1 Basic concept of Taylor expansion
A Taylor expansion expresses a function near a chosen point as a polynomial in the displacement from that point. The coefficients are determined by the function’s derivatives at the center, so the approximation captures local behavior up to a prescribed order.
1.1 Centered polynomial approximation
The core idea is to approximate a function near a fixed center point \(a\) by a polynomial in \((x-a)\).
1.1.1 Derivatives as coefficients
For a sufficiently smooth function \(f\), the Taylor polynomial of degree \(n\) at \(a\) has the form \[ T_n(x)=\sum_{k=0}^{n}\frac{f^{(k)}(a)}{k!}(x-a)^k. \] Each derivative \(f^{(k)}(a)\) contributes the coefficient of \((x-a)^k\), linking local rate-of-change information to the polynomial terms.
1.1.2 Displacement variable and power series form
Writing the expansion in terms of the displacement \(h=x-a\) yields \[ f(a+h)\approx \sum_{k=0}^{n}\frac{f^{(k)}(a)}{k!}h^k, \] and, under stronger conditions, the approximation extends to an infinite series. This power-series viewpoint emphasizes that the Taylor series is simply a special power series centered at \(a\).
1.2 Taylor polynomial vs. Taylor series
A Taylor polynomial is finite and used for approximation; a Taylor series is infinite and attempts to represent the function on a neighborhood or more.
1.2.1 Notation and common conventions
The degree-\(n\) truncation is usually denoted \(T_n(x)\), while the full infinite sum is called the Taylor series. In many texts, the Taylor series at \(a\) is written \[ \sum_{k=0}^{\infty}\frac{f^{(k)}(a)}{k!}(x-a)^k, \] with the understanding that it may or may not converge to \(f(x)\).
1.2.2 Order of approximation and truncation
Truncating after degree \(n\) controls the leading behavior of the error: if the function is smooth, the first omitted term is typically of order \((x-a)^{n+1}\). Higher-order truncations improve accuracy when the remainder becomes small in the region of interest.
2 Taylor’s theorem and remainder
Taylor’s theorem formalizes the relationship between the function, its Taylor polynomial, and a remainder term that quantifies the difference.
2.1 Statement of Taylor’s theorem
A common formulation states that if \(f\) has derivatives up to order \(n+1\) near \(a\), then for \(x\) close to \(a\), \[ f(x)=T_n(x)+R_n(x), \] where \(R_n(x)\) depends on derivatives of \(f\) and the distance \(x-a\).
2.1.1 Lagrange form of the remainder
In the Lagrange (mean-value) form, there exists a point \(\xi\) between \(a\) and \(x\) such that \[ R_n(x)=\frac{f^{(n+1)}(\xi)}{(n+1)!}(x-a)^{n+1}. \] This expression makes error behavior explicit in terms of the \((n+1)\)-st derivative.
2.1.2 Cauchy form of the remainder
The Cauchy form introduces a parameterized version of the same idea: \[ R_n(x)=\frac{(x-a)^{n+1}}{n!}\int_{0}^{1}(1-t)^n f^{(n+1)}(a+t(x-a))\,dt. \] It is especially useful for deriving bounds because it integrates information about the derivative along the segment from \(a\) to \(x\).
2.1.3 Integral form of the remainder
A closely related integral remainder commonly appears as \[ R_n(x)=\int_{a}^{x}\frac{(x-t)^n}{n!}\,f^{(n+1)}(t)\,dt, \] showing the error as a weighted average of higher derivatives over the interval. This form is central in rigorous analysis and in estimating remainders.
2.2 Error bounds and asymptotics
Once a remainder formula is available, it can be converted into practical estimates using inequalities or limiting arguments.
2.2.1 Big-O and little-o notation
Error terms are often expressed using asymptotic notation relative to a small parameter such as \(h=x-a\). For example, \(R_n(x)=O(h^{n+1})\) indicates that the remainder is bounded by a constant multiple of \(h^{n+1}\) near \(h=0\). Similarly, \(R_n(x)=o(h^{n+1})\) means it becomes negligible compared with \(h^{n+1}\) as \(h\to 0\).
2.2.2 Estimating approximation error
| To estimate \(R_n(x)\), one typically bounds the relevant derivative on an interval containing \(a\) and \(x\). For instance, in the Lagrange remainder, if \( | f^{(n+1)}(t) | \le M\) throughout the interval, then |
|---|
\[
| R_n(x) | \le \frac{M}{(n+1)!} | x-a | ^{n+1}. |
|---|
\] This converts theoretical remainder expressions into computable error guarantees.
2.3 Conditions for existence and differentiability
The ability to write a Taylor polynomial requires finite differentiability, while writing an infinite Taylor series requires stronger assumptions.
2.3.1 Smoothness requirements
A degree-\(n\) Taylor polynomial requires \(f^{(k)}(a)\) for \(k\le n\), so \(n\) derivatives are sufficient for defining the coefficients. Taylor’s theorem with remainder at order \(n\) uses one additional derivative in a typical statement, ensuring that the remainder can be controlled.
