Algebraic Rules of Differentiation

Suppose you know that f(x)=x2f(x) = x^2 and g(x)=sin⁡xg(x) = \sin x are both differentiable. You want the derivative of h(x)=3x2sin⁡x−7x2h(x) = 3x^2 \sin x - 7x^2. Must you return to the raw limit definition? Must you wrestle with a difference quotient for this whole complicated expression? The entire practical power of calculus rests on the answer being no. A small number of algebraic rules let you build the derivative of a complicated function. You do this mechanically, out of the derivatives of its simpler pieces. This matters because the difference-quotient definition, while foundational, is a blunt instrument. Computing f′(x)f'(x) from

lim⁡h→0f(x+h)−f(x)h\lim_{h\to 0}\frac{f(x+h)-f(x)}{h}

directly is often a genuine algebraic ordeal. This is true even for functions as simple as f(x)=xf(x) = \sqrt{x} or f(x)=x7f(x) = x^7. What makes differentiation into a usable calculus — a system of computation — is different. This limit behaves well under the basic operations of arithmetic: addition, scalar multiplication, multiplication, and division of functions. The big picture is that differentiation is almost, but not quite, a ring homomorphism from functions to functions. It is linear (it respects sums and scalar multiples exactly, the way a homomorphism should). But multiplication picks up an extra "cross term," giving the celebrated product rule. Division inherits a rule built from the product rule and a special case. This article proves, from the difference-quotient definition and nothing else, the three pillars of algebraic differentiation: linearity, the product rule, and the quotient rule. Each proof is elementary and self-contained. Each one exhibits the exact algebraic trick (adding and subtracting a term, or building a common denominator) that makes the limit tractable. These rules, together with The Chain Rule for composition, form the complete toolkit. This toolkit lets you differentiate any function built from elementary pieces, without ever touching the definition again.

Key takeaways
  • Differentiation is linear: (af+bg)′=af′+bg′(af+bg)' = af' + bg' for constants a,ba,b, proved directly from limit laws.
  • The product rule (fg)′=f′g+fg′(fg)' = f'g + fg' requires an "add and subtract" trick, not naive term-by-term differentiation.
  • The product rule's cross term appears because a difference quotient of a product does not factor cleanly.
  • The quotient rule (f/g)′=(f′g−fg′)/g2(f/g)' = (f'g - fg')/g^2 follows from the product rule applied to f⋅(1/g)f \cdot (1/g), plus a reciprocal-rule lemma.
  • Every rule requires the individual functions to be differentiable at the point in question; none work if a factor is merely continuous.
  • Key idea: algebraic operations on functions translate into algebraic operations on derivatives, but multiplication and division pick up correction terms that vanish only in the trivial case of constants.

Where this lives today

These three rules are the computational backbone of essentially all applied calculus. In physics, the product rule underlies the derivation of quantities like power (P=FvP = Fv, differentiated when FF and vv both vary with time) and appears constantly in deriving equations of motion and in quantum mechanics when differentiating products of wavefunctions and operators. In engineering, the quotient rule is indispensable for differentiating transfer functions and rational approximations in control theory and signal processing. In statistics and machine learning, the product and quotient rules combine with The Chain Rule to compute gradients of loss functions built from ratios and products of simpler terms — the backpropagation algorithm used to train neural networks is, at its core, a systematic bookkeeping application of these very rules to enormous compositions of functions. In pure mathematics, the linearity of differentiation is the first example students meet of a linear operator between vector spaces of functions, a concept that reappears throughout Higher-Order Derivatives, differential equations, and functional analysis, where derivatives are studied as elements of operator algebras satisfying algebraic identities directly descended from the product rule (such as the commutator relations in quantum mechanics, [x^,p^]∝1^[\hat x, \hat p] \propto \hat 1, whose derivation uses a Leibniz-rule-like identity for operators).

Ask the tutor: Can you show me how to combine the product rule and chain rule to differentiate x2sin⁡(3x)x^2 \sin(3x) step by step?

