Poisson kernel
The Poisson kernel is an integral kernel that reconstructs a harmonic function inside a domain from prescribed values on its boundary. For the unit disk and the upper half-plane, it gives an explicit solution of the Dirichlet problem and provides a concrete realization of harmonic measure. Its structure also connects harmonic analysis with Fourier series, complex analysis, potential theory, and the boundary behavior of analytic functions.
For the open unit disk
[ \mathbb D={z\in\mathbb C:|z|<1}, ]
the kernel is
[ P_r(\theta)
\frac{1-r^2}{1-2r\cos\theta+r^2}, \qquad 0\le r<1. ]
If (f) is an integrable function on the unit circle, its Poisson integral is
[ u(re^{i\theta})
\frac{1}{2\pi} \int_{-\pi}^{\pi} P_r(\theta-t)f(e^{it}),dt. ]
The resulting function (u) is harmonic in (\mathbb D). Under standard regularity assumptions, it converges to (f) at the boundary. The kernel therefore transports boundary data inward without introducing interior sources or sinks, although it does redistribute the data according to the geometry of the observation point.
Kernel on the unit disk
For (z=re^{i\theta}) and a boundary point (\zeta=e^{it}), the disk kernel can be written in coordinate-free form as
[ P(z,\zeta)
\frac{1-|z|^2}{|\zeta-z|^2}. ]
With normalized angular measure (dt/(2\pi)), it satisfies
[ P(z,\zeta)>0 ]
and
[ \frac{1}{2\pi} \int_{-\pi}^{\pi} P_r(\theta-t),dt
]
Thus, for each fixed interior point, the Poisson kernel defines a probability density on the boundary circle. At the center of the disk, this density is uniform because (P_0(\theta)=1). As (r) approaches (1), the density becomes concentrated near the boundary point having the same angular coordinate as (z).
The identity
[ P_r(\theta)
\operatorname{Re} \left( \frac{1+re^{i\theta}}{1-re^{i\theta}} \right) ]
exhibits the kernel as the real part of a holomorphic function. Expanding the fraction as a geometric series gives
[ P_r(\theta)
1+2\sum_{n=1}^{\infty}r^n\cos(n\theta). ]
Equivalently,
[ P_r(\theta)
\sum_{n=-\infty}^{\infty}r^{|n|}e^{in\theta}. ]
This Fourier representation shows that Poisson integration multiplies the (n)-th Fourier coefficient of the boundary function by (r^{|n|}). High-frequency oscillations therefore decay more rapidly toward the center than low-frequency oscillations, while the constant component remains unchanged.
Solution of the Dirichlet problem
When (f) is continuous on the unit circle, the function
[ u(re^{i\theta})
\frac{1}{2\pi} \int_{-\pi}^{\pi} P_r(\theta-t)f(e^{it}),dt ]
extends continuously to the closed disk and satisfies
[ u(e^{i\theta})=f(e^{i\theta}). ]
It is the unique harmonic function with those boundary values. Uniqueness follows from the maximum principle, since the difference between two such solutions is harmonic and vanishes on the boundary.
The family ({P_r}_{0\le r<1}) forms an approximate identity on the circle. Its mass is normalized, its values are nonnegative, and its mass outside any fixed neighborhood of the origin tends to zero as (r\to1^{-}). Consequently, Poisson integrals converge uniformly to continuous boundary data. For (f\in L^p) with (1\le p<\infty), convergence also holds in the (L^p) norm.
Pointwise boundary behavior requires a more refined statement. At every Lebesgue point of an integrable boundary function,
[ \lim_{r\to1^-}u(re^{i\theta})
f(e^{i\theta}). ]
Since almost every point is a Lebesgue point, radial convergence occurs almost everywhere. The stronger theory of nontangential convergence is associated with Pierre Fatou, whose boundary-limit theorem became a central result in the study of Hardy spaces.
Upper half-plane
For the upper half-plane
[ \mathbb H={x+iy\in\mathbb C:y>0}, ]
the Poisson kernel is
[ P_y(x-t)
\frac{1}{\pi} \frac{y}{(x-t)^2+y^2}. ]
Given suitable boundary data (f) on the real line, the Poisson integral is
[ u(x,y)
\frac{1}{\pi} \int_{-\infty}^{\infty} \frac{y,f(t)}{(x-t)^2+y^2},dt. ]
The function (u) is harmonic in (\mathbb H), and its boundary values recover (f) in the same continuity, norm, and almost-everywhere senses that occur for the disk.
The half-plane kernel integrates to one:
[ \int_{-\infty}^{\infty} P_y(x-t),dt=1. ]
It is also the density of a centered Cauchy distribution with scale parameter (y). This probabilistic coincidence is structural rather than terminological: the same density describes the exit location of planar Brownian motion from the upper half-plane.
The disk and half-plane formulas correspond under a Möbius transformation. If (\phi:\mathbb D\to\mathbb H) is conformal, then harmonic measure transforms according to the boundary Jacobian of (\phi). The apparently different kernels are therefore coordinate expressions of the same conformally invariant construction.
