CALCULUS 2 • SEQUENCES & SERIES

Taylor Polynomial Approximations — Finding Taylor Polynomial Approximations of Functions

Approximating complex functions with polynomials built from derivatives at a single point.

Historical Context & Motivation

Long before computers could evaluate transcendental functions like sin(x), eˣ, or ln(x) with the press of a button, mathematicians faced a fundamental challenge: how does one compute the value of a function that cannot be expressed through finite algebraic operations? The answer—one of the most elegant ideas in all of analysis—is to replace difficult functions with polynomials that mimic their behavior near a chosen point. This idea, now formalized as the Taylor polynomial approximation, grew out of centuries of effort to reconcile the infinite with the finite, and it remains indispensable in modern engineering, physics, and numerical computation.

1668
James Gregory's Series Expansions
Scottish mathematician James Gregory discovered infinite series representations for arctan(x) and other trigonometric functions, anticipating the formal theory by decades. His work demonstrated that transcendental functions could be captured by infinite sums of polynomial terms.
1715
Brook Taylor's Methodus Incrementorum Directa
Brook Taylor published his treatise containing the general formula for expanding a function as an infinite power series about any point a. Although the result had been used informally before, Taylor's systematic treatment earned the series its name.
1742
Colin Maclaurin's Special Case
Colin Maclaurin popularized the special case where the expansion is centered at a = 0, now known as the Maclaurin series. His Treatise of Fluxions applied the technique extensively to geometry and mechanics.
1797
Lagrange's Remainder Theorem
Joseph-Louis Lagrange provided a rigorous formula for the remainder term R_n(x), transforming Taylor polynomials from heuristic tools into rigorously bounded approximations. This remainder form allowed mathematicians—and later engineers—to guarantee the accuracy of truncated series.
1821
Cauchy's Rigorous Foundation
Augustin-Louis Cauchy established the modern ε-δ framework for convergence, placing Taylor series on firm analytic ground. His work clarified when and where a Taylor series actually converges to the original function, a subtlety that earlier treatments had glossed over.

The central question driving this theory is deceptively simple: given a smooth function f(x), can we construct a polynomial Pn(x) of degree n that agrees with f not only in value at some point a but also in its first n derivatives? If so, how good is the approximation, and how far from a can we trust it? These are the questions that Taylor polynomial approximations answer with remarkable precision.

Core Principles & Definitions

A Taylor polynomial is constructed so that it matches a target function's behavior—its value and the values of its successive derivatives—at a single point called the center of expansion. The more derivative information we encode into the polynomial, the better it tends to approximate the function near that center. Understanding the construction requires a few foundational ideas that connect derivatives, factorials, and polynomial algebra.

1

Derivative Matching

The nth-degree Taylor polynomial Pn(x) is the unique polynomial of degree ≤ n satisfying Pn(k)(a) = f(k)(a) for every k from 0 to n. Each derivative at a pins down one coefficient of the polynomial.
2

Factorial Scaling

The coefficient of the (x − a)ᵏ term is f(k)(a) / k!. The factorial in the denominator arises naturally: differentiating (x − a)ᵏ exactly k times produces k!, so dividing by k! ensures the derivatives match precisely.
3

Center of Expansion

The point a around which the polynomial is built is called the center. When a = 0, the result is a Maclaurin polynomial. The approximation is most accurate near x = a and generally degrades as |x − a| increases.
4

Degree vs. Order

The degree of the Taylor polynomial controls the fidelity of the approximation. A first-degree polynomial gives the tangent line; a second-degree polynomial captures curvature; higher degrees encode subtler shape features like inflection and oscillation.
5

Remainder & Error

The difference f(x) − Pn(x) is the remainder Rn(x). Lagrange's form bounds this error using the (n+1)th derivative, providing a practical tool for guaranteeing approximation accuracy within a specified tolerance.
KEY TAKEAWAY
Think of a Taylor polynomial as a custom-fitted lens focused on a single point of a function's graph. A low-degree polynomial is like a blurry lens that captures only the slope and curvature; increasing the degree sharpens the lens, bringing finer details—inflection, oscillation—into focus. The factorial scaling ensures each new "layer of detail" is calibrated so the polynomial's derivatives perfectly replicate the function's derivatives at the center.

