n}\big)^{m-1}\cdot\frac1n x^{(1/n)-1} \]
Add the exponents
Why: The exponents m minus 1 over n and one over n minus 1 add to m over n minus 1.
\[ \frac{d}{dx}x^{m/n} = \frac{m}{n}x^{(m/n)-1} \]
Check numerically at x equal to 8
Why: With m over n equal to two thirds the rule predicts one third; a difference quotient with h equal to one millionth gives 0.3333333.
\[ \frac23 \cdot 8^{-1/3} = \frac23\cdot\frac12 = \frac13, \qquad \frac{8.000001^{2/3} - 7.999999^{2/3}}{0.000002} \approx 0.3333333 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, pp. 261-262 — Theorem 3.12 and its proof
The proof has two stages, and it is worth seeing how little each one needs. The first stage is the cube-root example with a general n. The derivative of x to the n is n x to the n minus 1; evaluated at the nth root of x it becomes n x to the n minus 1 over n; and the reciprocal is one over n times x to the one minus n over n, which is the power rule for exponent one over n.
The second stage handles any fraction m over n by writing x to the m over n as the nth root raised to the mth power. That is a composition, so the chain rule applies: the integer power rule on the outside, the first stage on the inside.
The only fiddly part is adding the exponents. The outside contributes m minus 1 over n, the inside contributes one over n minus 1, and together they give m over n minus 1. The numerical check confirms the result at one point.
Notation
Annotate
On: \( \frac{d}{dx}\left(x^{m/n}\right) = \frac{m}{n}x^{(m/n)-1} \)
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 261 — equations 3.20 and 3.21
The main message of this annotation is that there is nothing new to memorise. The rule for fractional exponents is the rule you already use for whole-number exponents: bring the exponent down, then subtract one from it.
The third note is where the errors happen. Subtracting 1 from two thirds gives minus one third, because 1 is three thirds. A common slip is to subtract 1 from the numerator only, which gives one third, and that answer is wrong.
The last note connects back to the vertical tangents. When the new exponent is negative, as it is for every root, the derivative has x in the denominator and blows up at zero. The root curves really do leave the origin vertically.
Tweak it
Parameter explorer
The curve is the nth root of x. Slide n from 1 to 8. What happens to the steepness of the curve near x equal to 0, and how does the derivative formula predict it?
\[ y = x^{1/{n}}, \qquad y' = \frac{1}{{n}}x^{(1/{n}) - 1} \]
Start with n equal to 1, where the curve is simply the line y equals x. Then step n up to 2, 3 and beyond, and watch what happens near the origin.
For every n from 2 upward, the curve leaves the origin heading straight up, and the larger n is, the more abruptly it turns to run nearly flat. The formula predicts both features. The exponent one over n minus 1 is negative, so the derivative blows up as x approaches zero; and for x well above 1 the derivative is small, since it has the factor one over n.
Every one of these roots is the mirror image of a power curve that is flat at the origin. The higher the power, the flatter the power curve near zero, and the steeper its reflection.
Worked example
Find an equation of the line tangent to the graph at x equal to 8.
\[ y = x^{2/3} \]
Differentiate with the rational power rule
Why: Two thirds minus one is minus one third.
\[ \frac{dy}{dx} = \frac23 x^{-1/3} \]
Evaluate the slope at 8
Why: The cube root of 8 is 2.
\[ \frac{dy}{dx}\Big|_{x=8} = \frac23\cdot\frac12 = \frac13 \]
Find the point
Why: Square the cube root of 8.
\[ y(8) = 8^{2/3} = 2^2 = 4 \]
Write point-slope form
Why: Through (8, 4) with slope one third.
\[ y - 4 = \frac13(x - 8) \]
Solve for y
Why: Four minus eight thirds is four thirds.
\[ y = \frac13 x + \frac43 \]
Figure (svg): The curve y equals x to the two thirds from minus 2 to 12, with a sharp cusp at the origin, and the tangent line y equals x over 3 plus 4 over 3 touching it at (8, 4).
Check the line touches the curve
Why: At x equal to 8 the line gives 8 thirds plus 4 thirds, which is 4; one unit to the right the curve is at 4.327 and the line at 4.333, close but the line is just above, as a tangent to a curve that bends downward should be.
\[ \frac83 + \frac43 = 4, \qquad 9^{2/3} \approx 4.327 \quad\text{vs}\quad \frac{9+4}{3} \approx 4.333 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 262 — Example 3.62
A tangent line needs two things, a slope and a point, and you get them from the derivative and the function respectively. The derivative is two thirds times x to the minus one third, by the rational power rule, and at 8 the cube root is 2, so the slope is one third.
The point comes from the original function: 8 to the two thirds is the square of the cube root of 8, which is 4. Point-slope form through (8, 4) with slope one third gives the line.
Look at the figure. The curve has a cusp at the origin, a sharp point where the derivative does not exist; the formula agrees, since x to the minus one third is undefined at zero. Far from the cusp, at x equal to 8, the tangent line hugs the curve closely, and the numbers in the check show the line sitting just above it one unit away.
Worked example
\[ s(t) = \sqrt{2t+1} \]
Rewrite as a power
Why: A square root is the one-half power.
\[ s(t) = (2t+1)^{1/2} \]
Rational power rule outside
Why: One half minus one is minus one half.
\[ \frac12(2t+1)^{-1/2} \]
Times the inner derivative
Why: The derivative of 2t plus 1 is 2.
\[ s'(t) = \frac12(2t+1)^{-1/2}\cdot 2 \]
Simplify
Why: The halves cancel.
\[ s'(t) = (2t+1)^{-1/2} = \frac{1}{\sqrt{2t+1}} \]
Check by squaring first
Why: Square both sides and differentiate: 2 s times s prime equals 2, so s prime is one over s. At t equal to 4 both routes give one third.
\[ s^2 = 2t + 1 \;\Longrightarrow\; 2s\,s' = 2 \;\Longrightarrow\; s' = \frac1s, \qquad s'(4) = \frac{1}{\sqrt 9} = \frac13 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 262 — Checkpoint 3.44
The first move is always to rewrite a root as a power, because the power rule is stated for powers. The square root of 2t plus 1 becomes 2t plus 1 to the one half.
Then it is the chain rule as in Section 3.6: the power rule on the outside, which gives one half times the inside to the minus one half, and then multiply by the inside's derivative, 2. The one half and the 2 cancel, leaving one over the square root of 2t plus 1.
The check is a neat trick worth remembering. Squaring removes the root, and differentiating the squared equation gives 2 s times s prime equals 2, so s prime is one over s. That is the same answer with no fractional exponents at all, and it is a preview of the implicit differentiation in Section 3.8.
