Laplace's equation is $\frac{\partial^2f}{\partial x^2}+\frac{\partial^2f}{\partial y^2}=0$.
A function $f(x,y)$ is said to be harmonic if it satisfies this equation.
Roughly speaking, this means that for every point $(x,y)$, the value $f(x,y)$ is the
average of the values of $f$ at the points around $(x,y)$.
In the picture, we know the values of $f$ on the boundary of the disc: these are
represented by the dark band. We want to find the values of $f$ inside the disc;
these are represented by the mesh. Initially we take these values to be zero so
the mesh is flat.
Clicking the Step button replaces the values on the mesh by the average of the
values at the neighbouring points. If we do this repeatedly, the mesh will eventually
settle down to a shape that is close to the solution of Laplace's equation. However,
convergence is slow.
Clicking the Overstep button implements a faster algorithm. Suppose that
the value at position $(i,j)$ is $z_{ij}$, and the average of the neighbours
is $z'_{ij}$, and put $r_{ij}=z'_{ij}-z_{ij}$. The previous method is equivalent
to adding $r_{ij}$ to $z_{ij}$. For the overstep method we add $1.9r_{ij}$ to $z_{ij}$,
but only if $i+j$ is even (if the time step is even) or if $i+j$ is odd (if the time step
is odd).
Every point is updated using the average of the neighbours:
$$z_{ij} \leftarrow \frac{1}{4}(z_{i-1,j}+z_{i+1,j}+z_{i,j-1}+z_{i,j+1})$$
Every even point is updated using the average of the (odd) neighbours.
Every odd point is updated using the average of the (even) neighbours.