Motivation
An equilibrium potential is determined by spatial constraints rather than initial motion. Prescribing its boundary values should determine the interior, but two questions remain distinct: can we construct a function with those values, and can there be another one? The maximum principle answers the second question. On a disk, the Poisson kernel gives an explicit answer to the first.
The kernel is an averaging weight on the boundary. Near the center, every boundary direction has equal influence. Near a boundary point, the weight concentrates around that point. Positivity and unit integral of the normalized kernel control the size of the extension; concentration makes it attain the prescribed data. These properties must be proved together with harmonicity. Merely finding a function that solves the interior differential equation does not settle its boundary behavior.
A source inside the disk requires another object. Green's function has a controlled logarithmic singularity at the source and vanishes on the boundary. Its negative outward normal derivative is , the normalized Poisson kernel. Thus the boundary and interior-source formulas are two parts of one construction.
Definitions
Identify the plane with , write , and let . The real Laplacian is . A real function is harmonic when , and is subharmonic in this classical sense when . We write the Poisson equation as
The minus sign is a convention chosen so that our Dirichlet Green function is positive. For the alternative convention , the source term in the representation changes sign. Boundary data are real continuous functions on the unit circle; equivalently is a continuous periodic function of the angle. The desired classical solution belongs to , so boundary attainment is part of the problem.
Definition 5.1.1
(The disk's two kernels)
For , define the Poisson kernel
For distinct , define the Dirichlet Green function
Here will denote planar area, and is the derivative in the outward unit normal direction at the boundary point .
The boundary kernel is nonsingular for each fixed interior radius. The Green function is singular on its diagonal, but a logarithmic singularity is locally integrable in two dimensions: is finite. This permits a volume integral against a bounded source. Local integrability alone, however, does not justify twice differentiating that singular integral by a naive pointwise rule.
Concept lensGeometric
Green's function as a transported logarithm
For fixed , the map is a bijection of onto itself with . Since ,
Thus is the logarithmic kernel centered at , pulled back by a disk automorphism. Zero boundary values and positivity say exactly that on and in , which is the modulus identity used in the proof below.
Theorem/Proposition
Theorem 5.1.2
(Maximum principles and Dirichlet uniqueness)
Let be a nonempty bounded open subset of the plane. If and , then
Applying this to both signs of a harmonic function gives its boundary maximum and minimum bounds. Hence a classical Dirichlet solution of the Poisson equation is unique when both source and boundary data are fixed. If is connected, a harmonic function that attains its maximum or minimum at an interior point is constant.
Theorem 5.1.3
(Continuous boundary data have a unique harmonic extension)
For every real , the function
is smooth and harmonic in , extends continuously to the prescribed on the closed disk, and is the unique such solution. Moreover,
Theorem 5.1.4
(A Green representation with an explicit existence class)
Suppose is real and . Then the unique solution in of with boundary values is
The notation means smooth functions supported in a compact subset of the open disk. In particular, such a source vanishes in a neighborhood of the boundary.
This source class is deliberately explicit. A representation identity for an already known smooth solution can hold under weaker assumptions, but constructing a twice continuously differentiable solution for arbitrary rough sources requires further potential or regularity theory. The theorem does not turn mere integrability of a singular kernel into an unrestricted existence claim.
Proof sketch or proof idea
Proof
(Weak maximum principle and uniqueness)
For the weak maximum principle, replace by . Its Laplacian is strictly positive, so it cannot attain a maximum inside: at an interior maximum the Hessian is negative semidefinite and the Laplacian is nonpositive. Its maximum must therefore be on the boundary. Boundedness of the domain bounds on its closure. Letting gives the asserted bound for . The difference of two Poisson solutions with the same source is harmonic, so applying the result to both signs proves uniqueness.
The extension theorem is proved by an explicit series, and the same formula then gives the strong principle.
Proof
(Poisson series, positivity and boundary concentration)
To construct the harmonic extension, sum a geometric series:
The terms are harmonic polynomials; their two Cartesian second derivatives cancel. The series and all its spatial derivatives converge uniformly on every smaller closed disk. Indeed, differentiating these polynomials introduces only polynomial factors in , while their powers are geometrically bounded away from radius one. Integrating against bounded therefore gives a smooth harmonic function. The constant term shows , and the rational formula shows strict positivity. These facts prove the supremum estimate and show that the center value is the boundary average.
