Page Menu
Home
c4science
Search
Configure Global Search
Log In
Files
F91435167
fix_drude_transform.cpp
No One
Temporary
Actions
Download File
Edit File
Delete File
View Transforms
Subscribe
Mute Notifications
Award Token
Subscribers
None
File Metadata
Details
File Info
Storage
Attached
Created
Mon, Nov 11, 02:19
Size
10 KB
Mime Type
text/x-c++
Expires
Wed, Nov 13, 02:19 (2 d)
Engine
blob
Format
Raw Data
Handle
22262815
Attached To
rLAMMPS lammps
fix_drude_transform.cpp
View Options
/* ----------------------------------------------------------------------
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.
------------------------------------------------------------------------- */
/** Fix Drude Transform ******************************************************/
#include <math.h>
#include "fix_drude_transform.h"
#include "atom.h"
#include "domain.h"
#include "comm.h"
#include "error.h"
#include "modify.h"
#include "force.h"
#include <string.h>
using
namespace
LAMMPS_NS
;
using
namespace
FixConst
;
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
FixDrudeTransform
<
inverse
>::
FixDrudeTransform
(
LAMMPS
*
lmp
,
int
narg
,
char
**
arg
)
:
Fix
(
lmp
,
narg
,
arg
),
mcoeff
(
NULL
)
{
if
(
narg
!=
3
)
error
->
all
(
FLERR
,
"Illegal fix drude/transform command"
);
comm_forward
=
9
;
fix_drude
=
NULL
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
FixDrudeTransform
<
inverse
>::~
FixDrudeTransform
()
{
if
(
mcoeff
)
delete
[]
mcoeff
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
void
FixDrudeTransform
<
inverse
>::
init
()
{
int
ifix
;
for
(
ifix
=
0
;
ifix
<
modify
->
nfix
;
ifix
++
)
if
(
strcmp
(
modify
->
fix
[
ifix
]
->
style
,
"drude"
)
==
0
)
break
;
if
(
ifix
==
modify
->
nfix
)
error
->
all
(
FLERR
,
"fix drude/transform requires fix drude"
);
fix_drude
=
(
FixDrude
*
)
modify
->
fix
[
ifix
];
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
int
FixDrudeTransform
<
inverse
>::
setmask
()
{
int
mask
=
0
;
mask
|=
INITIAL_INTEGRATE
;
mask
|=
FINAL_INTEGRATE
;
return
mask
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
void
FixDrudeTransform
<
inverse
>::
setup
(
int
)
{
int
nlocal
=
atom
->
nlocal
;
int
ntypes
=
atom
->
ntypes
;
int
*
type
=
atom
->
type
;
double
*
rmass
=
atom
->
rmass
,
*
mass
=
atom
->
mass
;
tagint
*
drudeid
=
fix_drude
->
drudeid
;
int
*
drudetype
=
fix_drude
->
drudetype
;
if
(
!
rmass
)
{
if
(
!
mcoeff
)
mcoeff
=
new
double
[
ntypes
+
1
];
double
mcoeff_loc
[
ntypes
+
1
];
for
(
int
itype
=
0
;
itype
<=
ntypes
;
itype
++
)
mcoeff_loc
[
itype
]
=
2.
;
// an impossible value: mcoeff is at most 1.
for
(
int
i
=
0
;
i
<
nlocal
;
i
++
)
{
if
(
drudetype
[
type
[
i
]]
==
DRUDE_TYPE
)
{
int
j
=
atom
->
map
(
drudeid
[
i
]);
// i is drude, j is core
if
(
mcoeff_loc
[
type
[
i
]]
<
1.5
)
{
// already done
if
(
mcoeff_loc
[
type
[
j
]]
>
1.5
){
// not yet done ??