DefinitionDifferentiability and the derivative

Let ff be a real-valued function defined on an open interval containing the point a∈Ra \in \mathbb{R} (or on an interval, with aa an endpoint understood via the corresponding one-sided limit). We say ff is differentiable at aa if the limit

f′(a)  =  lim⁡h→0f(a+h)−f(a)hf'(a) \;=\; \lim_{h \to 0} \frac{f(a+h) - f(a)}{h}

exists as a finite real number. Equivalently, writing x=a+hx = a+h,

f′(a)=lim⁡x→af(x)−f(a)x−a.f'(a) = \lim_{x \to a} \frac{f(x) - f(a)}{x - a}.

If ff is differentiable at every point of a set SS, the function f′:S→Rf' : S \to \mathbb{R}, a↦f′(a)a \mapsto f'(a), is called the derivative of ff. All theorems below assume this limit definition and the basic limit laws (sum, product, and quotient laws for limits of functions), together with the fact established in Differentiability Implies Continuity that differentiability at aa forces lim⁡h→0f(a+h)=f(a)\lim_{h\to 0} f(a+h) = f(a).

Intuition

Think of a difference quotient as measuring an average rate of change over a shrinking interval. If you scale a function by a constant cc, its average rate of change over any interval scales by that same constant — nothing subtle happens, and the limit inherits the scaling exactly. If you add two functions, the average rate of change of the sum is exactly the sum of the average rates of change, because difference quotients of a sum split apart before any limit is taken. This is why linearity is "free": it is really a statement about how differences and quotients interact with addition, which is completely straightforward, and the limit laws simply carry that structure through to the limit. Multiplication is different, and the reason is geometric. Picture f(x)g(x)f(x)g(x) as the area of a rectangle with side lengths f(x)f(x) and g(x)g(x). As xx moves to x+hx+h, both sides change: the rectangle grows a thin strip along the top (from gg changing) and a thin strip along the side (from ff changing), and there is a tiny corner rectangle where both strips overlap. The corner is of order h2h^2 and vanishes in the limit, but the two strips survive, giving exactly two terms: f′(x)g(x)f'(x)g(x) from the vertical strip and f(x)g′(x)f(x)g'(x) from the horizontal strip. This is the geometric soul of the product rule.

Figure
The product rule as growing a rectangle: two strips survive, the corner vanishes
f(x) g(x) h g'(x)h corner ~ h² (vanishes) f'(x)h (top strip width) rectangle grows to (f+f'(x)h)(g+g'(x)h) Area gained ≈ f'(x)h·g(x) + f(x)·g'(x)h + O(h²)

Central theorems

TheoremLinearity of Differentiation

Let ff and gg be functions differentiable at a point aa, and let α,β∈R\alpha, \beta \in \mathbb{R} be constants. Then the function αf+βg\alpha f + \beta g is differentiable at aa, and

(αf+βg)′(a)=αf′(a)+βg′(a).(\alpha f + \beta g)'(a) = \alpha f'(a) + \beta g'(a).
Proof

Technique: direct manipulation of the difference quotient, then the sum and scalar-multiple laws for limits. By definition, for h≠0h \neq 0 small enough that a+ha+h lies in the domain of both ff and gg, form the difference quotient of αf+βg\alpha f + \beta g:

(αf+βg)(a+h)−(αf+βg)(a)h=αf(a+h)+βg(a+h)−αf(a)−βg(a)h.\frac{(\alpha f + \beta g)(a+h) - (\alpha f + \beta g)(a)}{h} = \frac{\alpha f(a+h) + \beta g(a+h) - \alpha f(a) - \beta g(a)}{h}.

Regroup the numerator by which function each term came from — this is legitimate because addition of real numbers is commutative and associative, so the terms may be reordered without changing the value:

=α(f(a+h)−f(a))+β(g(a+h)−g(a))h=α⋅f(a+h)−f(a)h+β⋅g(a+h)−g(a)h.= \frac{\alpha\big(f(a+h) - f(a)\big) + \beta\big(g(a+h) - g(a)\big)}{h} = \alpha \cdot \frac{f(a+h)-f(a)}{h} + \beta \cdot \frac{g(a+h)-g(a)}{h}.

