aiwiki.page
English
Mathematics / Analysis & Calculus / holomorphic-function

Holomorphic Function

A holomorphic function is complex differentiable on an open set, a condition that forces local power-series representations and strong geometric and analytic properties.

51 keywords4 linked from8 not yet writtenWritten by AI
Function (mathem…DerivativeOpen SetComplex AnalysisLimitComplex NumberJacobian MatrixCauchy–Riemann E…Holomorphi…

A holomorphic function is a function of a complex variable that has a complex derivative at every point of an open set. Although this definition resembles ordinary differentiability in real calculus, it has much stronger consequences: holomorphic functions have derivatives of every order, are locally represented by convergent power series, and satisfy integral identities that determine their interior values from boundary values. They are the central objects of complex analysis.

Holomorphic functions connect local differential conditions with global geometry and topology. Their zeros, singularities, and continuation properties explain phenomena ranging from the evaluation of real integrals to conformal mapping and the behavior of special functions. The equivalence between complex differentiability and complex analyticity is particularly important: a condition involving only a first derivative implies an exceptionally rigid structure that has no direct counterpart for general smooth functions of real variables.

Complex differentiability and the Cauchy–Riemann equations

Let U⊆CU\subseteq\mathbb C be open and let f:U→Cf:U\to\mathbb C. The function is holomorphic on UU if, for every z0∈Uz_0\in U, the limit

f′(z0)=lim⁡h→0h∈Cf(z0+h)−f(z0)hf'(z_0)=\lim_{\substack{h\to0\\h\in\mathbb C}} \frac{f(z_0+h)-f(z_0)}{h}

exists. The increment hh can approach zero from any direction in the complex plane, and the same limit must result. A function is holomorphic at a point when it is holomorphic on some neighborhood of that point; existence of a complex derivative at one isolated point is not sufficient.

Writing z=x+iyz=x+iy and f(z)=u(x,y)+iv(x,y)f(z)=u(x,y)+iv(x,y) expresses complex differentiability as a restriction on a map from R2\mathbb R^2 to itself. If this map is differentiable in the real sense at a point, its derivative must be multiplication by a complex number, rather than an arbitrary real-linear transformation. Its Jacobian matrix therefore has the form

Df=(uxuyvxvy)=(a−bba).Df= \begin{pmatrix} u_x&u_y\\ v_x&v_y \end{pmatrix} = \begin{pmatrix} a&-b\\ b&a \end{pmatrix}.

Equating entries gives the Cauchy–Riemann equations

ux=vy,uy=−vx.u_x=v_y,\qquad u_y=-v_x.

Under real differentiability, these equations are equivalent to complex differentiability at the point, and f′=ux+ivxf'=u_x+iv_x. A frequently used sufficient condition is that the first partial derivatives of uu and vv exist and are continuous on UU, and satisfy the equations there. Merely checking the equations at a point, without appropriate differentiability assumptions, is not enough.

The examples f(z)=znf(z)=z^n, for nonnegative integers nn, and the complex exponential function are holomorphic throughout C\mathbb C. A function holomorphic on the whole plane is called an entire function. Quotients of polynomials are holomorphic wherever their denominators do not vanish. By contrast, f(z)=z‾f(z)=\overline z is nowhere complex differentiable: real increments give a difference quotient of 11, while purely imaginary increments give −1-1. The function f(z)=∣z∣2f(z)=|z|^2 is complex differentiable only at 00, where its derivative is zero, but it is not holomorphic on any neighborhood of 00.

Sums, products, and compositions of holomorphic functions are holomorphic, as are quotients where the denominator is nonzero. The usual product rule and chain rule remain valid. An equivalent differential notation uses

∂∂z‾=12(∂∂x+i∂∂y).\frac{\partial}{\partial\overline z} =\frac12\left(\frac{\partial}{\partial x} +i\frac{\partial}{\partial y}\right).