error
->
all
(
FLERR
,
"There must be one Drude type per core type"
);}
continue
;
}
mcoeff_loc
[
type
[
i
]]
=
mass
[
type
[
i
]]
/
(
mass
[
type
[
i
]]
+
mass
[
type
[
j
]]);
mcoeff_loc
[
type
[
j
]]
=
-
mass
[
type
[
i
]]
/
mass
[
type
[
j
]];
}
}
MPI_Allreduce
(
mcoeff_loc
,
mcoeff
,
ntypes
+
1
,
MPI_DOUBLE
,
MPI_MIN
,
world
);
// mcoeff is 2 for non polarizable
// 0 < mcoeff < 1 for drude
// mcoeff < 0 for core
}
}
/* ---------------------------------------------------------------------- */
namespace
LAMMPS_NS
{
// required for specialization
template
<>
void
FixDrudeTransform
<
false
>::
initial_integrate
(
int
){
comm
->
forward_comm_fix
(
this
);
real_to_reduced
();
//comm->forward_comm_fix(this); // Normally not needed
}
template
<>
void
FixDrudeTransform
<
false
>::
final_integrate
(){
comm
->
forward_comm_fix
(
this
);
real_to_reduced
();
//comm->forward_comm_fix(this); // Normally not needed
}
template
<>
void
FixDrudeTransform
<
true
>::
initial_integrate
(
int
){
comm
->
forward_comm_fix
(
this
);
reduced_to_real
();
//comm->forward_comm_fix(this); // Normally not needed
}
template
<>
void
FixDrudeTransform
<
true
>::
final_integrate
(){
comm
->
forward_comm_fix
(
this
);
reduced_to_real
();
//comm->forward_comm_fix(this); // Normally not needed
}
}
// end of namespace
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
void
FixDrudeTransform
<
inverse
>::
real_to_reduced
()
{
int
nlocal
=
atom
->
nlocal
;
int
ntypes
=
atom
->
ntypes
;
int
dim
=
domain
->
dimension
;
int
*
mask
=
atom
->
mask
,
*
type
=
atom
->
type
;
double
**
x
=
atom
->
x
,
**
v
=
atom
->
v
,
**
f
=
atom
->
f
;
double
*
rmass
=
atom
->
rmass
,
*
mass
=
atom
->
mass
;
double
mcore
,
mdrude
,
coeff
;
int
icore
,
idrude
;
tagint
*
drudeid
=
fix_drude
->
drudeid
;
int
*
drudetype
=
fix_drude
->
drudetype
;
if
(
!
rmass
)
{
// TODO: maybe drudetype can be used instead?
for
(
int
itype
=
1
;
itype
<=
ntypes
;
itype
++
)
if
(
mcoeff
[
itype
]
<
1.5
)
mass
[
itype
]
*=
1.
-
mcoeff
[
itype
];
}
for
(
int
i
=
0
;
i
<
nlocal
;
i
++
)
{
if
(
mask
[
i
]
&
groupbit
&&
drudetype
[
type
[
i
]]
!=
NOPOL_TYPE
)
{
drudeid
[
i
]
=
(
tagint
)
domain
->
closest_image
(
i
,
atom
->
map
(
drudeid
[
i
]));
}
}
for
(
int
i
=
0
;
i
<
nlocal
;
i
++
)
{
if
(
mask
[
i
]
&
groupbit
&&
drudetype
[
type
[
i
]]
!=
NOPOL_TYPE
)
{
int
j
=
(
int
)
drudeid
[
i
];
if
(
drudetype
[
type
[
i
]]
==
DRUDE_TYPE
&&
j
<
nlocal
)
continue
;
if
(
drudetype
[
type
[
i
]]
==
DRUDE_TYPE
)
{
idrude
=
i
;
icore
=
j
;
}
else
{
icore
=
i
;
idrude
=
j
;
}
if
(
rmass
)
{
mcore
=
rmass
[
icore
];
mdrude
=
rmass
[
idrude
];
rmass
[
icore
