The derivative grew out of two of the four classic 17th-century calculus problems: the tangent line problem and the velocity problem. Both reduce to the same limiting construction.
From secant to tangent. For a curve y=f(x) and a point P(x0,f(x0)), a second point Q(x0+Δx,f(x0+Δx)) determines a secant line with slope (the difference quotient)
msec=ΔxΔy=Δxf(x0+Δx)−f(x0).
Letting Q→P (i.e. Δx→0) makes the secant line approach the tangent line, whose slope is
mtan=limΔx→0Δxf(x0+Δx)−f(x0),
provided this limit exists (Definition 10.1). Since a point and a slope fix a unique line, the tangent line — when it exists — is unique.
From average to instantaneous velocity. If s=f(t) is the position of an object moving on a line, the average velocity over [t0,t0+Δt] is vavg=Δtf(t0+Δt)−f(t0) — again a secant slope, now on the position–time graph. Shrinking Δt→0 gives the instantaneous velocity
v(t0)=limΔt→0Δtf(t0+Δt)−f(t0).
The derivative, in general. Both constructions are the same limit applied to different functions, so calculus gives it one name:
Definition 10.2. f is differentiable at x0 if f′(x0)=Δx→0limΔxf(x0+Δx)−f(x0) exists. For every x where the limit exists, f′(x) defines the derivative function, and finding it is called differentiation.
Notations. f′(x), y′, dxdy, dxdf(x), Dy all denote the same object. dxdy is read "derivative of y w.r.t. x" — it is a single symbol for a limit, not literally a fraction dy÷dx, even though later techniques (implicit and parametric differentiation) manipulate it formally as one.
One-sided derivatives. Splitting the limit by direction gives the left-hand derivative f′(x0−)=limΔx→0−Δxf(x0+Δx)−f(x0) and right-hand derivative f′(x0+)=limΔx→0+Δxf(x0+Δx)−f(x0). The full (two-sided) derivative f′(x0) exists iff both one-sided derivatives exist and are equal. On a closed interval [a,b], differentiability means differentiable on (a,b), with f′(a) taken as the right-hand derivative at a and f′(b) as the left-hand derivative at b (Definition 10.3) — there being no function on the other side of an endpoint to take a two-sided limit against.
Deriving formulas from this definition (computing f′ directly from the limit, with no shortcut rule) is called finding the derivative from first principle; it is how every standard derivative formula in this chapter is originally established.