A point source inside a ball produces the wrong value on the boundary if used alone. For a plane, a reflected source cancels it. A sphere requires a different geometry: inversion sends the source to an exterior point, and a carefully chosen weight cancels the original field on the sphere.
The inverted image
Let , , , and , so . For the point is the inversion of in the sphere ; it lies on the same ray as , at distance from the origin.
Definition 5.6.1
(Green function of the ball)
For and with , set
For the central source define, for ,
Expression (2) is independent of the unit vector because is radial. It is the limit of (1) as for fixed , since . The image lies outside the ball, so the second term is smooth and harmonic in , while the first has a unit source at .
Theorem 5.6.2
(Boundary cancellation and positivity)
For each , the function is harmonic in , for , and for every with .
Proof
(One identity for squared distances)
Let . For , expanding and yields
If , the right side is zero, so and ; the two terms in (1) then agree and . If , the right side is strictly positive, so . Since , this gives whenever , hence . Harmonicity of holds because . For , (2) is zero on , positive for because is radially decreasing, and is constant.
Proof X-ray
Boundary cancellation and interior positivity
The whole proof is the single algebraic identity (3), read in two ways. Its right side factors as a product of and , so its sign records only where and sit relative to the sphere. On the sphere the first factor vanishes, and this is exactly the boundary cancellation. Inside, both factors are positive, which compares the two distances. The comparison is turned into a comparison of values by the monotonicity of , and this is where is used. The weight is the factor that converts "distances proportional with ratio " into "equal values of ". The resulting sign matches the convention : positive sources produce positive zero-boundary solutions.
Both the inversion and the weight are needed, as a concrete check shows.
Counterexample mode
The image needs its weight
Claim: the unweighted image already vanishes on the sphere. By (3), on it equals . Take , , , so and the bracket is . At , and the boundary value is ; at , and the value is . The boundary trace is neither zero nor even constant. Inversion alone is not enough: the factor in (1) is essential.
The source problem
For set
Theorem 5.6.3
(Green representation for the source problem)
Let and . Then in (4) is smooth in , satisfies , extends continuously to with on , and is the only such solution in . If and , then in .
Proof
(Newtonian potential plus a harmonic correction)
The first term is the Newtonian potential of the zero extension of , and is smooth with . The image term is harmonic and smooth on a neighborhood of the closed ball: the support of is compactly contained in , so all noncentral image points stay a positive distance outside the sphere, and the expression has a smooth limit as . Therefore (4) is a smooth classical solution in the ball and extends continuously with boundary value zero, because on the sphere for every . The difference of two solutions is harmonic with zero boundary values, so the maximum principle shows uniqueness among functions in . If and is not identically zero, the positivity of gives inside.
From the Green kernel to boundary data
For and , differentiation with respect to the source variable along its outward normal produces the kernel for boundary data.
Definition 5.6.4
(Poisson kernel of the ball)
For and ,
To verify the constant, write with and . In the image term, . Differentiating both terms of (1) with respect to and setting gives (5). The derivative is taken from inside the ball. Green's identity applied on the punctured ball, followed by the shrinking-sphere limit, gives for every harmonic
In particular, using proves that integrates to one, and (5) shows that it is positive.
Theorem 5.6.5
(The Poisson integral solves the Dirichlet problem)
Let and define
Then is harmonic in , and setting on gives a function in . It is the unique harmonic function in that is continuous on with boundary values .
Proof
(Harmonic kernel, unit mass and concentration)
The kernel is harmonic in : this follows by differentiating (1) in and using that its singularity stays on the boundary. Differentiation under the integral on compact subsets then makes harmonic.
To see that as , use unit mass to write , and split the integral into and its complement. Continuity makes uniformly small on the first part, and positivity with unit mass bounds that part by the same small number. On the complement, for sufficiently close to , while the numerator tends to zero, so that part tends to zero. Hence extends continuously to . The maximum principle gives uniqueness.
Worked examples
Worked example 5.6.6
(Linear boundary data)
If , then : this function is harmonic and has the prescribed boundary values, so uniqueness identifies it with the Poisson integral. No integral needs to be evaluated; the theorem converts a guess into a proof.
Worked example 5.6.7
(Quadratic boundary data)
Let on . Try . Since and , the choice makes harmonic. On we need , so :
At the center . This agrees with the symmetry computation of the spherical mean: the averages of over the sphere are equal and add up to .
Worked example 5.6.8
(The center of the ball)
Take in (5). Then for every , so
which is one over the area of . Formula (6) becomes the mean value property: is the average of over the sphere. This is the degenerate case in which the kernel carries no directional information. As approaches a boundary point the kernel concentrates instead: at with , as .
Common mistakes
Common mistake
(Dropping the minus sign in the Poisson kernel)
The Green function vanishes on the sphere and is positive inside, so its outward normal derivative in is negative there. The Poisson kernel is ; omitting the sign gives a negative kernel of total mass and reverses every boundary value.
Common mistake
(Copying the weight into two dimensions)
For the weight equals , but the unweighted image with equals the nonzero constant on the circle. The planar Green function uses instead; the inversion survives, but the correction is additive rather than multiplicative.
Practice with solutions
Quick check
(Compute for .)
Use .
Solution · Answer
. The image contribution is , a constant; this agrees with (2) by the symmetry of the formula.
Quick check
(Explain why an interior source cannot be canceled by placing the image at itself.)
Ask what happens to the unit source.
Solution · Answer
Equal and opposite sources at the same point cancel the unit source as well as its boundary trace. The correction must be harmonic in the ball, so its singularity has to lie outside .
Quick check
(If , compute its Poisson integral.)
Use the case of (6).
Solution · Answer
It equals , by unit mass of ; equivalently, the constant is harmonic with boundary values , so uniqueness gives the same answer.
Quick check
(What changes if boundary data are merely bounded but not continuous?)
Look at which step of the boundary-limit proof uses continuity.
Solution · Answer
The integral remains harmonic for bounded measurable data and tends to at each continuity point by the same splitting argument. At a discontinuity that argument does not establish a pointwise boundary limit.