Finally updated the github tutorial
This commit is contained in:
BIN
doc/src/Eqs/pair_coul_shield.jpg
Normal file
BIN
doc/src/Eqs/pair_coul_shield.jpg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 186 KiB |
33
doc/src/Eqs/pair_coul_shield.tex
Normal file
33
doc/src/Eqs/pair_coul_shield.tex
Normal file
@ -0,0 +1,33 @@
|
||||
\documentclass[aps,pr,onecolumn,superscriptaddress,noshowpacs,a4paper,15pt]{revtex4}
|
||||
\pdfoutput=1
|
||||
\bibliographystyle{apsrev4}
|
||||
\usepackage{color}
|
||||
\usepackage{dcolumn} %Align table columns on decimal point
|
||||
\usepackage{amssymb}
|
||||
\usepackage{amsmath}
|
||||
\usepackage{amsthm}
|
||||
\usepackage{graphicx}
|
||||
\usepackage[pdftex]{hyperref}
|
||||
\hypersetup{colorlinks=true,citecolor=blue,linkcolor=red,urlcolor=blue}
|
||||
\usepackage[all]{hypcap}
|
||||
\newcommand{\red}{\color{red}}
|
||||
\newcommand{\blue}{\color{blue}}
|
||||
\definecolor{green}{rgb}{0,0.5,0}
|
||||
\newcommand{\green}{\color{green}}
|
||||
\newcommand{\white}{\color{white}}
|
||||
%\newcommand{\cite}[1]{\hspace{-1 ex} % \nocite{#1}\citenum{#1}}
|
||||
\thickmuskip=0.5\thickmuskip %shorter spaces in math
|
||||
|
||||
\begin{document}
|
||||
\begingroup
|
||||
\Large
|
||||
\begin{eqnarray*}
|
||||
E & = & \frac{1}{2} \sum_i \sum_{j \neq i} V_{ij} \\[15pt]
|
||||
V_{ij} & = & {\rm Tap}(r_{ij})\frac{\kappa q_i q_j}{\sqrt[3]{r_{ij}^3+(1/\lambda_{ij})^3}}\\[15pt]
|
||||
{\rm Tap}(r_{ij}) & = & 20\left ( \frac{r_{ij}}{R_{cut}} \right )^7 -
|
||||
70\left ( \frac{r_{ij}}{R_{cut}} \right )^6 +
|
||||
84\left ( \frac{r_{ij}}{R_{cut}} \right )^5 -
|
||||
35\left ( \frac{r_{ij}}{R_{cut}} \right )^4 + 1
|
||||
\end{eqnarray*}
|
||||
\endgroup
|
||||
\end{document}
|
||||
BIN
doc/src/Eqs/pair_ilp_gr_hBN.jpg
Normal file
BIN
doc/src/Eqs/pair_ilp_gr_hBN.jpg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 301 KiB |
42
doc/src/Eqs/pair_ilp_gr_hBN.tex
Normal file
42
doc/src/Eqs/pair_ilp_gr_hBN.tex
Normal file
@ -0,0 +1,42 @@
|
||||
\documentclass[aps,pr,onecolumn,superscriptaddress,noshowpacs,a4paper,15pt]{revtex4}
|
||||
\pdfoutput=1
|
||||
\bibliographystyle{apsrev4}
|
||||
\usepackage{color}
|
||||
\usepackage{dcolumn} %Align table columns on decimal point
|
||||
\usepackage{amssymb}
|
||||
\usepackage{amsmath}
|
||||
\usepackage{amsthm}
|
||||
\usepackage{graphicx}
|
||||
\usepackage[pdftex]{hyperref}
|
||||
\hypersetup{colorlinks=true,citecolor=blue,linkcolor=red,urlcolor=blue}
|
||||
\usepackage[all]{hypcap}
|
||||
\newcommand{\red}{\color{red}}
|
||||
\newcommand{\blue}{\color{blue}}
|
||||
\definecolor{green}{rgb}{0,0.5,0}
|
||||
\newcommand{\green}{\color{green}}
|
||||
\newcommand{\white}{\color{white}}
|
||||
%\newcommand{\cite}[1]{\hspace{-1 ex} % \nocite{#1}\citenum{#1}}
|
||||
\thickmuskip=0.5\thickmuskip %shorter spaces in math
|
||||
%
|
||||
\begin{document}
|
||||
%
|
||||
\begingroup
|
||||
\Large
|
||||
\begin{eqnarray*}
|
||||
E & = & \frac{1}{2} \sum_i \sum_{j \neq i} V_{ij} \\[15pt]
|
||||
V_{ij} & = & {\rm Tap}(r_{ij})\left \{ e^{-\alpha (r_{ij}/\beta -1)}
|
||||
\left [ \epsilon + f(\rho_{ij}) + f(\rho_{ji})\right ] -
|
||||
\frac{1}{1+e^{-d\left [ \left ( r_{ij}/\left (s_R \cdot r^{eff} \right ) \right )-1 \right ]}}
|
||||
\cdot \frac{C_6}{r^6_{ij}} \right \}\\[15pt]
|
||||
\rho_{ij}^2 & = & r_{ij}^2 - ({\bf r}_{ij} \cdot {\bf n}_i)^2 \\[15pt]
|
||||
\rho_{ji}^2 & = & r_{ij}^2 - ({\bf r}_{ij} \cdot {\bf n}_j)^2 \\[15pt]
|
||||
f(\rho) & = & C e^{ -( \rho / \delta )^2 }\\[15pt]
|
||||
{\rm Tap}(r_{ij}) & = & 20\left ( \frac{r_{ij}}{R_{cut}} \right )^7 -
|
||||
70\left ( \frac{r_{ij}}{R_{cut}} \right )^6 +
|
||||
84\left ( \frac{r_{ij}}{R_{cut}} \right )^5 -
|
||||
35\left ( \frac{r_{ij}}{R_{cut}} \right )^4 + 1
|
||||
\end{eqnarray*}
|
||||
\endgroup
|
||||
%
|
||||
\end{document}
|
||||
%
|
||||
BIN
doc/src/Eqs/pair_kolmogorov_crespi_full.jpg
Normal file
BIN
doc/src/Eqs/pair_kolmogorov_crespi_full.jpg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 178 KiB |
33
doc/src/Eqs/pair_kolmogorov_crespi_full.tex
Normal file
33
doc/src/Eqs/pair_kolmogorov_crespi_full.tex
Normal file
@ -0,0 +1,33 @@
|
||||
\documentclass[aps,pr,onecolumn,superscriptaddress,noshowpacs,a4paper,15pt]{revtex4}
|
||||
\pdfoutput=1
|
||||
\bibliographystyle{apsrev4}
|
||||
\usepackage{color}
|
||||
\usepackage{dcolumn} %Align table columns on decimal point
|
||||
\usepackage{amssymb}
|
||||
\usepackage{amsmath}
|
||||
\usepackage{amsthm}
|
||||
\usepackage{graphicx}
