The Laplace operator Δ of a function of n variables is defined by
If Δ f = 0 in some region Ω, f is said to be harmonic on Ω. In that case f takes on its maximum and minimum over Ω at locations on the boundary ∂Ω of Ω. Here is an example of a harmonic function over a square which clearly takes on its maximum on two sides of the boundary and its minimum on the other two sides.
![Plot of x^2 - y^2 over [-1,1] cross [-1,1].](https://www.johndcook.com/harmonic_example_plot.png)
The theorem above can be split into two theorems and generalized:
If Δ f ≥ 0, then f takes on its maximum on ∂Ω.
If Δ f ≤ 0, then f takes on its minimum on ∂Ω.
These two theorems are called the maximum principle and minimum principle respectively.
Now just as functions with Δf equal to zero are called harmonic, functions with Δf non-negative or non-positive are called subharmonic and superharmonic. Or is it the other way around?
If Δ f ≥ 0 in Ω, then f is called subharmonic in Ω. And if Δ f ≤ 0 then f is called superharmonic. Equivalently, f is superharmonic if −f is subharmonic.
The names subharmonic and superharmonic may seem backward: the theorem with the greater than sign is for subharmonic functions, and the theorem with the less than sign is for superharmonic functions. Shouldn’t the sub-thing be less than something and the super-thing greater?
Indeed they are, but you have to look f and not Δf. That’s the key.
If a function f is subharmonic on Ω, then f is below the harmonic function interpolating f from ∂Ω into the interior of Ω. That is, if g satisfies Laplace’s equation
then f ≤ g on Ω.
For example, let f(x) = ||x|| and let Ω be the unit ball in ℝn. Then Δ f ≥ 0 and so f is subharmonic. (The norm function f has a singularity at the origin, but this example can be made rigorous.) Now f is constantly 1 on the boundary of the ball, and the constant function 1 is the unique solution of Laplace’s equation on the unit ball with such boundary condition, and clearly f is less than 1 on the interior of the ball.
What about if f is subharmonic and g is superharmonic on a domain G , with (f-g)(x)<0 near boundary of a domain G. Then what can we say about (f-g) on whole of G. Does it remains negative?