LAMMPS

Dear LAMMPS,

when I replicate a system of linear polymers from a restart file, I  
receive errors on bonds too long. I guess that before replicating the  
system, the replicate command does not release the periodic boundary  
conditions such that bonded atoms in the same molecules are placed  
close together. Is that the case here? If yes, could you give a fix  
for that soon?

Thank you,
-Trung




-- 
Trung D. Nguyen
PhD student, Department of Chemical Engineering
3440 GGBB, 2350 Hayward St., Ann Arbor, MI 48109
University of Michigan

The replicate command should recognize bonds that cross
periodic boundaries and do the right thing.  But you need to
have the image flags correct.  Can you use it on a system you
read in from a data file?  I would start from that and see
if you can get it to work as you expect.

Steve

On Mon, Aug 11, 2008 at 10:44 AM, Trung Dac Nguyen <ndtrung@...> wrote:
> Dear LAMMPS,
>
> when I replicate a system of linear polymers from a restart file, I
> receive errors on bonds too long. I guess that before replicating the
> system, the replicate command does not release the periodic boundary
> conditions such that bonded atoms in the same molecules are placed
> close together. Is that the case here? If yes, could you give a fix
> for that soon?
>
> Thank you,
> -Trung
>
>
>
>
> --
> Trung D. Nguyen
> PhD student, Department of Chemical Engineering
> 3440 GGBB, 2350 Hayward St., Ann Arbor, MI 48109
> University of Michigan
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> lammps-users mailing list
> lammps-users@...
> https://lists.sourceforge.net/lists/listinfo/lammps-users
>

Thank you, Steve.

It works well when I use restart2data to create a data file and read it in.

Just my curiosity, in the atom_vec_bond.cpp (I use atom_style bond), I  
see that both pack_restart() and unpack_restart() involve the image  
flags. Why does read_restart still give errors when used with replicate?

Best regards,
-Trung

Quoting Steve Plimpton <sjplimp@...>:

> The replicate command should recognize bonds that cross
> periodic boundaries and do the right thing.  But you need to
> have the image flags correct.  Can you use it on a system you
> read in from a data file?  I would start from that and see
> if you can get it to work as you expect.
>
> Steve
>
> On Mon, Aug 11, 2008 at 10:44 AM, Trung Dac Nguyen <ndtrung@...> wrote:
>> Dear LAMMPS,
>>
>> when I replicate a system of linear polymers from a restart file, I
>> receive errors on bonds too long. I guess that before replicating the
>> system, the replicate command does not release the periodic boundary
>> conditions such that bonded atoms in the same molecules are placed
>> close together. Is that the case here? If yes, could you give a fix
>> for that soon?
>>
>> Thank you,
>> -Trung
>>
>>
>>
>>
>> --
>> Trung D. Nguyen
>> PhD student, Department of Chemical Engineering
>> 3440 GGBB, 2350 Hayward St., Ann Arbor, MI 48109
>> University of Michigan
>>
>> -------------------------------------------------------------------------
>> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
>> Build the coolest Linux based applications with Moblin SDK & win  
>> great prizes
>> Grand prize is a trip for two to an Open Source event anywhere in the world
>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>> _______________________________________________
>> lammps-users mailing list
>> lammps-users@...
>> https://lists.sourceforge.net/lists/listinfo/lammps-users
>>
>
>
>



-- 
Trung D. Nguyen
PhD student, Department of Chemical Engineering
3440 GGBB, 2350 Hayward St., Ann Arbor, MI 48109
University of Michigan

Not sure what you are asking.  Maybe it is that the system
has zero center-of-mass velocity in x,y initially but acquires it
later due to the Berendsen thermostat?  I don't think this should
happen, although it might due to roundoff.  It could be that
you are also using fix rigid.  I would try thermostatting the
non-rigid atoms (e.g. the solvent) if possible.  I've had more
success with using fix langevin as a thermostat on rigid bodies.