Section
Part 3
Concept
Figure (svg): The sine curve drawn faintly from minus 2.2 to 2.2, with the piece from minus pi over 2 to pi over 2 drawn bold. Its reflection in the diagonal is the inverse sine, running from (minus 1, minus pi over 2) to (1, pi over 2).
\[ f(x) = \sin x,\; -\tfrac{\pi}{2} \le x \le \tfrac{\pi}{2} \quad\longleftrightarrow\quad g(x) = \sin^{-1}x,\; -1 \le x \le 1 \]
Sine repeats, so it has no inverse on the whole line. On the interval from minus pi over 2 to pi over 2 it rises steadily from minus 1 to 1, and that piece has an inverse: the inverse sine, also written arcsin.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 262 — Example 3.63, first line
Sine fails the horizontal line test badly: every value between minus 1 and 1 is taken infinitely often. To get an inverse, you keep only one piece of the graph on which sine rises steadily through every value once. The standard choice is the bold piece from minus pi over 2 to pi over 2.
Reflect that piece in the diagonal and you get the inverse sine, shown in orange. Its inputs are the sine values, from minus 1 to 1, and its outputs are angles, from minus pi over 2 to pi over 2. Every point you know on the sine curve gives one on the inverse: sine of pi over 6 is one half, so the inverse sine of one half is pi over 6.
Notice the endpoints. Sine is flat at the top and bottom of its bold piece, so by the mirror argument the inverse sine will have vertical tangents at 1 and minus 1.
Trap
Reading the minus one as a power:
\[ \sin^{-1}x = \frac{1}{\sin x} \]
\[ \frac{d}{dx}\sin^{-1}x = -\frac{\cos x}{\sin^2 x} \]
Wrong. That differentiates the cosecant.
On a function name, the minus one means inverse function: the angle whose sine is x. One over sine x is the cosecant, a different function with a different graph. Write arcsin whenever the notation might mislead.
\[ \sin^{-1}x = \arcsin x \]
\[ \frac{1}{\sin x} = \csc x \]
This notation causes endless trouble because the same symbol, a raised minus one, means two different things. On a number or on a variable it means a reciprocal. On a function name it means the inverse function.
So sine inverse of x is the angle whose sine is x, while one over sine of x is the cosecant. They have completely different graphs: the cosecant has vertical asymptotes and repeats forever, while the inverse sine lives only between minus 1 and 1. Their derivatives are nothing alike either.
Confusingly, sine squared of x does mean the square of sine of x. The minus one is the exception. When in doubt, write arcsin, which cannot be misread.
Prediction
Predict first
The derivative of a polynomial is a polynomial; the derivative of a trigonometric function is trigonometric. What kind of function do you expect the derivative of the inverse sine to be?
Correct: An algebraic function: powers and roots of x
Why: The theorem produces one over the cosine of the inverse sine of x, and a right triangle turns that cosine into a square root of 1 minus x squared. For the first time, a derivative leaves the family of the function it came from.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 262 — the paragraph before Example 3.63
Make a guess before revealing; the point of this slide is the surprise. Everything you have differentiated so far has stayed in its family. Polynomials give polynomials, and sines and cosines give sines and cosines.
The inverse sine breaks that pattern. The theorem gives one over the cosine of the inverse sine of x, which looks trigonometric, but the cosine of an angle whose sine you already know can be computed from a right triangle without any trigonometry at all. The result is a square root of 1 minus x squared.
So the derivative of the inverse sine is algebraic. That is why these functions matter so much later: in the integration chapters, expressions like one over the square root of 1 minus x squared have to be integrated, and the answer is an inverse trigonometric function.
Worked example
\[ g(x) = \sin^{-1}x, \quad -1 < x < 1 \]
Name the function it undoes
Why: Sine, on the restricted interval.
\[ f(x) = \sin x, \quad -\tfrac{\pi}{2} \le x \le \tfrac{\pi}{2} \]
Differentiate f
Why: From Section 3.5.
\[ f'(x) = \cos x \]
Evaluate at g of x
Why: The cosine of the angle whose sine is x.
\[ f'\big(g(x)\big) = \cos\big(\sin^{-1}x\big) \]
Replace the cosine using the triangle
Why: The next slide builds the triangle; its adjacent side is the square root of 1 minus x squared.
\[ \cos\big(\sin^{-1}x\big) = \sqrt{1-x^2} \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = \frac{d}{dx}\big(\sin^{-1}x\big) = \frac{1}{\sqrt{1-x^2}} \]
Check at two points
Why: At 0 the slope is 1, matching sine's slope of 1 at the mirror point (0, 0). At x equal to 0.6 the formula gives 1.25, and a difference quotient gives 1.2500000.
\[ g'(0) = 1 = \frac{1}{\cos 0}, \qquad g'(0.6) = \frac{1}{\sqrt{0.64}} = \frac{1}{0.8} = 1.25 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, pp. 262-263 — Example 3.63
This follows the same four moves as the cube-root example. The inverse sine undoes sine on the restricted piece; the derivative of sine is cosine; evaluate the cosine at the inverse sine of x; take the reciprocal.
The third move leaves you with the cosine of the inverse sine of x, which looks like it needs a calculator. It does not. The next slide draws the right triangle that turns it into the square root of 1 minus x squared, and the argument there is the one piece of this lesson you should be able to reproduce from memory.
The checks confirm the formula at two points. At zero the inverse sine has slope 1, the reciprocal of sine's slope of 1 at the origin, the mirror point. At 0.6 the formula gives 1.25, since the square root of 0.64 is 0.8.
Concept
Figure (svg): A right triangle with acute angle theta at the lower left, hypotenuse 1, side opposite theta of length x, and adjacent side of length the square root of 1 minus x squared. The right angle is at the lower right.
\[ \theta = \sin^{-1}x \iff \sin\theta = x, \quad 0 < \theta < \tfrac{\pi}{2} \]
\[ \text{adjacent} = \sqrt{1^2 - x^2} = \sqrt{1-x^2} \]
\[ \cos\big(\sin^{-1}x\big) = \cos\theta = \frac{\sqrt{1-x^2}}{1} = \sqrt{1-x^2} \]
Name the angle, draw it in a right triangle with the given ratio, let Pythagoras supply the missing side, and read off the ratio you need. This one move handles every inverse trigonometric derivative.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 263 — Figure 3.29
Here is the key move. Give the angle a name, theta, so that the sine of theta is x. Sine is opposite over hypotenuse, so draw a right triangle with hypotenuse 1 and the side opposite theta equal to x.
Pythagoras then gives the adjacent side: 1 squared minus x squared, square-rooted. Cosine is adjacent over hypotenuse, which is that square root over 1. So the cosine of the inverse sine of x is the square root of 1 minus x squared, with no trigonometric function left anywhere.
