Liouville’s theorem is a fundamental result in complex analysis: a complex-valued function that is holomorphic throughout the complex plane and bounded there must be constant. Its importance lies in the strength of this conclusion. Boundedness alone imposes little rigidity on real differentiable functions, but together with complex differentiability everywhere it eliminates every nonconstant possibility.
The theorem connects local analytic properties with global behavior. It gives a short proof of the fundamental theorem of algebra, characterizes entire functions of polynomial growth, and underlies results about singularities, complex mappings, harmonic functions, and operator spectra. Its standard proof follows from Cauchy’s integral formula by estimating a derivative on circles of arbitrarily large radius.
The name commemorates Joseph Liouville (1809–1882). Several distinct results also bear his name, notably preservation of phase-space volume in Hamiltonian mechanics, a bound on rational approximation to algebraic numbers, and a structural theorem about elementary antiderivatives. These are separate theorems, not alternative formulations of the bounded-entire-function result.
Statement, hypotheses, and scope
A function is called an entire function if it is holomorphic at every complex number. Here holomorphic means complex differentiable in a neighborhood of each point. Liouville’s theorem states that if there is a finite constant such that
then for some fixed . The absolute value is the complex modulus, so boundedness means that the entire image lies inside a single disk of finite radius.
Both the global domain and the global bound are essential. The function is entire but unbounded. The function is holomorphic and bounded on the region , but it is not entire. The identity function is bounded on the unit disk without being constant there. Thus the theorem does not assert that every bounded holomorphic function on an arbitrary domain is constant. It uses the absence of any finite boundary: circles centered at a chosen point may be expanded indefinitely while remaining inside the domain.
Boundedness on the real axis is also insufficient. The exponential function is bounded on the imaginary axis, and is bounded on the real axis, yet both are nonconstant entire functions. In particular,
whose modulus grows without bound as . Growth in complex directions that are invisible on a real slice is often decisive.
The theorem can equivalently be stated as follows: every nonconstant entire function is unbounded. More strongly, it is unbounded outside every disk. Indeed, if it were bounded outside one disk, continuity would make it bounded on the compact disk as well, yielding a global bound. Nevertheless, unboundedness does not mean that the modulus tends to infinity in every direction. For example, tends to zero along the negative real axis and grows along the positive real axis.
The contrast with real analysis is substantial. The real function is bounded, differentiable, and nonconstant. Complex differentiability is much more restrictive than real differentiability: through the Cauchy–Riemann equations and integral formulas, it links behavior in different directions and controls derivatives by values on surrounding curves. Liouville’s theorem is a global expression of that rigidity.
Proof through Cauchy’s estimates
The standard proof uses Cauchy’s integral formula, which reconstructs a holomorphic function and its derivatives from boundary values. Fix . Because an entire function is holomorphic on and inside every circle centered at , the derivative formula gives
for every , with the circle traversed counterclockwise. If , the integrand has modulus at most , while the circle has length . Consequently,
The left side is independent of . Letting therefore yields . Since was arbitrary, the derivative vanishes throughout the plane. A holomorphic function with zero derivative on a connected domain is constant, completing the proof. No maximum value of needs to be attained; the uniform upper bound is enough.
This argument is an instance of the general Cauchy estimates. If is holomorphic on a neighborhood of the closed disk , and
then, for each nonnegative integer ,
The theorem follows by taking and observing that remains bounded as the radius increases. The same inequality explains why prescribed growth rates constrain higher derivatives and, ultimately, the degree of a function.
There is also a coefficient version. An entire function has a power series expansion
valid throughout . Cauchy’s estimates imply
for every . For each , sending to infinity forces . Only the constant coefficient survives. This proof makes clear that boundedness suppresses every positive-degree term, rather than merely ruling out some particular class of examples.
The proof also indicates why replacing the plane with a bounded domain changes the conclusion. Inside the unit disk, a circle centered at a fixed point cannot be expanded indefinitely without leaving the domain. The estimate remains valid, but its denominator cannot be made arbitrarily large. Liouville’s theorem therefore depends on both local analyticity, which supplies the estimate, and global geometry, which permits unlimited radii.