Visualizing Taylor Approximations

The power of Taylor polynomials becomes most vivid when you see successive approximations layered over the original function. The following diagram shows the function f(x) = eˣ alongside its first-, second-, third-, and fourth-degree Taylor polynomials centered at a = 0. Notice how each additional degree extends the region over which the polynomial closely follows the exponential curve.

Successive Taylor polynomials P1 through P4 for f(x) = eˣ centered at a = 0. The green dot marks the center of expansion where all polynomials agree with f. As the degree increases, the approximation hugs the exponential curve over a wider interval.

Several important features emerge from the diagram. First, observe that all four polynomials pass through the point (0, 1), since they are all centered at a = 0 and f(0) = e⁰ = 1. Second, the tangent-line approximation P1(x) = 1 + x captures only the slope at the origin and diverges quickly. The quadratic P2 adds curvature, matching the concavity of eˣ. By the time we reach P4, the polynomial is nearly indistinguishable from eˣ on the interval (−2, 2). This progressive tightening illustrates the fundamental trade-off: higher degree yields better accuracy but requires more derivative information.

Mathematical Framework

The construction of a Taylor polynomial is driven by a single organizing principle: choose the coefficients of a polynomial so that its derivatives at the center a match those of the target function f. We begin with the general formula and then examine the remainder that quantifies the error incurred by truncation.

NTH-DEGREE TAYLOR POLYNOMIAL
Pₙ(x) = f(a) + f′(a)(x − a) + f″(a)/2! × (x − a)² + f‴(a)/3! × (x − a)³ + ⋯ + f⁽ⁿ⁾(a)/n! × (x − a)ⁿ
Equivalently, Pₙ(x) = Σ from k = 0 to n of [ f(k)(a) / k! ] × (x − a)ᵏ. Here a is the center of expansion, n is the degree, and f(k)(a) denotes the kth derivative of f evaluated at a.

To verify that derivative matching holds, differentiate Pₙ(x) term by term. The kth derivative of (x − a)ᵏ / k! evaluated at x = a yields exactly 1, while every other term vanishes. Hence Pₙ(k)(a) = f(k)(a) for 0 ≤ k ≤ n, which is precisely the requirement. When the center of expansion is a = 0, the formula simplifies to the Maclaurin polynomial: Pₙ(x) = Σ [ f(k)(0) / k! ] xᵏ.

LAGRANGE REMAINDER (ERROR BOUND)
Rₙ(x) = f⁽ⁿ⁺¹⁾(c) / (n + 1)! × (x − a)ⁿ⁺¹
There exists some c between a and x such that the above holds. The remainder Rₙ(x) = f(x) − Pₙ(x) tells us exactly how much error the nth-degree polynomial incurs. To bound |Rₙ(x)|, find M = max |f(n+1)(t)| for t between a and x, giving |Rₙ(x)| ≤ M|x − a|ⁿ⁺¹ / (n + 1)!.
COMMON MACLAURIN POLYNOMIALS
eˣ ≈ 1 + x + x²/2! + x³/3! + ⋯ | sin x ≈ x − x³/3! + x⁵/5! − ⋯ | cos x ≈ 1 − x²/2! + x⁴/4! − ⋯
These three expansions are worth memorizing. Notice that sin x has only odd powers (odd function) and cos x has only even powers (even function). All three converge for every real number x, meaning the Taylor series converges to the function on (−∞, ∞).
💡 Derivation Insight
Why does k! appear in the denominator? When you differentiate xᵏ once, you get kxᵏ⁻¹. Differentiate again: k(k−1)xᵏ⁻². After k differentiations, you obtain k! × x⁰ = k!. So if the coefficient of xᵏ is f(k)(0) / k!, then the kth derivative of the polynomial at 0 is (k!) × f(k)(0) / k! = f(k)(0), exactly as required.

Error Behavior & Convergence

Understanding how quickly the error decreases as the degree n increases is essential for practical applications. The Lagrange remainder provides a worst-case bound, but examining actual error values for concrete functions reveals the remarkable efficiency of Taylor polynomials. The following diagram and table illustrate the error |eˣ − Pₙ(x)| at x = 1 for increasing polynomial degree.

