Historical Context & Motivation
Long before modern computing, mathematicians faced a fundamental challenge: how to evaluate transcendental functions like sin x, eˣ, and ln x at arbitrary points. There were no calculators, no lookup tables for every possible input — and yet astronomers, navigators, and physicists needed numerical answers. The breakthrough came from recognizing that any sufficiently smooth function can be represented as an infinite polynomial, whose coefficients are determined entirely by the function's derivatives at a single point. This idea, which we now call a Taylor series, transformed analysis and remains one of the most powerful tools in all of mathematics.
The central question these mathematicians pursued is remarkably concrete: given a function f(x) that may be complicated to evaluate directly, can we construct a polynomial — or an infinite series of polynomial terms — that reproduces f(x) exactly within some interval? And if so, how do we systematically determine each coefficient of that polynomial? The answer, as we shall see, lies entirely in the derivatives of f evaluated at a single base point.
Core Principles & Definitions
The construction of a Taylor series rests on a beautifully simple idea: if a function f(x) can be written as a power series, then the coefficients of that series are uniquely determined by the derivatives of f at the center of expansion. This principle connects the local behavior of a function — captured by its derivatives at a point — to its global behavior over an interval. Before diving into computation, it is essential to internalize the foundational concepts that make this machinery work.
Taylor Series at x = a
Maclaurin Series (a = 0)
Radius of Convergence
Taylor Polynomial of Degree n
Visual Explanation — Successive Taylor Approximations
The power of a Taylor series becomes viscerally clear when you plot successive Taylor polynomials alongside the original function. Each additional term in the expansion draws the polynomial approximation closer to the true curve, hugging it over a wider interval. The diagram below illustrates this for f(x) = eˣ centered at a = 0, showing how the Maclaurin polynomials T₁, T₃, and T₅ progressively improve the approximation.
Observe the behavior near the center point a = 0: all three polynomials pass through the point (0, 1) and share the same tangent line there, because every Taylor polynomial agrees with f through as many derivatives as it includes. The linear approximation T₁ = 1 + x captures only the slope, diverging quickly. The cubic T₃ adds curvature information from f″(0) and f‴(0), staying accurate on roughly (−1.5, 1.5). The quintic T₅ extends fidelity further still. In the limit, the full Maclaurin series ∑ xⁿ/n! converges to eˣ for all real x, giving it an infinite radius of convergence.
Mathematical Framework
We now formalize the construction. Suppose f is infinitely differentiable at x = a. Our goal is to determine the coefficients c₀, c₁, c₂, … in a power series representation f(x) = ∑ cₙ(x − a)ⁿ. The key insight is that substituting x = a into the series and its successive derivatives isolates each coefficient one at a time.
Deriving the Coefficients
To see why cₙ = f⁽ⁿ⁾(a)/n!, assume f(x) = c₀ + c₁(x − a) + c₂(x − a)² + c₃(x − a)³ + …. Setting x = a kills every term with (x − a) and leaves f(a) = c₀. Now differentiate once: f′(x) = c₁ + 2c₂(x − a) + 3c₃(x − a)² + …, and again set x = a to get f′(a) = c₁. Differentiating a second time gives f″(x) = 2c₂ + 6c₃(x − a) + …, so f″(a) = 2c₂, which means c₂ = f″(a)/2!. In general, the nth derivative produces n! · cₙ as the surviving constant, so cₙ = f⁽ⁿ⁾(a)/n!.
Essential Maclaurin Series & Their Patterns
Several functions appear so frequently that their Maclaurin series should be committed to memory. Each exhibits a distinctive pattern in its coefficients — alternating signs, missing terms, or factorial growth — that reflects the derivative structure of the underlying function. The table below collects the most important standard series, along with their radii of convergence.
| Function | Maclaurin Series | Radius R |
|---|---|---|
| eˣ | ∑ xⁿ/n! = 1 + x + x²/2! + x³/3! + … | R = ∞ |
| sin x | ∑ (−1)ⁿ x²ⁿ⁺¹/(2n+1)! = x − x³/3! + x⁵/5! − … | R = ∞ |
| cos x | ∑ (−1)ⁿ x²ⁿ/(2n)! = 1 − x²/2! + x⁴/4! − … | R = ∞ |
| 1/(1 − x) | ∑ xⁿ = 1 + x + x² + x³ + … | R = 1 |
| ln(1 + x) | ∑ (−1)ⁿ⁺¹ xⁿ/n = x − x²/2 + x³/3 − … | R = 1 |
| arctan x | ∑ (−1)ⁿ x²ⁿ⁺¹/(2n+1) = x − x³/3 + x⁵/5 − … | R = 1 |
A powerful technique for generating new series is algebraic manipulation of known series. For instance, replacing x with −x² in the geometric series 1/(1−x) = ∑ xⁿ immediately yields 1/(1+x²) = ∑ (−1)ⁿ x²ⁿ. Term-by-term integration of this result produces the Maclaurin series for arctan x. Similarly, multiplying, composing, or differentiating known series often provides the fastest route to a new expansion, bypassing tedious derivative calculations entirely.
