Skip to content

Mathematics · Ch 8 — Differentials and Partial Derivatives

Differentials

8.2.3

Differentials

Returning to the derivative dfdx=lim⁡Δx→0f(x+Δx)−f(x)Δx=f′(x)\dfrac{df}{dx}=\displaystyle\lim_{\Delta x\to0}\frac{f(x+\Delta x)-f(x)}{\Delta x}=f'(x) (Leibniz's notation for the differential coefficient): is it meaningful to treat dfdx\dfrac{df}{dx} as an actual quotient of two separate quantities dfdf and dxdx — not merely a single symbol for a limit?

When ff is linear, f(x)=mx+cf(x)=mx+c, the answer is unambiguously yes: Δy=f(x+Δx)−f(x)=mΔx=f′(x)Δx\Delta y = f(x+\Delta x)-f(x)=m\Delta x=f'(x)\Delta x exactly, for every xx and every Δx\Delta x (no limiting process needed, since a line's slope is the same everywhere). In this case

change in fchange in x=ΔyΔx=f′(x)=dfdx=dydx,\frac{\text{change in }f}{\text{change in }x} = \frac{\Delta y}{\Delta x} = f'(x) = \frac{df}{dx} = \frac{dy}{dx},

so df/dxdf/dx genuinely is the quotient of df=Δf=Δydf=\Delta f=\Delta y and dx=Δxdx=\Delta x.

Definition 8.4 (Differential). For a general differentiable f:(a,b)→Rf:(a,b)\to\mathbb R, x∈(a,b)x\in(a,b), and Δx\Delta x an increment given to xx, the differential of ff is

df=f′(x) Δx.(8)df = f'(x)\,\Delta x. \qquad(8)

Taking f(x)=xf(x)=x itself gives dx=f′(x)Δx=1⋅Δx=Δxdx=f'(x)\Delta x=1\cdot\Delta x=\Delta x — so the differential of the independent variable xx is simply its change, dx=Δxdx=\Delta x, and (8) can be rewritten df=f′(x) dxdf=f'(x)\,dx: precisely the "quotient" reading of df/dxdf/dx that motivated the whole discussion.

Geometric meaning. For y=f(x)y=f(x), let Δf=f(x+dx)−f(x)\Delta f=f(x+dx)-f(x) be the actual change in output along the curve, and let dydy (or dfdf) be the change along the tangent line instead: since the tangent line has slope f′(x)f'(x), dy=f′(x) dxdy=f'(x)\,dx. From the picture, Δf≈dy=df=f′(x)dx\Delta f\approx dy=df=f'(x)dx — so f′(x)f'(x) can be viewed (approximately) as the quotient of Δf\Delta f and Δx\Delta x, i.e. df/dxdf/dx is legitimately interpreted as a quotient of dfdf and dxdx, even though ff itself is nonlinear.

Note

Unlike the derivative f′(x)f'(x), which is a function of xx alone, the differential df=f′(x) dxdf=f'(x)\,dx is a function of two independent, freely-chosen quantities: the point xx and the increment dxdx. And Δf≈df\Delta f\approx df — an approximation, not an identity, except when ff is linear.

Differentials of standard functions (paired with their derivatives): f(x)=xn⇒df=nxn−1dxf(x)=x^n\Rightarrow df=nx^{n-1}dx;  f(x)=cos⁡(x2+7)⇒df=−sin⁡(x2+7)(2x) dx\ f(x)=\cos(x^2+7)\Rightarrow df=-\sin(x^2+7)(2x)\,dx;  f(x)=cot⁡(x2)⇒df=−csc⁡2(x2)(2x) dx\ f(x)=\cot(x^2)\Rightarrow df=-\csc^2(x^2)(2x)\,dx;  f(x)=sin⁡−1x⇒df=dx1−x2\ f(x)=\sin^{-1}x\Rightarrow df=\dfrac{dx}{\sqrt{1-x^2}};  f(x)=tan⁡−1x⇒df=dx1+x2\ f(x)=\tan^{-1}x\Rightarrow df=\dfrac{dx}{1+x^2};  f(x)=e3x2−5x+7⇒df=e3x2−5x+7(6x−5) dx\ f(x)=e^{3x^2-5x+7}\Rightarrow df=e^{3x^2-5x+7}(6x-5)\,dx;  f(x)=log⁡(x2+1)⇒df=2xx2+1dx\ f(x)=\log(x^2+1)\Rightarrow df=\dfrac{2x}{x^2+1}dx — in every case, exactly the derivative multiplied by dxdx.

Properties of Differentials (for f,gf,g differentiable, cc a real constant):

  1. If ff is constant, df=0df=0.
  2. If f(x)=xf(x)=x (identity), df=dxdf=dx.
  3. d(cf)=cf′(x) dx=c dfd(cf) = c f'(x)\,dx = c\,df.
  4. d(f+g)=f′(x)dx+g′(x)dx=df+dgd(f+g) = f'(x)dx+g'(x)dx = df+dg.
  5. Product rule: d(fg)=f dg+g dfd(fg) = f\,dg+g\,df (proved below).
  6. Quotient rule: d(f/g)=g df−f dgg2d(f/g) = \dfrac{g\,df-f\,dg}{g^2}, provided g(x)≠0g(x)\ne0.
  7. Chain rule: if h=f∘gh=f\circ g is defined, dh=f′(g(x)) g′(x) dxdh = f'(g(x))\,g'(x)\,dx.
  8. If h(x)=ef(x)h(x)=e^{f(x)}, dh=ef(x)f′(x) dxdh = e^{f(x)}f'(x)\,dx.
  9. If f(x)>0f(x)>0 and g(x)=log⁡(f(x))g(x)=\log(f(x)), dg=f′(x)f(x) dxdg = \dfrac{f'(x)}{f(x)}\,dx. …
Figure 8.4Fig. 8.4 Linear Approximation and Differential — the differential $dy=f'(x)\,dx$ measured along the tangent, compared with the true change $\Delta y$ of $y=f(x)$ over the increment $dx$
Fig. 8.4 — Fig. 8.4 Linear Approximation and Differential — the differential $dy=f'(x)\,dx$ measured along the tangent, compared with the true change $\Delta y$ of $y=f(x)$ over the increment $dx$

Drawn by us to help you understand the concept clearly, and verified to make sure it's accurate. For exams, practice from your textbook's own diagram.

What this figure shows. Fig. 8.4 Linear Approximation and Differential — the differential dy=f′(x) dxdy=f'(x)\,dx measured along the tangent, compared with the true change Δy\Delta y of y=f(x)y=f(x) over …