1 Definition and construction
1.1 Taylor polynomial of order n at a point
For a function \(f\) that is sufficiently differentiable at a point \(a\), the Taylor polynomial of order \(n\) is the degree-\(n\) polynomial that approximates \(f(x)\) near \(a\). It is built so that the polynomial reproduces the function’s local behavior as encoded by derivatives at \(a\). The approximation improves as \(n\) increases, subject to appropriate smoothness and convergence conditions.
1.2 Coefficients from derivatives
The coefficients of the Taylor polynomial come directly from the derivatives of \(f\) evaluated at the expansion point \(a\). In the standard formula, the \(k\)-th derivative determines the coefficient of \((x-a)^k\).
1.2.1 The n-th derivative term
The contribution of the \(n\)-th derivative is \[ \frac{f^{(n)}(a)}{n!}(x-a)^n. \] This term reflects how rapidly the function changes at order \(n\) around \(a\), scaled by the factorial factor \(n!\) that normalizes derivative growth in the series representation.
1.2.2 Notation and indexing conventions
A common convention denotes the Taylor polynomial of order \(n\) about \(a\) by \[ T_n(x)=\sum_{k=0}^{n}\frac{f^{(k)}(a)}{k!}(x-a)^k. \] Here \(k\) indexes the derivative order, and \(f^{(0)}(a)\) is interpreted as \(f(a)\). Alternative notational choices may shift indices or denote the polynomial by \(P_n\) or \(T_n^{(a)}\), but the structural rule—derivatives at \(a\) weighted by \(k!\) and powers of \((x-a)\)—remains the same.
1.3 Polynomial matching properties
The defining feature of Taylor polynomials is their ability to match the function at and near the expansion point through derivative information.
1.3.1 Matching function value
At the expansion point itself, the polynomial satisfies \[ T_n(a)=f(a), \] because every term with \((x-a)^k\) for \(k\ge 1\) vanishes when \(x=a\). Thus the approximation is exact at the center.
1.3.2 Matching derivatives up to order n
More strongly, the polynomial matches derivatives up to order \(n\): \[ T_n^{(k)}(a)=f^{(k)}(a)\quad \text{for } 0\le k\le n. \] This property is a direct consequence of the coefficient construction. It explains why Taylor polynomials are effective in local approximation: they align not just values but also the local slope, curvature, and higher-order shape captured by derivatives.
2 Remainder term and error estimates
2.1 Lagrange form of the remainder
A standard way to quantify the approximation error is to express the difference between the function and its Taylor polynomial as a “remainder” term. In Lagrange’s form, for \(x\) near \(a\), \[ f(x)=T_n(x)+R_n(x), \] with \[ R_n(x)=\frac{f^{(n+1)}(\xi)}{(n+1)!}(x-a)^{n+1} \] for some point \(\xi\) between \(a\) and \(x\) (assuming the relevant derivative exists). This formula turns error control into bounding a derivative on an interval.
2.1.1 Interpreting the mean-value theorem input
The existence of a point \(\xi\) is tied to the mean value theorem applied to a carefully constructed auxiliary function that subtracts the Taylor polynomial from \(f\). The result is that the remainder behaves like a controlled multiple of \((x-a)^{n+1}\), where the multiplier depends on the \((n+1)\)-st derivative somewhere between \(a\) and \(x\).
2.2 Cauchy form of the remainder
Another common representation is Cauchy’s form: \[ R_n(x)=\frac{f^{(n+1)}(a+\theta(x-a))}{n!}\,(1-\theta)^{n}\,(x-a)^{n+1} \] for some \(\theta\in(0,1)\) (with equivalent variants depending on the chosen parametrization). Like Lagrange’s form, it expresses the remainder via derivatives evaluated at an interior point, but the scaling factors are arranged differently, which can be advantageous for certain bounds.
