Mathematics · Ch 10 — Differential Calculus – Differentiability and Methods of Differentiation
Derivatives of basic elementary functions
10.4.1
Derivatives of basic elementary functions
With the general rules of §10.4 in hand, we now derive — once, from first principle, each time using the h-form f′(x)=limh→0hf(x+h)−f(x) — the derivatives of every basic elementary function. Once this table exists, no later problem in the chapter needs to return to a first-principle limit.
(1) Constant function. For f(x)=k (k constant), f(x+h)−f(x)=k−k=0 for every h, so the difference quotient is identically 0 and dxd(k)=0.
(2) Power function, y=xn (n a positive integer).f(x+h)−f(x)=(x+h)n−xn. Writing y=x+h and factoring the difference of nth powers, h(x+h)n−xn=y−xyn−xn→nxn−1 as h→0 (equivalently y→x), so dxd(xn)=nxn−1.
Corollary 10.1. When n=p/q (a rational exponent, p,q integers, q=0), the same power formula holds: dxd(xp/q)=qpxqp−1.
Corollary 10.2. For any real number α, dxd(xα)=αxα−1 — the power rule holds for every real exponent, not just integers.
For instance: dxd(5)=0 (a constant); dxd(x3)=3x2; dxd(x3/2)=23x1/2; dxd(x2)=2x; dxd(x2/3)=32x−1/3 (x=0); and, by the constant-multiple rule, dxd(100x9)=100⋅9x8=900x8.
(3) Logarithmic function. For f(x)=logx (natural log, base e), f(x+h)−f(x)=log(xx+h)=log(1+xh), so hf(x+h)−f(x)=x1⋅h/xlog(1+h/x). Using the standard limit limα→0αlog(1+α)=1 (with α=h/x→0) gives dxd(logx)=x1.
Corollary 10.3. For y=logax=logalogx (change of base, logx = natural log), the constant-multiple rule gives dxd(logax)=xloga1.
(4) Exponential function. For f(x)=ax (a>0), hf(x+h)−f(x)=ax(hah−1), and the standard limit limh→0hah−1=loga gives dxd(ax)=axloga. In the special case a=e (so loge=1), dxd(ex)=ex — the exponential function is its own derivative.
(5) The six trigonometric functions.
Sine.f(x+h)−f(x)=sin(x+h)−sinx=2sin(2h)cos(x+2h) (sum-to-product identity), so hf(x+h)−f(x)=h/2sin(h/2)⋅cos(x+2h)→1⋅cosx as h→0 (using limθ→0sinθ/θ=1 and continuity of cos). So dxd(sinx)=cosx.
Cosine. Using cosx=sin(2π+x) and the chain rule with inner function u=2π+x (u′=1): dxd(cosx)=cosu⋅1=cos(2π+x)=−sinx. So dxd(cosx)=−sinx.
Tangent.tanx=sinx/cosx; the quotient rule gives dxd(tanx)=cos2xcosx⋅cosx−sinx⋅(−sinx)=cos2xcos2x+sin2x=cos2x1=sec2x.
Secant.secx=(cosx)−1; the chain rule gives dxd(secx)=−(cosx)−2(−sinx)=cos2xsinx=secxtanx.
Cotangent.cotx=cosx/sinx; the quotient rule gives dxd(cotx)=sin2xsinx(−sinx)−cosx(cosx)=sin2x−(sin2x+cos2x)=−cosec2x.
(6) The six inverse trigonometric functions.
sin−1x. Let y=sin−1x, so x=siny; differentiating both increments and passing to the limit as in implicit differentiation (§10.4.3) gives dxdy=cosy1=1−sin2y1=1−x21 (taking the principal branch, cosy≥0). So dxd(sin−1x)=1−x21.
cos−1x. From the identity sin−1x+cos−1x=2π (a constant), differentiating both sides gives dxd(cos−1x)=−dxd(sin−1x)=−1−x21.
tan−1x. Let y=tan−1x, so x=tany; the same implicit-style limit argument gives dxdy=sec2y1=1+tan2y1=1+x21. So dxd(tan−1x)=1+x21. …