Polynomial growth and the fundamental theorem of algebra
Liouville’s theorem is the degree-zero case of a broader growth principle. Suppose is entire and, for some and real ,
For a positive integer , Cauchy’s estimate at the origin gives
The right side tends to zero as . Every Taylor coefficient of degree greater than therefore vanishes, so is a polynomial of degree at most .
This sharpens the relationship between growth and algebraic form. If an entire function grows at most linearly, it is affine; if it grows at most quadratically, it has degree at most two. An entire function satisfying a sublinear bound with must be constant. More generally, if its maximum modulus satisfies for a positive integer , then its degree is at most . These conclusions require control in all complex directions, not just along selected rays.
A celebrated application is the fundamental theorem of algebra, which asserts that every nonconstant complex polynomial has a complex root. Let
and suppose, for contradiction, that never vanishes. Then is entire. Since the leading term dominates for sufficiently large ,
so is bounded outside a sufficiently large disk.
Inside that disk, continuity and the absence of zeros imply that has a positive minimum. Hence is bounded there too. Liouville’s theorem makes constant, contradicting the assumption that is nonconstant. Once one root is known to exist, polynomial division and repetition give the factorization
with roots counted according to multiplicity.
The proof illustrates a recurring analytic strategy: transform the hypothetical absence of zeros into a bounded holomorphic reciprocal. The same strategy works whenever lower bounds on a function supply upper bounds on its inverse. It does not show that every nonconstant entire function has a zero, because a general entire function need not grow to infinity in all directions. The zero-free function , for example, has entire reciprocal , but that reciprocal is unbounded. Polynomial growth is what supplies the extra global control in the algebraic argument.
Infinity, removable singularities, and meromorphic functions
The behavior of an entire function at infinity becomes easier to classify after introducing the coordinate . If is entire, then
is holomorphic on a punctured neighborhood of . The isolated singularity of at zero is called the singularity of at infinity. This viewpoint packages global growth into a local question near one additional point.
If is bounded for sufficiently large , then is bounded near zero. The removable singularity theorem extends holomorphically to . Thus extends holomorphically over infinity to the Riemann sphere. A complex-valued holomorphic function on that compact surface must be constant: its modulus attains a maximum, and the maximum modulus principle forces constancy. This provides a geometric interpretation of Liouville’s theorem.
More generally, an entire function is a nonconstant polynomial precisely when infinity is a pole. If is a polynomial of degree , then has a pole of order at zero. Conversely, a pole of order at infinity supplies a bound of the form outside a disk. The polynomial-growth argument then shows that is a polynomial, with the pole order determining its degree. An entire function that is not a polynomial has an essential singularity at infinity.
Liouville’s theorem also helps characterize meromorphic functions on the sphere. Such a function has only finitely many poles, because poles cannot accumulate on a compact surface. Subtract the finite principal part at each pole in the finite plane. The remaining function is entire and has at most a pole at infinity, so it is polynomial. Restoring the principal parts expresses the original function as a rational function. Thus meromorphic functions on the sphere are exactly rational functions.
A qualification is important: nonconstant holomorphic maps from the sphere to itself do exist. Rational functions are examples, with poles interpreted as points mapped to infinity. The constancy statement concerns holomorphic functions taking values in , not arbitrary holomorphic maps into the sphere. Compactness prevents an everywhere-finite complex-valued holomorphic function from varying, but it does not prevent a sphere-valued map from passing through infinity. Distinguishing the target spaces avoids an apparent conflict between Liouville’s theorem and familiar rational maps such as .
Consequences for images and conformal mappings
A useful strengthening of unboundedness is that the image of a nonconstant entire function is dense in . Suppose, instead, that the image misses a disk centered at of radius . Then
for every , and the function
is entire and bounded by . Liouville’s theorem makes , and hence , constant. Therefore every nonempty open disk contains a value attained by a nonconstant entire function.
Density is not the same as surjectivity. The exponential function omits zero, although its image is dense in the plane. The much stronger little Picard theorem states that a nonconstant entire function can omit at most one complex value. Liouville’s theorem rules out an omitted disk; Picard’s theorem rules out even two omitted points. The difference marks a substantial increase in depth rather than a minor change in wording.