Notice that you never needed to know what theta actually is. You needed only its sine, which the problem gave you, and the fact that it sits in a right triangle. The same idea works for every inverse trigonometric function: put the given ratio on the right two sides, let Pythagoras supply the third, and read off the ratio you need.
Concept
The triangle needs an acute angle. The book's analysis checks that the same answer holds for every angle the inverse sine can produce.
\[ -\tfrac{\pi}{2} < \theta < 0: \quad \cos\theta = \cos(-\theta) = \sqrt{1-x^2} \]
\[ \theta = \pm\tfrac{\pi}{2}: \quad \cos\theta = 0 = \sqrt{1-(\pm 1)^2} \]
\[ \theta = 0: \quad \cos 0 = 1 = \sqrt{1-0^2} \]
The reason underneath: cosine is never negative from minus pi over 2 to pi over 2, so the positive square root is always the right one. At the two endpoints the cosine is zero, which is where the derivative formula breaks down.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 263 — Example 3.63, Analysis
The triangle argument only works for an acute angle, one between zero and pi over 2, because a right triangle cannot contain anything else. The inverse sine can also produce negative angles and the endpoints, so the book checks those separately.
For a negative angle, the cosine is unchanged when you flip the sign of the angle, since cosine is an even function, so the answer is still the positive square root. At the endpoints, where the angle is plus or minus pi over 2, the cosine is zero, and so is the square root of 1 minus 1. At zero both sides equal 1.
The single fact underneath all these cases is that cosine is never negative on the interval from minus pi over 2 to pi over 2. That is why the formula has a positive square root and no plus-or-minus.
Ranking
Put in order
Order the steps that turn cos of the inverse sine of x into an algebraic expression.
Why: The angle has to be named before it can be drawn, and the triangle has to exist before Pythagoras can fill it in. The sign check comes last because the triangle only covers acute angles; the range of the inverse sine is what guarantees the positive root for the rest.
Drag the steps into order before checking. Most of the order is forced: you cannot draw an angle before naming it, and Pythagoras needs two sides before it can supply the third.
The step people tend to leave out is the last one. The triangle gives the right answer for acute angles, and the sign check is what extends it to every angle the inverse sine can produce. On a written solution, one sentence noting that cosine is not negative on the range of the inverse sine is enough.
Intuition
The inverse sine is not a power, a root, or a ratio of them. Its derivative is. That surprise is exactly what makes these derivatives so useful later.
\[ \frac{d}{dx}\sin^{-1}x = \frac{1}{\sqrt{1-x^2}} \]
Read it backwards: an antiderivative of one over the square root of 1 minus x squared is the inverse sine. In the integration chapters, algebraic expressions like this one will be integrated by recognising them as inverse trigonometric derivatives.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 262 — the remark that the derivatives are algebraic
It is worth pausing on how odd this result is. The inverse sine is not built from powers and roots; you cannot write it with any finite combination of them. Yet its derivative is one over a square root, a perfectly ordinary algebraic expression.
The same will be true of all six inverse trigonometric functions. Their derivatives involve only x, squares and square roots. So when you read the formula backwards, it tells you that an antiderivative of one over the square root of 1 minus x squared is the inverse sine.
That reading backwards is how these derivatives are mostly used. In the integration chapters you will meet algebraic expressions whose antiderivatives are inverse trigonometric functions, and recognising them depends on knowing these formulas well.
Worked example
\[ g(x) = \tan^{-1}x \]
Name the function it undoes
Why: Tangent, restricted to the open interval where it rises through every real value.
\[ f(x) = \tan x, \quad -\tfrac{\pi}{2} < x < \tfrac{\pi}{2} \]
Differentiate f
Why: From Section 3.5.
\[ f'(x) = \sec^2 x \]
Evaluate at g of x
Why: The squared secant of the angle whose tangent is x.
\[ f'\big(g(x)\big) = \sec^2\big(\tan^{-1}x\big) \]
Use the Pythagorean identity
Why: Secant squared is 1 plus tangent squared, and the tangent of the inverse tangent of x is x.
\[ \sec^2\theta = 1 + \tan^2\theta = 1 + x^2 \]
Take the reciprocal
Why: Theorem 3.11.
\[ g'(x) = \frac{1}{1+x^2} \]
Figure (svg): A right triangle with acute angle theta at the lower left, adjacent side 1, opposite side x, and hypotenuse the square root of 1 plus x squared.
Check with the triangle and a number
Why: The triangle with sides 1 and x gives the same secant squared. At x equal to 2 the formula gives one fifth, and a difference quotient gives 0.2000000.
\[ \sec\theta = \frac{\sqrt{1+x^2}}{1} \;\Longrightarrow\; \sec^2\theta = 1 + x^2, \qquad g'(2) = \frac{1}{5} \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 263 — Checkpoint 3.45
The inverse tangent undoes tangent on the interval from minus pi over 2 to pi over 2, where tangent rises through every real number. The derivative of tangent is secant squared, so the theorem needs the secant squared of the inverse tangent of x.
There are two ways to compute that, and the slide shows both. The first uses the identity secant squared equals 1 plus tangent squared. The tangent of the inverse tangent of x is just x, so the result is 1 plus x squared. The second draws the triangle in the figure: opposite side x, adjacent side 1, hypotenuse the square root of 1 plus x squared, and the secant is hypotenuse over adjacent.
Either way, the derivative of the inverse tangent is one over 1 plus x squared. Notice that this denominator is never zero, which will matter when you look at domains.
Section
Part 4
Concept
The same argument, a named angle, a triangle and a reciprocal, gives all six. They come in three pairs.
\[ \frac{d}{dx}\sin^{-1}x = \frac{1}{\sqrt{1-x^2}} \qquad \frac{d}{dx}\cos^{-1}x = \frac{-1}{\sqrt{1-x^2}} \]
\[ \frac{d}{dx}\tan^{-1}x = \frac{1}{1+x^2} \qquad \frac{d}{dx}\cot^{-1}x = \frac{-1}{1+x^2} \]
\[ \frac{d}{dx}\sec^{-1}x = \frac{1}{|x|\sqrt{x^2-1}} \qquad \frac{d}{dx}\csc^{-1}x = \frac{-1}{|x|\sqrt{x^2-1}} \]
Learn the left column. Each function on the right is a co-function, and its derivative is the one on its left with a minus sign.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — Theorem 3.13, equations 3.22 to 3.27
The theorem lists all six. Rather than memorising six separate formulas, notice the structure: there are only three shapes, and each appears twice, once with a plus and once with a minus.
The left column is sine, tangent and secant. Their derivatives are one over the square root of 1 minus x squared, one over 1 plus x squared, and one over the absolute value of x times the square root of x squared minus 1. The right column holds the co-functions, cosine, cotangent and cosecant, and each has the same derivative as its partner with a minus sign.
