Computational Methods (Prev) Programs (Next)

The C++/CUDA version of PSCF that is described in this manual was initially designed starting in 2015-2016 as a complete rewrite of an older Fortran program of the same name. The Fortran version of PSCF is an SCFT program for incompressible liquids containing one or more species of linear block polymers and (optionally) one or more small molecule solvents. The acronym PSCF stands for "Polymer Self-Consistent Field", reflecting this origin of the package as an SCFT program.

The Fortran version of PSCF is maintained in a separate github.com repository at https://github.com/dmorse/pscf , but is no longer being actively developed. The name "pscfpp" that is that is used for the github repository containing the current version was chosen to avoid a clash with the name "pscf" that is used for the repository containing the older Fortran code.

The current C++/CUDA version of PSCF is intended to supersede the Fortran version. The current version provides almost all of the capabilities of the Fortran program, and some important new capabilities. New users of PSCF should use the C++/CUDA version documented here, and users of the Fortran version should consider switching to the current version whenever they can.

Features and advantages of current C++ version :

Features of the current version of PSCF that distinguish it from the legacy Fortran version include the following:

The current version was designed as an extensible package of several programs intended for different geometries, algorithms and/or computer hardware. The Fortran code is a single SCFT program designed for periodic structures and linear polymers.
The current version enables simulations of mixtures containing acyclic branched block polymers, in addition to the linear block polymers and linear homopolymers allowed by the Fortran code.
The current version enables the use of graphics processing units (GPUs) to dramatically accelerate some applications.
The current version enables partial-saddle point field theoretic simulation (PS-FTS) calculations in addition to SCFT.

Features unique to the legacy Fortran program :

The following features of the legacy Fortran program have not been ported to the C++/CUDA version:

The Fortran PSCF program provides a quasi-Newton-Broyden SCFT iteration algorithm that works very well for problems of modest size.
The Fortran PSCF program can perform RPA-like linear-reponse calculations for ordered structures.

These two features are described in more detail below;

The Fortran PSCF program provided a unique quasi-Newton/Broyden SCFT iteration algorithm that was based on a physically motivated initial approximation for the inverse Jacobian that is iteratively improved via a Broyden update algorithm. This was an excellent algorithm for the problems of moderate size for which it was originally designed. It was abandoned in the new version in favor of a variety of Anderson-mixing algorithms because the amount of memory required by the Newton/Broyden algorithm increases as the square of the number of spatial degrees of freedom (i.e., number of basis functions), which prevented the algorithm from being applied to larger problems.

The Fortran PSCF program has the ability to perform generalized RPA calculations to compute the self-consistent linear response of a periodic structure to an infinitesimal inhomogeneous external field. This algorithm allowed description of limits of stability of ordered phases and of composition fluctuations within such phases. This feature has not been ported to the new C++ version, but will remain available within the legacy Fortran program, via the github repository for that program. The required algorithm and some application are described in the following reference:

Ranjan, A. and Qin, J. and Morse, D.C., "Linear Response and Stability of Ordered Phases of Block Copolymers", 2008 Macromolecules 41, 942.

Computational Methods (Prev) About (Up) Programs (Next)