2.3 Integral form of the remainder
The remainder can also be written using an integral expression. One typical form is \[ R_n(x)=\frac{1}{n!}\int_{a}^{x} f^{(n+1)}(t)\,(x-t)^n\,dt, \] again presuming sufficient differentiability. This approach provides a direct way to estimate error using integral inequalities and bounds on \(f^{(n+1)}\).
2.3.1 Error bounds from integral estimates
If \(f^{(n+1)}\) is bounded on the interval between \(a\) and \(x\), say \(\lvert f^{(n+1)}(t)\rvert\le M\), then \[
| R_n(x) | \le \frac{M}{n!}\int_{a}^{x} | x-t | ^n\,dt |
|---|---|---|---|
| = \frac{M}{(n+1)!}\, | x-a | ^{n+1}. |
\] Thus the integral representation yields error bounds closely resembling the ones obtained from Lagrange’s form.
2.4 Big-O notation for remainder
In many contexts, it is convenient to summarize error behavior using asymptotic notation. If \(f\) is \((n+1)\)-times differentiable near \(a\) and the \((n+1)\)-st derivative does not grow too fast, then \[ R_n(x)=O\big((x-a)^{n+1}\big)\quad \text{as } x\to a. \] This captures the leading-order dependence on distance from the expansion point without specifying the exact constant.
2.4.1 Relating remainder order to smoothness
The order \(n+1\) in the remainder reflects the degree to which \(f\) can be expanded using derivatives up to order \(n+1\). If higher derivatives fail to exist or become unbounded, the remainder estimate can break down or change form. Consequently, smoothness assumptions are not merely technical—they directly influence the accuracy guarantees.
3 Existence, smoothness, and differentiability assumptions
3.1 Conditions for Taylor polynomial validity
To construct the Taylor polynomial of order \(n\) at \(a\), \(f\) must have derivatives up to order \(n\) at \(a\). For error representations like those in the previous section, it typically needs an additional derivative, often \(f^{(n+1)}\) on an interval between \(a\) and \(x\), plus conditions ensuring the derivative behaves well enough to apply the relevant theorem.
3.2 Analytic vs. merely differentiable functions
A function can be infinitely differentiable at a point yet still fail to equal its Taylor series expansion there. Analytic functions are those for which the Taylor series converges to the function in a neighborhood of the expansion point. By contrast, infinitely differentiable (but non-analytic) functions may have Taylor coefficients that do not capture the function’s true local behavior beyond a formal series.
3.3 Examples illustrating failure modes
Failure modes often occur when differentiability is insufficient, when derivatives exist but do not control the remainder properly, or when Taylor series coefficients do not lead to a convergent representation of the function.
3.3.1 Functions with limited differentiability
If \(f\) is only \(n\)-times differentiable at \(a\), then the Taylor polynomial \(T_n\) remains meaningful as a finite polynomial approximation. However, without enough extra smoothness, one may lack robust remainder estimates and cannot guarantee that increasing \(n\) improves approximation in a controlled manner.
4 Convergence and Taylor series
4.1 From Taylor polynomials to Taylor series
The Taylor series associated with \(f\) at \(a\) is obtained by letting \(n\to\infty\) in the Taylor polynomials: \[ \sum_{k=0}^{\infty}\frac{f^{(k)}(a)}{k!}(x-a)^k. \] The Taylor series may converge to \(f(x)\), converge to some other function, or diverge, depending on properties of \(f\).
4.2 Pointwise convergence criteria
Convergence at a particular \(x\) depends on the behavior of the coefficients \(f^{(k)}(a)/k!\) and the factor \((x-a)^k\). In practice, convergence can often be assessed using ratio tests, root tests, or known expansions. Even when the series converges, matching \(f(x)\) is a further issue tied to analyticity and uniqueness of power series.
