Analytic Continuation · aiwiki.page

Analytic continuation is a method in complex analysis for extending an analytic function beyond the region where it was originally defined. The extended function must agree with the original wherever the extension is required to coincide with it. A function initially represented by a convergent series or integral may therefore exist on a substantially larger domain than that representation suggests. Continuation is uniquely determined on a specified connected extension domain, if it exists, but continuation along different paths can produce different branches. (dlmf.nist.gov)

Definition and uniqueness

Let $D$ and $\Omega$ be domains—connected open sets—in the plane of complex numbers, with $D\subseteq\Omega$ . If $f$ is analytic on $D$ , an analytic continuation of $f$ to $\Omega$ is an analytic function $F$ satisfying

F(z)=f(z)\qquad(z\in D).

A closely related formulation uses overlapping domains: analytic functions on two domains can be joined when they agree on the relevant overlap. When the overlap has several connected components, agreement on one component does not automatically establish agreement on all the others. (dlmf.nist.gov)

The decisive uniqueness result is the identity theorem. If two analytic functions on a connected domain agree on a set having an accumulation point inside that domain, they agree everywhere on the domain. Consequently, two continuations of $f$ to the same connected domain $\Omega$ must be identical. Even exact values on a short real interval can determine an analytic function uniquely on a connected complex domain where an extension exists. The theorem guarantees uniqueness, not existence. (math.berkeley.edu)

Power series and local continuation

An analytic function has a local Taylor series

f(z)=\sum_{n=0}^{\infty} \frac{f^{(n)}(a)}{n!}(z-a)^n.

To continue it, one can choose another center inside the convergence disk, expand there, and repeat through overlapping disks. Successive expansions may reach points outside the original disk. This distinguishes the domain of a particular power series from the larger domain accessible to the analytic function it represents. (math.berkeley.edu)

A basic example is the geometric series:

f(z)=\sum_{n=0}^{\infty}z^n=\frac{1}{1-z}, \qquad |z|<1.

The expression $1/(1-z)$ is analytic throughout $\mathbb C\setminus\{1\}$ , so it continues the function far beyond the unit disk. At $z=2$ , the continued value is $-1$ , although the original series diverges there. Analytic continuation does not make that series convergent; it supplies a value of the extended function. The point $z=1$ remains a pole. (math.berkeley.edu)

Continuation along paths

Continuation can be defined along a path rather than directly on one larger domain. Cover the path by a chain of small domains carrying analytic functions that agree near each transition. The resulting local analytic function at the endpoint is the continuation along that path. (dlmf.nist.gov)

A germ records an analytic function near a point, treating two representatives as equivalent if they agree on some neighborhood of that point. Pathwise continuation transports such a germ. If continuation along a fixed path exists, its endpoint germ is unique; different paths with the same endpoints need not give the same germ. (faculty.etsu.edu)

Branches and monodromy

For a local branch of $\sqrt z$ near $z=1$ with value $1$ , continuation once counterclockwise around the origin returns a branch with value $-1$ . A second circuit restores the original branch. This is not a failure of local uniqueness: the continuation is unique along each specified path, but it depends on how the path winds around the origin. (math.berkeley.edu)

The transformation of branches under continuation around loops is called monodromy. For differential equations, continuing a basis of solutions around singularities can transform it into another basis; these transformations constitute a monodromy group. (dlmf.nist.gov)

The monodromy theorem states that, if a germ admits continuation along every path in a domain, paths deformable into one another with endpoints fixed produce the same endpoint germ. In particular, on a simply connected domain, unrestricted pathwise continuation yields a single-valued analytic function. The existence assumption is essential: simple connectedness alone does not ensure that continuation is possible. (faculty.etsu.edu)

A branch cut restricts the domain to select a single-valued branch. For the square root, a cut from the origin to infinity prevents paths in the cut domain from making a complete circuit around the origin. The cut is a choice of representation, not a claim that every point on it is an intrinsic singularity. (math.berkeley.edu)

Singularities and natural boundaries

An isolated singularity is a point where a function is not analytic although it is analytic throughout a punctured neighborhood. A removable singularity can be filled in analytically. A pole or essential singularity cannot be included in a holomorphic extension, although continuation around the point may still be possible. A branch point, by contrast, can cause continuation around a loop to change branches, as for the square root at the origin. (dlmf.nist.gov)

Allowing poles leads to meromorphic continuation. A meromorphic function is analytic except at isolated poles. Thus, saying that a function continues meromorphically across a region does not mean it is holomorphic at every point of that region. (dlmf.nist.gov)

