The outside derivative is evaluated at the inside function, and then multiplied by the derivative of the inside function.
Proof (Chain rule). Let u=g(x) and y=f(u). A small change in x creates a small change in u, and that creates a small change in y. The ratio can be split as
ΞxΞyβ=ΞuΞyββ ΞxΞuβ.
As Ξxβ0, the intermediate change Ξuβ0 for a differentiable inside function. The two ratios approach fβ²(u) and gβ²(x), so
dxdyβ=fβ²(g(x))gβ²(x).
The chain rule is about layered change. If x changes, then the inside function g(x) changes first. That change then causes the outside function f(g(x)) to change.
Symbolically,
dxdyβ=dudyββ dxduβ.
The notation makes the chain rule look like canceling fractions. That is a useful way to remember it, but derivatives are limits of ratios, so algebra with derivative notation still needs a rule that justifies it.
AP problems often hide the chain rule inside:
powers of expressions,
trig functions with nontrivial angles,
exponentials with nontrivial exponents,
logarithms of expressions,
inverse trig functions with expressions inside.
Whenever you differentiate an outside function, pause and ask what the inside function is.
Example. Differentiate y=(3x2+1)5.
The outside function is a fifth power and the inside function is 3x2+1. Differentiate the outside power, keeping the inside unchanged, then multiply by the derivative of the inside:
dxdyβ=5(3x2+1)4β dxdβ(3x2+1).
Since dxdβ(3x2+1)=6x,
dxdyβ=5(3x2+1)4(6x)=30x(3x2+1)4.
Note that the chain rule should be used for all layers of a function. For example, if you have a function f(g(h(x))), the derivative is equal to:
dxdβf(g(h(x)))=fβ²(g(h(x)))gβ²(h(x))hβ²(x).
This is correct because each layer contributes its own derivative factor, starting from the outside layer and moving inward.
Implicit equations describe a relationship rather than a solved function, meaning that the xs and ys of the equation are not fully separated. To take dxdyβ of such function we can use implicit differentiation.
After differentiating, collect every term containing dy/dx on one side. Factor out dy/dx only after all product and chain rules have been expanded. The remaining coefficient determines the slope. If that coefficient is zero while the other side is nonzero, the curve may have a vertical tangent rather than an ordinary finite slope.
Example. Find the slope of the tangent line of the circle x2+y2=25 at the point (3,4).
Differentiate both sides with respect to x, remembering that y depends on x:
2x+2ydxdyβ=0.
Solve for dxdyβ:
dxdyβ=βyxβ.
Now substitute the point (3,4):
dxdyββ(3,4)β=β43β.
So the tangent line at (3,4) has slope β43β.
Example. The curve
x2y+sin(xy)=x+sin1
passes through (1,1). Find the equation of its tangent line at that point.
Differentiate both sides with respect to x. The term x2y needs the product rule, while sin(xy) needs both the chain rule and another product rule:
Theorem (Inverse Function Derivative). If f is differentiable and invertible with fβ²(a)ξ =0, then
(fβ1)β²(b)=fβ²(a)1β
where b=f(a).
Equivalent formula:
(fβ1)β²(x)=fβ²(fβ1(x))1β.
The derivative of an inverse is a reciprocal slope, but the reciprocal is taken at the matching point on the original function. If f(a)=b, then the point (a,b) on f becomes (b,a) on fβ1. The slope fβ²(a) belongs to the original point, while (fβ1)β²(b) belongs to the reflected point.
The condition fβ²(a)ξ =0 matters. A horizontal tangent on f reflects to a vertical tangent on its inverse, so the inverse does not have a finite derivative there.
Proof (Inverse Function Derivative). If y=f(x) and x=fβ1(y), then composing the functions gives
f(fβ1(x))=x.
Differentiate both sides:
fβ²(fβ1(x))(fβ1)β²(x)=1.
Solving for the inverse derivative gives
(fβ1)β²(x)=fβ²(fβ1(x))1β.
Example. Let f(x)=x3+x. Given that f(1)=2, find (fβ1)β²(2).
Here a=1 and b=2, since f(1)=13+1=2. First compute the derivative of f:
fβ²(x)=3x2+1.
Evaluate it at a=1:
fβ²(1)=3(1)2+1=4.
By the inverse function rule,
(fβ1)β²(2)=fβ²(1)1β=41β.
Example. Suppose f(2)=5 and fβ²(2)=β3. Define
g(x)=fβ1(x2+1).
Find the equation of the tangent line to g at x=2.
For AP work, arcsin, arccos, and arctan are the most common inverse trig functions. All six formulas follow from the inverse-function derivative formula; one proof is shown below.
