Skip to content
homeaboutworkprojectsthesiswritingresume
Loading
~/blog/differentiation0%dark
  1. home/
  2. writing/
  3. Differentiation and Taylor's Theorem

30 May 2026 · 5 min read · updated 13 June 2026

Differentiation and Taylor's Theorem

The derivative measures instantaneous rate, and the mean value theorem converts local derivative information into global control of a function. We define the derivative, show differentiability implies continuity, prove Fermat's principle that an interior extremum has vanishing derivative, prove Rolle's theorem and the mean value theorem, and prove Taylor's theorem with the Lagrange form of the remainder. The Taylor expansion with its explicit remainder is the tool the characteristic-function moment expansion and every local approximation rely on.

  • 5 equations
  • 11 results
  • 9 connections
  • real-analysis
  • calculus
On this page▾
  • The derivative
  • The mean value theorem
  • Taylor's theorem

5 min left

  • The derivative1m
  • The mean value theorem2m
  • Taylor's theorem3m

The derivative replaces a function near a point by its best linear approximation, and the mean value theorem is the lever that turns that local statement into a global one, bounding how far a function can travel from its derivative. From it follow the sign tests for monotonicity and the vanishing of the derivative of a constant. It also gives Taylor's theorem, the polynomial approximation with an exact error term that the characteristic function expansion runs on. This post builds them from the continuity of the previous post [1], [2].

#The derivative

Definition1

A function fff is differentiable at an interior point ccc of its domain when the limit

f′(c)=lim⁡x→cf(x)−f(c)x−c(1)f'(c)=\lim_{x\to c}\frac{f(x)-f(c)}{x-c} \tag{1}f′(c)=x→clim​x−cf(x)−f(c)​(1)

exists. The number f′(c)f'(c)f′(c) is the derivative.

Differentiability implies continuity, but not conversely.

Proposition2

If fff is differentiable at ccc, then fff is continuous at ccc.

Proof

For x≠cx\neq cx=c, write f(x)−f(c)=f(x)−f(c)x−c (x−c)f(x)-f(c)=\dfrac{f(x)-f(c)}{x-c}\,(x-c)f(x)−f(c)=x−cf(x)−f(c)​(x−c). As x→cx\to cx→c the difference quotient tends to f′(c)f'(c)f′(c) and x−cx-cx−c tends to 000, so the product tends to f′(c)⋅0=0f'(c)\cdot 0=0f′(c)⋅0=0 by the product law for limits. Hence f(x)→f(c)f(x)\to f(c)f(x)→f(c), which is continuity at ccc.

The converse fails, as the absolute value function is continuous but not differentiable at the origin, where its difference quotients are ±1\pm 1±1.

#The mean value theorem

The bridge from local to global passes through extrema. An interior extremum pins the derivative to zero.

Proposition3

If fff has a local maximum or minimum at an interior point ccc and is differentiable there, then f′(c)=0f'(c)=0f′(c)=0.

Proof

Suppose ccc is a local maximum, so f(x)≤f(c)f(x)\le f(c)f(x)≤f(c) near ccc. For x>cx>cx>c near ccc the quotient f(x)−f(c)x−c≤0\frac{f(x)-f(c)}{x-c}\le 0x−cf(x)−f(c)​≤0, and letting x→c+x\to c^+x→c+ gives f′(c)≤0f'(c)\le 0f′(c)≤0. For x<cx<cx<c the quotient is ≥0\ge 0≥0, and x→c−x\to c^-x→c− gives f′(c)≥0f'(c)\ge 0f′(c)≥0. The two force f′(c)=0f'(c)=0f′(c)=0. A minimum is the maximum of −f-f−f.

Theorem4

If fff is continuous on [a,b][a,b][a,b], differentiable on (a,b)(a,b)(a,b), and f(a)=f(b)f(a)=f(b)f(a)=f(b), then f′(ξ)=0f'(\xi)=0f′(ξ)=0 for some ξ∈(a,b)\xi\in(a,b)ξ∈(a,b).

Proof

By the extreme value theorem fff attains a maximum and a minimum on [a,b][a,b][a,b]. If both are attained at the endpoints, then since f(a)=f(b)f(a)=f(b)f(a)=f(b) the maximum and minimum are equal, so fff is constant and f′=0f'=0f′=0 throughout (a,b)(a,b)(a,b). Otherwise an extremum is attained at an interior point ξ\xiξ, where Proposition 3 gives f′(ξ)=0f'(\xi)=0f′(ξ)=0.

Rolle's theorem tilts into the mean value theorem by subtracting the secant line.