Steve

On Mon, Aug 11, 2008 at 7:59 AM, jiwu liu <liujiwu@...> wrote:
>
> Hi,
>
>    I have found using Berendsen thermostat will cause the system to have a
> nonzero velocity. Is this due to the nature of this thermostat?
>    I used nve together with Berendsen. Below is an example of my input file.
> The nonzero velocity of center of mass appears in x and y direction, while in
> the z direction I used a LJ93 potential.
>
>  Thanks,
>
> jiwu
>
> units           real
> boundary        s s s
> atom_style      molecular
> read_data       ./sim_10_473_.data
> include         mass.bs
> include         pair.bs
> include         ff.bs
> special_bonds   0.0 0.0 1
> neigh_modify    delay 1
> velocity        all create 473 64289816 mom yes rot yes
> fix             1 all temp/berendsen 473 473 200
> fix             2 bcb rigid molecule
> fix             3 ch2 nve
> fix             4 all wall/lj93 zlo 0.0 0.44 3.7 7.65
> timestep        4
>
> thermo          10000
> dump            1 all atom 10000 sim_10_473_.equili.dump
> dump_modify     1 scale no
> run             2500000
>
> unfix           1
> fix             5 all temp/berendsen 473 300 200
> run             500000
>
> unfix           5
> fix             6 all temp/berendsen 300 300 400
> undump          1
> thermo          1000
> restart         3200000 sim_10_473_.restart
> dump            2 all atom 1000 sim_10_473_.dump
> dump_modify     2 scale no
> run             200000
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> lammps-users mailing list
> lammps-users@...
> https://lists.sourceforge.net/lists/listinfo/lammps-users
>

This turned out to be an interesting issue with the math libraries
different compilers use.  I saw no difference in old vs current versions
of Tersoff with g++.  But I saw something similar to your results with
the Intel compiler, which is also generates much faster Tersoff code
than g++.

The reason turned out to be that the pow() function was used as
pow(arg,m) in the
current version and as pow(arg,3.0) in the old version.  M is now a variable
which can be 3 or 1.  I think the Intel pow() must check for a constant
when compiling and avoid taking a log() or something like that when it
can shortcut
the answer.

Since all Tersoff variants we know of have m = 1 or 3, I just posted a patch
that uses m as a constant instead of a variable in the pow() and it
recovers nearly all of the old speed.

Thanks,
Steve

On Wed, Aug 6, 2008 at 6:11 PM, Romain Perriot <rperriot@...> wrote:
> Hi,
>
> I would have a question for the developers of the code about Tersoff
> potential. I am using an older version of LAMMPS, and running NVE  run on 1
> million Silicon atoms. I compared the parallel performances of the this
> version against the new on, on a quad-core machine. And it seems the older
> version runs much faster. The only difference I found are 2 coefficients for
> the 3 body problem, that are not present in my version. Would this only
> explain the difference in performances? And would the old version considered
> accurate?
> Here are the tables (in seconds, for 100 steps):
>
> LAMMPS, old version
> N 1 cpu 2 cpu 3 cpu 4 cpu
> tN 688 344 276 179
> t1/tN 1 2 2.49 3.84
>
> LAMMPS, latest version (26 Jul 08)
>
> N 1 cpu 2 cpu 3 cpu 4 cpu
> tN 981 628 368 260
> t1/tN 1 1.56 2.673.77
>
> Thanks in advance
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great
> prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> lammps-users mailing list
> lammps-users@...
> https://lists.sourceforge.net/lists/listinfo/lammps-users
>
>

On Sun, 10 Aug 2008, Jian Yao wrote:

JY> 
JY> Hello Lammps   
JY> 
JY> Suppose I want to use the tersoff potential(pair_tersoff.cpp), but in

please explain what you mean by "use".

JY> the SRC/main.cpp I could not find pair_tersoff.cpp or pair_tersoff.h are
JY> included as #include "pair_tersoff.cpp" and "pair_tersoff.h".
JY> 
JY> How can Lammps use the pair_tersoff.cpp ?

pair styles are activated through the Force:new_pair() method.

JY> Could anybody explain how the code is running? I think any c++ program
JY> must run from int main(). Why doesn't main.cpp include pair_tersoff.cpp?

why should it include pair_tersoff.cpp???

true, any c/c++ starts from main(), but you only need to include
(header) files that contain function prototypes or class definitions 
that are used by main(). correspondingly for other files.
please update your knowledge of c/c++.

also, please keep in mind that lammps achieves its modularity and 
extensiblity through smart use of the c++ class inheritance
mechanism and a few carefully crafted macros. also you have to 
understand how "virtual" and "pure" methods work.

thus the implementation is not your standard "hello world" 
type of code and you may have to work your way up through 
the advanced section of your c++ programming book(s).

cheers,
   axel. 


JY> 
JY> Thank you!
JY> 
JY> Jian Yao
JY> 
JY> 
JY> -------------------------------------------------------------------------
JY> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
JY> Build the coolest Linux based applications with Moblin SDK & win great prizes
JY> Grand prize is a trip for two to an Open Source event anywhere in the world
JY> http://moblin-contest.org/redirect.php?banner_id=100&url=/
JY> _______________________________________________
JY> lammps-users mailing list
JY> lammps-users@...
JY> https://lists.sourceforge.net/lists/listinfo/lammps-users
JY> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey@...   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.