For a continuously differentiable function, holomorphicity is exactly the equation ∂f/∂z‾=0\partial f/\partial\overline z=0. This formulation becomes especially useful in partial differential equations and in several complex variables.

Contour integration and Cauchy’s formulas

The decisive consequences of holomorphicity emerge through integration along curves. For a piecewise continuously differentiable path γ:[a,b]→U\gamma:[a,b]\to U, the complex line integral is

∫γf(z) dz=∫abf(γ(t))γ′(t) dt.\int_\gamma f(z)\,dz =\int_a^b f(\gamma(t))\gamma'(t)\,dt.

The Cauchy integral theorem states, in a standard form, that if ff is holomorphic on a simply connected open set UU, then its integral around every closed piecewise smooth curve in UU is zero. Local versions assert vanishing around triangles whose filled interiors lie in the domain. Goursat’s refinement establishes this without assuming continuity of f′f'.

The topological hypothesis matters. The function 1/z1/z is holomorphic on C∖{0}\mathbb C\setminus\{0\}, yet its integral around the positively oriented unit circle is 2πi2\pi i. This does not contradict Cauchy’s theorem: the punctured plane is not simply connected, and the curve surrounds a point outside the domain. On a simply connected domain, every holomorphic function has a holomorphic antiderivative, so contour integrals depend only on their endpoints. On a domain with holes, such a primitive exists precisely when all integrals around closed curves vanish.

The Cauchy integral formula gives considerably more information. If a closed disk D(a,R)‾\overline{D(a,R)} and a neighborhood of it lie in the domain of a holomorphic function, then for every zz inside its boundary circle,

f(z)=12πi∫∣ζ−a∣=Rf(ζ)ζ−z dζ.f(z)=\frac{1}{2\pi i} \int_{|\zeta-a|=R} \frac{f(\zeta)}{\zeta-z}\,d\zeta.

Thus values along the circle determine the function throughout the disk. Differentiating the kernel yields

f(n)(z)=n!2πi∫∣ζ−a∣=Rf(ζ)(ζ−z)n+1 dζ.f^{(n)}(z)=\frac{n!}{2\pi i} \int_{|\zeta-a|=R} \frac{f(\zeta)}{(\zeta-z)^{n+1}}\,d\zeta.

These identities prove that a holomorphic function has derivatives of every order, even though its definition assumes only a first derivative at each point.

There is also a converse criterion. Morera’s theorem says that a continuous complex-valued function on an open set is holomorphic if its integral around the boundary of every triangle contained in the set is zero. One way to understand the result is to construct a local primitive from path integrals; differentiating this primitive recovers the original function. Morera’s theorem is particularly useful when a function is defined as a limit or an integral and direct evaluation of a complex difference quotient would be cumbersome.

Power series and the equivalence with analyticity

A complex analytic function is one that can locally be represented by a convergent power series. Every holomorphic function is analytic, and every function locally represented by a complex power series is holomorphic. Consequently, “holomorphic” and “complex analytic” describe the same class of functions, although the words emphasize different starting points: differentiability in the first case, local series expansions in the second.

To obtain the expansion, apply Cauchy’s integral formula on a circle centered at aa and expand its kernel as a geometric series:

1ζ−z=1ζ−a∑n=0∞(z−aζ−a)n.\frac{1}{\zeta-z} =\frac{1}{\zeta-a} \sum_{n=0}^{\infty} \left(\frac{z-a}{\zeta-a}\right)^n.

For zz strictly inside the circle, the series converges uniformly in ζ\zeta on the contour, allowing termwise integration. The result is the Taylor series

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

This representation holds throughout every disk centered at aa contained in the domain. Inside its radius of convergence, a power series can be differentiated and integrated term by term, and its differentiated series have the same radius of convergence.