This last equality is just distributing division by hh over a sum, valid since h≠0h \neq 0. Now take h→0h \to 0. By hypothesis, ff is differentiable at aa, so

lim⁡h→0f(a+h)−f(a)h=f′(a)\lim_{h\to 0} \frac{f(a+h)-f(a)}{h} = f'(a)

exists, and likewise lim⁡h→0g(a+h)−g(a)h=g′(a)\lim_{h\to 0} \frac{g(a+h)-g(a)}{h} = g'(a) exists. By the constant-multiple law for limits (if lim⁡h→0φ(h)=L\lim_{h\to0} \varphi(h) = L exists, then lim⁡h→0αφ(h)=αL\lim_{h\to 0}\alpha \varphi(h) = \alpha L), each of the two terms above has a limit as h→0h \to 0, namely αf′(a)\alpha f'(a) and βg′(a)\beta g'(a) respectively. By the sum law for limits (if two individual limits exist, the limit of the sum exists and equals the sum of the limits), the limit of the whole difference quotient exists and equals

lim⁡h→0[α⋅f(a+h)−f(a)h+β⋅g(a+h)−g(a)h]=αf′(a)+βg′(a).\lim_{h \to 0}\left[ \alpha \cdot \frac{f(a+h)-f(a)}{h} + \beta \cdot \frac{g(a+h)-g(a)}{h}\right] = \alpha f'(a) + \beta g'(a).

Since the difference quotient of αf+βg\alpha f + \beta g converges to this value as h→0h \to 0, the function αf+βg\alpha f + \beta g is by definition differentiable at aa with derivative αf′(a)+βg′(a)\alpha f'(a) + \beta g'(a). ■\blacksquare

∎
CorollarySum rule and constant multiple rule

Taking α=β=1\alpha = \beta = 1 gives (f+g)′(a)=f′(a)+g′(a)(f+g)'(a) = f'(a) + g'(a). Taking β=0\beta = 0 gives (αf)′(a)=αf′(a)(\alpha f)'(a) = \alpha f'(a). Taking α=1,β=−1\alpha=1,\beta=-1 gives the difference rule (f−g)′(a)=f′(a)−g′(a)(f - g)'(a) = f'(a) - g'(a).

TheoremProduct Rule

Let ff and gg be functions differentiable at a point aa. Then the product fgfg is differentiable at aa, and

(fg)′(a)=f′(a)g(a)+f(a)g′(a).(fg)'(a) = f'(a)g(a) + f(a)g'(a).
Proof

Technique: add and subtract a well-chosen cross term inside the numerator, then split the limit using <a class="wiki-link" href="/wiki/math-analysis-differentiability-implies-continuity" data-track="wiki_article.body.wiki_link">Differentiability Implies Continuity</a>. Fix h≠0h \neq 0 small enough that a+ha + h is in the common domain of ff and gg. The difference quotient of fgfg at aa is

(fg)(a+h)−(fg)(a)h=f(a+h)g(a+h)−f(a)g(a)h.\frac{(fg)(a+h) - (fg)(a)}{h} = \frac{f(a+h)g(a+h) - f(a)g(a)}{h}.

The obstruction to splitting this immediately is that the numerator mixes f(a+h),f(a)f(a+h), f(a) with g(a+h),g(a)g(a+h), g(a) inseparably. The standard resolution is to insert the term f(a+h)g(a)f(a+h)g(a), which is subtracted immediately after being added, so the numerator's value is unchanged:

f(a+h)g(a+h)−f(a)g(a)=[f(a+h)g(a+h)−f(a+h)g(a)]+[f(a+h)g(a)−f(a)g(a)].f(a+h)g(a+h) - f(a)g(a) = \Big[f(a+h)g(a+h) - f(a+h)g(a)\Big] + \Big[f(a+h)g(a) - f(a)g(a)\Big].

Factor each bracket: the first factors f(a+h)f(a+h) out of both terms, the second factors g(a)g(a) out of both terms:

=f(a+h)(g(a+h)−g(a))+(f(a+h)−f(a))g(a).= f(a+h)\big(g(a+h) - g(a)\big) + \big(f(a+h) - f(a)\big)g(a).

Divide by hh (legitimate since h≠0h \neq 0), distributing over the sum:

(fg)(a+h)−(fg)(a)h=f(a+h)⋅g(a+h)−g(a)h+f(a+h)−f(a)h⋅g(a).\frac{(fg)(a+h)-(fg)(a)}{h} = f(a+h) \cdot \frac{g(a+h)-g(a)}{h} + \frac{f(a+h)-f(a)}{h} \cdot g(a).

