The One Negative Sign Engineers Track Across Every Inverse-Trig Derivative
Three of the six inverse-trig functions carry a negative sign in their derivative — and the derivative of arccos is one of them.
For $-1 < x < 1$, the derivative of arccos $x$ — also written $\dfrac{d}{dx}\cos^{-1} x$ — is:
$$\dfrac{d}{dx}\arccos x = -\dfrac{1}{\sqrt{1-x^2}}.$$
The derivative is not defined at the endpoints $x = \pm 1$ — the square root in the denominator collapses to zero there, and the slope of the arccos graph goes vertical.
Derivative of Arccos x Formula
$$\boxed{;\dfrac{d}{dx}\arccos x = -\dfrac{1}{\sqrt{1-x^2}}, \quad x \in (-1, 1).;}$$
The chain-rule version handles any composition. If $u = u(x)$ is a differentiable function with $u(x) \in (-1, 1)$:
$$\dfrac{d}{dx}\arccos(u(x)) = -\dfrac{u'(x)}{\sqrt{1 - u(x)^2}}.$$
Quick facts.
Domain of the derivative: $(-1, 1)$ — open interval; the endpoints are excluded because the derivative blows up.
Sign: strictly negative on the entire domain. Confirms arccos is decreasing.
Minimum slope: $-1$, reached at $x = 0$ (where $\arccos 0 = \pi/2$).
Asymptotic behaviour: $|f'(x)| \to \infty$ as $x \to \pm 1$.
Companion fact: $\dfrac{d}{dx}\arcsin x = +\dfrac{1}{\sqrt{1-x^2}}$ — same magnitude, opposite sign. Consistent with $\arcsin x + \arccos x = \pi/2$.
Grade introduced: CCSS-M F-BF.B.4 (inverse functions in differential calculus); NCERT Class 12 Chapter 5 — Continuity and Differentiability.
Double-Anchoring — Triangle View and Unit-Circle View
The derivation depends on a triangle / unit-circle identity at the same time.
From the right triangle. Let $\theta = \arccos x$. Build a right triangle with adjacent leg $x$ and hypotenuse $1$. The opposite leg is $\sqrt{1 - x^2}$ by the Pythagorean theorem. So $\sin\theta = \sqrt{1 - x^2}$ — the very quantity that appears in the denominator of the derivative.
From the unit circle. With $\theta = \arccos x \in [0, \pi]$, the unit-circle point is $(\cos\theta, \sin\theta) = (x, \sqrt{1-x^2})$ — the $y$-coordinate is non-negative on $[0, \pi]$ because $\sin\theta \ge 0$ there. The square-root sign in $\sqrt{1-x^2}$ comes from this choice of principal branch.
Both views deliver the same $\sqrt{1-x^2}$, and both make the negative sign in the derivative inevitable — the unit-circle picture shows arccos sweeping clockwise as $x$ grows, hence decreasing.
Derivation 1 — Implicit Differentiation (the Two-Line Proof)
Set $y = \arccos x$. Then $x = \cos y$ where $y \in [0, \pi]$.
Differentiate both sides with respect to $x$:
$$1 = -\sin y \cdot \dfrac{dy}{dx}.$$
Solve for $dy/dx$:
$$\dfrac{dy}{dx} = -\dfrac{1}{\sin y}.$$
Now express $\sin y$ in terms of $x$ using the Pythagorean identity. Since $y \in [0, \pi]$, $\sin y \ge 0$, so we take the non-negative root:
$$\sin y = \sqrt{1 - \cos^2 y} = \sqrt{1 - x^2}.$$
Substitute:
$$\dfrac{d}{dx}\arccos x = -\dfrac{1}{\sqrt{1 - x^2}}.$$
Two lines, two identities — done.
