Anderson acceleration method of finding steady-state particle size distribution for a wide class of aggregation–fragmentation models

Anderson acceleration method of finding steady-state particle size distribution for a wide class of aggregation–fragmentation models

A fast numerical method of finding steady-state distributions of particles sizes for a wide class of aggregation–fragmentation models, including the models with a source of monomers is developed. The method is based on a fast evaluation scheme for large sets of non-linear Smoluchowski-type ODE with...

Full description

Saved in:
Bibliographic Details
Published in:Computer physics communications Vol. 224; pp. 154 - 163
Main Authors: Matveev, S.A., Stadnichuk, V.I., Tyrtyshnikov, E.E., Smirnov, A.P., Ampilogova, N.V., Brilliantov, N.V.
Format: Journal Article
Language:English
Published: Elsevier B.V 01-03-2018
Subjects:
Aggregation
Anderson acceleration
Fragmentation
Low-rank matrices
Smoluchowski equation
Steady-state distributions
Aggregation
Anderson acceleration
Smoluchowski equation
Low-rank matrices
Steady-state distributions
Fragmentation
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A fast numerical method of finding steady-state distributions of particles sizes for a wide class of aggregation–fragmentation models, including the models with a source of monomers is developed. The method is based on a fast evaluation scheme for large sets of non-linear Smoluchowski-type ODE with an application of the Anderson acceleration method of the fixed point iterations. In the numerical tests of the suggested approach we demonstrate that huge sets of non-linear ODE may be solved with high precision in terms of Euclidian norm of the residual in modest times with use of a regular desktop computer. We compare our numerical solutions with the known analytical results for the steady-state distributions as well as with the other fast numerical schemes and prove the high accuracy of the novel method and its significant superiority with respect to the existing fast numerical method of solution of the addressed problems.
ISSN:0010-4655
1879-2944
DOI:10.1016/j.cpc.2017.11.002