Calculus › Section F · Logarithmic, exponential, and other transcendental functions › Topic F.7

Inverse trigonometric functions: derivatives

Sine is not one-to-one, so an inverse sine can only exist on a restricted piece. Built that way, arcsin and arctan work perfectly, and their derivatives turn out to be clean algebraic fractions.

Inverting an oscillating function

The sine keeps repeating values, so it is not one-to-one, and F.3 showed that only a one-to-one function has an inverse. The repair is the one we used for \(x^2\): restrict the domain. Chop sine to the interval \(\left[-\frac{\pi}{2}, \frac{\pi}{2}\right]\), one full monotonic sweep from \(-1\) to \(1\), and the restricted function is strictly increasing, hence invertible. Its inverse is arcsine, also written \(\sin^{-1}\) or \(\arcsin\): \(\arcsin x\) is the angle in \(\left[-\frac{\pi}{2}, \frac{\pi}{2}\right]\) whose sine is \(x\). The other trig functions are inverted the same way, each on its own chosen interval. The value an inverse returns from that interval is called the principal value.

Definitions and principal values

The three to know. \(\arcsin x\): domain \([-1, 1]\), range \(\left[-\frac{\pi}{2}, \frac{\pi}{2}\right]\). \(\arccos x\): domain \([-1, 1]\), range \([0, \pi]\), cosine decreasing through its sweep. \(\arctan x\): domain all real numbers, range \(\left(-\frac{\pi}{2}, \frac{\pi}{2}\right)\), which makes it the only one of the three accepting every input. So \(\arcsin\frac{1}{2} = \frac{\pi}{6}\), \(\arctan 1 = \frac{\pi}{4}\), and \(\arctan\) carries two built-in horizontal asymptotes at \(\pm\frac{\pi}{2}\). Every answer an inverse trig function returns is required to land inside its range, and that discipline settles sign questions all by itself.

Worked example: deriving arcsine's derivative

Let \(y = \arcsin x\), so \(\sin y = x\) with the principal range on \(y\). Differentiate implicitly (C.5's machine): \(\cos y \cdot \frac{dy}{dx} = 1\), hence \(\frac{dy}{dx} = \frac{1}{\cos y}\). Now translate back to \(x\). On the principal range, cosine is non-negative, so \(\cos y = \sqrt{1 - \sin^2 y} = \sqrt{1 - x^2}\):

\frac{d}{dx} \arcsin x = \frac{1}{\sqrt{1 - x^2}}, \qquad -1 < x < 1

The sign choice is decided by the principal range, which is precisely why the restriction in the first card was not decoration. Note the endpoints: the slope grows without bound as \(|x| \to 1\). There sine has slope 0 (at \(y = \pm\frac{\pi}{2}\)), so by F.3's reciprocal rule the inverse has vertical tangents.

The three derivative rules

The same implicit manoeuvre produces the others. For arctangent, differentiating \(\tan y = x\) gives \(\sec^2 y \cdot \frac{dy}{dx} = 1\), and since \(\sec^2 y = 1 + \tan^2 y = 1 + x^2\),

\frac{d}{dx} \arctan x = \frac{1}{1 + x^2}

The remaining one to remember: \(\frac{d}{dx}\arccos x = -\frac{1}{\sqrt{1 - x^2}}\), minus arcsine's, because \(\arccos x = \frac{\pi}{2} - \arcsin x\) on their shared domain (an identity explained two cards below). All three derivatives are algebraic, with no trig left in them. Read backwards, they give antiderivatives of \(\frac{1}{1 + x^2}\) and \(\frac{1}{\sqrt{1 - x^2}}\), which F.8 develops.

Worked example: chain forms

Differentiate \(y = \arctan(3x)\). The chain form is \(\frac{d}{dx}\arctan u = \frac{u'}{1 + u^2}\), so \(y' = \frac{3}{1 + 9x^2}\). Second: differentiate \(y = \arcsin(x^2)\). Same reading: \(y' = \frac{2x}{\sqrt{1 - x^4}}\), and the domain works out cleanly: the inside must lie in \((-1, 1)\), so \(|x| < 1\), and the derivative exists on exactly that interval. Both follow the same pattern: apply the basic rule to the inside, then multiply by the derivative of the inside.