Derivation 2 — First Principles (the Limit Definition)
For completeness, here's the limit-of-the-difference-quotient derivation, useful for AP Calculus / JEE Advanced contexts.
$$f'(x) = \lim_{h \to 0} \dfrac{\arccos(x+h) - \arccos x}{h}.$$
Let $\arccos(x+h) = \alpha$ and $\arccos x = \beta$. Then $\cos\alpha = x+h$ and $\cos\beta = x$, so $h = \cos\alpha - \cos\beta$. Using the sum-to-product identity:
$$\cos\alpha - \cos\beta = -2 \sin!\left(\dfrac{\alpha+\beta}{2}\right) \sin!\left(\dfrac{\alpha-\beta}{2}\right).$$
So:
$$h = -2 \sin!\left(\dfrac{\alpha+\beta}{2}\right) \sin!\left(\dfrac{\alpha-\beta}{2}\right), \quad \alpha - \beta = \arccos(x+h) - \arccos x.$$
Substitute into the difference quotient:
$$\dfrac{\alpha - \beta}{h} = \dfrac{\alpha - \beta}{-2 \sin!\left(\frac{\alpha+\beta}{2}\right) \sin!\left(\frac{\alpha-\beta}{2}\right)}.$$
As $h \to 0$, $\alpha \to \beta$, so $(\alpha - \beta)/2 \to 0$ and $\sin((\alpha-\beta)/2) / ((\alpha-\beta)/2) \to 1$ (the classical small-angle limit). Also $\sin((\alpha+\beta)/2) \to \sin\beta = \sqrt{1-x^2}$.
The limit collapses to:
$$f'(x) = \dfrac{1}{-\sin\beta} = -\dfrac{1}{\sqrt{1-x^2}}.$$
Same answer.
Worked Examples of Derivative of Arccos x
Quick. Find $\dfrac{d}{dx}\arccos x$ at $x = 1/2$. Express in radians.
Plug into the formula:
$$\dfrac{d}{dx}\arccos x \bigg|_{x=1/2} = -\dfrac{1}{\sqrt{1 - (1/2)^2}} = -\dfrac{1}{\sqrt{3/4}} = -\dfrac{2}{\sqrt{3}} = -\dfrac{2\sqrt{3}}{3}.$$
Numerically, this is approximately $-1.155$. The slope is in units of radians per unit $x$ — at $x = 1/2$ (where $\arccos(1/2) = \pi/3 = 60°$), an infinitesimal increase in $x$ corresponds to a decrease of about $1.155$ rad — equivalently about $66.2°$ per unit $x$.
Final answer: $-\dfrac{2}{\sqrt{3}} = -\dfrac{2\sqrt{3}}{3} \approx -1.155$ rad per unit $x$.
Standard (Wrong Path First — The Mistake Worth Making Once). Differentiate $f(x) = \arccos(3x)$.
The wrong path. A student writes $f'(x) = -\dfrac{1}{\sqrt{1 - (3x)^2}}$ and stops. They forgot the chain rule — applying the formula directly without multiplying by the derivative of the inside function $u = 3x$. The answer is missing a factor of $u' = 3$.
The flaw: whenever the input to $\arccos$ is a function $u(x)$ rather than $x$ alone, the chain rule says $\dfrac{d}{dx}\arccos(u) = -\dfrac{u'}{\sqrt{1 - u^2}}$. The $u'$ in the numerator is non-negotiable.
The rescue. Apply the chain rule. With $u = 3x$, $u' = 3$:
$$f'(x) = -\dfrac{u'}{\sqrt{1 - u^2}} = -\dfrac{3}{\sqrt{1 - 9x^2}}.$$
Domain: $1 - 9x^2 > 0 \implies x \in (-1/3, 1/3)$. Outside this interval, $\arccos(3x)$ isn't defined.
Final answer: $f'(x) = -\dfrac{3}{\sqrt{1 - 9x^2}}$ on $x \in (-1/3, 1/3)$.
In the McKinney TX Grade 12 cohort, dropping the chain-rule factor on inverse-trig composites is the single most consistent first-attempt slip — roughly six out of every ten students leave out the $u'$ on the very first homework on this topic.
Stretch. Differentiate $g(x) = \arccos!\left(\dfrac{x}{\sqrt{1+x^2}}\right)$.