]
+=
mdrude
;
rmass
[
idrude
]
*=
mcore
/
rmass
[
icore
];
coeff
=
mdrude
/
(
mcore
+
mdrude
);
}
else
{
// TODO check that all atoms of this types are in the group
coeff
=
mcoeff
[
type
[
idrude
]];
}
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
{
x
[
idrude
][
k
]
-=
x
[
icore
][
k
];
x
[
icore
][
k
]
+=
coeff
*
x
[
idrude
][
k
];
v
[
idrude
][
k
]
-=
v
[
icore
][
k
];
v
[
icore
][
k
]
+=
coeff
*
v
[
idrude
][
k
];
f
[
icore
][
k
]
+=
f
[
idrude
][
k
];
f
[
idrude
][
k
]
-=
coeff
*
f
[
icore
][
k
];
}
}
}
fix_drude
->
is_reduced
=
true
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
void
FixDrudeTransform
<
inverse
>::
reduced_to_real
()
{
int
nlocal
=
atom
->
nlocal
;
int
ntypes
=
atom
->
ntypes
;
int
dim
=
domain
->
dimension
;
int
*
mask
=
atom
->
mask
,
*
type
=
atom
->
type
;
double
**
x
=
atom
->
x
,
**
v
=
atom
->
v
,
**
f
=
atom
->
f
;
double
*
rmass
=
atom
->
rmass
,
*
mass
=
atom
->
mass
;
double
mcore
,
mdrude
,
coeff
;
int
icore
,
idrude
;
tagint
*
drudeid
=
fix_drude
->
drudeid
;
int
*
drudetype
=
fix_drude
->
drudetype
;
for
(
int
i
=
0
;
i
<
nlocal
;
i
++
)
{
if
(
mask
[
i
]
&
groupbit
&&
drudetype
[
type
[
i
]]
!=
NOPOL_TYPE
)
{
int
j
=
(
int
)
drudeid
[
i
];
// local index of drude partner because drudeid is in reduced form
if
(
drudetype
[
type
[
i
]]
==
DRUDE_TYPE
&&
j
<
nlocal
)
continue
;
if
(
drudetype
[
type
[
i
]]
==
DRUDE_TYPE
)
{
idrude
=
i
;
icore
=
j
;
}
else
{
icore
=
i
;
idrude
=
j
;
}
if
(
rmass
)
{
double
s
=
sqrt
(
1.
-
rmass
[
idrude
]
/
rmass
[
icore
]);
rmass
[
idrude
]
=
0.5
*
rmass
[
icore
]
*
(
1.
-
s
);
mdrude
=
rmass
[
idrude
];
rmass
[
icore
]
-=
mdrude
;
mcore
=
rmass
[
icore
];
coeff
=
mdrude
/
(
mcore
+
mdrude
);
}
else
{
if
(
!
mcoeff
[
type
[
icore
]])
{
// TODO: should it be > 1.5 ?
double
s
=
sqrt
(
1.
-
mass
[
type
[
idrude
]]
/
mass
[
type
[
icore
]]);
mass
[
type
[
idrude
]]
=
0.5
*
mass
[
type
[
icore
]]
*
(
1.
-
s
);
mdrude
=
mass
[
type
[
idrude
]];
mass
[
type
[
icore
]]
-=
mdrude
;
mcore
=
mass
[
type
[
icore
]];
mcoeff
[
type
[
icore
]]
=
mdrude
/
(
mcore
+
mdrude
);
}
coeff
=
mcoeff
[
type
[
idrude
]];
}
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
{
x
[
icore
][
k
]
-=
coeff
*
x
[
idrude
][
k
];
x
[
idrude
][
k
]
+=
x
[
icore
][
k
];
v
[
icore
][
k
]
-=
coeff
*
v
[
idrude
][
k
];
v
[
idrude
][
k
]
+=
v
[
icore
][
k
];
f
[
idrude
][
k
]
+=
coeff
*
f
[
icore
][
k
];
f
[
icore
][
k
]
-=
f
[
idrude
][
k
];
}
}
}
for
(
int
i
=
0
;
i
<
nlocal
;
i
++
)
{
if
(
mask
[
i
]
&
groupbit
&&
drudetype
[
type
[
i
]]
!=
NOPOL_TYPE
)
{
drudeid
[
i
]
=
atom
->
tag
[(
int
)
drudeid
[
i
]];
}
}
if
(
!