The theorem also shows that no bounded plane domain can be biholomorphically equivalent to the whole complex plane. A biholomorphic map from onto such a domain would itself be a bounded entire function, hence constant, whereas a biholomorphism must be bijective. In particular, the plane and the unit disk are not conformally equivalent. They are homeomorphic as topological spaces, so this obstruction belongs to complex geometry rather than to topology alone.
A related consequence concerns half-planes. If an entire function satisfies
everywhere, then
is entire and takes values in the unit disk. Indeed, for a complex number with positive real part,
Thus , and Liouville’s theorem makes , and consequently , constant. The same conclusion holds whenever the image lies in any open half-plane.
These transformations reveal that a bound on the full modulus is not always needed in the original formulation of a problem. A geometric restriction on the image may be converted into boundedness by a suitable holomorphic transformation. For example, an upper bound implies that is bounded by . Its constancy gives , so . An entire function whose real part is bounded on just one side is therefore already constant.
Harmonic functions and elliptic Liouville principles
A real-valued function on a region in is harmonic if it satisfies Laplace’s equation,
The real and imaginary parts of a holomorphic function are harmonic in two dimensions. There is a corresponding harmonic Liouville theorem: every bounded harmonic function on all of is constant. In dimension one this follows directly from ; in higher dimensions it follows from mean-value formulas or interior derivative estimates.
For a harmonic function on a ball of radius , a typical interior estimate has the form
where depends only on the dimension. If is globally bounded and harmonic everywhere, can tend to infinity, forcing the gradient to vanish. This is structurally the same argument as the complex proof: a local estimate becomes a global rigidity statement because the domain admits arbitrarily large balls.
In the plane, the complex theorem yields an even stronger formulation. Every harmonic function on has a global harmonic conjugate because the plane is simply connected. Write for the associated entire function. If , then
Liouville’s theorem makes constant; differentiation gives , so is constant. Thus an entire planar harmonic function bounded on only one side is constant. In every dimension, the corresponding assertion for nonnegative harmonic functions follows from Harnack inequalities.
Results of this kind are often called Liouville properties for partial differential equations. Their content depends on the operator, the domain, and the allowable growth. For example, admits the unbounded solution , so a growth assumption cannot simply be discarded. Polynomial-growth versions identify harmonic polynomials under suitable hypotheses, much as Cauchy estimates identify holomorphic polynomials.
The ambient geometry matters as well. On a curved noncompact manifold, bounded nonconstant harmonic functions may exist; on the hyperbolic disk, harmonic extensions of nonconstant bounded boundary data provide examples. Consequently, a harmonic Liouville theorem on a manifold requires additional geometric assumptions or specific analytic estimates. In applications, these theorems often establish uniqueness: if two solutions have the same forcing, their difference is harmonic, and boundedness may show that the only remaining freedom is an additive constant.
Several complex variables and Banach-valued extensions
The direct extension to several complex variables states that every bounded entire function
is constant. One proof restricts to complex affine lines. Given , define
This is a bounded entire function of one complex variable, so . Hence , and arbitrary points have the same value. The proof shows that the one-variable theorem controls the higher-dimensional statement without requiring new integral formulas.
Another proof holds all but one coordinate fixed. Liouville’s theorem makes the function independent of the remaining coordinate, and repetition removes every coordinate. Multivariable Cauchy estimates give corresponding polynomial-growth statements: an entire function on with a global bound , for a nonnegative integer , is a polynomial of total degree at most . High-order partial derivatives vanish because their Cauchy bounds decay as the surrounding polydisks expand.
There is also a vector-valued version. Let be a complex Banach space, and let be holomorphic in the norm sense. If everywhere, then is constant. For every continuous linear functional , the composition is bounded and entire, hence constant. The Hahn–Banach theorem ensures that continuous linear functionals separate points, so constancy of all these scalar compositions implies constancy of .
An important application in functional analysis is the nonemptiness of the spectrum of a bounded operator on a nonzero complex Banach space. Suppose a bounded linear operator had empty spectrum. Then
would exist and be operator-valued holomorphic for every . For , the Neumann series gives
Thus at infinity.