|
||||
\usepackage[pdftex]{hyperref}
|
||||
\hypersetup{colorlinks=true,citecolor=blue,linkcolor=red,urlcolor=blue}
|
||||
\usepackage[all]{hypcap}
|
||||
\newcommand{\red}{\color{red}}
|
||||
\newcommand{\blue}{\color{blue}}
|
||||
\definecolor{green}{rgb}{0,0.5,0}
|
||||
\newcommand{\green}{\color{green}}
|
||||
\newcommand{\white}{\color{white}}
|
||||
%\newcommand{\cite}[1]{\hspace{-1 ex} % \nocite{#1}\citenum{#1}}
|
||||
\thickmuskip=0.5\thickmuskip %shorter spaces in math
|
||||
|
||||
\begin{document}
|
||||
|
||||
\begingroup
|
||||
\Large
|
||||
\begin{eqnarray*}
|
||||
E & = & \frac{1}{2} \sum_i \sum_{j \neq i} V_{ij} \\[15pt]
|
||||
V_{ij} & = & e^{-\lambda (r_{ij} -z_0)} \left [ C + f(\rho_{ij}) + f(\rho_{ji}) - A \left ( \frac{r_{ij}}{z_0}\right )^{-6} \right ] \\
|
||||
\rho_{ij}^2 & = & r_{ij}^2 - ({\bf r}_{ij}\cdot {\bf n}_{i})^2 \\[15pt]
|
||||
\rho_{ji}^2 & = & r_{ij}^2 - ({\bf r}_{ij}\cdot {\bf n}_{j})^2 \\[15pt]
|
||||
f(\rho) & = & e^{-(\rho/\delta)^2} \sum_{n=0}^2 C_{2n} { \rho/\delta }^{2n}
|
||||
\end{eqnarray*}
|
||||
\endgroup
|
||||
\end{document}
|
||||
83
doc/src/pair_coul_shield.txt
Normal file
83
doc/src/pair_coul_shield.txt
Normal file
@ -0,0 +1,83 @@
|
||||
"LAMMPS WWW Site"_lws - "LAMMPS Documentation"_ld - "LAMMPS Commands"_lc :c
|
||||
|
||||
:link(lws,http://lammps.sandia.gov)
|
||||
:link(ld,Manual.html)
|
||||
:link(lc,Section_commands.html#comm)
|
||||
|
||||
:line
|
||||
|
||||
pair_style coul/shield command :h3
|
||||
|
||||
[Syntax:]
|
||||
|
||||
pair_style coul/shield cutoff tap_flag :pre
|
||||
|
||||
cutoff = global cutoff (distance units)
|
||||
tap_flag = 0/1 to turn off/on the Tapper function
|
||||
|
||||
[Examples:]
|
||||
|
||||
pair_style coul/shield 16.0 1
|
||||
pair_coeff 1 2 0.70 :pre
|
||||
|
||||
|
||||
[Description:]
|
||||
|
||||
Style {coul/shield} computes a Coulomb interaction for boron and nitrigon
|
||||
atoms locate in different layers of hexagonal boron nitride. This potential must
|
||||
be used in combination with the potential "pair_style pair_ilp_gr_hBN"_pair_ilp_gr_hBN.html
|
||||
|
||||
NOTE: This potential is intended for electrostatic interactions between two different
|
||||
layers of hexagonal boron nitride. Therefore, to avoid interaction within the same layers,
|
||||
each layer should have a separate molecule id and is recommended to use
|
||||
"full" atom style in the data file.
|
||||
|
||||
:c,image(Eqs/pair_coul_shield.jpg)
|
||||
|
||||
Where Tap(r_ij) is the tapper function which provides a continuous cutoff (up to third derivative)
|
||||
for interatomic separations larger than r_c "(Maaravi)"_#Maaravi. Here \lambda is the shielding
|
||||
parameter that eliminates the short-range sigularity of the classical monopolar electrostatic
|
||||
interaction expression "(Maaravi)"_#Maaravi.
|
||||
|
||||
The shielding parameter \lambda (1/distance units) must be defined for each pair of atom
|
||||
types via the "pair_coeff"_pair_coeff.html command as in the example
|
||||
above, or in the data file or restart files read by the
|
||||
"read_data"_read_data.html or "read_restart"_read_restart.html
|
||||
commands:
|
||||
|
||||
The global cutoff (r_c) specified in the pair_style command is used.
|
||||
|
||||
:line
|
||||
|
||||
[Mixing, shift, table, tail correction, restart, rRESPA info]:
|
||||
|
||||
This pair style does not support parameter mixing. Coefficients must be given explicitly for each type of particle pairs.
|
||||
|
||||
The "pair_modify"_pair_modify.html table option is not relevant
|
||||
for this pair style.
|
||||
|
||||
This pair style does not support the "pair_modify"_pair_modify.html
|
||||
tail option for adding long-range tail corrections to energy and
|
||||
pressure.
|
||||
|
||||
This pair style can only be used via the {pair} keyword of the
|
||||
"run_style respa"_run_style.html command. It does not support the
|
||||
{inner}, {middle}, {outer} keywords.
|
||||
|
||||
[Restrictions:]
|
||||
|
||||
This style is part of the "user-misc" package. It is only enabled
|
||||
if LAMMPS was built with that package. See the "Making
|
||||
LAMMPS"_Section_start.html#start_2_3 section for more info.
|
||||
|
||||
[Related commands:]
|
||||
|
||||
"pair_coeff"_pair_coeff.html
|
||||
"pair_style pair_ilp_gr_hBN"_pair_ilp_gr_hBN.html
|
||||
|
||||
[Default:] tap_flag = 1
|
||||
|
||||
:line
|
||||
|
||||
:link(Maaravi)
|
||||
[(Maaravi)] T. Maaravi et al, J. Phys. Chem. C 121, 22826-22835 (2017).