Theorem5

If fff is continuous on [a,b][a,b][a,b] and differentiable on (a,b)(a,b)(a,b), then there is ξ∈(a,b)\xi\in(a,b)ξ∈(a,b) with

f′(ξ)=f(b)−f(a)b−a.(2)f'(\xi)=\frac{f(b)-f(a)}{b-a}. \tag{2}f′(ξ)=b−af(b)−f(a)​.(2)
Proof

Apply Theorem 4 to g(x)=f(x)−f(a)−f(b)−f(a)b−a(x−a)g(x)=f(x)-f(a)-\dfrac{f(b)-f(a)}{b-a}(x-a)g(x)=f(x)−f(a)−b−af(b)−f(a)​(x−a), which is continuous on [a,b][a,b][a,b], differentiable on (a,b)(a,b)(a,b), and has g(a)=g(b)=0g(a)=g(b)=0g(a)=g(b)=0. The conclusion g′(ξ)=0g'(\xi)=0g′(ξ)=0 reads f′(ξ)−f(b)−f(a)b−a=0f'(\xi)-\frac{f(b)-f(a)}{b-a}=0f′(ξ)−b−af(b)−f(a)​=0, which is Equation (2).

The mean value theorem is the source of the qualitative calculus. A function with f′=0f'=0f′=0 on an interval is constant, because Equation (2) makes every difference f(b)−f(a)f(b)-f(a)f(b)−f(a) vanish, and a function with f′>0f'>0f′>0 is strictly increasing by the same identity.

#Taylor's theorem

The mean value theorem is the first-order case of a general approximation. A function with n+1n+1n+1 derivatives is approximated near a point by its degree-nnn Taylor polynomial, with an error governed by the next derivative.

Theorem6

Suppose fff has n+1n+1n+1 derivatives on an interval containing aaa and xxx. Then there is a point ξ\xiξ strictly between aaa and xxx with

f(x)=∑k=0nf(k)(a)k!(x−a)k+f(n+1)(ξ)(n+1)!(x−a)n+1.(3)f(x)=\sum_{k=0}^n\frac{f^{(k)}(a)}{k!}(x-a)^k+\frac{f^{(n+1)}(\xi)}{(n+1)!}(x-a)^{n+1}. \tag{3}f(x)=k=0∑n​k!f(k)(a)​(x−a)k+(n+1)!f(n+1)(ξ)​(x−a)n+1.(3)
Proof

Assume x≠ax\neq ax=a; for x=ax=ax=a both sides of Equation (3) equal f(a)f(a)f(a) and the remainder vanishes, so the claim is trivial. Fix xxx and define F(t)=f(x)−∑k=0nf(k)(t)k!(x−t)kF(t)=f(x)-\sum_{k=0}^n\dfrac{f^{(k)}(t)}{k!}(x-t)^kF(t)=f(x)−∑k=0n​k!f(k)(t)​(x−t)k, the error of the Taylor polynomial expanded at the moving base point ttt. The sum telescopes under differentiation. Write Tk(t)=f(k)(t)k!(x−t)kT_k(t)=\frac{f^{(k)}(t)}{k!}(x-t)^kTk​(t)=k!f(k)(t)​(x−t)k, so F=f(x)−∑k=0nTkF=f(x)-\sum_{k=0}^n T_kF=f(x)−∑k=0n​Tk​. For k≥1k\geq 1k≥1, Tk′(t)=f(k+1)(t)k!(x−t)k−f(k)(t)(k−1)!(x−t)k−1T_k'(t)=\frac{f^{(k+1)}(t)}{k!}(x-t)^k-\frac{f^{(k)}(t)}{(k-1)!}(x-t)^{k-1}Tk′​(t)=k!f(k+1)(t)​(x−t)k−(k−1)!f(k)(t)​(x−t)k−1, while T0′(t)=f′(t)T_0'(t)=f'(t)T0′​(t)=f′(t). Reindexing the second pieces by j=k−1j=k-1j=k−1 turns ∑k=1nf(k)(t)(k−1)!(x−t)k−1\sum_{k=1}^n\frac{f^{(k)}(t)}{(k-1)!}(x-t)^{k-1}∑k=1n​(k−1)!f(k)(t)​(x−t)k−1 into ∑j=0n−1f(j+1)(t)j!(x−t)j\sum_{j=0}^{n-1}\frac{f^{(j+1)}(t)}{j!}(x-t)^j∑j=0n−1​j!f(j+1)(t)​(x−t)j, which cancels all but the k=nk=nk=n first piece, leaving