The complementary identities

Two facts smooth the algebra later. First: \(\arcsin x + \arccos x = \frac{\pi}{2}\) for every \(x\) in \([-1, 1]\). For \(0 < x < 1\) you can see it in a right triangle with hypotenuse 1 and one side \(x\): one acute angle has sine \(x\), the other has cosine \(x\), and the two acute angles sum to a right angle. Second: as \(x\) grows, \(\tan y = x\) forces \(y\) towards \(\frac{\pi}{2}\), so \(\lim_{x \to \infty} \arctan x = \frac{\pi}{2}\) and \(\lim_{x \to -\infty} \arctan x = -\frac{\pi}{2}\): two horizontal asymptotes in D.5's sense. The identity has a practical use: since \(-\arccos x = \arcsin x - \frac{\pi}{2}\), an answer written with \(-\arccos x\) and one written with \(\arcsin x\) differ only by a constant.

Worked example: an evaluation chain

Compute exactly: \(\sin\left(\arctan\frac{3}{4}\right)\). Set \(\theta = \arctan\frac{3}{4}\), an angle in the first quadrant with opposite \(3\), adjacent \(4\); the hypotenuse is \(5\), so \(\sin\theta = \frac{3}{5}\). The whole composition collapsed to a triangle. This is the evaluation style that turns up in trigonometric substitution (I.4), where an inverse trig answer feeds straight back into a trig function, and a sketch of the triangle is always the right first move. The principal-range discipline did quiet work again: arctangent's output is a first-quadrant angle for positive inputs, so no sign ambiguity ever arose.

Inverse trig in the physical world

Whenever an angle must be recovered from a ratio, an inverse trig function does the recovering: a robot arm's joint angles come from arctangents of link-position ratios, ship bearings from componentwise velocities, and optics' refraction angles from arcsines. The derivative rules then answer the engineer's follow-up: how sensitive is the computed angle to measurement error in the ratio? Since \(\frac{d}{dz}\arctan z = \frac{1}{1 + z^2}\), D.9's error propagation gives \(\Delta\theta \approx \frac{\Delta z}{1 + z^2}\): the bigger the measured ratio, the more the \((1 + z^2)\) denominator shrinks the resulting angle error, because arctangent's graph is nearly flat for large \(|z|\).

Common mistakes to avoid

The classics: treating \(\sin^{-1} x\) as \(\frac{1}{\sin x}\), the superscript means inverse, like F.3, not reciprocal; forgetting the restricted ranges, so \(\arcsin\) of something should never answer an angle outside \(\left[-\frac{\pi}{2}, \frac{\pi}{2}\right]\); choosing the negative square root when converting \(\cos y\) to \(x\), the range fixes the sign; dropping chain factors in \(\arctan(3x)\); and feeding arcsine arguments outside \([-1, 1]\), where it has no real value. The domain discipline also answers a classic exam trap: \(\arcsin(\sin \pi)\) is \(0\), not \(\pi\), because the range wins.

Key takeaways

Inverting trig requires restriction: sine to \(\left[-\frac{\pi}{2}, \frac{\pi}{2}\right]\), cosine to \([0, \pi]\), tangent to \(\left(-\frac{\pi}{2}, \frac{\pi}{2}\right)\), and outputs of the inverses are confined to those ranges. Implicit differentiation then gives algebraic derivatives: \(\frac{d}{dx}\arcsin x = \frac{1}{\sqrt{1 - x^2}}\), \(\frac{d}{dx}\arctan x = \frac{1}{1 + x^2}\), \(\frac{d}{dx}\arccos x = -\frac{1}{\sqrt{1 - x^2}}\), with chain factors for composed insides. Triangle sketches evaluate compositions like \(\sin(\arctan \frac{3}{4})\). Whenever a square root or a sign must be chosen, the principal range decides it.

Flashcards

Flashcard

arcsin x

The angle in [−π/2, π/2] whose sine is x

Domain −1 to 1.

Test your understanding

Question 1 of 5

What is the derivative of arcsin x?

Question 2 of 5

What is the derivative of arctan(3x)?

Question 3 of 5

What is the exact value of sin(arctan(3/4))?

Question 4 of 5

What is arcsin(sin π)?

Question 5 of 5

The graph of y = arctan x approaches what as x → ∞?

Standards: 3

← Previous topicNext topic →