|
||||
106
doc/src/pair_ilp_gr_hBN.txt
Normal file
106
doc/src/pair_ilp_gr_hBN.txt
Normal file
@ -0,0 +1,106 @@
|
||||
"LAMMPS WWW Site"_lws - "LAMMPS Documentation"_ld - "LAMMPS Commands"_lc :c
|
||||
|
||||
:link(lws,http://lammps.sandia.gov)
|
||||
:link(ld,Manual.html)
|
||||
:link(lc,Section_commands.html#comm)
|
||||
|
||||
:line
|
||||
|
||||
pair_style ILP/graphene/hBN command :h3
|
||||
|
||||
[Syntax:]
|
||||
|
||||
pair_style hybrid/overlay ILP/graphene/hBN cutoff tap_flag :pre
|
||||
|
||||
cutoff = global cutoff (distance units)
|
||||
tap_flag = 0/1 to turn off/on the Tapper function
|
||||
|
||||
[Examples:]
|
||||
|
||||
pair_style hybrid/overlay ILP/graphene/hBN 16.0 1
|
||||
pair_coeff * * ILP/graphene/hBN BNCH.ILP B N C :pre
|
||||
|
||||
pair_style hybrid/overlay rebo tersoff ILP/graphene/hBN 16.0 coul/shield 16.0
|
||||
pair_coeff * * rebo CH.airebo NULL NULL C
|
||||
pair_coeff * * tersoff BNC.tersoff B N NULL
|
||||
pair_coeff * * ILP/graphene/hBN BNCH.ILP B N C :pre
|
||||
pair_coeff 1 1 coul/shield 0.70
|
||||
pair_coeff 1 2 coul/shield 0.69498201415576216335
|
||||
pair_coeff 2 2 coul/shield 0.69
|
||||
|
||||
[Description:]
|
||||
|
||||
The {ILP/graphene/hBN} style computes the registry-dependent interlayer potential (RDILP)
|
||||
potential as described in "(Leven)"_#Leven and "(Maaravi)"_#Maaravi. The normals are calculated in
|
||||
the way as described in "(Kolmogorov)"_#Kolmogorov.
|
||||
|
||||
:c,image(Eqs/pair_ilp_gr_hBN.jpg)
|
||||
|
||||
Where Tap(r_ij) is the tapper function which provides a continuous cutoff (up to third derivative)
|
||||
for interatomic separations larger than r_c "(Maaravi)"_#Maaravi. The definitons of each parameter
|
||||
in the above equation can be found in "(Leven)"_#Leven and "(Maaravi)"_#Maaravi.
|
||||
|
||||
It is important to include all the pairs to build the neighbor list for
|
||||
calculating the normals.
|
||||
|
||||
NOTE: This potential is intended for interactions between two different layers
|
||||
of graphene or hexagonal boron nitride. Therefore, to avoid interaction within
|
||||
the same layers, each layer should have a separate molecule id and is recommended
|
||||
to use "full" atom style in the data file.
|
||||
|
||||
The parameter file (e.g. BNCH.ILP), is intended for use with metal
|
||||
"units"_units.html, with energies in meV. Two additional parameters, {S},
|
||||
and {rcut} are included in the parameter file. {S} is designed to
|
||||
facilitate scaling of energies. {rcut} is designed to build the neighbor
|
||||
list for calculating the normals for each atom pair.
|
||||
|
||||
NOTE: The parameters presented in the parameter file (e.g. BNCH.ILP), are
|
||||
fitted with tapper function by setting the cutoff equal to 16.0 Angstrom.
|
||||
Using different cutoff or tapper function should be careful.
|
||||
|
||||
This potential must be used in combination with hybrid/overlay.
|
||||
Other interactions can be set to zero using pair_style {none}.
|
||||
|
||||
:line
|
||||
|
||||
[Mixing, shift, table, tail correction, restart, rRESPA info]:
|
||||
|
||||
This pair style does not support the pair_modify mix, shift, table, and tail options.
|
||||
|
||||
This pair style does not write their information to binary restart files,
|
||||
since it is stored in potential files. Thus, you need to re-specify the
|
||||
pair_style and pair_coeff commands in an input script that reads a restart file.
|
||||
|
||||
[Restrictions:]
|
||||
|
||||
This fix is part of the USER-MISC package. It is only enabled if
|
||||
LAMMPS was built with that package. See the "Making
|
||||
LAMMPS"_Section_start.html#start_3 section for more info.
|
||||
|
||||
This pair potential requires the newton setting to be “on” for pair interactions.
|
||||
|
||||
The BNCH.ILP potential file provided with LAMMPS (see the potentials directory) are parameterized for metal units.
|
||||
You can use this potential with any LAMMPS units, but you would need to create your BNCH.ILP potential file with
|
||||
coefficients listed in the appropriate units, if your simulation doesn’t use “metal” units.
|
||||
|
||||
[Related commands:]
|
||||
|
||||
"pair_coeff"_pair_coeff.html
|
||||
"pair_none"_pair_none.html
|
||||
"pair_style hybrid/overlay"_pair_hybrid.html
|
||||
"pair_style pair_kolmogorov_crespi_z"_pair_kolmogorov_crespi_z.html
|
||||
"pair_style pair_kolmogorov_crespi_full"_pair_kolmogorov_crespi_full.html
|
||||
"pair_style pair_coul_shield"_pair_coul_shield.html
|
||||
|
||||
[Default:] tap_flag = 1
|
||||
|
||||
:line
|
||||
|
||||
:link(Leven)
|
||||
[(Leven)] I. Leven et al, J. Chem.Theory Comput. 12, 2896-905 (2016)
|
||||
|
||||
:link(Maaravi)
|
||||
[(Maaravi)] T. Maaravi et al, J. Phys. Chem. C 121, 22826-22835 (2017).
|
||||
|
||||
:link(Kolmogorov)
|
||||
[(Kolmogorov)] A. N. Kolmogorov, V. H. Crespi, Phys. Rev. B 71, 235415 (2005)
|
||||
90
doc/src/pair_kolmogorov_crespi_full.txt
Normal file
90
doc/src/pair_kolmogorov_crespi_full.txt
Normal file
@ -0,0 +1,90 @@
|
||||
"LAMMPS WWW Site"_lws - "LAMMPS Documentation"_ld - "LAMMPS Commands"_lc :c
|
||||
|
||||
:link(lws,http://lammps.sandia.gov)
|
||||
:link(ld,Manual.html)
|
||||
:link(lc,Section_commands.html#comm)
|
||||
|
||||
:line
|
||||
|
||||
pair_style kolmogorov/crespi/full command :h3
|
||||
|
||||
[Syntax:]
|
||||
|
||||
pair_style hybrid/overlay kolmogorov/crespi/full cutoff tap_flag:pre
|
||||
|
||||
cutoff = global cutoff (distance units)
|
||||
tap_flag = 0/1 to turn off/on the Tapper function
|
||||
|
||||
[Examples:]
|
||||
|
||||
pair_style hybrid/overlay kolmogorov/crespi/full 20.0 0
|
||||
pair_coeff * * none
|
||||
pair_coeff * * kolmogorov/crespi/full CC.KC C C :pre
|
||||
|
||||
pair_style hybrid/overlay rebo kolmogorov/crespi/full 16.0
|
||||
pair_coeff * * rebo CH.airebo C C
|
||||
pair_coeff * * kolmogorov/crespi/full CC.KC C C :pre
|
||||
|
||||
[Description:]
|
||||
|
||||
The {kolmogorov/crespi/full} style computes the Kolmogorov-Crespi interaction
|
||||
potential as described in "(Kolmogorov)"_#Kolmogorov. No simplification is made,
|
||||
|
||||
:c,image(Eqs/pair_kolmogorov_crespi_full.jpg)
|
||||
|
||||
It is important to have a suffiently large cutoff to ensure smooth forces and
|
||||
to include all the pairs to build the neighbor list for calculating the normals.
