Integro-partial differential equations occur in many contexts in mathematic
al physics. Typical examples include time-dependent diffusion equations con
taining a parameter (e.g., the temperature) that depends on integrals of th
e unknown distribution function. The standard approach to solving the resul
ting nonlinear partial differential equation involves the use of predictor-
corrector algorithms, which often require many iterations to achieve an acc
eptable level of convergence. In this paper we present an alternative proce
dure that allows us to separate a family of integro-partial differential eq
uations into two related problems, namely (i) a perturbation equation for t
he temperature, and (ii) a linear partial differential equation for the dis
tribution function. We demonstrate that the variation of the temperature ca
n be determined by solving the perturbation equation before solving for the
distribution function. Convergent results for the temperature are obtained
by recasting the divergent perturbation expansion as a continued fraction.
Once the temperature variation is determined, the self-consistent solution
for the distribution function is obtained by solving the remaining, linear
partial differential equation using standard techniques. The validity of t
he approach is confirmed by comparing the (input) continued-fraction temper
ature profile with the (output) temperature computed by integrating the res
ulting distribution function. (C) 1999 American Institute of Physics. [S002
2-2488(99)03410-6].