Bar chart showing the absolute error |e¹ − Pₙ(1)| for n = 0 through 5. The dashed pink curve connecting the bar tops illustrates the rapid, super-linear decay of error as degree increases. By n = 5 the error is approximately 0.002—less than 0.1% of the true value.
Error in approximating e = e¹ using Maclaurin polynomials of increasing degree
Degree nPₙ(1)|e − Pₙ(1)|Lagrange Bound
011.71828e ≈ 2.718
120.71828e/2 ≈ 1.359
22.50.21828e/6 ≈ 0.453
32.66670.05161e/24 ≈ 0.113
42.70830.00994e/120 ≈ 0.023
52.71670.00161e/720 ≈ 0.004

The table reveals two important patterns. First, the actual error is always smaller than the Lagrange bound, confirming that the bound is conservative but reliable. Second, the error decreases roughly by a factor of 1/(n+1) with each additional degree—a consequence of the factorial growth in the denominator of each successive term. This factorial growth is the mathematical engine behind the remarkable accuracy of Taylor approximations: even for modest values of n, the factorial dominates the power (x − a)ⁿ⁺¹, driving the remainder toward zero.

Worked Example

Let us construct the third-degree Taylor polynomial for f(x) = ln(x) centered at a = 1, and then use it to approximate ln(1.2).

Finding P₃(x) for f(x) = ln(x) centered at a = 1
1
Step 1 — Compute derivatives of f(x) = ln(x)We need f(a), f′(a), f″(a), and f‴(a) where a = 1. Compute each derivative in turn: f(x) = ln(x), so f′(x) = 1/x, f″(x) = −1/x², and f‴(x) = 2/x³.
f′(x) = x⁻¹, f″(x) = −x⁻², f‴(x) = 2x⁻³
2
Step 2 — Evaluate each derivative at a = 1Substituting x = 1 into each expression: f(1) = ln(1) = 0, f′(1) = 1/1 = 1, f″(1) = −1/1² = −1, and f‴(1) = 2/1³ = 2.
f(1) = 0, f′(1) = 1, f″(1) = −1, f‴(1) = 2
3
Step 3 — Assemble the Taylor polynomial formulaSubstitute into P₃(x) = f(a) + f′(a)(x − a) + f″(a)/2! × (x − a)² + f‴(a)/3! × (x − a)³. With a = 1, this gives P₃(x) = 0 + 1 × (x − 1) + (−1)/2 × (x − 1)² + 2/6 × (x − 1)³.
P₃(x) = (x − 1) − (x − 1)²/2 + (x − 1)³/3
4
Step 4 — Approximate ln(1.2)Set x = 1.2, so (x − 1) = 0.2. Then P₃(1.2) = 0.2 − (0.2)²/2 + (0.2)³/3 = 0.2 − 0.02 + 0.00267 = 0.18267.
P₃(1.2) ≈ 0.18267
5
Step 5 — Assess accuracyThe true value is ln(1.2) ≈ 0.18232. The absolute error is |0.18267 − 0.18232| ≈ 0.00035. Using the Lagrange remainder with f⁽⁴⁾(x) = −6/x⁴, the bound gives |R₃(1.2)| ≤ 6 × (0.2)⁴ / 4! = 6 × 0.0016 / 24 = 0.0004, which is consistent with our observed error.
Absolute error ≈ 0.00035 (within Lagrange bound of 0.0004)

Strengths & Limitations

Taylor polynomials are an extraordinarily versatile tool, but they are not without limitations. Understanding when to deploy them—and when alternative methods may be preferable—is as important as knowing how to construct them.

Comparison of strengths and limitations of Taylor polynomial approximations
StrengthsLimitations
Polynomials are trivial to evaluate—only addition, multiplication, and exponentiation are needed, making them ideal for numerical computation.Accuracy degrades as x moves away from the center a; the approximation is inherently local.
The Lagrange remainder provides rigorous, computable error bounds, enabling engineers and scientists to certify approximation quality.High-degree polynomials require computing many higher-order derivatives, which can be algebraically complex or computationally expensive.
Widely applicable: any function that is sufficiently differentiable admits a Taylor expansion.Not all Taylor series converge to their generating function. The classic counterexample is f(x) = e^(−1/x²), whose Maclaurin series is identically zero despite f being nonzero for x ≠ 0.
Taylor expansions facilitate symbolic manipulation: integrating, differentiating, or composing power series is often easier than working with the original function.Convergence may be limited to a finite radius. For instance, the Taylor series for 1/(1 + x²) about a = 0 converges only on (−1, 1).
PRACTICAL PERSPECTIVE
In engineering practice, Taylor polynomials are the workhorse behind everything from your calculator's sin button (which uses a truncated Maclaurin polynomial internally) to linearization in control theory, where complex nonlinear systems are approximated by their first-degree Taylor polynomial (the tangent-line approximation) near an operating point. The key is choosing the right degree and the right center for the problem at hand: too few terms and you lose accuracy; too many and you waste computation.

