[yohey44]

Hi,

I tried to solove Poisson equation by spectral method.
Referring this note(http://www.physics.buffalo.edu/phy41...6/ch6-lec2.pdf), I pluged FFT for discretized form as,
(U_j+1,k + U_j-1,k + U_j,k+1, Uj, K-1 - 4U_j,k)/h^2 = -f_j,k.

Fourier transform is defined as,
U'_m,n=1/N*Sum_{j=0}^{N-1}(W^-{mj+nk}U_j,k),
W = exp(2*i*pi/N)

Then I obtain,
U'_m,n = -h^2*f_j,k/(W^m+W^-m+W^n+W^-n-4).

With inverse transfom, seemingly correct result is obtained.

However, I think fourier transform before discretization is also correct.
I mean,
d^2U/dx^2 + d^2U/dy^2 = -f,
is transformed as,
(-m^2-n^2)U'_m,n = -f'_m,n.

Then, inverse transforming for above equation must return correct result.
However, no correct answer is obtained.

What is missing in the second approach?
And how I can solve it?

Thanks in advance.