F′(t)=−f(n+1)(t)n!(x−t)n.(4)F'(t)=-\frac{f^{(n+1)}(t)}{n!}(x-t)^n. \tag{4}F′(t)=−n!f(n+1)(t)​(x−t)n.(4)

Now set G(t)=F(t)−(x−tx−a)n+1F(a)G(t)=F(t)-\Big(\dfrac{x-t}{x-a}\Big)^{n+1}F(a)G(t)=F(t)−(x−ax−t​)n+1F(a), which is well defined since (x−a)n+1≠0(x-a)^{n+1}\neq 0(x−a)n+1=0. Then G(a)=F(a)−F(a)=0G(a)=F(a)-F(a)=0G(a)=F(a)−F(a)=0 and G(x)=F(x)−0=0G(x)=F(x)-0=0G(x)=F(x)−0=0, since F(x)=0F(x)=0F(x)=0. Because f(n+1)f^{(n+1)}f(n+1) exists at every point of the interval (endpoints included), f(n)f^{(n)}f(n) is differentiable and hence continuous there, so FFF, a combination of f(0),…,f(n)f^{(0)},\dots,f^{(n)}f(0),…,f(n) with polynomial coefficients, and GGG are continuous on the closed interval between aaa and xxx and differentiable on its interior. By Rolle's theorem there is ξ\xiξ strictly between aaa and xxx with G′(ξ)=0G'(\xi)=0G′(ξ)=0. Differentiating, G′(t)=F′(t)+(n+1)(x−t)n(x−a)n+1F(a)G'(t)=F'(t)+(n+1)\dfrac{(x-t)^n}{(x-a)^{n+1}}F(a)G′(t)=F′(t)+(n+1)(x−a)n+1(x−t)n​F(a), so G′(ξ)=0G'(\xi)=0G′(ξ)=0 gives

f(n+1)(ξ)n!(x−ξ)n=(n+1)(x−ξ)n(x−a)n+1F(a)(5)\frac{f^{(n+1)}(\xi)}{n!}(x-\xi)^n=(n+1)\frac{(x-\xi)^n}{(x-a)^{n+1}}F(a) \tag{5}n!f(n+1)(ξ)​(x−ξ)n=(n+1)(x−a)n+1(x−ξ)n​F(a)(5)

after substituting Equation (4). Cancelling (x−ξ)n(x-\xi)^n(x−ξ)n, which is nonzero because ξ≠x\xi\neq xξ=x, solves for F(a)=f(n+1)(ξ)(n+1)!(x−a)n+1F(a)=\frac{f^{(n+1)}(\xi)}{(n+1)!}(x-a)^{n+1}F(a)=(n+1)!f(n+1)(ξ)​(x−a)n+1. Since F(a)=f(x)−∑k=0nf(k)(a)k!(x−a)kF(a)=f(x)-\sum_{k=0}^n\frac{f^{(k)}(a)}{k!}(x-a)^kF(a)=f(x)−∑k=0n​k!f(k)(a)​(x−a)k by definition, rearranging is Equation (3).

Taking n=0n=0n=0 recovers the mean value theorem, so Taylor's theorem is its higher-order extension, each derivative buying one more order of polynomial accuracy. The remainder is an explicit multiple of the (n+1)(n+1)(n+1)-st derivative, so it bounds the approximation error. When the derivatives are bounded it shrinks like (x−a)n+1(x-a)^{n+1}(x−a)n+1, giving the small-ooo remainder the moment expansion of a characteristic function uses to read the central limit theorem off a second-order Taylor approximation. Differentiation turns a function into its derivatives, and Taylor's theorem turns the derivatives back into the function, recovering it up to the explicit remainder.

[1]
W. Rudin, Principles of Mathematical Analysis, 3rd ed. McGraw-Hill, 1976.
[2]
S. Abbott, Understanding Analysis, 2nd ed. Springer, 2015.

Part 4 of 6 in Real Analysis

← previousMetric and Normed Spacesnext →The Riemann Integral

Explore connections

see in the atlas →

related

  • Continuity and Limits of Functions
  • Series and Power Series
  • Characteristic Functions

referenced by (6)

  • Convex Sets and Functions
  • Eigenvalues and the Spectral Theorem
  • Holomorphic Functions and Cauchy's Theorem
  • Ito's Formula
  • Series and Power Series
  • The Riemann Integral
cite
@misc{differentiation,
  author = {Zac Kienzle},
  title  = {Differentiation and Taylor's Theorem},
  year   = {2026},
  month  = {05},
  url    = {https://zackienzle.com/blog/differentiation}
}