For boundary attainment, fix a small angular distance . Outside that distance from zero, the denominator of stays bounded away from zero as , while its numerator tends to zero. The kernel mass outside this arc therefore tends to zero. Inside the arc, uniform continuity of controls the difference of boundary values. Splitting the integral into these two parts proves uniform radial convergence. Combining it with continuity of proves continuity along any approach to a boundary point, not just along a fixed radius. Uniqueness follows from the weak maximum principle.
This disk formula also supplies the mean-value property needed for the strong principle. Restrict a harmonic function to any closed disk lying inside its domain, then translate and rescale the Poisson formula. Its center value is the circle average. If the center is a global interior maximum, every point on each sufficiently small surrounding circle must have that same value: otherwise continuity would make the average smaller. Thus the maximum set is open as well as relatively closed. Connectedness forces it to be the whole domain. Apply the argument to the negative function for an interior minimum.
The Green representation needs the logarithmic singularity to carry exactly one unit of source.
Proof
(Green function: symmetry, sign, existence and representation)
For Green's function, fix and write . The possible pole of this numerator term lies outside the closed disk; for the term is zero. Direct calculation gives zero Laplacian for a logarithmic distance away from its pole, proving the needed harmonicity without additional complex analysis. The same is true of the denominator term away from . The modulus formula also gives , so the harmonicity statements apply in either variable. On the unit circle , so vanishes there. At the pole has the form plus a smooth harmonic term. Positivity follows from
The sign and normalization can be read from a small circle around the pole. The outward radial derivative of has integral . Consequently the negative Laplacian corresponds to a unit source there. More concretely, for every smooth compactly supported test function on the plane, integration by parts outside a small circle and passage to the limit give
The logarithm times the first-derivative boundary term vanishes because its size is at most a constant times ; the normal-derivative term supplies the center value.
This identity also justifies the existence claim without differentiating a nonintegrable second derivative of the kernel. Extend smoothly by zero to the plane and put . In , derivatives can be transferred to the smooth compactly supported factor. Locally in the observation point, all resulting integrals are controlled by the locally integrable logarithm on one fixed bounded set. The displayed test-function identity then gives . The remaining term in is harmonic and smooth near the closed disk: the support of stays a positive distance from the boundary, so its numerator never vanishes there. The volume potential is zero on the boundary. Adding the already constructed Poisson extension proves existence; the maximum principle proves uniqueness.
For a function smooth up to the boundary, Green's identity on the disk with a small circle about removed explains the full representation:
On the inner circle the punctured domain's outward normal points toward the pole. Its normal-derivative contribution tends to , while the logarithm times a bounded derivative contributes zero. On the outer circle . Direct radial differentiation at gives
Thus the boundary part of Green's identity is exactly the Poisson integral. For merely continuous boundary data, the preceding construction established the solution separately; it did not assume unavailable boundary derivatives.
Proof X-ray
Three different uses of the boundary
Boundary continuity allows a maximum argument on the closed domain. Concentration of the Poisson kernel proves that a constructed function actually has those values. Vanishing of the Green function eliminates the unknown normal derivative of the solution from the representation. These are separate steps: interior harmonicity alone supplies none of the missing boundary information.
Worked examples
Worked example 5.1.5
(A finite boundary profile becomes harmonic polynomials)
For , the Poisson extension is
In Cartesian coordinates this is . Differentiating verifies zero Laplacian, and setting radius one verifies the boundary values. The center value is two, the average boundary value. The factors of radius suppress angular variation near the center without changing the prescribed boundary trace.
Worked example 5.1.6
(A constant positive source and its sign)
For in the disk with zero boundary data,
is a direct solution because . It is positive inside and has its maximum at the center. This does not contradict the harmonic maximum principle: its Laplacian is negative. For this smooth explicit solution, the punctured-domain representation applies directly and yields
The compact-source existence theorem was not needed for this example. The identity provides a useful normalization check on the Green function and its sign.
Worked example 5.1.7
(Separate source error from boundary error)
Suppose two classical continuous-boundary solutions have sources differing by at most and boundary values differing by at most . Their difference satisfies . Set
Then and on the boundary, so the maximum principle gives . Apply the same argument to to obtain
This is an estimate for solutions assumed to exist. It distinguishes boundary mismatch from interior source mismatch and shows why source error has less influence close to the fixed boundary. An estimate alone does not construct either solution.