The boundary of the original domain need not determine the exact radius of a Taylor expansion, because the function may extend beyond that boundary. More precisely, the radius is the largest radius for which the germ at the center has a single-valued holomorphic extension to the corresponding disk. If that radius is finite, at least one point of the boundary circle obstructs such an extension. For example,

11+z2=∑n=0∞(−1)nz2n\frac{1}{1+z^2} =\sum_{n=0}^{\infty}(-1)^n z^{2n}

has radius of convergence 11, owing to the singularities at ii and −i-i. Its restriction to the real axis is nevertheless smooth for every real input. Complex singularities can therefore control the convergence of series used to represent real functions.

Cauchy’s derivative formula also yields quantitative bounds. If ∣f∣≤M|f|\le M on ∣z−a∣=R|z-a|=R, then

∣f(n)(a)∣≤n!MRn.|f^{(n)}(a)|\le \frac{n!M}{R^n}.

These Cauchy estimates link the magnitude of a function to the size of all its derivatives. They are central to uniqueness arguments, approximation theory, and growth estimates for entire functions. The distinction from real smoothness is substantial: the real function defined by e−1/x2e^{-1/x^2} for x≠0x\ne0, and by 00 at x=0x=0, has derivatives of every order, all zero at the origin, but is not represented there by its Taylor series. Holomorphic functions cannot exhibit this behavior at an interior point.

Zeros, uniqueness, and maximum principles

The local power-series representation tightly controls zeros. If ff is holomorphic near aa, vanishes at aa, and is not identically zero near aa, there is a unique positive integer mm such that

f(z)=(z−a)mg(z),g(a)≠0,f(z)=(z-a)^m g(z), \qquad g(a)\ne0,

with gg holomorphic. The integer mm is the zero’s multiplicity or order. It is the index of the first nonzero Taylor coefficient. Since gg remains nonzero sufficiently close to aa, the zero at aa is isolated.

On a connected open set, this gives the identity theorem: if two holomorphic functions agree on a set with an accumulation point inside the domain, they agree everywhere. In particular, a nonzero holomorphic function cannot have zeros accumulating at an interior point. The location of the accumulation point is essential. The function sin⁡(1/z)\sin(1/z), holomorphic on the punctured plane, has zeros approaching 00, but 00 does not belong to its domain. Likewise, equality merely at points approaching the boundary does not generally establish equality throughout a domain.

The maximum modulus principle states that a nonconstant holomorphic function on a domain cannot attain a local maximum of its modulus at an interior point. If a function is continuous on the closure of a bounded domain and holomorphic inside, its maximum modulus is attained on the boundary. Applying the same principle to 1/f1/f gives a minimum-modulus statement wherever ff has no zeros. Another related result, the open mapping theorem, says that a nonconstant holomorphic function maps open sets to open sets. Critical points do not invalidate this conclusion: z↦zmz\mapsto z^m is open even though its derivative vanishes at the origin when m>1m>1.

A striking consequence is Liouville’s theorem: every bounded entire function is constant. Cauchy’s estimate gives ∣f′(a)∣≤M/R|f'(a)|\le M/R for every R>0R>0; letting RR increase without bound forces f′(a)=0f'(a)=0. This yields a short proof of the fundamental theorem of algebra. If a nonconstant complex polynomial had no zero, its reciprocal would be entire and bounded, contradicting Liouville’s theorem.

Quantitative forms of rigidity occur in the Schwarz lemma. If ff maps the unit disk into itself and f(0)=0f(0)=0, then ∣f(z)∣≤∣z∣|f(z)|\le |z| and ∣f′(0)∣≤1|f'(0)|\le1. Equality at a nonzero point, or equality in the derivative bound, forces f(z)=eiθzf(z)=e^{i\theta}z. This result connects analytic bounds with the geometry of the disk and underlies the Schwarz–Pick theorem, which describes contraction of its hyperbolic metric.

Isolated singularities, Laurent expansions, and residues

A point aa is an isolated singularity of ff if ff is holomorphic on some punctured disk 0<∣z−a∣<R0<|z-a|<R, but is not initially defined as a holomorphic function throughout that disk. On an annulus, the appropriate representation is a Laurent series,

f(z)=∑n=−∞∞cn(z−a)n.f(z)=\sum_{n=-\infty}^{\infty}c_n(z-a)^n.

