P3.4 - Differentiation

Syllabus
2019
Topic
P3.4
Level
A2

Learning objectives

Differentiate exponential, logarithmic and trigonometric terms

Differentiate a sum or difference term by term. For each exponential, logarithmic or trigonometric term, multiply by the derivative of its linear inner expression.

f(x)f(x) f(x)f'(x)
ekxe^{kx} kekxke^{kx}
ln(kx)\ln(kx) 1/x1/x on its domain
sin(kx)\sin(kx) kcos(kx)k\cos(kx)
cos(kx)\cos(kx) ksin(kx)-k\sin(kx)
tan(kx)\tan(kx) ksec2(kx)k\sec^2(kx)

For y=3e2x+4ln(5x)2cos(3x),x>0,y=3e^{-2x}+4\ln(5x)-2\cos(3x),\qquad x>0, differentiate each term: dydx=6e2x+4x+6sin(3x).\frac{dy}{dx}=-6e^{-2x}+\frac4x+6\sin(3x). The domain condition comes from the logarithm.

These trigonometric derivatives assume angles are measured in radians. A derivative may then be evaluated at a point to obtain a tangent gradient or set equal to zero to locate stationary candidates.

For ln(kx)\ln(kx), the chain factor cancels algebraically to 1/x1/x, but the original condition kx>0kx>0 remains. Do not omit the inner factor kk from ekxe^{kx}, sine, cosine or tangent.

Choose and nest the product, quotient and chain rules

The structure of the expression selects the rule: multiplication needs the product rule, division needs the quotient rule, and a function inside another function needs the chain rule. More than one rule may be nested in the same derivative.

Structure Derivative
uvuv uv+uvu'v+uv'
u/vu/v (vuuv)/v2(vu'-uv')/v^2
f(g(x))f(g(x)) f(g(x))g(x)f'(g(x))g'(x)

ddx(secx)=secxtanx,ddx(cosecx)=cosecxcotx,ddx(cotx)=cosec2x\frac{d}{dx}(\sec x)=\sec x\tan x,\qquad \frac{d}{dx}(\operatorname{cosec}x)=-\operatorname{cosec}x\cot x,\qquad \frac{d}{dx}(\cot x)=-\operatorname{cosec}^2x

Examples expose the chosen structure: ddx(2x4sinx)=8x3sinx+2x4cosx,\frac{d}{dx}(2x^4\sin x)=8x^3\sin x+2x^4\cos x, ddx(e3xx)=e3x(3x1)x2,\frac{d}{dx}\left(\frac{e^{3x}}x\right)=\frac{e^{3x}(3x-1)}{x^2}, ddx(tan2(2x))=4tan(2x)sec2(2x).\frac{d}{dx}\big(\tan^2(2x)\big)=4\tan(2x)\sec^2(2x).

For a nested expression, identify the outermost operation first, write its rule without simplifying, and then differentiate each inner part. Factor common terms afterwards when solving a stationary-point equation.

The quotient numerator order is denominator times derivative of numerator minus numerator times derivative of denominator. Do not differentiate a product by multiplying derivatives, and retain every original domain restriction.

Differentiate when x is given as a function of y

If a curve is given as x=f(y)x=f(y), differentiate with respect to yy first and then take the reciprocal. This works on a local inverse branch wherever dx/dy0dx/dy\ne0.

dydx=1dxdy\frac{dy}{dx}=\frac{1}{\dfrac{dx}{dy}}

Use this order: find dx/dy; invert it; simplify in terms of y; then, only if required, use the original relation and the stated interval to rewrite the result in terms of x. The interval controls any square-root sign.

Suppose x=sin(3y)x=\sin(3y) on π6<y<π6-\frac\pi6<y<\frac\pi6. Then dxdy=3cos(3y),dydx=13cos(3y).\frac{dx}{dy}=3\cos(3y),\qquad \frac{dy}{dx}=\frac1{3\cos(3y)}. On this branch cosine is positive and cos(3y)=1x2\cos(3y)=\sqrt{1-x^2}, so dydx=131x2.\frac{dy}{dx}=\frac1{3\sqrt{1-x^2}}.

After evaluating dy/dxdy/dx at a point, the normal gradient is its negative reciprocal. Keep the point coordinates with the gradient when forming a tangent or normal equation.

The symbols dy/dxdy/dx and dx/dydx/dy are reciprocals only where a differentiable local inverse exists. If dx/dy=0dx/dy=0, this calculation does not produce a finite dy/dxdy/dx; do not divide by zero or discard the branch condition.

Interpret and test exponential growth or decay models

An exponential model describes change at a rate proportional to its current displacement from a limiting level. Read the initial value at t=0, the sign of the rate, and the long-term behaviour before trusting a numerical prediction.

ddx(ax)=axlna(a>0),ddt(Aekt+C)=kAekt=k(yC)\frac{d}{dx}(a^x)=a^x\ln a\quad(a>0),\qquad \frac{d}{dt}(Ae^{kt}+C)=kAe^{kt}=k(y-C)

For A>0A>0:

Model feature Interpretation
y(0)=A+Cy(0)=A+C initial value
k>0k>0 growth away from CC
k<0k<0 decay towards CC as tt\to\infty
y=Abty=Ab^t growth if b>1b>1; decay if 0<b<10<b<1

For a temperature model T=20+80e0.3t,t0,T=20+80e^{-0.3t},\qquad t\ge0, the initial temperature is 100100, the limiting temperature is 2020, and dTdt=24e0.3t=0.3(T20).\frac{dT}{dt}=-24e^{-0.3t}=-0.3(T-20). Thus the temperature falls quickly at first and the magnitude of its rate tends to zero.

A model's algebra may be correct but its prediction may be inappropriate outside the observed time range: unlimited growth may exceed a physical capacity, while a decay limit may be unrealistic. Compare an improved model by its fit in the relevant range and by whether its initial value, limit and allowed outputs make sense.

Time models normally use t0t\ge0. Do not call every decreasing exponential a decay to zero: an outside constant CC changes the limit, and the sign of AA also affects whether the curve approaches CC from above or below.