Counterexample mode
Harmonic away from a missing point is insufficient
The function is harmonic on the punctured disk and has zero values on the outer circle, yet is nonzero inside. There is no contradiction with Dirichlet uniqueness on the full disk: the function is singular at the origin and is not continuous on the closed disk. An omitted pole cannot be ignored when stating the domain and boundary conditions. Green's function uses a prescribed singularity; an ordinary harmonic Dirichlet solution does not have one.
The continuity hypotheses, at interior points and on the boundary, can each be tested by a concrete function.
Counterexample mode
A jump in boundary data is not attained continuously
Set the boundary value equal to one on the open upper semicircle and zero on the lower semicircle and its endpoints. Its Poisson integral is harmonic inside, but along the radius toward the boundary point it has value at every radius. The kernel is even in angle, so the upper and lower semicircles have equal weight there. The assigned endpoint value zero is not attained. Continuous boundary data were an explicit hypothesis of the extension theorem; an interior smoothing formula does not repair a discontinuous trace into a continuous one.
Common mistakes
Common mistake
(Expecting every solution to peak on the boundary)
The maximum principle does not say every solution of every elliptic equation takes its largest value on the boundary. The sign of the Laplacian matters. With our convention, a nonnegative source and zero boundary data produce a nonnegative solution, which can have a positive interior maximum. Applying the weak principle to its negative gives the appropriate lower bound.
Common mistake
(Differentiating the logarithmic kernel twice)
Likewise, differentiating the logarithmic volume kernel twice under the integral without cancellation is not justified just because the original integral converges. The proof transferred derivatives to a smooth compactly supported source. Rougher source classes require their own estimates. The two-dimensional area element near a pole is also essential; treating a logarithm as a one-dimensional singular integral loses the geometry of the calculation.
Common mistake
(Confusing the normals on the hole and the outer circle)
Finally, the normal on the small hole points inward toward the pole, whereas the normal on the disk boundary points outward. Confusing them reverses the source normalization. State the sign of the Poisson equation before borrowing a Green-function formula from another convention.
Summary
The maximum principle gives uniqueness and comparison, while the Poisson kernel constructs the continuous harmonic extension on a disk. Positivity, unit integral of the normalized kernel , and boundary concentration are as important as harmonicity. A positive Dirichlet Green function handles the convention ; its negative outward normal derivative is the boundary kernel . The source regularity and treatment of the logarithmic pole determine what existence and differentiation claims are justified.
Exercises
Quick check
(Extend the boundary data harmonically into the unit disk and find the center value.)
Each angular mode acquires its corresponding power of radius.
Solution
The extension is , or . These harmonic polynomials have the prescribed boundary values, and the center value is zero. Uniqueness follows from the maximum principle.
Quick check
(Solve in the disk with zero boundary data using a radial polynomial, and verify smoothness at the origin.)
In two dimensions the Laplacian of the fourth power of radius is sixteen times radius squared.
Solution
The function has zero boundary values and satisfies . Since , it is an ordinary polynomial in Cartesian coordinates and is smooth at the origin. There is no radial-coordinate singularity in the solution itself. The center value is .
Quick check
(Verify the identity proving is positive, then use it to show that a nonzero nonnegative smooth compactly supported source with zero boundary data gives a strictly positive interior solution.)
Continuity makes the source positive on an open set; one point of the integration domain does not determine its integral.
Solution
Expansion of the two squared moduli gives as their difference. Thus the numerator modulus is larger than the denominator and off the pole. A nonzero nonnegative continuous source is strictly positive on a small open ball. Removing the single observation point does not remove that positive-area contribution, so its Green integral is strictly positive at every interior point.
Quick check
(Can a solution on the closed disk satisfy and zero outward normal derivative? What constant normal derivative is compatible, and what solution family realizes it?)
Integrate the equation over the disk before trying to solve it.
Solution
The divergence theorem requires . Zero normal derivative would give , which is impossible. A constant normal derivative must be , since the circumference is . The family realizes it. Two smooth solutions with the same source and normal derivative differ by a harmonic function with . Green's identity gives , so connectedness makes the difference constant. Neumann data therefore leave this additive freedom.