Suppose the temperature on the edge of a circular plate is prescribed. Which steady temperature can occur inside? The mathematical question is to find a harmonic function with those boundary values. The Poisson kernel gives a weighted average of the boundary data; as the observation point approaches the edge, the weight concentrates near that point.
Write and let be a continuous real function on the unit circle.
Definition 3.10.1
(Dirichlet problem on the disk)
A solution of the Dirichlet problem with data is a real function continuous on , twice continuously differentiable and harmonic on , with for every .
The continuity on the closed disk is part of the problem, not a consequence of harmonicity inside it.
Constructing the solution
Definition 3.10.2
(Poisson kernel)
For , define
The denominator is . Moreover,
The series follows from the geometric series and is uniformly convergent in for fixed ; termwise integration gives the last identity. Thus is a positive probability density after division by .
Theorem 3.10.3
(Poisson formula for continuous data)
For , the unique solution of the disk Dirichlet problem is
In particular, the right side extends continuously to the boundary with value there.
Proof
(Harmonicity, boundary convergence, uniqueness)
For fixed , the kernel in (2) is the real part of , which is holomorphic in . Expanding it in powers of and integrating term by term, uniformly on each compact subdisk, shows that (2) is the real part of a holomorphic function and is harmonic.
Let approach . Using (1), subtract inside the integral. Given , continuity of gives such that the difference of the two boundary values is below whenever the circular distance from to is below . For sufficiently close to , this applies whenever the circular distance from to is below . On the remaining arc, the denominator of is bounded below by a positive constant when is near , while its numerator tends to zero. Hence the integral of the kernel over that arc tends to zero. Positivity and total mass one then bound the difference in (2) by plus a term tending to zero. This proves convergence along arbitrary interior approaches, not just radii.
If and are two solutions, is harmonic inside and continuous on the closed disk, with zero boundary values. The maximum and minimum principles give . Thus the solution is unique.
Proof X-ray
Where the boundary condition enters
The kernel has total mass one at every interior point. Its mass outside any fixed angular neighborhood tends to zero as the point approaches the boundary. These two facts turn continuity of into boundary convergence; harmonicity alone does not supply it.
Reading the formula
Worked example 3.10.4
(Trigonometric boundary data)
For , orthogonality in (1) gives
This polynomial is harmonic and has the prescribed boundary values. More generally, each boundary frequency is multiplied by : fine oscillations are suppressed rapidly near the center. Formula (2) also shows that ; here , the mean of .
Common mistake
(Extending the data without the factor )
For , the function has the right boundary values but is not a solution. It equals , which takes every value in near and so has no continuous value at the center, and in polar form . The Poisson integral multiplies frequency by , giving the solution .
Worked example 3.10.5
(The center value is the boundary average)
At the kernel is , so (2) gives . For the continuous data ,
No closed form for is needed for this value. As a check, , in agreement with .
Worked example 3.10.6
(Evaluating the solution for )
Since , the factor gives
using . The polynomial is harmonic, being the real part of , and at it returns . At its value is , which lies between and .
The theorem asks for continuity on the whole closed disk, and that requirement cannot be weakened by one point.
Counterexample mode
Boundary limits at all but one point do not give uniqueness
Claim: two harmonic functions on the disk with the same boundary limits at all but one boundary point coincide. The harmonic function
has radial limit zero at every boundary point except , yet it is positive inside and diverges along the radius toward . It has no continuous extension to the closed disk. It and the zero function have the same boundary limits at every boundary point except , yet they differ. The repair is the hypothesis of the theorem: continuity on the whole closed disk, with no exceptional point.
Practice with solutions
Quick check
(Find the solution for and its value at .)
Lower the power with a double-angle identity.
Solution · Answer
Since , the solution is ; its center value is .
Quick check
(Find the solution for .)
Rewrite the data as a single frequency.
Solution · Answer
, so the solution is , the imaginary part of .
Quick check
(Show that no function satisfying the stated Dirichlet conditions can have boundary values zero on the upper semicircle and one on the lower semicircle, including a prescribed value at each endpoint.)
What does continuity on the closed disk imply for the boundary values?
Solution · Answer
Such boundary data are discontinuous at both endpoints regardless of the values assigned there. The restriction of a continuous function on the closed disk to its boundary must be continuous, so no solution in the stated class exists.
A conformal map from another domain onto the disk can transfer harmonic functions by composition. Transferring prescribed boundary values additionally requires control of how the map extends to that boundary; a conformal map of interiors alone does not automatically provide this.