|
||||
Energies are shifted so that they go continously to zero at the cutoff assuming
|
||||
that the exponential part of {Vij} (first term) decays sufficiently fast.
|
||||
This shift is achieved by the last term in the equation for {Vij} above.
|
||||
|
||||
NOTE: This potential is intended for interactions between two different graphene
|
||||
layers. Therefore, to avoid interaction within the same layers, each layer
|
||||
should have a separate molecule id and is recommended to use
|
||||
"full" atom style in the data file.
|
||||
|
||||
The parameter file (e.g. CC.KC), is intended for use with metal
|
||||
"units"_units.html, with energies in meV. Two additional parameters, {S},
|
||||
and {rcut} are included in the parameter file. {S} is designed to
|
||||
facilitate scaling of energies. {rcut} is designed to build the neighbor
|
||||
list for calculating the normals for each atom pair.
|
||||
|
||||
This potential must be used in combination with hybrid/overlay.
|
||||
Other interactions can be set to zero using pair_style {none}.
|
||||
|
||||
:line
|
||||
|
||||
[Mixing, shift, table, tail correction, restart, rRESPA info]:
|
||||
|
||||
This pair style does not support the pair_modify mix, shift, table, and tail options.
|
||||
|
||||
This pair style does not write their information to binary restart files,
|
||||
since it is stored in potential files. Thus, you need to re-specify the
|
||||
pair_style and pair_coeff commands in an input script that reads a restart file.
|
||||
|
||||
[Restrictions:]
|
||||
|
||||
This fix is part of the USER-MISC package. It is only enabled if
|
||||
LAMMPS was built with that package. See the "Making
|
||||
LAMMPS"_Section_start.html#start_3 section for more info.
|
||||
|
||||
This pair potential requires the newton setting to be “on” for pair interactions.
|
||||
|
||||
The CC.KC potential file provided with LAMMPS (see the potentials directory) are parameterized for metal units.
|
||||
You can use this potential with any LAMMPS units, but you would need to create your CC.KC potential file with
|
||||
coefficients listed in the appropriate units, if your simulation doesn’t use “metal” units.
|
||||
|
||||
[Related commands:]
|
||||
|
||||
"pair_coeff"_pair_coeff.html
|
||||
"pair_none"_pair_none.html
|
||||
"pair_style hybrid/overlay"_pair_hybrid.html
|
||||
"pair_style pair_kolmogorov_crespi_z"_pair_kolmogorov_crespi_z.html
|
||||
"pair_style pair_ilp_gr_hBN"_pair_ilp_gr_hBN.html
|
||||
|
||||
[Default:] tap_flag = 0
|
||||
|
||||
:line
|
||||
|
||||
:link(Kolmogorov)
|
||||
[(Kolmogorov)] A. N. Kolmogorov, V. H. Crespi, Phys. Rev. B 71, 235415 (2005)
|
||||
370
src/USER-MISC/pair_coul_shield.cpp
Normal file
370
src/USER-MISC/pair_coul_shield.cpp
Normal file
@ -0,0 +1,370 @@
|
||||
/* ----------------------------------------------------------------------
|
||||
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
||||
http://lammps.sandia.gov, Sandia National Laboratories
|
||||
Steve Plimpton, sjplimp@sandia.gov
|
||||
|
||||
Copyright (2003) Sandia Corporation. Under the terms of Contract
|
||||
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
|
||||
certain rights in this software. This software is distributed under
|
||||
the GNU General Public License.
|
||||
|
||||
See the README file in the top-level LAMMPS directory.
|
||||
------------------------------------------------------------------------- */
|
||||
/* ----------------------------------------------------------------------
|
||||
Contributing author: Wengen Ouyang (Tel Aviv University)
|
||||
e-mail: w.g.ouyang at gmail dot com
|
||||
|
||||
This is a Coulomb potential described in
|
||||
[Maaravi et al, J. Phys. Chem. C 121, 22826-22835 (2017)]
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
#include <math.h>
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include "pair_coul_shield.h"
|
||||
#include "atom.h"
|
||||
#include "comm.h"
|
||||
#include "force.h"
|
||||
#include "neighbor.h"
|
||||
#include "neigh_list.h"
|
||||
#include "memory.h"
|
||||
#include "error.h"
|
||||
|
||||
using namespace LAMMPS_NS;
|
||||
|
||||
/* ---------------------------------------------------------------------- */
|
||||
|
||||
PairCoulshield::PairCoulshield(LAMMPS *lmp) : Pair(lmp) {
|
||||
Tapflag = 1;
|
||||
}
|
||||
|
||||
/* ---------------------------------------------------------------------- */
|
||||
|
||||
PairCoulshield::~PairCoulshield()
|
||||
{
|
||||
if (allocated) {
|
||||
memory->destroy(setflag);
|
||||
memory->destroy(sigmae);
|
||||
memory->destroy(offset);
|
||||
memory->destroy(cutsq);
|
||||
memory->destroy(cut);
|
||||
allocated = 0;
|
||||
}
|
||||
}
|
||||
|
||||
/* ---------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::compute(int eflag, int vflag)
|
||||
{
|
||||
int i,j,ii,jj,inum,jnum,itype,jtype;
|
||||
double qtmp,xtmp,ytmp,ztmp,delx,dely,delz,ecoul,fpair,Tap,dTap;
|