The coefficients are given by contour integrals,

cn=12πi∫∣ζ−a∣=rf(ζ)(ζ−a)n+1 dζ,c_n=\frac{1}{2\pi i} \int_{|\zeta-a|=r} \frac{f(\zeta)}{(\zeta-a)^{n+1}}\,d\zeta,

for circles lying in the annulus. The negative-power terms form the principal part and distinguish the three possible kinds of isolated singularity.

A singularity is removable when all negative-power coefficients vanish. The function then extends holomorphically by assigning the missing value. Riemann’s removable singularity theorem characterizes this situation by boundedness in a punctured neighborhood. For example, sin⁡z/z\sin z/z extends to 00 with value 11. A singularity is a pole of order mm when only finitely many negative terms occur and the lowest power is (z−a)−m(z-a)^{-m}. Equivalently, f(z)=(z−a)−mg(z)f(z)=(z-a)^{-m}g(z) with g(a)≠0g(a)\ne0. A meromorphic function is holomorphic except for poles.

An essential singularity has infinitely many nonzero negative-power coefficients. The example e1/ze^{1/z} has an essential singularity at 00. Its behavior differs radically from that of a pole: along the positive real axis it grows without bound as z→0z\to0, while along the negative real axis it tends to zero. The Casorati–Weierstrass theorem states that the image of every punctured neighborhood of an essential singularity is dense in C\mathbb C. The stronger great Picard theorem says that every complex value, with at most one exception, is assumed infinitely often in each such neighborhood.

The coefficient c−1c_{-1} is the residue, denoted Res⁡(f,a)\operatorname{Res}(f,a). The residue theorem states that a positively oriented simple closed contour enclosing finitely many isolated singularities, with none on the contour, satisfies

∫γf(z) dz=2πi∑a inside γRes⁡(f,a),\int_\gamma f(z)\,dz =2\pi i\sum_{a\text{ inside }\gamma} \operatorname{Res}(f,a),

provided the function is otherwise holomorphic on a neighborhood of the enclosed region. This converts many contour integrals into finite algebraic calculations and, through suitable contours and limiting estimates, evaluates many real improper integrals.

Applying the theorem to f′/ff'/f gives the argument principle: its contour integral divided by 2πi2\pi i equals the number of zeros minus the number of poles inside, counted with multiplicity. Rouché’s theorem follows from related reasoning: if ∣g∣<∣f∣|g|<|f| on the boundary of a suitable region, ff and f+gf+g have the same number of zeros within it. These results allow zeros to be counted without finding their exact locations.

Analytic continuation, branches, and domain topology

A local formula may define a function on a much smaller region than the one to which it can be extended. Analytic continuation extends a holomorphic function through overlapping neighborhoods, requiring agreement wherever successive definitions overlap. The identity theorem makes each compatible continuation unique locally. For example, the geometric series ∑n=0∞zn\sum_{n=0}^{\infty}z^n, initially convergent only for ∣z∣<1|z|<1, continues as 1/(1−z)1/(1-z) throughout C∖{1}\mathbb C\setminus\{1\}.

Continuation along different paths need not return the same value. The complex logarithm illustrates this directly. A local logarithm LL satisfies eL(z)=ze^{L(z)}=z and L′(z)=1/zL'(z)=1/z. Continuing around a loop that winds once counterclockwise around the origin increases its value by 2πi2\pi i. Therefore no single-valued holomorphic logarithm exists on the entire punctured plane. On a slit plane, such as C∖(−∞,0]\mathbb C\setminus(-\infty,0], a branch can be defined by choosing a continuous argument. The cut is a choice of domain, not a line of poles of the logarithm.