The next slide explains why that minus sign appears, and after that you will derive the secant formula yourself, which is the one with the unusual absolute value.
Concept
Figure (svg): The inverse sine rising from minus pi over 2 to pi over 2 and the inverse cosine falling from pi to 0 over the interval from minus 1 to 1, with a dashed line at height pi over 2. Short tangent segments at x equals 0.6 have slopes 1.25 and minus 1.25.
\[ \sin^{-1}x + \cos^{-1}x = \frac{\pi}{2} \]
\[ \frac{d}{dx}\cos^{-1}x = \frac{d}{dx}\left(\frac{\pi}{2} - \sin^{-1}x\right) = -\frac{1}{\sqrt{1-x^2}} \]
The angle whose sine is x and the angle whose cosine is x are the two acute angles of the same right triangle, so they add to a right angle. A constant minus a function has the negative of its derivative. The same identity links the inverse tangent with the inverse cotangent, and the inverse secant with the inverse cosecant.
Stewart, Calculus: Early Transcendentals 8e, §3.5 Implicit Differentiation §3.5, pp. 208-217 — the inverse trigonometric derivatives in Stewart
Look at the figure. The orange inverse cosine is the blue inverse sine flipped upside down about the dashed line at pi over 2. At every x the two heights add to exactly pi over 2.
The reason is a right triangle again. If the sine of one acute angle is x, then the cosine of the other acute angle is also x, and the two acute angles of a right triangle add to a right angle. So the inverse cosine of x equals pi over 2 minus the inverse sine of x.
Differentiate that and the constant disappears, leaving the negative of the inverse sine's derivative. The two short tangent segments at 0.6 show it: slopes of 1.25 and minus 1.25. The same identity, with tangent and cotangent or with secant and cosecant, explains the other two minus signs.
Worked example
Theorem 3.13 states this one; derive it the same way to see where the absolute value comes from.
\[ y = \sec^{-1}x \iff \sec y = x, \quad 0 \le y \le \pi,\; y \ne \tfrac{\pi}{2} \]
Differentiate the original
Why: The derivative of the secant is secant times tangent.
\[ f(y) = \sec y, \quad f'(y) = \sec y\,\tan y \]
Take the reciprocal
Why: Theorem 3.11.
\[ \frac{dy}{dx} = \frac{1}{\sec y\,\tan y} \]
Replace the secant
Why: By definition of y.
\[ \sec y = x \]
Replace the tangent, up to sign
Why: From tangent squared plus 1 equals secant squared.
\[ \tan^2 y = \sec^2 y - 1 = x^2 - 1 \;\Longrightarrow\; \tan y = \pm\sqrt{x^2-1} \]
Settle the sign
Why: Below pi over 2 the secant and tangent are both positive; above it both are negative. Either way their product is positive.
\[ \sec y\,\tan y = |x|\sqrt{x^2-1} \]
Assemble
Why: Put the product in the denominator.
\[ \frac{d}{dx}\sec^{-1}x = \frac{1}{|x|\sqrt{x^2-1}} \]
Check at x equal to minus 2
Why: The inverse secant of minus 2 is two thirds of pi, where the secant is minus 2 and the tangent is minus root 3; their product is 2 root 3, positive. A difference quotient gives 0.2886751.
\[ \frac{1}{|-2|\sqrt{3}} = \frac{1}{2\sqrt3} \approx 0.2886751 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — Theorem 3.13, equation 3.26
This derivation follows the same pattern as the inverse sine, and it shows you where the absolute value in the formula comes from. The derivative of the secant is secant times tangent, so the theorem gives one over the secant of y times the tangent of y, where y is the inverse secant of x.
The secant of y is x, by definition. The tangent needs the identity tangent squared equals secant squared minus 1, which gives x squared minus 1, but only up to sign: the tangent could be the positive or the negative square root.
Here is the subtle part. The inverse secant's outputs run from 0 to pi, skipping pi over 2. Below pi over 2 the secant and the tangent are both positive. Above it they are both negative. So their product is always positive, and the way to write a positive product is with the absolute value of x. The check at minus 2 confirms a positive slope.
Notation
Annotate
On: \( \frac{d}{dx}\sec^{-1}x = \frac{1}{|x|\sqrt{x^2-1}} \)
Step through the annotations and connect each piece of the formula to something you can see in the graph of the inverse secant, which is on a later slide.
The absolute value is the piece most often forgotten. Both branches of the inverse secant rise from left to right, so its derivative is positive everywhere it exists, including for negative x. Without the absolute value the formula would give negative slopes on the left branch, which the graph flatly contradicts.
The square root of x squared minus 1 is the triangle's third side. It vanishes at x equal to plus or minus 1, which is where the graph has vertical tangents, and for x between minus 1 and 1 the function is not defined at all.
Trap
Remembering the formula without its bars:
\[ \frac{d}{dx}\sec^{-1}x = \frac{1}{x\sqrt{x^2-1}} \]
\[ x = -2: \quad \frac{1}{-2\sqrt 3} < 0 \]
Wrong for every negative x.
The inverse secant rises on both of its branches, so its slope is positive everywhere it exists. A negative value at x equal to minus 2 contradicts the graph. The absolute value is what fixes the sign.
\[ \frac{1}{|-2|\sqrt 3} = \frac{1}{2\sqrt3} > 0 \]
This error is easy to make and easy to catch, if you have a picture of the graph in mind. Without the absolute value the formula gives a negative derivative at minus 2, which would mean the inverse secant is falling there.
It is not. The inverse secant's left branch climbs from pi over 2 toward pi as x moves from far left toward minus 1. Rising means a positive slope. The absolute value is the formula's way of recording that fact.
A good general habit: whenever you use a derivative formula, test it against the sign you expect from the graph at one convenient point. That one-second check catches a surprising number of mistakes.
Matching
Match the pairs
Why: Three shapes, each appearing twice. The sine, tangent and secant take the positive versions; each co-function takes the same expression with a minus sign, because it equals pi over 2 minus its partner.
Pair each function with its derivative before checking. You do not need six separate facts: identify which of the three shapes goes with each pair of functions, then decide the sign.
Sine and cosine share the shape with the square root of 1 minus x squared, because their triangles have hypotenuse 1. Tangent and cotangent share 1 plus x squared, because their triangles have an adjacent side of 1 and no square root of a difference. Secant and cosecant share the absolute value form.
Then the sign: the plain functions get plus, the co-functions get minus. If you remember that the inverse cosine falls as x increases, you can always reconstruct which one carries the minus sign.