Now let h→0h \to 0 and examine each of the two terms. Second term. Since ff is differentiable at aa, f(a+h)−f(a)h→f′(a)\dfrac{f(a+h)-f(a)}{h} \to f'(a) as h→0h \to 0. Since g(a)g(a) is a fixed constant (it does not depend on hh), the constant-multiple limit law gives

f(a+h)−f(a)h⋅g(a)  ⟶  f′(a) g(a).\frac{f(a+h)-f(a)}{h}\cdot g(a) \;\longrightarrow\; f'(a)\,g(a).

First term. We need two facts: g(a+h)−g(a)h→g′(a)\dfrac{g(a+h)-g(a)}{h} \to g'(a), which holds since gg is differentiable at aa; and f(a+h)→f(a)f(a+h) \to f(a) as h→0h \to 0. The latter is precisely the continuity of ff at aa, which follows from the differentiability of ff at aa by Differentiability Implies Continuity (this is the one place that theorem is needed — differentiability alone does not immediately hand us f(a+h)→f(a)f(a+h) \to f(a) without invoking it). By the product law for limits (the limit of a product of two functions, each possessing a limit at aa, is the product of the limits), we get

f(a+h)⋅g(a+h)−g(a)h  ⟶  f(a) g′(a).f(a+h) \cdot \frac{g(a+h)-g(a)}{h} \;\longrightarrow\; f(a)\,g'(a).

By the sum law for limits, since both individual limits above exist, the full difference quotient converges as h→0h\to 0 to the sum:

lim⁡h→0(fg)(a+h)−(fg)(a)h=f(a)g′(a)+f′(a)g(a).\lim_{h\to 0}\frac{(fg)(a+h) - (fg)(a)}{h} = f(a)g'(a) + f'(a)g(a).

Hence fgfg is differentiable at aa with (fg)′(a)=f′(a)g(a)+f(a)g′(a)(fg)'(a) = f'(a)g(a) + f(a)g'(a). ■\blacksquare

∎
LemmaReciprocal Rule

Let gg be differentiable at aa with g(a)≠0g(a) \neq 0. Then 1/g1/g is differentiable at aa, and

(1g)′(a)=−g′(a)g(a)2.\left(\frac{1}{g}\right)'(a) = -\frac{g'(a)}{g(a)^2}.
Proof

Technique: continuity to guarantee gg is nonzero nearby, then a single common-denominator algebraic manipulation. Since gg is differentiable at aa, it is continuous at aa by Differentiability Implies Continuity, so lim⁡x→ag(x)=g(a)≠0\lim_{x\to a} g(x) = g(a) \neq 0. By the definition of a limit (applied with ε=∣g(a)∣/2>0\varepsilon = |g(a)|/2 > 0), there is δ>0\delta > 0 such that for all xx with 0<∣x−a∣<δ0 < |x-a| < \delta (and also at x=ax=a itself, trivially), ∣g(x)−g(a)∣<∣g(a)∣/2|g(x) - g(a)| < |g(a)|/2, which by the reverse triangle inequality forces ∣g(x)∣>∣g(a)∣/2>0|g(x)| > |g(a)|/2 > 0. So g(x)≠0g(x) \neq 0 for all xx within δ\delta of aa, and 1/g1/g is well-defined on this punctured neighborhood together with aa. For h≠0h \neq 0 with ∣h∣<δ|h| < \delta, form the difference quotient of 1/g1/g:

1g(a+h)−1g(a)h=1h⋅g(a)−g(a+h)g(a+h)g(a)=−1g(a+h)g(a)⋅g(a+h)−g(a)h,\frac{\frac{1}{g(a+h)} - \frac{1}{g(a)}}{h} = \frac{1}{h}\cdot \frac{g(a) - g(a+h)}{g(a+h)g(a)} = -\frac{1}{g(a+h)g(a)} \cdot \frac{g(a+h)-g(a)}{h},

where the middle equality is a single common-denominator combination of the two fractions 1g(a+h)\frac1{g(a+h)} and 1g(a)\frac1{g(a)}, and the last step just factors out a sign and separates the difference quotient of gg. As h→0h \to 0: g(a+h)−g(a)h→g′(a)\dfrac{g(a+h)-g(a)}{h} \to g'(a) since gg is differentiable at aa; and g(a+h)→g(a)g(a+h) \to g(a) by continuity of gg at aa, so g(a+h)g(a)→g(a)2≠0g(a+h)g(a) \to g(a)^2 \neq 0, whence −1g(a+h)g(a)→−1g(a)2-\dfrac{1}{g(a+h)g(a)} \to -\dfrac{1}{g(a)^2} by the quotient law for limits (valid since the denominator's limit g(a)2g(a)^2 is nonzero). By the product law for limits, the whole expression converges to

−1g(a)2⋅g′(a)=−g′(a)g(a)2.-\frac{1}{g(a)^2}\cdot g'(a) = -\frac{g'(a)}{g(a)^2}.

Hence 1/g1/g is differentiable at aa with the claimed derivative. ■\blacksquare

∎
TheoremQuotient Rule

Let ff and gg be functions differentiable at a point aa, with g(a)≠0g(a) \neq 0. Then f/gf/g is differentiable at aa, and

(fg)′(a)=f′(a)g(a)−f(a)g′(a)g(a)2.\left(\frac{f}{g}\right)'(a) = \frac{f'(a)g(a) - f(a)g'(a)}{g(a)^2}.
Proof

Technique: reduce to the already-proved product rule and reciprocal rule (composition of previously established results, not the raw limit definition again). Write f/g=f⋅(1/g)f/g = f \cdot (1/g) as an equality of functions on the neighborhood of aa where gg is nonzero (which exists by continuity of gg at aa, as established inside the proof of the Reciprocal Rule above). By the Reciprocal Rule, 1/g1/g is differentiable at aa with (1/g)′(a)=−g′(a)/g(a)2(1/g)'(a) = -g'(a)/g(a)^2. Since ff is differentiable at aa by hypothesis, both factors in the product f⋅(1/g)f \cdot (1/g) are differentiable at aa, so the Product Rule applies directly:

(fg)′(a)=(f⋅(1/g))′(a)=f′(a)⋅1g(a)+f(a)⋅(1g)′(a)=f′(a)g(a)−f(a)g′(a)g(a)2.\left(\frac{f}{g}\right)'(a) = (f \cdot (1/g))'(a) = f'(a)\cdot \frac{1}{g(a)} + f(a)\cdot \left(\frac{1}{g}\right)'(a) = \frac{f'(a)}{g(a)} - \frac{f(a)g'(a)}{g(a)^2}.

Combine the two fractions over the common denominator g(a)2g(a)^2 (multiplying the first fraction's numerator and denominator by g(a)g(a), which is nonzero, so this is a valid algebraic step):

=f′(a)g(a)g(a)2−f(a)g′(a)g(a)2=f′(a)g(a)−f(a)g′(a)g(a)2.= \frac{f'(a)g(a)}{g(a)^2} - \frac{f(a)g'(a)}{g(a)^2} = \frac{f'(a)g(a) - f(a)g'(a)}{g(a)^2}.

This is precisely the claimed formula. ■\blacksquare Note that this proof is legitimately non-circular: it invokes the Product Rule and the Reciprocal Rule, both already proved directly from the difference-quotient definition, and does not smuggle in the Quotient Rule itself or anything depending on it.

∎

Worked examples

Example 1 (linearity and product rule together, asymmetric coefficients). Let f(x)=3x2−5f(x) = 3x^2 - 5 and g(x)=xg(x) = \sqrt{x} for x>0x > 0. We take as previously established (from the definition, in the article on the derivative) that

ddx[x2]=2x,ddx[1]=0,ddx[x]=12x.\frac{d}{dx}[x^2] = 2x, \qquad \frac{d}{dx}[1] = 0, \qquad \frac{d}{dx}[\sqrt{x}] = \frac{1}{2\sqrt{x}}.