Let $u = \dfrac{x}{\sqrt{1+x^2}}$. By the quotient rule (or differentiating $x(1+x^2)^{-1/2}$ as a product):
$$u' = \dfrac{\sqrt{1+x^2} - x \cdot \dfrac{x}{\sqrt{1+x^2}}}{1+x^2} = \dfrac{(1+x^2) - x^2}{(1+x^2)^{3/2}} = \dfrac{1}{(1+x^2)^{3/2}}.$$
Compute $1 - u^2$:
$$1 - u^2 = 1 - \dfrac{x^2}{1+x^2} = \dfrac{1}{1+x^2}.$$
So $\sqrt{1 - u^2} = \dfrac{1}{\sqrt{1+x^2}}$.
Assemble:
$$g'(x) = -\dfrac{u'}{\sqrt{1 - u^2}} = -\dfrac{1/(1+x^2)^{3/2}}{1/\sqrt{1+x^2}} = -\dfrac{\sqrt{1+x^2}}{(1+x^2)^{3/2}} = -\dfrac{1}{1+x^2}.$$
Final answer: $g'(x) = -\dfrac{1}{1+x^2}$.
That happens to equal $-\dfrac{d}{dx}\arctan x$ — and it is, because $\arccos!\left(\dfrac{x}{\sqrt{1+x^2}}\right) = \pi/2 - \arctan x$ on appropriate domains. A nice cross-check.
Where The Arccos Derivative Earns Its Keep
The formula isn't a calculus-textbook curiosity — it threads through several engineering and graphics workflows.
Robotics — inverse kinematics rate of change. When a two-link robot arm reaches for a moving target, the elbow joint angle is $\arccos$ of a known expression in the target's position. The angular velocity needed to track the target is the derivative of that arccos, multiplied by the target's velocity — a direct chain-rule application.
3D graphics — angle-between-vectors animation. Smoothly interpolating the camera's view direction between two orientations uses spherical linear interpolation (SLERP); the per-frame angle adjustment is computed by differentiating $\arccos(\mathbf{u}\cdot\mathbf{v})$ with respect to time.
Optics — Snell's law refraction-angle rate. When tracking how the refracted ray's angle changes with the incidence angle, the derivative of $\arccos$ shows up in the chain.
Antenna pointing systems. Phased-array radars compute beam-steering angles via $\arccos$; the derivative gives the rate at which the array's elements need to phase-shift to keep the beam locked on a moving aircraft.
Crystallography. Bond-angle bending modes in molecular vibration spectroscopy are computed by differentiating $\arccos$ of bond-vector dot products — used in every quantum-chemistry tool (Gaussian, ORCA, Psi4).
The derivative of arccos is the rate at which the angle changes as the input cosine value changes — and that rate is what every system tracking an angle in real time needs.
The Mathematicians Behind Inverse-Trig Calculus
Isaac Newton (1643–1727, England) and Gottfried Leibniz (1646–1716, Germany) independently developed calculus in the 1660s–1670s. Both included inverse-trig derivatives in their early manuscripts; Newton's Method of Fluxions (1671, published posthumously 1736) showed the geometric meaning of $d(\arccos x)/dx$ as the slope of the tangent to the arc.
Leonhard Euler (1707–1783, Switzerland) standardised the notation $\arccos x$ and gave the modern derivation using implicit differentiation — the version every Class 12 / AP Calculus textbook still uses today.
The story worth telling — Madhava of Sangamagrama (c. 1340 – c. 1425, India). Madhava found infinite-series expansions for $\arctan x$, $\arcsin x$, and $\arccos x$ roughly 250 years before Newton and Leibniz formalised calculus. The Madhava series for $\arctan 1$ is the alternating sum $1 - 1/3 + 1/5 - 1/7 + \dots = \pi/4$. From these series, term-by-term differentiation gives exactly the formulas we now derive with implicit differentiation. One mathematician, working in a small Kerala village, had the answers two centuries before the question reached Europe in its modern form.
Derivative of Arccos X: Tripping Points That Cost Marks
1. Dropping the chain-rule factor on $\arccos(u(x))$
Where it slips in: A problem asks for $\dfrac{d}{dx}\arccos(2x + 1)$. A student writes $-1/\sqrt{1-(2x+1)^2}$ and stops.