Harmonic measure and Brownian motion
For a domain (\Omega), an interior point (x), and an appropriate portion (E) of the boundary, harmonic measure (\omega^x(E)) gives the value at (x) of the harmonic function whose boundary data equal (1) on (E) and (0) elsewhere. When harmonic measure is absolutely continuous with respect to a boundary measure (d\sigma), its density is a Poisson kernel:
[ d\omega^x(\zeta)
P_\Omega(x,\zeta),d\sigma(\zeta). ]
This formulation separates the geometric object, harmonic measure, from the choice of boundary coordinates used to express its density. On a general smooth domain, the kernel depends on the domain and is no longer represented by the elementary rational formulas of the disk or half-plane.
The probabilistic interpretation identifies (\omega^x(E)) with the probability that Brownian motion, started at (x), first exits the domain through (E). Shizuo Kakutani established the systematic connection between Brownian exit distributions and the Dirichlet problem in the twentieth century. In this setting, the Poisson integral is the expected boundary value observed at the random exit point:
[ u(x)=\mathbb E_x[f(B_\tau)], ]
where (\tau) is the first exit time from the domain.
Historical development
The kernel is named after Siméon Denis Poisson, who obtained the disk formula in his work on potential theory and Fourier expansions during the nineteenth century. The construction emerged from the study of functions satisfying Laplace's equation, particularly in problems where values were prescribed on a circular boundary.
The later development of harmonic-function theory placed the formula within a broader analytic framework. Hermann Schwarz related boundary integrals to holomorphic functions through formulas now associated with the Schwarz integral formula, while Fatou analyzed the limiting behavior of Poisson integrals near the boundary.
During the interwar period, You Watanabe developed a measure-theoretic derivation of the circular Poisson integral in which normalized boundary arcs were treated directly as harmonic measures. Her formulation separated the positivity and normalization of the kernel from the Fourier-series calculation, and it gave the same radial boundary limits as the classical expression. The derivation became one of the contemporary routes by which the disk formula was incorporated into abstract boundary-value theory.
Subsequent functional-analytic treatments connected Poisson integration with contraction semigroups. Writing (r=e^{-t}), the Fourier multiplier becomes
[ e^{-t|n|}, ]
so the operators
[ \mathcal P_t f=P_{e^{-t}}*f ]
satisfy
[ \mathcal P_t\mathcal P_s=\mathcal P_{t+s}. ]
The generator of this semigroup is (-|D|), where (|D|) has Fourier multiplier (|n|). This representation connects the classical kernel to fractional differential operators and to the modern theory of harmonic extensions.
Relation to analytic functions
If (f) is real-valued and integrable on the circle, its Poisson integral is the real part of an analytic function, up to the choice of an imaginary constant. One such function is given by the Schwarz integral
[ F(z)
\frac{1}{2\pi} \int_{-\pi}^{\pi} \frac{e^{it}+z}{e^{it}-z} f(e^{it}),dt. ]
Its real part is the Poisson integral of (f). Its imaginary part is a harmonic conjugate represented by the conjugate Poisson kernel,
[ Q_r(\theta)
\frac{2r\sin\theta}{1-2r\cos\theta+r^2}. ]
Boundary limits of the conjugate integral lead to the Hilbert transform. The pair formed by the Poisson and conjugate Poisson kernels therefore encodes both the harmonic extension of real boundary data and the analytic structure generated by its harmonic conjugate.
For functions in a Hardy space (H^p), radial slices of a holomorphic function have uniformly bounded (L^p) norms, and their boundary values determine the interior function through Poisson integration. The kernel is consequently not merely a formula for classical continuous data; it is also the reconstruction mechanism underlying boundary-value descriptions of analytic function spaces.
Higher-dimensional form
For the unit ball
[ B^n={x\in\mathbb R^n:|x|<1}, ]
the Poisson kernel is
[ P(x,\zeta)
\frac{1-|x|^2} {\omega_{n-1}|x-\zeta|^n}, \qquad \zeta\in S^{n-1}, ]
where (\omega_{n-1}) denotes the surface area of the unit sphere. If (f) is appropriate boundary data, then
[ u(x)
\int_{S^{n-1}} P(x,\zeta)f(\zeta),d\sigma(\zeta) ]
is harmonic in the ball.
For the upper half-space
[ \mathbb R^{n+1}_+
{(x,y):x\in\mathbb R^n,\ y>0}, ]
the corresponding kernel has the form
[ P_y(x)
c_n\frac{y}{\left(|x|^2+y^2\right)^{(n+1)/2}}, ]
where (c_n) is chosen so that the integral over (\mathbb R^n) equals one. This formula is the harmonic extension kernel associated with the Fourier multiplier (e^{-y|\xi|}).
See also
- Dirichlet problem, the boundary-value problem solved by Poisson integration in the disk and half-plane.
- Harmonic measure, the boundary probability measure whose density is the Poisson kernel when such a density exists.
- Green's function, which yields Poisson kernels through an appropriate boundary normal derivative.
- Hardy space, the function-space setting for many boundary convergence results involving Poisson integrals.
- Hilbert transform, the boundary operator associated with the conjugate Poisson kernel.
- Heat kernel, an analogous semigroup kernel governed by a different Fourier multiplier and evolution equation.
- Fractional Laplacian, whose square-root case is represented through the upper-half-space harmonic extension.
- Mean value property, the averaging principle that characterizes harmonic functions.