By linearity, f′(x)=3(2x)−5(0)=6xf'(x) = 3(2x) - 5(0) = 6x. Let h(x)=f(x)g(x)=(3x2−5)xh(x) = f(x)g(x) = (3x^2-5)\sqrt{x}. By the Product Rule,

h′(x)=f′(x)g(x)+f(x)g′(x)=6xx+(3x2−5)⋅12x.h'(x) = f'(x)g(x) + f(x)g'(x) = 6x\sqrt{x} + (3x^2-5)\cdot \frac{1}{2\sqrt{x}}.

Simplify by writing x=x1/2\sqrt x = x^{1/2} and putting everything over 2x2\sqrt{x}:

h′(x)=12x⋅x+(3x2−5)2x=12x2+3x2−52x=15x2−52x.h'(x) = \frac{12x\cdot x + (3x^2 - 5)}{2\sqrt{x}} = \frac{12x^2 + 3x^2 - 5}{2\sqrt{x}} = \frac{15x^2 - 5}{2\sqrt{x}}.

Check at x=4x=4: directly, h(x)=3x5/2−5x1/2h(x) = 3x^{5/2} - 5x^{1/2}, so h′(x)=152x3/2−52x−1/2h'(x) = \frac{15}{2}x^{3/2} - \frac{5}{2}x^{-1/2}. At x=4x=4: x3/2=8x^{3/2} = 8, x−1/2=1/2x^{-1/2} = 1/2, giving h′(4)=152(8)−52(12)=60−54=2354h'(4) = \frac{15}{2}(8) - \frac{5}{2}\left(\frac12\right) = 60 - \frac{5}{4} = \frac{235}{4}. From our formula: 15(16)−52(2)=240−54=2354\dfrac{15(16)-5}{2(2)} = \dfrac{240-5}{4} = \dfrac{235}{4}. The two computations agree, confirming the rule was applied correctly.

Example 2 (quotient rule with a genuinely asymmetric numerator and denominator). Let

F(x)=2x3+1x2−4,x≠±2.F(x) = \frac{2x^3 + 1}{x^2 - 4}, \qquad x \neq \pm 2.

Set f(x)=2x3+1f(x) = 2x^3+1, so f′(x)=6x2f'(x) = 6x^2, and g(x)=x2−4g(x) = x^2 - 4, so g′(x)=2xg'(x) = 2x. At a=3a = 3: f(3)=2(27)+1=55f(3) = 2(27)+1 = 55, f′(3)=6(9)=54f'(3) = 6(9) = 54, g(3)=9−4=5g(3) = 9-4=5, g′(3)=6g'(3) = 6. By the Quotient Rule,

F′(3)=f′(3)g(3)−f(3)g′(3)g(3)2=54⋅5−55⋅652=270−33025=−6025=−125.F'(3) = \frac{f'(3)g(3) - f(3)g'(3)}{g(3)^2} = \frac{54\cdot 5 - 55\cdot 6}{5^2} = \frac{270 - 330}{25} = \frac{-60}{25} = -\frac{12}{5}.

Sanity check via direct expansion near a=3a=3: using the general formula,

F′(x)=6x2(x2−4)−(2x3+1)(2x)(x2−4)2=6x4−24x2−4x4−2x(x2−4)2=2x4−24x2−2x(x2−4)2.F'(x) = \frac{6x^2(x^2-4) - (2x^3+1)(2x)}{(x^2-4)^2} = \frac{6x^4 - 24x^2 - 4x^4 - 2x}{(x^2-4)^2} = \frac{2x^4 - 24x^2 - 2x}{(x^2-4)^2}.

At x=3x=3: numerator =2(81)−24(9)−2(3)=162−216−6=−60= 2(81) - 24(9) - 2(3) = 162 - 216 - 6 = -60, denominator =(9−4)2=25=(9-4)^2 = 25, giving −60/25=−12/5-60/25 = -12/5, matching the direct substitution above.

CounterexampleWhy differentiability of each factor is essential

Weaken the Product Rule's hypothesis by allowing gg to be merely continuous at aa rather than differentiable there, and see the conclusion fail. Let

f(x)=x,g(x)=∣x∣,a=0.f(x) = x, \qquad g(x) = |x|, \qquad a = 0.