Continuity bounds the resolvent on every fixed compact disk, and the displayed estimate bounds it outside a sufficiently large disk. The Banach-valued Liouville theorem would therefore make constant, with its limit at infinity forcing that constant to be zero. This contradicts its being an inverse. The spectrum must consequently be nonempty. The argument generalizes the analytic proof that a complex polynomial has a root, replacing scalar reciprocals with operator inverses.
The distinct theorem in Hamiltonian mechanics
In classical mechanics, Liouville’s theorem usually means conservation of phase-space volume under Hamiltonian evolution. For degrees of freedom, canonical coordinates are
and Hamilton’s equations are
Assuming sufficient differentiability, the divergence of the phase-space velocity field vanishes:
For a smooth flow , the Jacobian determinant obeys
Since the divergence is zero and is the identity, the determinant remains one. Thus, wherever the flow is defined,
for measurable phase-space sets . The theorem remains valid for a sufficiently smooth time-dependent Hamiltonian; conservation of energy is not required for preservation of phase-space volume.
The coordinate-free description belongs to symplectic geometry. On a -dimensional symplectic manifold with symplectic form , Hamiltonian flow preserves , and therefore preserves the volume form
This makes clear that the invariant object is canonical phase-space volume, not necessarily ordinary coordinate volume after an arbitrary change of variables. A noncanonical coordinate system may require a nonconstant density in its volume element.
For a smoothly transported probability density , the continuity equation becomes
This is the Liouville equation. It states that the fine-grained density remains constant along each trajectory. A distribution may stretch and fold into intricate structures, but Hamiltonian transport alone does not compress its full phase-space volume.
This theorem is central to statistical mechanics, yet it does not by itself establish thermal equilibrium, mixing, or ergodicity. Those involve additional properties of the dynamics. Nor does volume preservation mean that a cloud retains its shape: expansion in one direction can be balanced by contraction in another. Frictional equations generally fail to preserve phase-space volume. The bounded-entire-function theorem and the mechanical theorem share Liouville’s name but concern different objects and have independent proofs.
Other namesakes in number theory, integration, and geometry
In number theory, Liouville’s approximation theorem restricts how closely an algebraic irrational number can be approximated by rational numbers. If is a real algebraic number of degree , there is a constant such that
for integers and positive integers . To see the source of the exponent, take an integer polynomial of degree with . Since , clearing denominators gives . A bound on near , together with the mean value theorem, then yields the inequality; rational numbers farther away are handled by reducing the constant.
Liouville used exceptionally accurate rational approximation to establish the existence of transcendental numbers in 1844. A standard example is
Truncation after the -th term gives a rational number with denominator , while the remaining tail is less than . These approximations eventually violate the bound for every possible algebraic degree, proving that is transcendental. This arithmetic theorem is unrelated to bounded entire functions, despite sharing a theme of global restrictions.
In differential algebra, Liouville’s theorem on elementary integration describes the structure of elementary antiderivatives. In a standard differential-field formulation, if in a differential field of characteristic zero has an antiderivative in an elementary extension introducing no new constants, then, in a suitable algebraic extension, it can be written
where the are constants. The corresponding antiderivative is an algebraic-field contribution plus a finite sum of logarithms. This result underlies algorithms for symbolic integration and proofs that some elementary integrands have no elementary antiderivative.
A further Liouville theorem concerns conformal mappings in real dimensions . Under suitable regularity assumptions, a conformal local diffeomorphism between connected Euclidean domains is the restriction of a Möbius transformation, built from similarities and inversions. This is much more rigid than the two-dimensional situation, where holomorphic functions with nonzero derivative supply large families of local conformal maps.
These usages require explicit identification of the subject. In complex analysis, “Liouville’s theorem” ordinarily denotes bounded-entire-function rigidity; in mechanics it denotes invariant phase-space volume; in Diophantine approximation it denotes the algebraic approximation bound. Phrases such as “Liouville property” may describe further rigidity statements, but they do not imply that those statements follow from a single common theorem.