Worked Example — Maclaurin Series for f(x) = cos x
Let us build the Maclaurin series for f(x) = cos x from scratch, illustrating each step of the systematic procedure introduced above. This example demonstrates how the cyclical nature of cosine's derivatives produces the characteristic alternating-sign, even-power pattern.
Strategies for Finding Taylor Series — Strengths & Limitations
There is rarely a single best way to find the Taylor series of a function. Depending on the function's complexity, one strategy may be far more efficient than another. The direct derivative-and-substitute method is universally applicable but can become computationally brutal for functions whose derivatives grow in complexity. Algebraic manipulation of known series — substitution, multiplication, differentiation, or integration — is often much faster but requires recognizing the connection to a standard series. The table below compares the principal methods.
| Method | When to Use | Limitations |
|---|---|---|
| Direct computation | When no known series applies; when only a few terms are needed. Always works if f is smooth. | Derivatives can become extremely complicated (e.g., f(x) = tan x). Identifying the general pattern for cₙ may be difficult. |
| Substitution into a known series | Replace x with g(x) in a standard series, e.g., e^(−x²) from eˣ by substituting u = −x². | Only works when the function can be expressed as a composition with a function whose series is known. |
| Term-by-term differentiation | If ∫f(x)dx or a derivative of f has a known series, differentiate term by term. Useful for 1/(1−x)² from the geometric series. | Radius of convergence is preserved, but behavior at endpoints must be re-checked. |
| Term-by-term integration | If f′(x) has a known series, integrate to get f. Key for ln(1+x), arctan x. | Must determine the constant of integration (usually by evaluating at x = 0). Endpoint convergence may change. |
| Multiplication / division of series | Multiply known series (e.g., eˣ sin x) or perform long division (e.g., tan x = sin x / cos x). | Produces terms one at a time; finding the general coefficient formula can be impractical. Best for a finite number of terms. |
Connection to Advanced Theory — Taylor Series at Arbitrary Centers & Complex Analysis
While Maclaurin series (centered at a = 0) are the most frequently encountered in a first course, many applications demand expansion about an arbitrary point. Approximating f(x) = √x near x = 4, for instance, requires a Taylor series centered at a = 4 because √x is not even defined at x = 0 (in the sense of having a convergent Maclaurin series). More broadly, Taylor series serve as the gateway to analytic function theory, where complex analysis reveals the deep reason behind radii of convergence: R equals the distance from the center a to the nearest singularity in the complex plane.
| Concept | Calculus 2 Perspective | Advanced Perspective |
|---|---|---|
| Radius of convergence | Found via ratio or root test; a mechanical computation. | Equals the distance to the nearest complex singularity. Example: 1/(1+x²) has R = 1 because of singularities at x = ±i. |
| Taylor polynomial error | Bounded by the Lagrange remainder; useful for estimating approximation accuracy. | Cauchy's integral formula provides exact remainder as a contour integral, sharpening error bounds. |
| Non-polynomial center | Taylor series about a ≠ 0 with (x−a)ⁿ terms; computed via the same derivative formula. | Laurent series generalize Taylor series to allow negative powers, enabling expansion near singularities. |
| Applications | Approximation, limit evaluation, definite integrals of non-elementary functions. | ODE solutions (Frobenius method), asymptotic analysis, quantum field theory perturbation expansions. |
In your next courses — complex analysis, differential equations, and numerical methods — you will encounter Taylor series as a ubiquitous tool. The Frobenius method solves differential equations by assuming a Taylor-like series solution and determining coefficients recursively. Numerical algorithms such as Runge-Kutta methods are derived by matching Taylor expansion terms to achieve desired accuracy orders. The foundation you build here — systematically computing derivatives, recognizing patterns, and testing convergence — will serve as the computational backbone of those more advanced techniques.
Practice Problems
Lesson Summary
A Taylor series represents a smooth function f(x) as the infinite power series ∑ f⁽ⁿ⁾(a)/n! · (x − a)ⁿ, where the coefficients are determined by the derivatives at the center a. When a = 0, the series is called a Maclaurin series. The standard procedure involves computing successive derivatives at the center, dividing by the corresponding factorial to form each coefficient, identifying the general pattern, expressing the result in sigma notation, and verifying the radius of convergence via the ratio or root test.
In practice, the most efficient approach is often to manipulate known series — substituting, differentiating, integrating, or multiplying standard expansions for eˣ, sin x, cos x, 1/(1−x), and ln(1+x) — rather than computing derivatives from scratch. The Lagrange remainder Rₙ(x) = f⁽ⁿ⁺¹⁾(c)/(n+1)! · (x−a)ⁿ⁺¹ provides rigorous error bounds, and showing Rₙ → 0 confirms that the series truly equals the function, not merely that it converges to something. These techniques underpin applications from numerical integration of non-elementary functions to differential equation solutions and perturbation methods in physics.