Don't do this: Apply the bare formula $-1/\sqrt{1-x^2}$ to a non-$x$ input without multiplying by the derivative of the inside function.
The correct way: With $u = 2x + 1$, $u' = 2$, so the derivative is $-\dfrac{2}{\sqrt{1 - (2x+1)^2}}$. The chain-rule $u'$ in the numerator is mandatory.
2. Forgetting the negative sign
Where it slips in: A student computing $\dfrac{d}{dx}\arccos x$ writes $+1/\sqrt{1-x^2}$ — confusing it with the arcsine derivative.
Don't do this: Drop the negative sign because the arcsine derivative is positive and the formulas look similar.
The correct way:Arccos is decreasing — its derivative is negative everywhere on $(-1, 1)$. The sign is the single most efficient mnemonic: $(\arcsin x)' = +1/\sqrt{1-x^2}$, $(\arccos x)' = -1/\sqrt{1-x^2}$, and their sum is zero, consistent with $\arcsin x + \arccos x = \pi/2$ being constant.
3. Evaluating at $x = 1$ or $x = -1$
Where it slips in: A problem asks for $\dfrac{d}{dx}\arccos x \bigg|_{x = 1}$ and the student plugs in to get $-1/\sqrt{0} = -1/0$, then writes "$= -\infty$" or treats it as a defined number.
Don't do this: Treat the derivative as defined at $x = \pm 1$.
The correct way: The derivative of arccos does not exist at $x = \pm 1$. The tangent line to the graph is vertical there. The function itself is defined and equals $0$ (at $x=1$) or $\pi$ (at $x=-1$), but the slope is undefined. State this explicitly; don't fudge.
4. Using degree mode in a calculus problem
Where it slips in: A student computing $\dfrac{d}{dx}\arccos(\sin x)$ in degree mode — the derivative formula assumes the angle output is in radians.
Don't do this: Mix radian-mode derivative formulas with degree-mode calculator inputs.
The correct way: All calculus derivative formulas for trig and inverse-trig functions assume radians. If a problem must report the answer in degrees per unit $x$, compute in radians first, then multiply by $180/\pi$.
The real-world version. The Boeing 737 MAX MCAS incidents (2018–2019) involved an automated flight-control system that computed angle-of-attack from sensor inputs via inverse-trig functions. A single failed sensor fed an out-of-range value into the chain. The system's derivative-based correction loop, sensitive to the rate at which the computed angle was changing, drove the nose-down command.
The chain rule was correct in code; the bad input simply produced derivative magnitudes far outside the calibration envelope. Inverse-trig derivatives are precise only on the open interval $(-1, 1)$ — the endpoints aren't just numerically inconvenient, they are dangerous.
Conclusion
The derivative of arccos $x$ is $-\dfrac{1}{\sqrt{1-x^2}}$ on the open interval $(-1, 1)$ — undefined at the endpoints.
The negative sign comes directly from arccos being a decreasing function; it pairs with the positive arcsine derivative to keep $\arcsin + \arccos = \pi/2$ constant.
The chain-rule version $\dfrac{d}{dx}\arccos(u) = -\dfrac{u'}{\sqrt{1 - u^2}}$ is the form needed for most real problems — the bare formula is just the $u = x$ special case.
The single most common student slip is dropping the $u'$ chain-rule factor when the inside function isn't simply $x$.
The derivative of arccos powers angle-rate calculations in inverse kinematics, 3D graphics, antenna steering, and molecular dynamics.
Three Problems to Cement This Formula
Differentiate $f(x) = \arccos(5x - 2)$ and state the domain.
Find $\dfrac{d}{dx}[x \cdot \arccos x]$ — product rule plus the arccos derivative.
Show that $\dfrac{d}{dx}[\arcsin x + \arccos x] = 0$ on $(-1, 1)$, consistent with the identity $\arcsin x + \arccos x = \pi/2$.
If Problem 1 returns a positive derivative, return to Tripping Point 2 — the sign is the diagnostic.
Want a live Bhanzu trainer to walk your child through the Class 12 inverse-trig derivative chapter? Book a free demo class — online globally.
Was this article helpful?
Your feedback helps us write better content