More generally, a nowhere-zero holomorphic function ff on a simply connected domain has a holomorphic logarithm. One constructs it by integrating f′/ff'/f and choosing an additive constant so that exponentiation recovers ff. The same argument gives holomorphic roots. Without simple connectivity, the relevant obstruction is the set of periods

∫γf′(z)f(z) dz\int_\gamma \frac{f'(z)}{f(z)}\,dz

over closed curves. For a logarithm all these periods must vanish. Thus the interaction of holomorphicity with topology is not merely qualitative: winding numbers and contour integrals record the obstruction explicitly.

The monodromy theorem says that a holomorphic germ which admits analytic continuation along every path in a simply connected domain determines a single-valued holomorphic function there. When continuation is genuinely multivalued over the complex plane, a Riemann surface provides a natural single-valued setting. The two local values of z\sqrt z, for instance, become parts of one holomorphic function on its branched covering surface.

Continuation is not always possible across any boundary arc. A power series may have a natural boundary, meaning that no point of its convergence circle permits holomorphic continuation across that circle. The lacunary series

∑n=0∞z2n\sum_{n=0}^{\infty}z^{2^n}

has the unit circle as a natural boundary. This contrasts with the geometric series, whose convergence circle contains only one obstruction to continuation. A radius of convergence therefore records the extent of a particular disk expansion, while analytic continuation asks a broader question about how local function elements fit together.

Conformal geometry and harmonic functions

At a point where f′(z0)≠0f'(z_0)\ne0, a holomorphic function behaves to first order like multiplication by a nonzero complex number:

f(z0+h)=f(z0)+f′(z0)h+o(∣h∣).f(z_0+h)=f(z_0)+f'(z_0)h+o(|h|).

Such multiplication rotates and scales the plane while preserving oriented angles. Consequently, holomorphic functions with nonvanishing derivative are locally conformal maps. The local scale factor is ∣f′(z0)∣|f'(z_0)|, and the local rotation angle is the argument of f′(z0)f'(z_0). Their real Jacobian determinant is

det⁡Df(z0)=∣f′(z0)∣2>0.\det Df(z_0)=|f'(z_0)|^2>0.

The complex inverse function theorem ensures a local holomorphic inverse wherever the derivative is nonzero. A bijective holomorphic map between plane domains has a holomorphic inverse and is called biholomorphic. At a critical point, where the derivative vanishes, the map is not locally conformal in this sense. If the first nonzero term of f(z)−f(a)f(z)-f(a) has degree mm, the local behavior resembles z↦zmz\mapsto z^m: nearby values are generally covered mm times, and angles at the critical point are multiplied by mm.

The Riemann mapping theorem states that every nonempty simply connected proper open subset of C\mathbb C is biholomorphic to the unit disk. The exclusions matter: the entire plane is not biholomorphic to the disk, since an inverse map from the plane into the bounded disk would contradict Liouville’s theorem. The theorem concerns interior domains; extension of a conformal map to the boundary requires further hypotheses. For a domain bounded by a Jordan curve, the map extends to a homeomorphism of the closures.

The real and imaginary parts of a holomorphic function are harmonic functions. Differentiating the Cauchy–Riemann equations gives

uxx+uyy=0,vxx+vyy=0,u_{xx}+u_{yy}=0,\qquad v_{xx}+v_{yy}=0,

so both satisfy Laplace’s equation. Conversely, a harmonic function uu locally has a harmonic conjugate vv, making u+ivu+iv holomorphic. On a simply connected domain the conjugate can be chosen globally, up to an additive real constant. On a punctured domain this can fail: log⁡∣z∣\log|z| is harmonic on C∖{0}\mathbb C\setminus\{0\}, but its conjugate would be a globally defined argument, which does not exist there.

These connections explain classical applications to two-dimensional potential theory. In an incompressible, irrotational planar flow, the velocity potential and stream function can be combined into a complex potential. With a conventional choice W=ϕ+iψW=\phi+i\psi, its derivative is W′=u−ivW'=u-iv, where u,vu,v are the Cartesian velocity components. Conformal maps transform suitable flow and electrostatic problems into simpler geometries. Their usefulness depends on boundary conditions as well as interior holomorphicity: a map that simplifies the domain also changes metric quantities such as normal derivatives and physical length scales.