Worked example
Apply the chain rule to the inverse sine formula, then use the result on the inverse sine of 2x cubed.
\[ h(x) = \sin^{-1}\big(g(x)\big) \]
Chain rule: outer derivative at the inside, times the inner derivative
Why: The outer function is the inverse sine.
\[ h'(x) = \frac{1}{\sqrt{1-\big(g(x)\big)^2}}\cdot g'(x) \]
Choose the inside function
Why: For the second part.
\[ g(x) = 2x^3, \quad g'(x) = 6x^2 \]
Square the inside
Why: Two squared is four, and x cubed squared is x to the sixth.
\[ \big(2x^3\big)^2 = 4x^6 \]
Substitute
Why: Put both pieces into the chain-rule formula.
\[ h'(x) = \frac{1}{\sqrt{1-4x^6}}\cdot 6x^2 = \frac{6x^2}{\sqrt{1-4x^6}} \]
Check at x equal to one half
Why: The book's text says the formula came from Example 3.61; it came from Example 3.63, the inverse sine. The value checks against a difference quotient, 1.5491933.
\[ h'\left(\tfrac12\right) = \frac{6/4}{\sqrt{1 - 4/64}} = \frac{1.5}{\sqrt{0.9375}} \approx 1.5491933 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 263 — Example 3.64
The six formulas are stated for a bare x. When the input is some other function, the chain rule adds the inside's derivative as a factor. The first line of the solution writes that out once in general, so you can use it for any inside function.
For the inverse sine of 2x cubed, the inside is 2x cubed and its derivative is 6 x squared. The square of the inside is 4 x to the sixth, which goes under the square root. Multiply, and the derivative is 6 x squared over the square root of 1 minus 4 x to the sixth.
One small correction to the book. Its text says to apply the chain rule to the formula derived in Example 3.61, but Example 3.61 is the cube root. The inverse sine formula is from Example 3.63, and that is the one used here.
Trap
Using the formula as if the inside were just x:
\[ \frac{d}{dx}\sin^{-1}\big(2x^3\big) = \frac{1}{\sqrt{1-4x^6}} \]
Wrong. The chain rule has been left out.
Every one of the six formulas is stated for a bare x. When the input is a function, multiply by that function's derivative, here 6x squared. A quick test: at x equal to 0 the wrong answer gives 1, but the function is flat there.
\[ \frac{d}{dx}\sin^{-1}u = \frac{u'}{\sqrt{1-u^2}} \]
This is the most common error with inverse trigonometric derivatives, and it is the same chain-rule slip you have seen with every other function. The formula for the outer function gets written down correctly, with the inside substituted, and then the inside's derivative is forgotten.
A quick test catches it here. At x equal to zero, 2x cubed is flat, so the inverse sine of it is flat too, and the derivative must be zero. The wrong answer gives 1 at zero; the right answer, with its factor of 6 x squared, gives zero. Whenever the inside has a flat spot, the whole derivative must vanish there, so it makes a good checkpoint.
Worked example
\[ f(x) = \tan^{-1}\big(x^2\big) \]
Name the inside
Why: The inverse tangent's input is x squared.
\[ g(x) = x^2, \quad g'(x) = 2x \]
Substitute into equation 3.24 with the chain rule
Why: Outer derivative at the inside, times the inner derivative.
\[ f'(x) = \frac{1}{1 + \big(x^2\big)^2}\cdot 2x \]
Simplify
Why: The square of x squared is x to the fourth.
\[ f'(x) = \frac{2x}{1+x^4} \]
Check at x equal to 2
Why: The formula gives 4 over 17, about 0.2352941, and a difference quotient agrees; at 0 the slope is 0, where x squared has its minimum.
\[ f'(2) = \frac{4}{17} \approx 0.2352941, \qquad f'(0) = 0 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — Example 3.65
Name the inside first: x squared, with derivative 2x. Then use the inverse tangent formula with the inside substituted for x, and multiply by the inside's derivative.
The only simplification is that the square of x squared is x to the fourth. So the derivative is 2x over 1 plus x to the fourth.
The checks are quick. At 2 the formula gives 4 over 17, which a difference quotient confirms. At zero it gives zero, which makes sense: the inside, x squared, has a minimum at zero, so the whole function does too, and its tangent there is flat. Notice that the denominator is never zero, so this derivative exists everywhere, like the inverse tangent's own.
Worked example
\[ h(x) = x^2\sin^{-1}x \]
Recognise a product
Why: Product rule: first's derivative times second, plus first times second's derivative.
\[ h'(x) = \big(x^2\big)'\sin^{-1}x + x^2\big(\sin^{-1}x\big)' \]
Differentiate the square
Why: Power rule.
\[ \big(x^2\big)' = 2x \]
Differentiate the inverse sine
Why: Equation 3.22.
\[ \big(\sin^{-1}x\big)' = \frac{1}{\sqrt{1-x^2}} \]
Assemble
Why: Put the pieces back into the product rule.
\[ h'(x) = 2x\sin^{-1}x + \frac{x^2}{\sqrt{1-x^2}} \]
Check at x equal to one half
Why: The inverse sine of one half is pi over 6. The formula gives 0.8122739, matching a difference quotient.
\[ h'\left(\tfrac12\right) = \frac{\pi}{6} + \frac{1/4}{\sqrt{3/4}} \approx 0.5235988 + 0.2886751 = 0.8122739 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — Example 3.66
Before reaching for any inverse trigonometric formula, look at the outermost structure. Here it is a product of two functions, x squared and the inverse sine of x, so the product rule comes first.
The product rule needs both derivatives. The derivative of x squared is 2x, and the derivative of the inverse sine is the formula from Example 3.63. Put them in the product-rule pattern and the answer is 2x times the inverse sine of x, plus x squared over the square root of 1 minus x squared.
There is no chain rule this time, because the inverse sine's input is plain x. The check at one half uses the fact that the inverse sine of one half is pi over 6, a value worth knowing by heart.
Worked example
\[ h(x) = \cos^{-1}(3x-1) \]
Name the inside
Why: A linear inside, so the inner derivative is a constant.
\[ u = 3x - 1, \quad u' = 3 \]
Use equation 3.23 with the chain rule
Why: The inverse cosine's derivative carries a minus sign.
\[ h'(x) = \frac{-1}{\sqrt{1-(3x-1)^2}}\cdot 3 \]
Expand under the root
Why: One minus the square of 3x minus 1.
\[ 1 - (9x^2 - 6x + 1) = 6x - 9x^2 \]
Simplify
Why: Collect the constant.
\[ h'(x) = -\frac{3}{\sqrt{6x-9x^2}} \]
Check the domain and a value
Why: The inside must lie strictly between minus 1 and 1, which means x between 0 and two thirds. At x equal to one half the formula gives minus 3.4641016, matching a difference quotient.
\[ 0 < x < \tfrac23, \qquad h'\left(\tfrac12\right) = -\frac{3}{\sqrt{0.75}} \approx -3.4641016 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — Checkpoint 3.46
This combines two things from this part of the lesson: the minus sign that comes with a co-function, and the chain rule. The inside, 3x minus 1, has derivative 3, so the answer is minus 3 over the square root of 1 minus the square of 3x minus 1.