||||
double rsq,r,r3,rarg,th,depsdr,epsr,forcecoul,factor_coul,Vc,fvc;
|
||||
int *ilist,*jlist,*numneigh,**firstneigh;
|
||||
|
||||
ecoul = 0.0;
|
||||
if (eflag || vflag) ev_setup(eflag,vflag);
|
||||
else evflag = vflag_fdotr = 0;
|
||||
|
||||
double **x = atom->x;
|
||||
double **f = atom->f;
|
||||
double *q = atom->q;
|
||||
int *type = atom->type;
|
||||
int nlocal = atom->nlocal;
|
||||
double *special_coul = force->special_coul;
|
||||
int newton_pair = force->newton_pair;
|
||||
double qqrd2e = force->qqrd2e;
|
||||
|
||||
inum = list->inum;
|
||||
ilist = list->ilist;
|
||||
numneigh = list->numneigh;
|
||||
firstneigh = list->firstneigh;
|
||||
|
||||
// loop over neighbors of my atoms
|
||||
|
||||
for (ii = 0; ii < inum; ii++) {
|
||||
i = ilist[ii];
|
||||
qtmp = q[i];
|
||||
xtmp = x[i][0];
|
||||
ytmp = x[i][1];
|
||||
ztmp = x[i][2];
|
||||
itype = type[i];
|
||||
jlist = firstneigh[i];
|
||||
jnum = numneigh[i];
|
||||
|
||||
for (jj = 0; jj < jnum; jj++) {
|
||||
j = jlist[jj];
|
||||
|
||||
factor_coul = special_coul[sbmask(j)];
|
||||
j &= NEIGHMASK;
|
||||
|
||||
delx = xtmp - x[j][0];
|
||||
dely = ytmp - x[j][1];
|
||||
delz = ztmp - x[j][2];
|
||||
rsq = delx*delx + dely*dely + delz*delz;
|
||||
jtype = type[j];
|
||||
|
||||
// only include the interation between different layers
|
||||
if (rsq < cutsq[itype][jtype] && atom->molecule[i] != atom->molecule[j]) {
|
||||
r = sqrt(rsq);
|
||||
r3 = rsq*r;
|
||||
rarg = 1.0/sigmae[itype][jtype];
|
||||
th = r3 + pow(rarg,3.0);
|
||||
epsr = 1.0/pow(th,0.333333333333333333333333);
|
||||
depsdr = pow(epsr,4.0);
|
||||
Vc = qqrd2e*qtmp*q[j]*epsr;
|
||||
|
||||
// turn on/off Tapper function
|
||||
if (Tapflag) {
|
||||
Tap = calc_Tap(r,sqrt(cutsq[itype][jtype]));
|
||||
dTap = calc_dTap(r,sqrt(cutsq[itype][jtype]));
|
||||
} else {Tap = 1.0; dTap = 0.0;}
|
||||
|
||||
forcecoul = qqrd2e*qtmp*q[j]*r*depsdr;
|
||||
fvc = forcecoul*Tap - Vc*dTap/r;
|
||||
fpair = factor_coul*fvc;
|
||||
|
||||
f[i][0] += delx*fpair;
|
||||
f[i][1] += dely*fpair;
|
||||
f[i][2] += delz*fpair;
|
||||
if (newton_pair || j < nlocal) {
|
||||
f[j][0] -= delx*fpair;
|
||||
f[j][1] -= dely*fpair;
|
||||
f[j][2] -= delz*fpair;
|
||||
}
|
||||
|
||||
if (eflag) {
|
||||
if (Tapflag) ecoul = Vc*Tap;
|
||||
else ecoul = Vc - offset[itype][jtype];
|
||||
ecoul *= factor_coul;
|
||||
}
|
||||
|
||||
if (evflag) ev_tally(i,j,nlocal,newton_pair,0.0,
|
||||
ecoul,fpair,delx,dely,delz);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (vflag_fdotr) virial_fdotr_compute();
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
allocate all arrays
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::allocate()
|
||||
{
|
||||
allocated = 1;
|
||||
int n = atom->ntypes;
|
||||
|
||||
memory->create(setflag,n+1,n+1,"pair:setflag");
|
||||
for (int i = 1; i <= n; i++)
|
||||
for (int j = i; j <= n; j++)
|
||||
setflag[i][j] = 0;
|
||||
|
||||
memory->create(cutsq,n+1,n+1,"pair:cutsq");
|
||||
memory->create(cut,n+1,n+1,"pair:cut");
|
||||
memory->create(sigmae,n+1,n+1,"pair:sigmae");
|
||||
//memory->create(rme,n+1,n+1,"pair:rme");
|
||||
memory->create(offset,n+1,n+1,"pair:offset");
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
global settings
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::settings(int narg, char **arg)
|
||||
{
|
||||
if (narg < 1 || narg > 2) error->all(FLERR,"Illegal pair_style command");
|
||||
|
||||
cut_global = force->numeric(FLERR,arg[0]);
|
||||
if (narg == 2) Tapflag = force->numeric(FLERR,arg[1]);
|
||||
|
||||
// reset cutoffs that have been explicitly set
|
||||
|
||||
if (allocated) {
|
||||
int i,j;
|
||||
for (i = 1; i <= atom->ntypes; i++)
|
||||
for (j = i+1; j <= atom->ntypes; j++)
|
||||
if (setflag[i][j]) cut[i][j] = cut_global;
|
||||
}
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
set coeffs for one or more type pairs
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::coeff(int narg, char **arg)
|
||||
{
|
||||
if (narg < 3 || narg > 4) error->all(FLERR,"Incorrect args for pair coefficients");
|
||||
if (!allocated) allocate();
|
||||
|
||||
int ilo,ihi,jlo,jhi;
|
||||
force->bounds(FLERR,arg[0],atom->ntypes,ilo,ihi);
|
||||
force->bounds(FLERR,arg[1],atom->ntypes,jlo,jhi);
|
||||
|
||||
double sigmae_one = force->numeric(FLERR,arg[2]);
|
||||
|
||||
double cut_one = cut_global;
|
||||
if (narg == 4) cut_one = force->numeric(FLERR,arg[3]);
|
||||
|
||||
int count = 0;
|
||||
for (int i = ilo; i <= ihi; i++) {
|
||||
for (int j = MAX(jlo,i); j <= jhi; j++) {
|
||||
sigmae[i][j] = sigmae_one;
|
||||
cut[i][j] = cut_one;
|
||||
setflag[i][j] = 1;
|
||||
count++;
|
||||
}
|
||||
}
|
||||
|
||||
if (count == 0) error->all(FLERR,"Incorrect args for pair coefficients");
|
||||
}
|
||||
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
init specific to this pair style
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::init_style()