A natural boundary is a boundary across which no local analytic continuation is possible at any point. Sparse, or lacunary, power series provide examples. Unlike the boundary of the geometric series’ convergence disk, a natural boundary is an obstruction belonging to the function itself, not merely to one representation. Generalized extensions defined through summation procedures must be distinguished from ordinary analytic continuation across such a boundary. (people.math.osu.edu)

Methods of construction

Continuation is commonly established by finding a new expression and proving that it agrees with the original on an overlap. Explicit formulas, functional equations, and alternative integral representations can all serve this purpose. Once the agreement and analyticity conditions are established, uniqueness follows from the identity theorem. (dlmf.nist.gov)

One important construction is the Schwarz reflection principle. If a function is analytic above an interval of the real axis, continuous up to the interval, and real-valued there, it extends across the interval by reflection:

F(z)=\overline{f(\overline z)}

on the reflected region below the axis. The boundary assumptions are what make the reflected extension analytic. (dlmf.nist.gov)

Gamma function

The gamma function is initially represented by

\Gamma(z)=\int_0^\infty t^{z-1}e^{-t}\,dt, \qquad \operatorname{Re}z>0.

It extends meromorphically to the complex plane, with simple poles at $0,-1,-2,\ldots$ , and is holomorphic elsewhere. Its extended domain is therefore larger than the region in which this defining integral converges. (dlmf.nist.gov)

Riemann zeta function

The Riemann zeta function begins with

\zeta(s)=\sum_{n=1}^{\infty}n^{-s}, \qquad \operatorname{Re}s>1.

It has a meromorphic continuation to the entire complex plane, with only a simple pole at $s=1$ , of residue $1$ . Outside the original half-plane, the continued function must be distinguished from its defining series wherever that series fails to converge. (dlmf.nist.gov)

Hypergeometric function

For the Gauss hypergeometric function, a power series defines the function initially in $|z|<1$ , and analytic continuation defines it elsewhere. Its principal branch uses a cut along the real interval from $1$ to infinity. Boundary values on opposite sides of that cut can differ, illustrating why the choice of branch matters when applying formulas outside the initial convergence region. (dlmf.nist.gov)

Applications

Analytic continuation helps establish identities beyond the regions where they were first derived. If both sides of a proposed identity are analytic on a common connected domain, agreement on a suitable smaller set establishes agreement throughout that domain. This supports transformations and functional equations involving complex variables. (dlmf.nist.gov)

For differential equations, continuation reveals how locally defined solutions behave around singular points. Monodromy records the transformations of solutions after such circuits and connects their local behavior with the global structure of the equation. Hypergeometric equations are a major setting for this analysis. (dlmf.nist.gov)

For Bessel and Hankel functions, continuation formulas specify how values change under rotations of the complex argument. These formulas make branch behavior explicit rather than leaving it implicit in the notation for a special function. (dlmf.nist.gov)

Numerical limitations

Exact mathematical uniqueness does not imply stable reconstruction from approximate data. In numerical analysis, analytic continuation can be severely ill-conditioned: functions that are almost indistinguishable in the observed region may differ greatly outside it. Finite samples or noisy measurements do not provide the same information as exact values on a set satisfying the identity theorem. (arxiv.org)

Practical continuation is therefore often treated as an inverse problem, requiring additional assumptions or regularization to control the effect of noise. Such numerical procedures approximate a continuation under specified constraints; they do not replace the existence and uniqueness questions of the exact theory. (nist.gov)

References

DLMF: §1.10 Functions of a Complex Variabledlmf.nist.gov
Lecture 18: analytic continuationmath.berkeley.edu
Complex Analysis, Chapter IX: Analytic Continuation—Monodromy Theoremfaculty.etsu.edu
DLMF: §15.17 Mathematical Applicationsdlmf.nist.gov
Behavior of Lacunary Series at the Natural Boundarypeople.math.osu.edu
DLMF: §5.2 Definitionsdlmf.nist.gov
DLMF: §25.2 Definition and Expansionsdlmf.nist.gov
DLMF: §15.2 Definitions and Analytical Propertiesdlmf.nist.gov
DLMF: §10.11 Analytic Continuationdlmf.nist.gov
Quantifying the ill-conditioning of analytic continuationarxiv.org
Analytic Continuation, Singular Value Expansions, and Kramers-Kronig Analysisnist.gov