Expanding under the root tidies it up. The square of 3x minus 1 is 9 x squared minus 6x plus 1, and subtracting that from 1 leaves 6x minus 9 x squared.
The domain check is worth doing. The inverse cosine's derivative needs its input strictly between minus 1 and 1, so 3x minus 1 must be in that range, which means x between zero and two thirds. You can see the same thing in the simplified form: 6x minus 9 x squared is positive exactly on that interval.
Sorting
Sort into buckets
Sort each function by the rule you apply first when differentiating it.
Sort each item by the first rule you would reach for. The skill being practised is reading the outermost structure of an expression before differentiating anything, which is what decides the rule.
Two of the items are fractional powers. The cube root of x to the fifth is just x to the five thirds, a direct use of the rational power rule. The two-thirds power of x squared plus 1 also needs the chain rule, but the power is still the outermost operation.
Two are inverse trigonometric functions of something other than x, so the formula with the chain rule applies. The last two are products, so the product rule comes first, and the inverse trigonometric formula appears inside it as one of the pieces.
Section
Part 5
Picture it
Figure (svg): The inverse sine from minus 1 to 1 together with its derivative, 1 over the square root of 1 minus x squared, which equals 1 at the centre and shoots up without bound near x equals minus 1 and x equals 1; dashed vertical lines mark the endpoints.
The inverse sine is defined at minus 1 and at 1, but its derivative is not: the square root in the denominator is zero there. The graph shows why, with a tangent that turns vertical at each end, the mirror image of sine's flat tangents at minus pi over 2 and pi over 2.
\[ \text{domain of } \sin^{-1}: [-1, 1], \qquad \text{domain of its derivative}: (-1, 1) \]
Look at the two curves together. The blue inverse sine is defined at both ends, at minus 1 and 1, where it reaches minus pi over 2 and pi over 2. But the orange curve, its derivative, heads off the top of the picture as x approaches either end.
The mirror argument explains it. Sine is flat at the top and bottom of its restricted piece, and flat tangents reflect into vertical ones. The formula agrees, since the square root of 1 minus x squared is zero at the ends and you cannot divide by zero.
So the inverse sine is a function whose derivative has a smaller domain than the function itself: the closed interval for the function, but only the open interval for the derivative.
Estimation
\[ \frac{d}{dx}\sin^{-1}x \Big|_{x=0.99} = ? \]
Predict first
Guess before computing: how steep is the inverse sine at x equal to 0.99, one hundredth from its right-hand end?
Correct: About 7
Why: One minus 0.99 squared is 0.0199, whose square root is about 0.141, so the slope is about 7.09. Steep, but still finite: the derivative grows without bound only as x reaches 1 itself, and it grows like one over a square root, slowly at first.
\[ \frac{1}{\sqrt{1 - 0.99^2}} = \frac{1}{\sqrt{0.0199}} \approx 7.09 \]
Make your guess before revealing. It is tempting to think that 0.99 is so close to 1 that the slope must be enormous.
In fact 1 minus 0.99 squared is about 0.02, and its square root is about 0.14, so the slope is about 7. Steep, but not wildly so. The derivative does grow without bound, but it grows like one over a square root of the distance to the end, which is slow at first.
To make the slope 50, you would need to get within about two ten-thousandths of 1. Blowing up at an endpoint does not mean being large everywhere near it.
Concept
Figure (svg): The inverse secant, defined only for x at most minus 1 and x at least 1: the right branch rises from (1, 0) toward the dashed line pi over 2, the left branch climbs from the dashed line toward (minus 1, pi). Its derivative, drawn in a second colour, is positive on both branches and unbounded near the two endpoints. The strip between minus 1 and 1 is shaded: no graph there.
| function | domain of the function | domain of the derivative |
|---|---|---|
| sin⁻¹ x, cos⁻¹ x | −1 ≤ x ≤ 1 | −1 < x < 1 |
| tan⁻¹ x, cot⁻¹ x | all real x | all real x |
| sec⁻¹ x, csc⁻¹ x | |x| ≥ 1 | |x| > 1 |
In every case the endpoints are lost for the same reason: the original trigonometric function has a horizontal tangent at the matching angle, and the reflection turns it vertical. That is Theorem 3.11's condition at work again.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, pp. 263-264 — the endpoint cases of Example 3.63 and Theorem 3.13
Look at the graph of the inverse secant first. It has two branches, one for x at least 1 and one for x at most minus 1, and nothing in between, which is the shaded strip. Both branches rise, and both start with a vertical tangent: at (1, 0) and at (minus 1, pi).
The table collects the domains for all six functions. The inverse sine and cosine lose their two endpoints; the inverse secant and cosecant lose theirs too; and the inverse tangent and cotangent lose nothing, because they have no endpoints to lose.
The pattern has one explanation. At each lost endpoint, the original trigonometric function has a horizontal tangent at the matching angle, and Theorem 3.11 needs a nonzero slope. Where the original is flat, the inverse is vertical and has no derivative.
Concept
Figure (svg): The inverse tangent rising from near minus pi over 2 to near pi over 2 across x from minus 6 to 6, between two dashed horizontal asymptotes, and its derivative 1 over 1 plus x squared, a smooth bell of height 1 at the origin.
\[ 1 + x^2 \ge 1 > 0 \quad \text{for every real } x \]
The inverse tangent's derivative has no square root and no zero denominator, so it exists at every real number. The slope is 1 at the origin and fades toward zero as the curve levels off at its two asymptotes, the mirror images of tangent's vertical asymptotes.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 264 — equation 3.24
The inverse tangent is the well-behaved member of the family. Its domain is every real number, its outputs stay between minus pi over 2 and pi over 2, and it has a derivative at every single point.
The formula shows why: 1 plus x squared is at least 1 for every x, so the denominator never vanishes and there is no square root to go undefined. The orange bell curve is the derivative, largest at the origin where the inverse tangent climbs most steeply.
The horizontal asymptotes are the mirror images of tangent's vertical asymptotes at plus and minus pi over 2. As the inverse tangent levels off toward them, its slope fades toward zero, which is exactly what the bell curve does in its tails.