|
||||
{
|
||||
if (!atom->q_flag)
|
||||
error->all(FLERR,"Pair style coul/shield requires atom attribute q");
|
||||
|
||||
neighbor->request(this,instance_me);
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
init for one type pair i,j and corresponding j,i
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
double PairCoulshield::init_one(int i, int j)
|
||||
{
|
||||
if (setflag[i][j] == 0) {
|
||||
error->all(FLERR,"for pair style coul/shield, parameters need to be set explicitly for all pairs.");
|
||||
}
|
||||
|
||||
double *q = atom->q;
|
||||
double qqrd2e = force->qqrd2e;
|
||||
double r,r3,rarg,th,epsr;
|
||||
|
||||
if (offset_flag) {
|
||||
r = cut[i][j];
|
||||
r3 = r*r*r;
|
||||
rarg = 1.0/sigmae[i][j];
|
||||
th = r3 + pow(rarg,3.0);
|
||||
epsr = 1.0/pow(th,0.333333333333333333);
|
||||
offset[i][j] = qqrd2e*q[i]*q[j]*epsr;
|
||||
} else offset[i][j] = 0.0;
|
||||
|
||||
|
||||
sigmae[j][i] = sigmae[i][j];
|
||||
offset[j][i] = offset[i][j];
|
||||
cut[j][i] = cut[i][j];
|
||||
|
||||
return cut[i][j];
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
proc 0 writes to restart file
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::write_restart(FILE *fp)
|
||||
{
|
||||
write_restart_settings(fp);
|
||||
|
||||
int i,j;
|
||||
for (i = 1; i <= atom->ntypes; i++)
|
||||
for (j = i; j <= atom->ntypes; j++) {
|
||||
fwrite(&setflag[i][j],sizeof(int),1,fp);
|
||||
if (setflag[i][j]) {
|
||||
fwrite(&sigmae[i][j],sizeof(double),1,fp);
|
||||
fwrite(&cut[i][j],sizeof(double),1,fp);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
proc 0 reads from restart file, bcasts
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::read_restart(FILE *fp)
|
||||
{
|
||||
read_restart_settings(fp);
|
||||
allocate();
|
||||
|
||||
int i,j;
|
||||
int me = comm->me;
|
||||
for (i = 1; i <= atom->ntypes; i++)
|
||||
for (j = i; j <= atom->ntypes; j++) {
|
||||
if (setflag[i][j]) {
|
||||
if (me == 0) {
|
||||
fread(&sigmae[i][j],sizeof(double),1,fp);
|
||||
fread(&cut[i][j],sizeof(double),1,fp);
|
||||
}
|
||||
MPI_Bcast(&sigmae[i][j],1,MPI_DOUBLE,0,world);
|
||||
MPI_Bcast(&cut[i][j],1,MPI_DOUBLE,0,world);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
proc 0 writes to restart file
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::write_restart_settings(FILE *fp)
|
||||
{
|
||||
fwrite(&cut_global,sizeof(double),1,fp);
|
||||
fwrite(&offset_flag,sizeof(int),1,fp);
|
||||
fwrite(&mix_flag,sizeof(int),1,fp);
|
||||
}
|
||||
|
||||
/* ----------------------------------------------------------------------
|
||||
proc 0 reads from restart file, bcasts
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
void PairCoulshield::read_restart_settings(FILE *fp)
|
||||
{
|
||||
if (comm->me == 0) {
|
||||
fread(&cut_global,sizeof(double),1,fp);
|
||||
fread(&offset_flag,sizeof(int),1,fp);
|
||||
fread(&mix_flag,sizeof(int),1,fp);
|
||||
}
|
||||
MPI_Bcast(&cut_global,1,MPI_DOUBLE,0,world);
|
||||
MPI_Bcast(&offset_flag,1,MPI_INT,0,world);
|
||||
MPI_Bcast(&mix_flag,1,MPI_INT,0,world);
|
||||
}
|
||||
|
||||
/* ---------------------------------------------------------------------- */
|
||||
|
||||
double PairCoulshield::single(int i, int j, int itype, int jtype,
|
||||
double rsq, double factor_coul, double factor_lj,
|
||||
double &fforce)
|
||||
{
|
||||
double r, rarg,Vc,fvc,forcecoul,phishieldec;
|
||||
double r3,th,epsr,depsdr,Tap,dTap;
|
||||
double *q = atom->q;
|
||||
double qqrd2e = force->qqrd2e;
|
||||
|
||||
r = sqrt(rsq);
|
||||
r3 = rsq*r;
|
||||
rarg = 1.0/sigmae[itype][jtype];
|
||||
th = r3 + pow(rarg,3.0);
|
||||
epsr = 1.0/pow(th,0.333333333333333333);
|
||||
depsdr = pow(epsr,4.0);
|
||||
Vc = qqrd2e*q[i]*q[j]*epsr;
|
||||
|
||||
// turn on/off Tapper function
|
||||
if (Tapflag) {
|
||||
Tap = calc_Tap(r,sqrt(cutsq[itype][jtype]));
|
||||
dTap = calc_dTap(r,sqrt(cutsq[itype][jtype]));
|
||||
} else {Tap = 1.0; dTap = 0.0;}
|
||||
|
||||
forcecoul = qqrd2e*q[i]*q[j]*r*depsdr;
|
||||
fvc = forcecoul*Tap - Vc*dTap/r;
|
||||
fforce = factor_coul*fvc;
|
||||
|
||||
if (Tapflag) phishieldec = factor_coul*Vc*Tap;
|
||||
else phishieldec = Vc - offset[itype][jtype];
|
||||
return factor_coul*phishieldec;
|
||||
}
|
||||
97
src/USER-MISC/pair_coul_shield.h
Normal file
97
src/USER-MISC/pair_coul_shield.h
Normal file
@ -0,0 +1,97 @@
|
||||
/* -*- c++ -*- ----------------------------------------------------------
|
||||
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
||||
http://lammps.sandia.gov, Sandia National Laboratories
|
||||
Steve Plimpton, sjplimp@sandia.gov
|
||||
|
||||
Copyright (2003) Sandia Corporation. Under the terms of Contract
|
||||
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
|
||||
certain rights in this software. This software is distributed under
|
||||
the GNU General Public License.
|
||||
|
||||
See the README file in the top-level LAMMPS directory.
