Consider the standard multipole expansion for the coulomb potential often used in the fast multipole method algorithm (Theorem 5.2). A translation theorem is used in the algorithm (Theorem 5.3).

enter image description here

Is O(n,m) the same as M(n,m) in Theorem 5.2 except shifted with coordinates rho', alpha', and beta'? This makes sense but I wanted to double check.

When evaluating the expansion in Theorem 5.3, the factorial terms in A(j-n,k-m) will contain negative integers (namely from k-m), giving an undefined result or complex infinity. Additionally, the O(j-n,k-m) term will be impossible to evaluate due to the spherical harmonic indices being impossible (in Y(n,m), n must be greater or equal to abs(m) and this is not the case for indices (j-n,k-m)). What do we do about this? How is Theorem 5.3 evaluated correctly?

source for Theorems: https://math.nyu.edu/faculty/greengar/shortcourse_fmm.pdf

  • $\begingroup$ Consider migration of this question to Computational Science SE. If it does not get enough attention, flag for the moderator. $\endgroup$ – Anton Menshov May 2 '20 at 22:45

This question was answered in SE! The indices of the summation need to be adjusted:



Not the answer you're looking for? Browse other questions tagged or ask your own question.