Comparison
Comparison matrix
| function | derivative | where the derivative exists |
|---|---|---|
| sin⁻¹ x | 1/√(1 − x²) | -1 < x < 1 |
| cos⁻¹ x | -1/√(1 − x²) | −1 < x < 1 |
| tan⁻¹ x | 1/(1 + x²) | all real x |
| sec⁻¹ x | 1/(|x|√(x² − 1)) | |x| > 1 |
| sin⁻¹(x/2) | 1/√(4 − x²) | -2 < x < 2 |
Fill in each blank before checking. For the derivatives, use the three shapes and the sign rule. For the domains, ask where the derivative's formula makes sense: a square root must have a positive quantity under it when it sits in a denominator, and there must be no division by zero.
The last row is a new function, the inverse sine of x over 2. Its derivative, by the chain rule, simplifies to one over the square root of 4 minus x squared, and that needs x squared less than 4, so x between minus 2 and 2. Stretching the input stretches the domain of the derivative with it.
Worked example
The position of a particle at time t is given below, for t at least one half. Find the velocity at t equal to 1.
\[ s(t) = \tan^{-1}\left(\frac1t\right), \quad t \ge \tfrac12 \]
Differentiate with the chain rule
Why: Inverse tangent outside, one over t inside, whose derivative is minus one over t squared.
\[ v(t) = s'(t) = \frac{1}{1 + \left(\frac1t\right)^2}\cdot\frac{-1}{t^2} \]
Multiply the denominators
Why: t squared times 1 plus one over t squared is t squared plus 1.
\[ v(t) = -\frac{1}{t^2 + 1} \]
Evaluate at t equal to 1
Why: One plus one is two.
\[ v(1) = -\frac12 \]
Figure (svg): The position s of t equals the inverse tangent of 1 over t for t from one half to 6, falling from about 1.11 toward 0, with the tangent line of slope minus one half at the point (1, pi over 4).
Check with an identity
Why: For positive t the angles whose tangents are t and one over t add to pi over 2, so s is pi over 2 minus the inverse tangent of t, and its derivative is minus one over 1 plus t squared.
\[ s(t) = \frac{\pi}{2} - \tan^{-1}t \;\Longrightarrow\; s'(t) = -\frac{1}{1+t^2}, \quad s'(1) = -\frac12 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, pp. 264-265 — Example 3.67
Velocity is the derivative of position, so differentiate s with the chain rule. The outer function is the inverse tangent, and its input is one over t, whose derivative is minus one over t squared.
The simplification is worth doing carefully. Multiply the t squared into the other denominator: t squared times 1 plus one over t squared is t squared plus 1. So the velocity is minus one over t squared plus 1, and at t equal to 1 it is minus one half.
The figure shows the position falling, and the tangent line at t equal to 1 with slope minus one half. The check uses an identity: for positive t, the inverse tangent of one over t equals pi over 2 minus the inverse tangent of t, because the two angles are the acute angles of one right triangle. Differentiating that gives the same velocity in one line.
Prediction
\[ f(x) = \sin^{-1}x, \quad x = 0 \]
Predict first
Which line is tangent to the graph of the inverse sine at x equal to 0?
Correct: y = x
Why: The inverse sine of 0 is 0, so the point is the origin, and the slope there is one over the square root of 1, which is 1. The tangent is the line y equals x, the mirror image of sine's own tangent at the origin, which is also y equals x.
\[ f(0) = 0, \quad f'(0) = \frac{1}{\sqrt{1-0}} = 1 \;\Longrightarrow\; y = x \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 265 — Checkpoint 3.47
Commit to one option before revealing. You need the point and the slope. The inverse sine of zero is zero, so the point is the origin, and the slope there is one over the square root of 1, which is 1.
So the tangent line is y equals x. That is also sine's own tangent line at the origin, and it is no coincidence: the diagonal is its own reflection, so a tangent along the diagonal reflects onto itself. Near zero, both sine and the inverse sine are very close to x, which is why engineers often replace the inverse sine of a small number by the number itself.
Worked example
A camera on the ground is 2000 feet from a launch pad. The angle of elevation to a rocket at height x is given below. Find how fast the angle changes with height when the camera and the rocket are 5000 feet apart.
\[ \theta = \tan^{-1}\left(\frac{x}{2000}\right) \]
Differentiate with the chain rule
Why: The inside is x over 2000, whose derivative is one over 2000.
\[ \frac{d\theta}{dx} = \frac{1}{1 + \left(\frac{x}{2000}\right)^2}\cdot\frac{1}{2000} \]
Clear the compound fraction
Why: Multiply the top and bottom by 2000 squared.
\[ \frac{d\theta}{dx} = \frac{2000}{2000^2 + x^2} \]
Read the distance from the triangle
Why: The line of sight is the hypotenuse, so 2000 squared plus x squared is 5000 squared; x itself is not needed.
\[ 2000^2 + x^2 = 5000^2 = 25{,}000{,}000 \]
Figure (svg): A to-scale sketch: a camera on the ground 2000 feet from the launch pad, a rocket straight above the pad at height x of about 4583 feet, and the 5000-foot line of sight from camera to rocket making angle theta with the ground.
Evaluate
Why: Two thousand over twenty-five million.
\[ \frac{d\theta}{dx} = \frac{2000}{25{,}000{,}000} = 0.00008 \text{ rad/ft} \]
Check with a small step
Why: The height is the square root of 21 million, about 4582.576 feet. Raising the rocket one more foot changes the angle by 0.0000800 radians, as predicted.
\[ \tan^{-1}\tfrac{4583.576}{2000} - \tan^{-1}\tfrac{4582.576}{2000} \approx 0.0000800 \]
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 267 — Exercise 298
This is a rate of change question in a physical setting. The angle depends on the rocket's height through an inverse tangent, and the question asks how quickly the angle changes as the height changes.
The chain rule gives one over 1 plus x over 2000 squared, times one over 2000. Clearing the compound fraction by multiplying top and bottom by 2000 squared gives 2000 over 2000 squared plus x squared.
Now look at the figure. The expression 2000 squared plus x squared is the square of the line of sight, by Pythagoras, and the question tells you that distance is 5000 feet. So you never need to find the height itself. The rate is 2000 over 25 million, eight hundred-thousandths of a radian per foot, and the check with a one-foot step confirms it.
Real world
Figure (svg): The viewing angle theta, in radians, plotted against the seat distance x from 0 to 60 feet: it rises steeply, peaks at about 0.6435 at x equals 20, marked by a dashed vertical line, then falls slowly.
Discussion prompt
A 30-foot screen starts 10 feet above your eye level. Sitting x feet from the screen, your viewing angle is the inverse cotangent of x over 40 minus the inverse cotangent of x over 10. Differentiate, and find the seat distance where the angle stops growing.
OpenStax Calculus Volume 1, §3.7 Derivatives of Inverse Functions §3.7, p. 267 — Exercise 299
The picture shows how your viewing angle changes as you move back from the screen. Sitting very close, the screen is high above you and appears thin; sitting far away, the whole screen shrinks. Somewhere in between the angle is largest.