4.3 Radius of convergence and analytic continuation conceptual
| For many standard functions expressed by power series, the Taylor series has a radius of convergence \(R\) such that it converges for \( | x-a | <R\) and diverges for \( | x-a | >R\). Within that disk, convergence is governed by the power series theory. Beyond it, one may still recover function values through analytic continuation—conceptually extending the domain via alternative representations—though the details depend on the function class. |
|---|
4.4 Uniform convergence on intervals
Convergence may be uniform on compact subintervals contained within the radius of convergence. When such uniformity holds, it becomes possible to exchange limits with other operations under appropriate hypotheses. Uniform convergence also supports reliable approximation: the error bound does not deteriorate unexpectedly as the point varies within the interval.
4.4.1 Implications for approximation quality
If partial sums converge uniformly and the remainder is controlled, then truncated Taylor polynomials provide dependable approximations across an interval, not just at a single nearby point. This is important in applications where one needs stable accuracy across multiple evaluation points.
5 Core examples and standard expansions
5.1 Polynomial approximations via known derivatives
For many frequently used functions, derivatives at a point are known explicitly, enabling standard Taylor expansions. These expansions often produce familiar series with interpretable coefficients.
5.1.1 Exponential function
At a point \(a\), the exponential function satisfies \(f^{(k)}(a)=e^a\). This yields the expansion \[ e^{x}=e^{a}\sum_{k=0}^{\infty}\frac{(x-a)^k}{k!}, \] and in particular around \(0\), \[ e^{x}=\sum_{k=0}^{\infty}\frac{x^k}{k!}. \]
| The resulting Taylor polynomials approximate \(e^x\) with an error that decreases rapidly as \(n\) grows for modest \( | x-a | \). |
|---|
5.1.2 Trigonometric functions
Sine and cosine have derivatives that cycle among \(\pm\sin\) and \(\pm\cos\). About \(0\), \[ \sin x=\sum_{k=0}^{\infty}(-1)^k\frac{x^{2k+1}}{(2k+1)!},\quad \cos x=\sum_{k=0}^{\infty}(-1)^k\frac{x^{2k}}{(2k)!}. \] These series lead to polynomial approximations that preserve parity (odd/even structure) and can be truncated to compute approximate values.
5.1.3 Logarithm and reciprocal functions (where defined)
For functions with domain restrictions, Taylor expansions are taken only where the function is defined and derivatives exist. For instance, around \(0\), \[
| \frac{1}{1-x}=\sum_{k=0}^{\infty}x^k \quad ( | x | <1), |
|---|
\] and the logarithm has expansions such as \[
| \ln(1+x)=\sum_{k=1}^{\infty}(-1)^{k+1}\frac{x^k}{k}\quad ( | x | <1), |
|---|
\] which are used in approximations near the expansion point.
5.2 Choosing the expansion point
The quality of a Taylor approximation depends on how far the evaluation point lies from the center \(a\), and on the convergence properties of the corresponding series.
5.2.1 Center at zero (Maclaurin polynomials)
When \(a=0\), the Taylor polynomial is often called a Maclaurin polynomial. This choice aligns naturally with many standard series and simplifies formulas by replacing \((x-a)\) with \(x\).
5.2.2 Center at a nonzero point
Expanding about \(a\ne 0\) can improve approximation when \(x\) is closer to \(a\) than to \(0\). It also can better respect domain features; for example, a logarithm or reciprocal function may be expanded around a point where the expression remains analytic and the series converges.
6 Approximation applications
6.1 Local approximation and error control
Taylor polynomials provide a systematic method to approximate a function near a chosen point while quantifying the discrepancy through remainder terms. In analysis and computation, this supports both qualitative reasoning (how the function behaves locally) and quantitative estimation (how accurate a truncated polynomial will be).
6.2 Estimating function values
A practical use is approximating \(f(x)\) by truncating the Taylor series after finitely many terms.
6.2.1 Numerical evaluation with a truncated polynomial
One computes derivatives at the expansion point (analytically or numerically), forms the polynomial \(T_n(x)\), and uses the remainder estimate to judge truncation error. When \(x\) is close to \(a\), the error typically shrinks quickly with \(n\).
6.3 Approximating derivatives and integrals
Since Taylor polynomials match derivatives at the expansion point, they can also serve as local surrogates for derivative and integral computations.