Here ff is differentiable everywhere with f′(0)=1f'(0) = 1, and gg is continuous at 00 but not differentiable at 00 (its left and right difference quotients are −1-1 and 11 respectively, so no two-sided limit exists). The product is

(fg)(x)=x∣x∣={x2x≥0−x2x<0.(fg)(x) = x|x| = \begin{cases} x^2 & x \geq 0 \\ -x^2 & x < 0.\end{cases}

Compute (fg)′(0)(fg)'(0) directly from the definition:

lim⁡h→0h∣h∣−0h=lim⁡h→0∣h∣=0,\lim_{h\to 0} \frac{h|h| - 0}{h} = \lim_{h \to 0} |h| = 0,

so (fg)′(0)=0(fg)'(0) = 0 actually does exist in this particular case — but notice the "formula" f′(0)g(0)+f(0)g′(0)f'(0)g(0) + f(0)g'(0) cannot even be written down, since g′(0)g'(0) does not exist. The theorem's conclusion (that the stated formula computes the derivative) is simply inapplicable, not merely numerically wrong, once the hypothesis is dropped: there is no valid substitute expression to fall back on. For a case where the product genuinely fails to be differentiable, take g(x)=∣x∣g(x) = |x| again but f(x)=1f(x) = 1 (a nonzero constant, differentiable everywhere with f′≡0f'\equiv 0); then fg=g=∣x∣fg = g = |x| itself is not differentiable at 00, so the conclusion "fgfg is differentiable at aa" fails outright — confirming that differentiability of every factor is not a cosmetic hypothesis but load-bearing.

Scope, edge cases & common traps

SituationWhat happens
One factor is differentiable, the other only continuousProduct Rule does not apply; product may fail to be differentiable (see counterexample)
g(a)=0g(a) = 0 in the Quotient Rulef/gf/g is not even defined at aa; the rule requires g(a)≠0g(a) \neq 0 as a standing hypothesis
Both f,gf,g differentiable but gg vanishes at some other point nearbyNo problem for the derivative at aa itself, since only local behavior near aa matters
Trying to differentiate fgfg "term by term" as f′g′f'g'A common student error; (fg)′≠f′g′(fg)' \neq f'g' in general — verify on f=g=xf=g=x: (x⋅x)′=2x≠1⋅1(x\cdot x)' = 2x \neq 1\cdot 1
Applying linearity to infinitely many summandsLinearity as proved here is only for finite linear combinations; infinite sums require uniform convergence hypotheses, a separate topic

The mathematician behind it

Origins · 1 minGottfried Wilhelm Leibniz and how this was discovered
GW
Gottfried Wilhelm Leibniz
1646–1716 · Development of calculus and its notation in late 17th-century Europe

Leibniz, working independently of and roughly contemporaneously with Newton, developed the differential calculus with an emphasis on symbolic notation and formal rules of manipulation — the dxdx, dydy, and ∫\int notation still used today are his. The product rule for differentials, d(uv)=u dv+v dud(uv) = u\,dv + v\,du, appears explicitly in his 1675–1676 manuscripts, derived from an infinitesimal argument almost identical in spirit to the "add and subtract the cross term" trick used in the modern epsilon-limit proof above. Leibniz's insistence on treating calculus as an algebraic calculus of symbols — a system of formal rules that could be applied mechanically once justified — is precisely the philosophy realized rigorously by the limit-based proofs in this article, two centuries later, once Cauchy and Weierstrass supplied the rigorous notion of limit that Leibniz's infinitesimals lacked.

History
- 1675 — Leibniz privately derives an early form of the product rule for differentials.
- 1684 — Leibniz publishes Nova Methodus, the first printed calculus paper, including differentiation rules.
- 1821 — Cauchy gives the first rigorous limit-based definition of the derivative in Cours d'Analyse.
- 1854–1861 — Weierstrass formalizes the ε\varepsilon–δ\delta definition of limit, making today's proofs fully rigorous.
- Present — the algebraic rules proved here remain the standard first computational toolkit in every calculus course.
Why can't I just differentiate ff and gg separately and multiply the results to get (fg)′(fg)'?