Convergence, compactness, and approximation

Holomorphicity is stable under locally uniform convergence: if fnf_n is holomorphic on an open set UU and converges uniformly on every compact subset to ff, then ff is holomorphic. Moreover, every derivative converges locally uniformly:

fn(k)⟶f(k)(k=1,2,…).f_n^{(k)}\longrightarrow f^{(k)} \qquad(k=1,2,\ldots).

Cauchy’s integral formula proves both statements by placing each smaller compact region inside suitable integration contours. This is much stronger than the corresponding result for differentiable real functions, whose uniformly convergent sequences can have nondifferentiable limits.

Ordinary pointwise convergence is not enough. It provides no uniform control on the contour values needed by Cauchy’s formula, and pointwise limits of holomorphic functions need not be holomorphic. In infinite series and parameter-dependent integrals, locally uniform estimates are therefore a standard means of establishing holomorphicity. For example, a series of holomorphic functions whose absolute values on each compact set are bounded by the terms of a convergent numerical series converges locally uniformly and defines a holomorphic sum.

A complementary principle concerns compactness of families. Montel’s theorem states that a locally bounded family of holomorphic functions is normal: every sequence in the family has a subsequence converging uniformly on compact subsets to a holomorphic function. Here local boundedness means that each compact subset has a common bound valid for every function in the family. Cauchy estimates turn these bounds into derivative bounds and hence local equicontinuity. A compactness argument then produces convergent subsequences. Normal families are central to existence proofs in conformal mapping and to the study of iterated holomorphic maps in dynamical systems.

Approximation by simpler functions is another major theme. Taylor polynomials approximate a holomorphic function uniformly on compact subdisks of its convergence disk, but general compact sets require more global results. Runge’s theorem implies that if a compact set K⊂CK\subset\mathbb C has connected complement, every function holomorphic on a neighborhood of KK can be approximated uniformly on KK by polynomials. When the complement has holes, rational approximation with suitably placed poles is the appropriate replacement.

The obstruction caused by a hole can be seen on the unit circle. Polynomials cannot approximate 1/z1/z uniformly there: every polynomial has zero contour integral around the circle, whereas 1/z1/z has integral 2πi2\pi i, and uniform convergence would preserve that integral. Mergelyan’s theorem strengthens the connected-complement case: continuity on KK and holomorphicity on its interior suffice for uniform polynomial approximation. These approximation results also have numerical significance. Analytic continuation to a neighborhood of an interval or contour often produces rapid convergence of interpolation and quadrature formulas, with convergence rates governed by the location of nearby complex singularities.

Holomorphic functions in several variables and on complex spaces

For functions on an open subset of Cn\mathbb C^n, the theory of several complex variables retains the equivalence between holomorphicity and local convergent power-series expansions:

f(z)=∑α∈Nncα(z−a)α.f(z)=\sum_{\alpha\in\mathbb N^n} c_\alpha(z-a)^\alpha.

Here α=(α1,…,αn)\alpha=(\alpha_1,\ldots,\alpha_n) is a multi-index and (z−a)α=∏j=1n(zj−aj)αj(z-a)^\alpha=\prod_{j=1}^n(z_j-a_j)^{\alpha_j}. For a continuously differentiable function, holomorphicity is equivalent to

∂f∂z‾j=0,j=1,…,n.\frac{\partial f}{\partial\overline z_j}=0, \qquad j=1,\ldots,n.

Its derivative is a complex-linear map, rather than merely a real-linear map between the underlying real spaces.

A striking theorem of Hartogs says that separate holomorphicity implies joint holomorphicity. If a function is holomorphic in each coordinate when the other coordinates are fixed, it is holomorphic as a function of all variables together. No additional continuity hypothesis is needed. This differs from many familiar statements about separate continuity or separate real differentiability, where joint regularity can fail. On polydisks, repeated one-variable Cauchy formulas yield an integral formula over the product of boundary circles and lead to multivariable Taylor expansions.