rmass
)
{
for
(
int
itype
=
1
;
itype
<=
ntypes
;
itype
++
)
if
(
mcoeff
[
itype
]
<
1.5
)
mass
[
itype
]
/=
1.
-
mcoeff
[
itype
];
}
fix_drude
->
is_reduced
=
false
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
int
FixDrudeTransform
<
inverse
>::
pack_forward_comm
(
int
n
,
int
*
list
,
double
*
buf
,
int
pbc_flag
,
int
*
pbc
)
{
double
**
x
=
atom
->
x
,
**
v
=
atom
->
v
,
**
f
=
atom
->
f
;
int
*
type
=
atom
->
type
,
*
drudetype
=
fix_drude
->
drudetype
;
double
dx
,
dy
,
dz
;
int
dim
=
domain
->
dimension
;
int
m
=
0
;
for
(
int
i
=
0
;
i
<
n
;
i
++
)
{
int
j
=
list
[
i
];
if
(
pbc_flag
==
0
||
(
fix_drude
->
is_reduced
&&
drudetype
[
type
[
j
]]
==
DRUDE_TYPE
))
{
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
buf
[
m
++
]
=
x
[
j
][
k
];
}
else
{
if
(
domain
->
triclinic
!=
0
)
{
dx
=
pbc
[
0
]
*
domain
->
xprd
+
pbc
[
5
]
*
domain
->
xy
;
dy
=
pbc
[
1
]
*
domain
->
yprd
;
if
(
dim
==
3
)
{
dx
+=
+
pbc
[
4
]
*
domain
->
xz
;
dy
+=
pbc
[
3
]
*
domain
->
yz
;
dz
=
pbc
[
2
]
*
domain
->
zprd
;
}
}
else
{
dx
=
pbc
[
0
]
*
domain
->
xprd
;
dy
=
pbc
[
1
]
*
domain
->
yprd
;
if
(
dim
==
3
)
dz
=
pbc
[
2
]
*
domain
->
zprd
;
}
buf
[
m
++
]
=
x
[
j
][
0
]
+
dx
;
buf
[
m
++
]
=
x
[
j
][
1
]
+
dy
;
if
(
dim
==
3
)
buf
[
m
++
]
=
x
[
j
][
2
]
+
dz
;
}
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
buf
[
m
++
]
=
v
[
j
][
k
];
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
buf
[
m
++
]
=
f
[
j
][
k
];
}
return
m
;
}
/* ---------------------------------------------------------------------- */
template
<
bool
inverse
>
void
FixDrudeTransform
<
inverse
>::
unpack_forward_comm
(
int
n
,
int
first
,
double
*
buf
)
{
double
**
x
=
atom
->
x
,
**
v
=
atom
->
v
,
**
f
=
atom
->
f
;
int
dim
=
domain
->
dimension
;
int
m
=
0
;
int
last
=
first
+
n
;
for
(
int
i
=
first
;
i
<
last
;
i
++
)
{
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
x
[
i
][
k
]
=
buf
[
m
++
];
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
v
[
i
][
k
]
=
buf
[
m
++
];
for
(
int
k
=
0
;
k
<
dim
;
k
++
)
f
[
i
][
k
]
=
buf
[
m
++
];
}
}
/* ---------------------------------------------------------------------- */
template
class
FixDrudeTransform
<
false
>
;
template
class
FixDrudeTransform
<
true
>
;
Event Timeline
Log In to Comment