The listed formulas give the derivative of the inverse trig function itself. If its input is another function u(x), the chain rule adds a factor of uβ²(x). For example,
dxdβarcsin(u)=1βu2βuβ²β,
and
dxdβarctan(u)=1+u2uβ²β.
Domain restrictions still matter after differentiating. The derivative of arcsinx becomes unbounded at x=Β±1, and the absolute value in the arcsec and arccsc formulas cannot be dropped without knowing the sign of the input.
Proof (Derivative of arcsinx).
Method 1 (Implicit Differentiation). Let y=arcsinx.
This means
siny=x,β2Οββ€yβ€2Οβ.
Differentiate implicitly with respect to x:
cosydxdyβ=1.
So
dxdyβ=cosy1β.
Since siny=x and y is in the principal range of arcsine, cosyβ₯0. Using sin2y+cos2y=1,
cosy=1βsin2yβ=1βx2β.
Therefore
dxdβarcsinx=1βx2β1β.
Method 2 (Inverse Function Formula). Let f(y)=siny, so fβ1(x)=arcsinx. The inverse derivative formula gives
(fβ1)β²(x)=fβ²(fβ1(x))1β.
Since fβ²(y)=cosy,
dxdβarcsinx=cos(arcsinx)1β.
If ΞΈ=arcsinx, then sinΞΈ=x and ΞΈ is in the principal range where cosΞΈβ₯0. Therefore
cos(arcsinx)=1βx2β,
so
dxdβarcsinx=1βx2β1β.
Example. Differentiate y=arctan(x2).
Use the chain-rule form with inside function u=x2, so uβ²=2x:
dxdyβ=1+u2uβ²β=1+(x2)22xβ.
Simplifying the square gives
dxdyβ=1+x42xβ.
Example. Differentiate and simplify
y=arcsin(1+x2βxβ).
Let
u=1+x2βxβ=x(1+x2)β1/2.
Differentiate u using the product and chain rules:
uβ²=(1+x2)β1/2βx2(1+x2)β3/2=(1+x2)3/21β.
Also,
1βu2=1β1+x2x2β=1+x21β.
Since 1+x2>0,
1βu2β=1+x2β1β.
Apply the chain-rule form of the arcsine derivative:
Logarithmic differentiation is useful when a function has products, quotients, powers, or variables in both the base and exponent. The idea is to take the natural logarithm of both sides, use log laws to simplify, and then differentiate implicitly.
For example, if
y=(x2+1)3xβ4β,
then taking logs gives
lny=3ln(x2+1)+21βln(xβ4).
Differentiate both sides:
y1βdxdyβ=3β x2+12xβ+21ββ xβ41β.
Then multiply by y:
dxdyβ=(x2+1)3xβ4β(x2+16xβ+2(xβ4)1β).
Logarithmic differentiation is a strategy, not a new derivative rule. It works because logarithms turn complicated multiplication, division, and powers into simpler operations:
ln(ab)=lna+lnb,ln(baβ)=lnaβlnb,ln(ar)=rlna.
This is especially helpful when a function has many factors or when a variable appears in both the base and the exponent.
For a variable power u(x)v(x) with u(x)>0, logarithmic differentiation turns the exponent into a factor:
lny=v(x)ln(u(x)).
Differentiating gives
yyβ²β=vβ²(x)ln(u(x))+v(x)u(x)uβ²(x)β.
This contains both the derivative of the exponent and the derivative of the base. Treating v(x) as a constant would miss the first term.
After differentiating, remember that differentiating lny gives
y1βdxdyβ,
so the final derivative usually comes from multiplying by y, where you resubstitude the original function (if y=f(x)).
Example. Use logarithmic differentiation to find dxdyβ for y=xx (with x>0).
The variable appears in both the base and the exponent, so take the natural logarithm of both sides:
lny=ln(xx)=xlnx.
Differentiate both sides with respect to x. The left side uses the chain rule, and the right side uses the product rule:
y1βdxdyβ=lnx+xβ x1β=lnx+1.
Multiply both sides by y and substitute y=xx:
dxdyβ=xx(lnx+1).
Example. Use logarithmic differentiation to find dy/dx for
y=x3xβ1β(x2+1)sinxβ,x>1.
Take the natural logarithm and expand using log laws:
lny=sinxln(x2+1)β3lnxβ21βln(xβ1).
Differentiate both sides. The first term requires the product rule:
This unit is about recognizing when the derivative is hidden inside another relationship.
Chain rule problems hide a changing input inside an outer function.
Implicit differentiation hides y as a function of x.
Inverse-function problems hide the slope relationship between a function and its inverse.
Related rates hide time dependence inside geometry or context.
Before moving on, make sure you can explain which rule is being used at each step. On AP-style questions, the hardest part is often choosing the rule order, not doing the algebra afterward.