2.3.2 Analyticity vs. merely differentiable functions
A function that is infinitely differentiable does not necessarily equal its Taylor series. Analyticity is the property that ensures the Taylor series, when formed, converges to the function in a neighborhood of the center. Without analyticity, derivatives may exist to all orders yet still fail to determine the function uniquely through the Taylor series.
3 Computing Taylor expansions
In practice, Taylor expansions are computed either by directly evaluating derivatives at the center or by manipulating known expansions.
3.1 Expansion about a general point
Expanding about \(a\neq 0\) can be reduced to expansions around zero.
3.1.1 Shifting the center
Given \(h=x-a\), one considers \(f(a+h)\) and expands in powers of \(h\). After computing the series in \(h\), substituting \(h=x-a\) yields the expansion in \(x\).
3.1.2 Change of variables techniques
For complicated functions, a change of variables may simplify the form. Typical strategies include rewriting expressions so that standard series (such as those for \(\exp\), \(\sin\), \(\cos\), or \(\log(1+u)\)) can be applied with \(u\) expressed as a function of \(x-a\).
3.2 Expansions from known series
A large fraction of Taylor computations rely on standard power series and algebraic operations.
3.2.1 Using standard power series
Known expansions near zero include, for example, \[ e^u=\sum_{k=0}^{\infty}\frac{u^k}{k!},\quad \sin u=\sum_{k=0}^{\infty}\frac{(-1)^k u^{2k+1}}{(2k+1)!},\quad \cos u=\sum_{k=0}^{\infty}\frac{(-1)^k u^{2k}}{(2k)!}. \] Substituting \(u\) with a shifted expression generates a Taylor series at the corresponding center.
3.2.2 Linear combinations, scaling, and composition
If two functions have Taylor expansions, their sum and scalar multiples can be expanded term-by-term. Scaling the argument corresponds to adjusting powers, while composition can be handled by substituting the inner series into the outer one and re-expanding, taking care with truncation order to control errors.
3.3 Symbolic differentiation approach
When the function is presented explicitly, computing derivatives can be the most direct method.
3.3.1 Derivative evaluation at the center
One computes \(f^{(k)}(a)\) for \(k=0,\dots,n\) and substitutes into \(\sum_{k=0}^{n}\frac{f^{(k)}(a)}{k!}(x-a)^k\). This approach is systematic but can become laborious for large \(n\).
3.3.2 Efficient bookkeeping for higher derivatives
Efficient computation often uses algebraic simplifications, recurrence relations, or symbolic identities such as product rules and chain rules organized to avoid redundant work. In algorithmic settings, expression trees and automatic differentiation techniques can generate derivatives more reliably than manual differentiation.
4 Special forms and related expansions
Taylor expansions appear in several standardized variants, including expansions at zero and extensions to multiple variables.
4.1 Maclaurin series
A Maclaurin series is a Taylor series centered at \(0\).
4.1.1 Expansion at zero
It is written as \[ f(x)=\sum_{k=0}^{\infty}\frac{f^{(k)}(0)}{k!}x^k \] when the series converges appropriately. Using the center at zero often streamlines both computation and presentation.
4.1.2 Common examples and templates
Many frequently used expansions are given in Maclaurin form, such as those for trigonometric functions and exponential-type functions. They serve as templates for generating Taylor series around other points via shifting.
4.2 Multivariable Taylor expansions
For functions of several variables, the polynomial approximation uses partial derivatives and multi-term sums.
4.2.1 Gradient and higher partial derivatives
Given \(f:\mathbb{R}^m\to\mathbb{R}\), a multivariable Taylor polynomial around \(a\) expands in powers of \((x-a)\) with coefficients involving partial derivatives. The first-order truncation corresponds to linearization using the gradient, while higher orders incorporate curvature information through second and higher partial derivatives.
4.2.2 Multi-index notation
Multi-index notation compactly represents combinations of partial derivatives: \[
| \partial^\alpha f(a) = \frac{\partial^{ | \alpha | }}{\partial x_1^{\alpha_1}\cdots \partial x_m^{\alpha_m}}f(a), |
|---|
\]
| where \(\alpha=(\alpha_1,\dots,\alpha_m)\) and \( | \alpha | =\alpha_1+\cdots+\alpha_m\). The Taylor polynomial becomes a sum over all multi-indices with \( | \alpha | \le n\). |
|---|
4.2.1.1 Remainder in the multivariable setting
Remainder terms in multiple variables can be expressed using integral forms or mean-value-type bounds. These formulations typically involve the \((n+1)\)-st order partial derivatives evaluated along a path or within a region connecting \(a\) and \(x\), enabling estimates in terms of suprema of derivatives.
4.3 Taylor expansions for vector-valued functions
If the target is a vector space, Taylor expansions apply componentwise or via tensor-valued derivatives.