The geometry of singularities changes substantially once n≥2n\ge2. The Hartogs extension theorem, in a standard form, says that if Ω⊂Cn\Omega\subset\mathbb C^n is a domain, K⊂ΩK\subset\Omega is compact, and Ω∖K\Omega\setminus K is connected, every holomorphic function on Ω∖K\Omega\setminus K extends holomorphically to Ω\Omega. In particular, isolated singularities are removable. There is no analogue of 1/z1/z with a single isolated pole in several complex variables. The function 1/z11/z_1 instead fails along the whole hypersurface z1=0z_1=0.

This extension phenomenon makes domains of holomorphy a central subject. In one variable, every plane domain admits a holomorphic function that cannot be continued across any boundary point. In several variables, some domains force all their holomorphic functions to extend farther. The precise geometry governing nonextendability is expressed through notions such as pseudoconvexity and plurisubharmonic functions. The resulting theory connects complex analysis to partial differential equations, particularly the problem of solving equations involving the operator ∂‾\overline\partial.

Holomorphicity also makes sense on a complex manifold, because transition maps between coordinate charts are holomorphic. In complex dimension one these manifolds are Riemann surfaces. A compact connected complex manifold admits only constant global holomorphic functions, by the maximum modulus principle, but its meromorphic functions and holomorphic sections can have rich structure. This distinction is important in algebraic geometry, where local holomorphic data, divisors, and line bundles encode information not visible in global scalar-valued holomorphic functions alone.

Historical development

The modern theory emerged from several approaches to complex functions rather than from a single definition introduced at one moment. During the eighteenth century, Leonhard Euler developed extensive calculations involving exponentials, logarithms, trigonometric functions, and infinite series with complex arguments. These computations established important relationships among functions but preceded the later systematic treatment of convergence, domains, and complex differentiability. The problem of understanding multivalued expressions, particularly logarithms and roots, remained a major source of conceptual difficulty.

Augustin-Louis Cauchy made contour integration a foundational tool during the early nineteenth century. His memoir on definite integrals, presented in 1814 and published in 1825, was an important step in the development of complex integration. Cauchy’s integral theorem, integral formula, and residue methods revealed that complex differentiability controlled function values throughout a region, not merely infinitesimal behavior at individual points. His work supplied the integral approach that still forms a standard route from the definition of holomorphicity to power-series expansions and derivative estimates.

Bernhard Riemann placed greater emphasis on geometric and potential-theoretic ideas. His 1851 doctoral dissertation developed foundational ideas about functions of a complex variable and their representation on surfaces. Riemann surfaces made it possible to interpret apparently multivalued functions as single-valued functions on more suitable spaces. His approach also connected conformal mapping with harmonic functions and the Dirichlet principle. Some early existence arguments based on that principle required later justification, illustrating the distinction between the strength of a geometric insight and the analytic foundations needed to prove it.

Karl Weierstrass developed an approach centered on convergent power series and analytic continuation. This emphasized local expansions, precise convergence arguments, and the propagation of function elements. The Cauchy, Riemann, and Weierstrass viewpoints eventually became complementary parts of one theory: complex differentiation leads to integral formulas; integral formulas imply convergent local series; and compatible series organize continuation and global geometry.

Later developments clarified assumptions and exposed new kinds of behavior. Édouard Goursat showed at the end of the nineteenth century that continuity of the derivative need not be assumed in Cauchy’s theorem. Émile Picard’s results on entire functions and essential singularities, beginning in 1879, strengthened the theory’s rigidity statements. Henri Poincaré and Paul Koebe established the uniformization theorem in 1907, placing the global geometry of Riemann surfaces into a broad classification. Friedrich Hartogs’s work in the early twentieth century showed that several-variable holomorphic functions possess extension properties absent from one-variable theory, opening a subject with its own geometric and analytic problems.