Infinite Resistor Grid
This problem was inspired by this XKCD comc:
Upon reading it, my first thought was that we could approach the problem with a discrete-space Fourier transform (as used in digital image processing theory), which would move from the infinite discrete space of the grid points of the array to the frequency domain (not actually “frequency”, but technically wavenumber as we are transforming space and not time coordinates, though I will still use 'frequency'), which would be continuous and finite:
If we have grid coordinates m,n and corresponding frequency coordinates u,v; then for a property of the grid nodes , we have:
and the inverse transform:
Consider a current injected (or removed) at point (m,n). By Kirchoff’s first law (conservation of electric charge), must equal the sum of the currents flowing out through the four resistors connecting to that node. Using Ohm’s Law V=IR and letting be the potential at node (m,n) we find those currents are , , , (as R=1 Ω). Summing these,
Using the inverse transform definition,
For this to be true for any point (or frequencies), the integrands must be equal, and so:
Putting in a current source of 1 A at the origin and a sink of -1 A at the point (M,N) gives us an . Note that the equivalent resistance between (0,0) and (M,N) is equal to for this current load. ().
Transforming to frequency,
Using the inverse transform,
Now, , thus
This gives our resistance as a (real) definite integral.
For the example in the comic, M=2, N=1, and with a lot of mathematical work, we can compute the integral to find , which is approximately 0.77324 Ω