Differentiate each inverse cotangent with the chain rule. Each contributes a term like the inverse cotangent's derivative, which carries a minus sign, times the inside's derivative, one over 40 or one over 10. Simplifying gives the two fractions on the slide.
Setting the derivative to zero and cross-multiplying gives 30 x squared equal to twelve thousand, so x is 20 feet. The signs of the derivative on either side confirm that it is a maximum: positive before 20, negative after. You will learn to treat questions like this systematically in Chapter 4.
Section
Part 6
Pattern
Figure (svg): A flow diagram: start from the input a of the inverse; solve f of b equals a for the matching input b; compute f prime of b; if it is zero, the inverse has a vertical tangent and no derivative there; otherwise the answer is one over f prime of b.
This is the procedure in one place. The flow diagram shows the version for the slope at a single point, which is what most exam questions ask for. The middle box is the one that matters: find the input b that the original sends to a, because that is where the original's derivative is evaluated.
When you need a formula for the inverse's derivative everywhere, use the second form of the theorem and simplify the original's derivative at g of x. For inverse trigonometric functions, a right triangle does that simplification.
Once the formulas are established, you rarely need the theorem directly. Rational powers follow the ordinary power rule, and the six inverse trigonometric functions have their formulas. The chain rule, the co-function minus signs and the secant pair's absolute value are the three details to watch.
Check
Check your understanding
A differentiable one-to-one function has f(2) = 7 and f′(2) = 4. What is (f⁻¹)′(7)?
Answer: A
Why: Since f sends 2 to 7, the inverse sends 7 to 2, so the theorem uses the original's slope at 2: the inverse's slope at 7 is one over 4, which is one quarter.
The data give you a point, (2, 7), on the graph of f, and the slope there. The mirror point on the inverse is (7, 2), and the inverse's slope there is the reciprocal of 4.
The distractors are the errors from earlier in the lesson. The answer 4 forgets to take the reciprocal. The option saying it cannot be found is the wrong-input error: it assumes the theorem needs the derivative at 7, when in fact it needs the derivative at 2, the inverse's output, which is given.
Check
Check your understanding
What is the derivative of tan⁻¹(3x)?
Answer: B
Why: The inverse tangent's derivative evaluated at the inside is one over 1 plus 9 x squared, since 3x squared is 9 x squared, and the chain rule multiplies by the inside's derivative, 3.
The derivative of the inverse tangent at the inside is one over 1 plus the inside squared, and the inside here is 3x. Its square is 9 x squared, not 3 x squared, since both the 3 and the x get squared.
Then the chain rule contributes the inside's derivative, 3, as a factor in the numerator. The option with secant squared is the notation trap from earlier: it treats the inverse tangent as if it were the tangent.
Check
Check your understanding
On which set is sin⁻¹ x differentiable?
Answer: B
Why: The derivative is one over the square root of 1 minus x squared, which needs 1 minus x squared to be positive. At the endpoints the denominator is zero and the graph has vertical tangents, so they are excluded.
The question is about the derivative, not the function. The inverse sine is defined on the closed interval from minus 1 to 1, but its derivative has the square root of 1 minus x squared in its denominator, which is zero at both ends.
So the derivative exists only strictly between minus 1 and 1. Geometrically, the graph has vertical tangents at the two endpoints, the reflections of sine's flat tangents at minus pi over 2 and pi over 2. The interval where the absolute value of x exceeds 1 belongs to the inverse secant.
Explain it to yourself
Discussion prompt
In two or three sentences, explain why reflecting a graph in the line y = x turns a slope into its reciprocal, and what that means at a point where the original has a horizontal tangent.
Write your explanation before revealing. Being able to say this in plain words means you can rebuild the theorem, and its condition, without memorising either.
The explanation has three parts. Reflection swaps coordinates, so it swaps rise and run and turns every slope into its reciprocal. Tangent lines reflect to tangent lines, so this applies to the slope of the curve itself at matching points. And a slope of zero has no reciprocal, which is why a horizontal tangent on the original becomes a vertical tangent on the inverse, where there is no derivative.
Exit ticket
\[ \text{(a) } f(x) = x^5 + x + 1: \; \text{find } \big(f^{-1}\big)'(3) \qquad \text{(b) } \frac{d}{dx}\tan^{-1}\sqrt{x} \]
Discussion prompt
Answer both parts. In (a), say which number the original's derivative is evaluated at, and why.
Both parts use this lesson's main ideas in their most common forms. In part (a), you cannot realistically invert a fifth-degree polynomial, so find the input that produces 3 by trying small integers. f of 1 is 3, so the original's derivative is evaluated at 1, where it equals 6, and the answer is one sixth.
In part (b), the inverse tangent formula at the inside gives one over 1 plus x, since the square of the square root of x is x, and the chain rule multiplies by the derivative of the square root, one over twice the square root of x.
If you evaluated the derivative in part (a) at 3, go back to the error-analysis slide; that one slip is the thing this lesson most wants you to stop making.
Recap
| result | formula | watch for |
|---|---|---|
| inverse function theorem | (f⁻¹)′(a) = 1/f′(b), where f(b) = a | evaluate f′ at b, not at a; needs f′(b) ≠ 0 |
| rational power rule | d/dx x^(m/n) = (m/n) x^((m/n) − 1) | subtract 1 from the whole fraction |
| inverse sine and cosine | ±1/√(1 − x²) | only for −1 < x < 1 |
| inverse tangent and cotangent | ±1/(1 + x²) | defined for every x |
| inverse secant and cosecant | ±1/(|x|√(x² − 1)) | the absolute value; only for |x| > 1 |
\[ \big(f^{-1}\big)'(x) = \frac{1}{f'\big(f^{-1}(x)\big)} \]
Next, Section 3.8 differentiates equations that never get solved for y at all, using the same chain-rule move you used on the cancellation equation.
Stewart, Calculus: Early Transcendentals 8e, §3.5 Implicit Differentiation §3.5, pp. 208-217 — the same material in Stewart
The whole lesson grew from one line: differentiate f of the inverse of x equals x with the chain rule, and divide. That gives the inverse function theorem, with its picture of reflected tangent lines and reciprocal slopes, and its one condition, that the original's slope at the matching point is not zero.
Two families of results followed. The roots gave the power rule for every rational exponent, now proved rather than assumed. The inverse trigonometric functions gave six derivatives in three pairs, all of them algebraic, with domains that lose any endpoint where the original trigonometric function is flat.
Section 3.8 takes the move you used on the cancellation equation, differentiating both sides of an equation that has not been solved for y, and turns it into a general method called implicit differentiation.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.