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
#ifdef PAIR_CLASS
|
||||
|
||||
PairStyle(coul/shield,PairCoulshield)
|
||||
|
||||
#else
|
||||
|
||||
#ifndef LMP_PAIR_COUL_SHIELD_H
|
||||
#define LMP_PAIR_COUL_SHIELD_H
|
||||
|
||||
#include "pair.h"
|
||||
|
||||
namespace LAMMPS_NS {
|
||||
|
||||
class PairCoulshield : public Pair {
|
||||
public:
|
||||
PairCoulshield(class LAMMPS *);
|
||||
virtual ~PairCoulshield();
|
||||
|
||||
virtual void compute(int, int);
|
||||
|
||||
virtual void settings(int, char **);
|
||||
virtual void coeff(int, char **);
|
||||
virtual void init_style();
|
||||
virtual double init_one(int, int);
|
||||
virtual void write_restart(FILE *);
|
||||
virtual void read_restart(FILE *);
|
||||
virtual void write_restart_settings(FILE *);
|
||||
virtual void read_restart_settings(FILE *);
|
||||
|
||||
virtual double single(int, int, int, int, double, double, double, double &);
|
||||
|
||||
protected:
|
||||
double cut_global;
|
||||
double **cut;
|
||||
double **sigmae, **offset;
|
||||
//double a_eps, b_eps, eps_s;
|
||||
int Tapflag;
|
||||
|
||||
void allocate();
|
||||
|
||||
/* ----Calculate the long-range cutoff term */
|
||||
inline double calc_Tap(double r_ij, double Rcut) {
|
||||
double Tap,r;
|
||||
//int Tap_coeff[8] = {1,0,0,0,-35,84,-70,20};
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
Tap = 0.0;
|
||||
|
||||
Tap = Tap_coeff[7] * r + Tap_coeff[6];
|
||||
Tap = Tap * r + Tap_coeff[5];
|
||||
Tap = Tap * r + Tap_coeff[4];
|
||||
Tap = Tap * r + Tap_coeff[3];
|
||||
Tap = Tap * r + Tap_coeff[2];
|
||||
Tap = Tap * r + Tap_coeff[1];
|
||||
Tap = Tap * r + Tap_coeff[0];
|
||||
|
||||
return(Tap);
|
||||
}
|
||||
|
||||
/* ----Calculate the derivatives of long-range cutoff term */
|
||||
inline double calc_dTap(double r_ij, double Rcut) {
|
||||
double dTap,r;
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
dTap = 0.0;
|
||||
|
||||
dTap = 7.0*Tap_coeff[7] * r + 6.0*Tap_coeff[6];
|
||||
dTap = dTap * r + 5.0*Tap_coeff[5];
|
||||
dTap = dTap * r + 4.0*Tap_coeff[4];
|
||||
dTap = dTap * r + 3.0*Tap_coeff[3];
|
||||
dTap = dTap * r + 2.0*Tap_coeff[2];
|
||||
dTap = dTap * r + Tap_coeff[1];
|
||||
dTap = dTap/Rcut;
|
||||
|
||||
return(dTap);
|
||||
}
|
||||
};
|
||||
|
||||
}
|
||||
|
||||
#endif
|
||||
#endif
|
||||
1117
src/USER-MISC/pair_ilp_gr_hBN.cpp
Normal file
1117
src/USER-MISC/pair_ilp_gr_hBN.cpp
Normal file
File diff suppressed because it is too large
Load Diff
143
src/USER-MISC/pair_ilp_gr_hBN.h
Normal file
143
src/USER-MISC/pair_ilp_gr_hBN.h
Normal file
@ -0,0 +1,143 @@
|
||||
/* ----------------------------------------------------------------------
|
||||
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
||||
http://lammps.sandia.gov, Sandia National Laboratories
|
||||
Steve Plimpton, sjplimp@sandia.gov
|
||||
|
||||
Copyright (2003) Sandia Corporation. Under the terms of Contract
|
||||
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
|
||||
certain rights in this software. This software is distributed under
|
||||
the GNU General Public License.
|
||||
|
||||
See the README file in the top-level LAMMPS directory.
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
#ifdef PAIR_CLASS
|
||||
|
||||
PairStyle(ILP/graphene/hBN,PairILP)
|
||||
|
||||
#else
|
||||
|
||||
#ifndef LMP_PAIR_ILP_H
|
||||
#define LMP_PAIR_ILP_H
|
||||
|
||||
#include "pair.h"
|
||||
#include "my_page.h"
|
||||
#include <math.h>
|
||||
|
||||
namespace LAMMPS_NS {
|
||||
|
||||
class PairILP : public Pair {
|
||||
public:
|
||||
PairILP(class LAMMPS *);
|
||||
virtual ~PairILP();
|
||||
|
||||
virtual void compute(int, int);
|
||||
void settings(int, char **);
|
||||
void coeff(int, char **);
|
||||
double init_one(int, int);
|
||||
void init_style();
|
||||
void calc_normal();
|
||||
int pack_forward_comm(int, int *, double *, int, int *);
|
||||
void unpack_forward_comm(int, int, double *);
|
||||
double single(int, int, int, int, double, double, double, double &);
|
||||
|
||||
protected:
|
||||
int me;
|
||||
int maxlocal; // size of numneigh, firstneigh arrays
|
||||
int pgsize; // size of neighbor page
|
||||
int oneatom; // max # of neighbors for one atom
|
||||
MyPage<int> *ipage; // neighbor list pages
|
||||
int *ILP_numneigh; // # of pair neighbors for each atom
|
||||
int **ILP_firstneigh; // ptr to 1st neighbor of each atom
|
||||
int tap_flag; // flag to turn on/off Tapper function
|
||||
|
||||
struct Param {
|
||||
double z0,alpha,epsilon,C,delta,d,sR,reff,C6,S;
|
||||
double delta2inv,seff,lambda,rcut;
|
||||
int ielement,jelement;
|
||||
};
|
||||
Param *params; // parameter set for I-J interactions
|
||||
char **elements; // names of unique elements
|
||||
int **elem2param; // mapping from element pairs to parameters
|
||||
int *map; // mapping from atom types to elements
|
||||
int nelements; // # of unique elements
|
||||
int nparams; // # of stored parameter sets
|
||||
int maxparam; // max # of parameter sets
|
||||
int nmax; // max # of atoms
|
||||
|
||||
double cut_global;
|
||||
double cut_normal;
|
||||
double **cut;
|
||||
double **cutILPsq; // mapping the cutoff from element pairs to parameters
|
||||
double **offset;
|
||||
double **normal;
|
||||
double ***dnormdri;
|
||||
double ****dnormal;
|
||||
|
||||
void read_file( char * );
|
||||
void allocate();
|
||||
void ILP_neigh();
|
||||
|
||||
/* ----Calculate the long-range cutoff term */
|
||||
inline double calc_Tap(double r_ij, double Rcut) {
|
||||
double Tap,r;
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
if(r >= 1.0) {Tap = 0.0;}
|
||||
else {
|
||||
Tap = Tap_coeff[7] * r + Tap_coeff[6];
|
||||
Tap = Tap * r + Tap_coeff[5];
|
||||
Tap = Tap * r + Tap_coeff[4];
|
||||
Tap = Tap * r + Tap_coeff[3];
|
||||
Tap = Tap * r + Tap_coeff[2];
|
||||
Tap = Tap * r + Tap_coeff[1];
|
||||
Tap = Tap * r + Tap_coeff[0];
|
||||
}
|
||||
return(Tap);
|
||||
}
|
||||
|
||||
/* ----Calculate the derivatives of long-range cutoff term */
|
||||
inline double calc_dTap(double r_ij, double Rcut) {
|
||||
double dTap,r;
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
if(r >= 1.0) {dTap = 0.0;}
|
||||
else {
|
||||
dTap = 7.0*Tap_coeff[7] * r + 6.0*Tap_coeff[6];
|
||||
dTap = dTap * r + 5.0*Tap_coeff[5];
|
||||
dTap = dTap * r + 4.0*Tap_coeff[4];
|
||||
dTap = dTap * r + 3.0*Tap_coeff[3];
|
||||
dTap = dTap * r + 2.0*Tap_coeff[2];
|
||||
dTap = dTap * r + Tap_coeff[1];
|
||||
dTap = dTap/Rcut;
|
||||
}
|
||||
return(dTap);
|
||||
}
|
||||
};
|
||||
|
||||
}
|
||||
|
||||
#endif
|
||||
#endif
|
||||
|
||||
/* ERROR/WARNING messages:
|
||||
|
||||
E: Illegal ... command
|
||||
|
||||
Self-explanatory. Check the input script syntax and compare to the
|
||||
documentation for the command. You can use -echo screen as a
|
||||
command-line option when running LAMMPS to see the offending line.