6.3.1 Differentiating a Taylor polynomial
Differentiating \(T_n(x)\) produces another polynomial that approximates \(f'(x)\) near \(a\). More generally, the derivative of order \(m\le n\) can be approximated by differentiating the Taylor polynomial of order \(n\), yielding a controlled local approximation.
6.3.2 Integrating a Taylor polynomial
Integrating \(T_n(x)\) gives a polynomial approximation to an antiderivative or to definite integrals over an interval near \(a\). The integral remainder can be estimated using the corresponding integral form of the Taylor remainder or related inequalities.
6.4 Solving equations approximately
Taylor methods also help solve nonlinear equations by constructing local approximations.
6.4.1 Root-finding near a known point
When one seeks a solution to \(g(x)=0\) near a point where an initial guess \(x_0\) is available, expanding \(g\) and truncating provides an approximate polynomial equation. Solving that polynomial yields a refined estimate of the root, often in harmony with methods related to Newton’s iteration and its variants (though the specific algorithm depends on how the approximation is used).
7 Theoretical connections in analysis
7.1 Taylor’s theorem and its variants
Taylor’s theorem formalizes the idea that a function can be approximated locally by its polynomial expansion plus a remainder term. Variants differ in the remainder representation (Lagrange, Cauchy, integral), but the overarching principle—local polynomial approximation with quantified error—remains the same.
7.1.1 Mean-value theorem viewpoint
Many proofs of Taylor’s theorem use repeated applications of the mean-value theorem to an auxiliary function. This viewpoint highlights the derivative-driven nature of the remainder: the error depends on higher derivatives evaluated somewhere between the center and the evaluation point.
7.2 Remainder bounds and smoothness classes
The size and form of remainder estimates connect to how smooth the function is and how large its derivatives can be. As a result, one can relate approximation quality to membership in smoothness or regularity classes, where derivatives satisfy bounds or integrability conditions.
7.3 Links to differential equations (formal motivation)
In differential equations, solutions are often studied locally. Taylor expansions provide formal power-series descriptions that can be substituted into differential equations, producing recursive relations for coefficients. Even when convergence is not immediate, such expansions guide existence proofs, perturbation analyses, and qualitative understanding.
7.4 Role in proving limit and continuity results
Taylor polynomials underpin many arguments in analysis. By controlling remainders, one can show that certain functions have limits, are continuous, or possess derivatives of a given order. In this way, Taylor’s theorem serves as a bridge between differentiability assumptions and local behavior.
8 Computation and practical considerations
8.1 Determining the order n needed
| Choosing \(n\) depends on the desired accuracy, the distance \( | x-a | \), and bounds on higher derivatives. Remainder estimates convert these ingredients into a stopping criterion, guiding how many terms to include. |
|---|
8.2 Managing numerical stability
Even when the mathematical remainder is small, numerical computation can be unstable if coefficients are large or if terms nearly cancel. Stable evaluation strategies, such as using nested multiplication (e.g., Horner’s method), can reduce rounding error and improve reliability.
8.3 Symbolic vs. numeric truncation
Truncation may occur in different ways. Symbolically, one keeps exact derivatives and forms a polynomial. Numerically, derivatives might be approximated from data or computed with finite precision, and truncation then interacts with approximation error from those derivative computations.
8.3.1 Error propagation in computations
Errors in derivative values or in floating-point arithmetic propagate through the polynomial coefficients and powers of \((x-a)\). As a result, it is important to treat truncation error and computational error jointly rather than assuming only the remainder term contributes.
8.4 Common pitfalls (domain restrictions, remainder misestimation)
A frequent issue is applying a Taylor formula outside its domain of validity, such as using an expansion that assumes the function is defined and differentiable on an interval. Another pitfall is underestimating the remainder by using insufficient derivative bounds, leading to overconfidence in accuracy. Careful attention to smoothness assumptions, convergence range, and derivative estimates is essential.