4.3.1 Componentwise expansions
For \(F:\mathbb{R}\to\mathbb{R}^p\), one expands each component \(F_i\) separately: \[ F_i(x)=\sum_{k=0}^{n}\frac{F_i^{(k)}(a)}{k!}(x-a)^k+R_{n,i}(x). \] This yields a polynomial approximation for the vector function by assembling the component results.
4.3.2 Jacobian and higher-order tensors
For multivariable vector-valued functions \(F:\mathbb{R}^m\to\mathbb{R}^p\), the first derivative is the Jacobian matrix, and higher-order derivatives form higher-order tensors. The Taylor polynomial then combines these tensor coefficients with monomials in \((x-a)\).
5 Convergence and analytic continuation considerations
Taylor polynomials always make sense when derivatives exist to the required order, but Taylor series representation depends on convergence behavior and regularity.
5.1 Convergence of Taylor series
Convergence is a property of the infinite series and typically depends on the function and the center.
5.1.1 Radius of convergence in one variable
| In one variable, a Taylor series around \(a\) can often be viewed as a power series in \((x-a)\), with a radius \(R\) such that the series converges for \( | x-a | <R\) and diverges for \( | x-a | >R\). Whether it converges at \( | x-a | =R\) may vary by endpoint. |
|---|
5.1.2 Criteria for convergence
Standard tests for power series and analytic function theory are used to determine convergence regions. For instance, the ratio test and root test can be applied to the coefficients when they can be estimated. Alternatively, complex-analytic methods can supply strong conclusions about where a Taylor series represents a function.
5.2 When Taylor series fail to represent the function
Even when the Taylor series converges, it may not coincide with the original function.
5.2.1 Smooth but non-analytic examples (conceptual)
There exist smooth functions with all derivatives defined at a point but whose Taylor series converges to something different from the function itself, including cases where the Taylor series can be identically zero while the function is nonzero away from the center. Such examples demonstrate that infinite differentiability alone does not guarantee representation.
5.2.2 Practical implications for approximation
In numerical or modeling contexts, truncated Taylor polynomials can still perform well locally even if the Taylor series fails to represent the function globally. The key requirement is that the remainder becomes small in the region where the approximation is used, which can hold even if analytic continuation is absent.
5.3 Relation to power series and analytic functions
Taylor series are intimately connected to analytic functions through power-series representation.
5.3.1 Uniqueness and matching coefficients
If two analytic functions share the same Taylor series coefficients at a point within a connected neighborhood of convergence, they coincide there. This “coefficient matching” property underpins uniqueness results in analytic continuation.
5.3.2 Power series as Taylor series
Conversely, any power series centered at \(a\) can be interpreted as a Taylor series of some analytic function on its disk of convergence. This equivalence clarifies how Taylor expansions fit into the broader framework of power series.
6 Applications in analysis
Taylor expansions provide local approximations and asymptotic descriptions that support both qualitative reasoning and quantitative estimates.
6.1 Local approximation and linearization
The simplest nontrivial truncation gives a linear approximation that reflects first-order behavior.
6.1.1 First-order Taylor approximation
For differentiable \(f\), the first-order Taylor polynomial at \(a\) is \[ f(x)\approx f(a)+f'(a)(x-a). \] The approximation is reliable when \(x\) is close to \(a\) and the remainder term is small.
6.1.2 Newton’s method intuition (conceptual)
Newton’s method can be motivated by replacing a function near a point by its first-order approximation and solving for where the approximation crosses zero. This connects the derivative information captured by Taylor expansion to iterative root-finding.
6.2 Asymptotic expansions
Taylor-type expansions are also used as asymptotic tools, especially when exact convergence is not available.
6.2.1 Asymptotic vs. convergent series
An asymptotic series is intended to approximate a function in a limiting regime, such as for large arguments, without necessarily converging to the function for all values. Truncating at an optimal order can yield high accuracy before the omitted terms grow in magnitude.
6.2.2 Error interpretation for large variables
In asymptotic settings, the “small parameter” might be \(1/x\) rather than \(x-a\). The remainder is interpreted in terms of how the error scales as the parameter approaches its limit, often yielding practical guidance for choosing truncation orders.
6.3 Inequalities and estimates using expansions
Truncated Taylor polynomials can be used to bound functions, control monotonicity, and derive quantitative estimates.
6.3.1 Bounding functions via truncated series
If the remainder has a controlled sign or magnitude, one can compare \(f(x)\) to \(T_n(x)\) to obtain upper and lower bounds. Integral remainder forms are particularly helpful because they can lead to rigorous inequalities based on bounds for derivatives.
6.3.2 Monotonicity and remainder sign (where applicable)
In situations where the \((n+1)\)-st derivative retains a consistent sign on an interval, the remainder inherits that sign, allowing one to infer whether \(T_n(x)\) underestimates or overestimates \(f(x)\). Such reasoning is common in classical inequality proofs derived from Taylor expansions.