|
||||
|
||||
E: Incorrect args for pair coefficients
|
||||
|
||||
Self-explanatory. Check the input script or data file.
|
||||
|
||||
E: All pair coeffs are not set
|
||||
|
||||
All pair coefficients must be set in the data file or by the
|
||||
pair_coeff command before running a simulation.
|
||||
|
||||
*/
|
||||
|
||||
1121
src/USER-MISC/pair_kolmogorov_crespi_full.cpp
Normal file
1121
src/USER-MISC/pair_kolmogorov_crespi_full.cpp
Normal file
File diff suppressed because it is too large
Load Diff
147
src/USER-MISC/pair_kolmogorov_crespi_full.h
Normal file
147
src/USER-MISC/pair_kolmogorov_crespi_full.h
Normal file
@ -0,0 +1,147 @@
|
||||
/* ----------------------------------------------------------------------
|
||||
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
||||
http://lammps.sandia.gov, Sandia National Laboratories
|
||||
Steve Plimpton, sjplimp@sandia.gov
|
||||
|
||||
Copyright (2003) Sandia Corporation. Under the terms of Contract
|
||||
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
|
||||
certain rights in this software. This software is distributed under
|
||||
the GNU General Public License.
|
||||
|
||||
See the README file in the top-level LAMMPS directory.
|
||||
------------------------------------------------------------------------- */
|
||||
|
||||
#ifdef PAIR_CLASS
|
||||
|
||||
PairStyle(kolmogorov/crespi/full,PairKolmogorovCrespiFull)
|
||||
|
||||
#else
|
||||
|
||||
#ifndef LMP_PAIR_KolmogorovCrespi_FULL_H
|
||||
#define LMP_PAIR_KolmogorovCrespi_FULL_H
|
||||
|
||||
#include "pair.h"
|
||||
#include "my_page.h"
|
||||
#include <math.h>
|
||||
|
||||
namespace LAMMPS_NS {
|
||||
|
||||
class PairKolmogorovCrespiFull : public Pair {
|
||||
public:
|
||||
PairKolmogorovCrespiFull(class LAMMPS *);
|
||||
virtual ~PairKolmogorovCrespiFull();
|
||||
|
||||
virtual void compute(int, int);
|
||||
void settings(int, char **);
|
||||
void coeff(int, char **);
|
||||
double init_one(int, int);
|
||||
void init_style();
|
||||
void calc_normal();
|
||||
int pack_forward_comm(int, int *, double *, int, int *);
|
||||
void unpack_forward_comm(int, int, double *);
|
||||
double single(int, int, int, int, double, double, double, double &);
|
||||
|
||||
protected:
|
||||
int me;
|
||||
int maxlocal; // size of numneigh, firstneigh arrays
|
||||
int pgsize; // size of neighbor page
|
||||
int oneatom; // max # of neighbors for one atom
|
||||
MyPage<int> *ipage; // neighbor list pages
|
||||
int *KC_numneigh; // # of pair neighbors for each atom
|
||||
int **KC_firstneigh; // ptr to 1st neighbor of each atom
|
||||
int tap_flag; // flag to turn on/off Tapper function
|
||||
|
||||
|
||||
struct Param {
|
||||
double z0,C0,C2,C4,C,delta,lambda,A,S;
|
||||
double delta2inv,z06,rcut;
|
||||
int ielement,jelement;
|
||||
};
|
||||
Param *params; // parameter set for I-J interactions
|
||||
char **elements; // names of unique elements
|
||||
int **elem2param; // mapping from element pairs to parameters
|
||||
int *map; // mapping from atom types to elements
|
||||
int nelements; // # of unique elements
|
||||
int nparams; // # of stored parameter sets
|
||||
int maxparam; // max # of parameter sets
|
||||
int nmax; // max # of atoms
|
||||
|
||||
double cut_global;
|
||||
double cut_normal;
|
||||
double **cut;
|
||||
double **cutKCsq;
|
||||
double **offset;
|
||||
double **normal;
|
||||
double ***dnormdri;
|
||||
double ****dnormal;
|
||||
|
||||
void read_file( char * );
|
||||
void allocate();
|
||||
void KC_neigh();
|
||||
|
||||
|
||||
/* ----Calculate the long-range cutoff term */
|
||||
inline double calc_Tap(double r_ij, double Rcut) {
|
||||
double Tap,r;
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
if(r >= 1.0) {Tap = 0.0;}
|
||||
else{
|
||||
Tap = Tap_coeff[7] * r + Tap_coeff[6];
|
||||
Tap = Tap * r + Tap_coeff[5];
|
||||
Tap = Tap * r + Tap_coeff[4];
|
||||
Tap = Tap * r + Tap_coeff[3];
|
||||
Tap = Tap * r + Tap_coeff[2];
|
||||
Tap = Tap * r + Tap_coeff[1];
|
||||
Tap = Tap * r + Tap_coeff[0];
|
||||
}
|
||||
|
||||
return(Tap);
|
||||
}
|
||||
|
||||
/* ----Calculate the derivatives of long-range cutoff term */
|
||||
inline double calc_dTap(double r_ij, double Rcut) {
|
||||
double dTap,r;
|
||||
double Tap_coeff[8] = {1.0,0.0,0.0,0.0,-35.0,84.0,-70.0,20.0};
|
||||
|
||||
r = r_ij/Rcut;
|
||||
if(r >= 1.0) {dTap = 0.0;}
|
||||
else {
|
||||
dTap = 7.0*Tap_coeff[7] * r + 6.0*Tap_coeff[6];
|
||||
dTap = dTap * r + 5.0*Tap_coeff[5];
|
||||
dTap = dTap * r + 4.0*Tap_coeff[4];
|
||||
dTap = dTap * r + 3.0*Tap_coeff[3];
|
||||
dTap = dTap * r + 2.0*Tap_coeff[2];
|
||||
dTap = dTap * r + Tap_coeff[1];
|
||||
dTap = dTap/Rcut;
|
||||
}
|
||||
|
||||
return(dTap);
|
||||
}
|
||||
};
|
||||
|
||||
}
|
||||
|
||||
#endif
|
||||
#endif
|
||||
|
||||
/* ERROR/WARNING messages:
|
||||
|
||||
E: Illegal ... command
|
||||
|
||||
Self-explanatory. Check the input script syntax and compare to the
|
||||
documentation for the command. You can use -echo screen as a
|
||||
command-line option when running LAMMPS to see the offending line.
|
||||
|
||||
E: Incorrect args for pair coefficients
|
||||
|
||||
Self-explanatory. Check the input script or data file.
|
||||
|
||||
E: All pair coeffs are not set
|
||||
|
||||
All pair coefficients must be set in the data file or by the
|
||||
pair_coeff command before running a simulation.
|
||||
|
||||
*/
|
||||
|
||||
Reference in New Issue
Block a user