The radius of convergence of a power series is a number , with , that separates absolute convergence from divergence according to distance from the series’ center. For a series
the series converges absolutely whenever and diverges whenever . At points satisfying , convergence must be investigated separately. The concept is fundamental to mathematical analysis, especially the representation of functions by Taylor series and the study of analytic functions.
Definition and convergence region
Let the coefficients and center be complex numbers. The radius of convergence can be defined by
At , the expression is understood as .
If , the region is called the disk of convergence. For a power series in a real variable , its interior convergence region is the interval
Either, both, or neither of the two endpoints may also belong to the full interval of convergence.
The extreme cases have straightforward interpretations:
- : the series converges only at its center.
- : the series converges absolutely at every finite value of the variable.
Every polynomial, viewed as a power series with all sufficiently high coefficients equal to zero, has infinite radius of convergence.
The circular shape of the interior convergence region is a special property of power series. A general series of functions need not have a convergence region determined solely by distance from one point.
Why a radius exists
The essential fact is that convergence at one point forces absolute convergence at every point closer to the center.
Suppose the power series converges at , and let . Its terms must be bounded, so there is a constant such that
for every . If , then
The expression on the right forms a convergent geometric series. Comparison therefore establishes absolute convergence at .
Conversely, if the series diverges at a point, it cannot converge at any point farther from the center: such convergence would force absolute convergence at the original point. Together, these statements establish the existence of a unique radius separating the interior and exterior behaviors.
Calculating the radius
Cauchy–Hadamard formula
The general coefficient formula is the Cauchy–Hadamard formula:
The conventions are
The limit superior is essential because the sequence of roots need not have an ordinary limit. In particular, the formula remains valid when infinitely many coefficients vanish or when coefficient sizes oscillate.
To see the connection with the root test, write
A value less than gives absolute convergence; a value greater than gives divergence. Equality corresponds to the boundary, where the test is inconclusive.
Ratio formula
When the coefficients are eventually nonzero and the relevant limit exists, the ratio test gives
Equivalently, if
then , using the same conventions for zero and infinity.
The ratio formula is often simpler to calculate, but it is less general than the Cauchy–Hadamard formula. Failure of the ratio limit to exist does not mean that the power series lacks a radius of convergence.
Examples and boundary behavior
Geometric series
For
all coefficients equal , so . Inside the disk,
At every point with , the terms have absolute value and do not tend to zero. The series therefore diverges everywhere on its boundary.
A series with some boundary convergence
Consider
Since , its radius is . At the real endpoint , it becomes the divergent harmonic series. At , it becomes a convergent alternating series, but does not converge absolutely.
Thus its real interval of convergence is
More generally, on the complex unit circle it converges at every point except , and the convergence at those other boundary points is conditional.
A series converging on the entire boundary
For
the radius is again . At every boundary point,
so the series converges absolutely throughout the unit circle.
These examples show that the radius alone does not determine boundary convergence, even absolute convergence there.
Infinite and zero radii
The exponential function has the expansion
Here
so .
By contrast,
has , since the successive coefficient ratio used to calculate equals . These examples illustrate how rapid coefficient decay enlarges the convergence region, while rapid coefficient growth shrinks it.
Analytic functions and singularities
Inside its disk of convergence, a complex power series defines an analytic function. Conversely, a function holomorphic on a disk is represented throughout that disk by its Taylor series about the center.
This yields a geometric interpretation in complex analysis: the radius of a Taylor series is the radius of the largest open disk centered at the expansion point to which its local analytic function extends holomorphically. If the radius is finite, at least one point on the boundary obstructs such an extension.
In familiar cases, the radius is therefore described as the distance to the nearest complex singularity. For example,
The singularities at and are each one unit from the origin, so the radius is . The corresponding real function is defined and smooth for every real , yet its Taylor series about zero converges only for . Complex singularities can thus limit a real Taylor expansion even when no singularity appears on the real axis.
A finite radius does not mean that the represented function cannot be extended anywhere beyond the disk. In the geometric-series example, is holomorphic near every boundary point except . Analytic continuation can extend a function through accessible boundary regions, although the original series retains its original radius.
Nor does convergence at a boundary point imply analyticity there. A power series may have a finite boundary value at a point through which its sum has no holomorphic extension.
Differentiation, integration, and uniform convergence
If
has radius , it can be differentiated term by term inside the disk:
The differentiated series has the same radius . A termwise antiderivative,
also has radius . Repeated differentiation and integration preserve the radius, although they can change convergence at individual boundary points.
For every , the original series converges absolutely and uniformly on the closed disk
Indeed, each term is bounded there by , whose sum is finite. More generally, convergence is uniform on every compact subset of the open disk.
These properties justify local operations on power-series representations. They do not imply uniform convergence over the entire open disk or its boundary.
Uses and limitations
The radius of convergence determines where a power-series representation can be used directly. In ordinary differential equations, it helps describe the neighborhood in which a solution constructed by a series actually converges. In the study of generating functions, it connects coefficient growth with the analytic behavior of the generating function.
For numerical approximation, the radius identifies a valid region but is not by itself an error estimate. After truncation at degree , the error satisfies
A useful numerical bound requires additional information about the coefficients or the function.
Finally, a Taylor series need not represent the function from which its coefficients were calculated merely because all real derivatives exist. For example, the smooth real function
has every derivative equal to zero at the origin. Its Taylor series there is the zero series, with infinite radius of convergence, but it does not equal at any nonzero real point. The radius describes convergence of the series itself; equality with a specified function is a separate question.