From Taylor Polynomials to Taylor Series

A Taylor polynomial of degree n is a finite approximation. The natural question is: what happens as n → ∞? If the infinite sum converges, we obtain the Taylor series of f centered at a. The Taylor series represents f(x) exactly (not approximately) within a region called the interval of convergence. This transition from polynomial to series is the bridge connecting this lesson to power series theory and, ultimately, to the full machinery of analytic functions.

Taylor polynomial vs. Taylor series
FeatureTaylor Polynomial Pₙ(x)Taylor Series
Number of termsFinite (n + 1 terms)Infinite (Σ from k = 0 to ∞)
Relationship to fApproximation with computable error Rₙ(x)Exact representation (when convergent)
Convergence concernNo convergence issue—always a finite sumMust verify convergence; may diverge outside a radius R
Typical useNumerical evaluation, linearization, error estimationExact symbolic representation, proof techniques, generating functions
Key formulaPₙ(x) = Σ_{k=0}^{n} f⁽ᵏ⁾(a)/k! × (x−a)ᵏf(x) = Σ_{k=0}^{∞} f⁽ᵏ⁾(a)/k! × (x−a)ᵏ, |x−a| < R

The transition to infinite series introduces new mathematical questions that form the core of the next several topics in this course: Does the series converge? If so, for which x? Does convergence actually mean the series equals f(x), or could the series converge to something else? These questions lead to the ratio and root tests for determining the radius of convergence, and to the concept of analytic functions—functions that equal their Taylor series everywhere within the interval of convergence. For now, the key insight is that every Taylor polynomial is a partial sum of the Taylor series, and the quality of the polynomial approximation is governed by how quickly the series converges.

Practice Problems

PROBLEM 1CONCEPTUAL
Explain why the coefficient of the (x − a)ᵏ term in the Taylor polynomial must be f(k)(a) / k! rather than simply f(k)(a). What role does the factorial play?
PROBLEM 2BASIC CALCULATION
Find the second-degree Maclaurin polynomial P₂(x) for f(x) = cos(x). Then use it to approximate cos(0.3).
PROBLEM 3INTERMEDIATE
Find the third-degree Taylor polynomial for f(x) = √x centered at a = 4. Use it to estimate √4.5 and bound the error using the Lagrange remainder.
PROBLEM 4APPLIED
A physicist needs to compute sin(θ) for small angles θ (in radians) in a pendulum simulation. The simulation requires |error| < 10⁻⁶. If the maximum angle is θ = 0.5 rad, what is the minimum degree Maclaurin polynomial needed? Justify your answer using the Lagrange remainder.
PROBLEM 5CRITICAL THINKING
Consider the function g(x) = 1/(1 − x). Its Maclaurin polynomial of degree n is Pₙ(x) = 1 + x + x² + ⋯ + xⁿ (a geometric partial sum). Using the Lagrange remainder, explain why the Taylor series converges to g(x) only for |x| < 1, even though g(x) is perfectly well-defined at, say, x = −2. What does this reveal about the limitations of Taylor series?

Lesson Summary

A Taylor polynomial Pₙ(x) = Σ_{k=0}^{n} f⁽ᵏ⁾(a)/k! × (x − a)ᵏ is the unique polynomial of degree ≤ n that matches a function f and its first n derivatives at a single point a, called the center of expansion. When a = 0 the result is called a Maclaurin polynomial. The factorial in each denominator arises from the differentiation rule for power functions and ensures perfect derivative matching. The most important Maclaurin expansions—for eˣ, sin x, and cos x—converge for all real x.

The Lagrange remainder Rₙ(x) = f⁽ⁿ⁺¹⁾(c)/(n+1)! × (x − a)ⁿ⁺¹ provides a rigorous error bound that enables practitioners to certify approximation quality. As the degree n increases, the factorial growth in the denominator generally forces the error toward zero, extending the region of accurate approximation. In the limit n → ∞, the Taylor polynomial becomes the Taylor series, which represents f exactly within its interval of convergence—a topic explored in depth through power series and convergence tests later in this course.

Varsity Tutors • Calculus 2 • Taylor Polynomial Approximations