Because a difference quotient of a product does not decompose into a product of the individual difference quotients — the algebra genuinely does not allow it. Geometrically, fgfg is like the area of a rectangle whose two side lengths both change simultaneously; the change in area comes from two independent "strips" (one from each side changing) plus a negligible corner term, giving exactly f′g+fg′f'g + fg', not f′g′f'g'. You can verify this fails with the simplest possible example: f(x)=g(x)=xf(x)=g(x)=x gives (x⋅x)′=(x2)′=2x(x\cdot x)'=(x^2)'=2x, while f′(x)g′(x)=1⋅1=1f'(x)g'(x) = 1\cdot 1 = 1, and these agree only at x=1/2x=1/2, confirming the two expressions are not generally equal.

Do I need g(a)≠0g(a)\neq 0 for the whole domain, or just at the point aa?

Just at the point aa (and, as the proof shows, this automatically forces gg to be nonzero on some neighborhood of aa too, by continuity). The Quotient Rule as stated is a pointwise theorem: it computes (f/g)′(a)(f/g)'(a) assuming g(a)≠0g(a)\neq 0, and says nothing about points where gg vanishes elsewhere. If you want a derivative formula valid on an entire interval, you need gg nonzero throughout that interval, and then the formula (f/g)′=(f′g−fg′)/g2(f/g)' = (f'g-fg')/g^2 holds pointwise everywhere on it.

Is the reciprocal rule really necessary, or could I prove the quotient rule directly from the difference quotient without it?

You could prove it directly, and it's a good exercise, but structuring the proof through the reciprocal rule is cleaner and exemplifies good mathematical practice: build small, fully-proved lemmas, then combine them. Writing f/g=f⋅(1/g)f/g = f\cdot(1/g) and invoking the already-proved Product Rule avoids re-deriving a common-denominator limit argument from scratch and makes the logical dependency between the rules completely transparent and non-circular.

Does linearity extend to more than two functions, like af+bg+chaf+bg+ch?

Yes, by a straightforward induction on the number of terms, using the two-term case proved here at each step. If af+bgaf+bg is differentiable with the expected derivative, and hh is differentiable, then (af+bg)+ch(af+bg)+ch is again a sum of two differentiable functions, so the two-term Linearity Theorem applies again, one summand at a time. This is why calculus textbooks state linearity for a "finite linear combination" without separately proving the three-term, four-term, etc. cases.

Why does the counterexample use f(x)=xf(x)=x, g(x)=∣x∣g(x)=|x| specifically, rather than something else?

It is chosen because ∣x∣|x| is the simplest standard example of a function that is continuous everywhere but fails to be differentiable at exactly one point, 00, due to the corner in its graph (its left- and right-hand difference quotients are −1-1 and +1+1, which disagree). Multiplying by f(x)=xf(x)=x produces x∣x∣x|x|, which turns out, by a small coincidence of this particular example, to still be differentiable at 00 — which is itself an instructive point: dropping a hypothesis does not always produce a dramatic failure, but it always voids your right to use the theorem's formula, since g′(0)g'(0) literally does not exist to plug in.

How do these rules relate to the Chain Rule?

The rules proved here handle functions built by the arithmetic operations +,−,×,÷+, -, \times, \div on simpler functions. The Chain Rule separately handles composition, f(g(x))f(g(x)), which is a fundamentally different way of combining functions and requires its own proof (typically via a linear-approximation or "little-oo" argument, since the naive difference-quotient manipulation runs into a 0/00/0 issue when g(a+h)=g(a)g(a+h)=g(a) for some hh near 00). Together, linearity, product rule, quotient rule, and chain rule are jointly sufficient to differentiate every function built from polynomials, roots, trigonometric, exponential, and logarithmic functions via finitely many algebraic operations and compositions.

What is the analogous higher-order product rule for (fg)(n)(fg)^{(n)}?

That is the general Leibniz rule, (fg)(n)=∑k=0n(nk)f(k)g(n−k)(fg)^{(n)} = \sum_{k=0}^n \binom{n}{k} f^{(k)} g^{(n-k)}, which generalizes the single product rule proved here to nn-th derivatives; it is proved by induction on nn using the ordinary Product Rule and Pascal's identity for binomial coefficients at each inductive step. This connects directly to the topic of Higher-Order Derivatives, where repeated differentiation of products, quotients, and compositions is studied systematically.

Before this

Where this leads