Not every function comes prepackaged as y=f(x). Many relations between x and y are implicit: x2+y2=1, xy+siny=x, and so on. Implicit differentiation treats y as a (possibly multi-valued) function of x, applies dxd to both sides of the equation, and solves for dxdy.
A related technique uses the chain rule to find the derivative of an inverse function: if y=f−1(x), then dxdy=f′(y)1.
Implicit differentiation: the procedure
Given F(x,y)=0:
Apply dxd to both sides, treating y as an implicit function of x.
Use the chain rule for any function of y: dxd(yn)=nyn−1dxdy, etc.
Collect dxdy terms on one side; solve algebraically.
Worked example: a circle
x2+y2=25. Differentiate: 2x+2ydxdy=0, so dxdy=−yx.
At (3,4): dxdy=−3/4.
Another example
xy+siny=x. Differentiate: y+xdxdy+cosy⋅dxdy=1. Solve: dxdy(x+cosy)=1−y, so dxdy=x+cosy1−y.
Derivative of an inverse
If y=f−1(x), then x=f(y). Differentiate both sides with respect to x: 1=f′(y)dxdy, so
dxdy=f′(y)1=f′(f−1(x))1.
This is the inverse function theorem: the derivative of the inverse is the reciprocal of the derivative of the original (evaluated at the right point).
Application: derivatives of inverse trig
For y=sin−1x: x=siny, so dxdy=cosy1=1−sin2y1=1−x21 (positive square root, since y∈[−π/2,π/2]).
For y=tan−1x: x=tany, so dxdy=sec2y1=1+tan2y1=1+x21.
For y=cos−1x: x=cosy with y∈[0,π], so siny≥0 and dxdy=−siny1=−1−x21.
Worked examples
Example 1. Find dxdy from x3+y3=3axy (the folium of Descartes).
Differentiate: 3x2+3y2dxdy=3a(y+xdxdy). Rearrange: dxdy(y2−ax)=ay−x2, so dxdy=y2−axay−x2.
Example 2. Find dxdy from ex+ey=ex+y.
Differentiate: ex+eydxdy=ex+y(1+dxdy). Rearrange: dxdy(ey−ex+y)=ex+y−ex, so dxdy=ey−ex+yex+y−ex=ey(1−ex)ex(ey−1)=−ey(1−ex)ex(1−ey).
Simplify: dxdy=−ey(ex−1)ex(ey−1) after cancellation.
Example 3. If y=cos−1(cosx), find dxdy on (0,π).
On this interval y=x, so dxdy=1.
Example 4. If y=tan−1(sinx1−cosx) for x∈(0,π), find dxdy.
Simplify the inside: sinx1−cosx=2sin(x/2)cos(x/2)2sin2(x/2)=tan(x/2). So y=tan−1(tan(x/2))=x/2 on the principal branch. Hence dxdy=1/2.
Example 5. If x=acosθ,y=asinθ, find dxdy.
Method 1: parametric (next subtopic). Method 2: x2+y2=a2, so dxdy=−x/y=−cosθ/sinθ=−cotθ.
Example 6. If y=f−1(x) and f′(y)=1/(1+y2), find dxdy in terms of x.
dxdy=1/f′(y)=1+y2. This is general. (Here f=tan−1, so y=tanx and indeed dxdy=sec2x=1+tan2x=1+y2.)
Try it yourself
Find dxdy from x2+y2=16 at (2,23).
Find dxdy from xy=4.
Find dxdy from sin(xy)=1 at any allowed point.
Find dxdy from y=log(tan(x/2+π/4)).
Find dxdy from y=tan−1(1+cosxsinx).
Derive the derivative of sec−1 using the inverse-function theorem.
Find dxdy from xey+yex=1.
Find dxdy from x3+y3−3xy=0.
Find dxdy from tan−1(y/x)=logx2+y2.
Find dxdy from cos−1(1+x21−x2)=2tan−1x.
If x1+y+y1+x=0, find dxdy.
Find the slope of the tangent to y2=4x at the point (1,2).
If 1−x2+1−y2=a(x−y), find dxdy (Hint: use sin−1 trick).
If y=x+x+x+…, find dxdy. (Square both sides first.)
Pitfalls / Tricks
Don't forget that derivatives of y pick up a factor of dxdy , this is the chain rule applied implicitly.
Solve algebraically for dxdy before plugging in numerical points.
The inverse-function rule 1/f′(y) requires f′(y)=0.
For nested or trig-identity-laden expressions, look for a simplification before differentiating.