Dear all,
I am working right now on metals and ceramics using EAM and MEAM potentials with one “fs” timestep, and I appreciate it if someone can provide me some answers regarding fixes npt/cauchy, setforce, and balance.
-
If I understood correctly, fix npt/cauchy attempts to enforce true stress or Cauchy stress. Since LAMMPS calculates Second Piola-Kirchhoff stress (PK2) by default, using the Lagrangian description of the system, isn’t it possible to mimic Cauchy stress by tweaking “nreset” keyword in “fix npt” command in an update Lagrangian fashion? If the answer is yes, what would be the appropriate number for “nreset”? (I am using “nreset 1000” for my simulations). I skimmed the paper by “(Miller) Miller, Tadmor, Gibson, Bernstein and Pavia, J Chem Phys, 144, 184107 (2016).” , and I didn’t see that the authors mention “nreset” using LAMMPS and its effect on the results.
-
Is there any chance to see the same implementation for Cauchy stress in “compute stress/atom” soon?
-
I apologize if I am asking a cliche question. Still, I couldn’t find on LAMMPS forum that if I want to shear a material by freezing atoms in two blocks on the top and bottom of it using “fix steforce 0.0 0.0 0.0” commend what should be the thickness of each block. Some people say two times of cutoff radius of the potential; some say freezing atoms up to the cutoff radius of the potential will suffice. I appreciate it if someone can elaborate on this for me if there is any physical or implementation reasoning behind it.
-
Is it rigorous to use the “fix balance” commend every 100 steps like this “fix 1 all balance 100 1.0 shift xyz 10 1.0 weight time 0.8”
Many thanks in advance for your time and consideration,
Bests,
Reza.
Dear all,
I am working right now on metals and ceramics using EAM and MEAM potentials with one “fs” timestep, and I appreciate it if someone can provide me some answers regarding fixes npt/cauchy, setforce, and balance.
- If I understood correctly, fix npt/cauchy attempts to enforce true stress or Cauchy stress. Since LAMMPS calculates Second Piola-Kirchhoff stress (PK2) by default, using the Lagrangian description of the system, isn’t it possible to mimic Cauchy stress by tweaking “nreset” keyword in “fix npt” command in an update Lagrangian fashion? If the answer is yes, what would be the appropriate number for “nreset”? (I am using “nreset 1000” for my simulations). I skimmed the paper by “(Miller) Miller, Tadmor, Gibson, Bernstein and Pavia, J Chem Phys, 144, 184107 (2016).” , and I didn’t see that the authors mention “nreset” using LAMMPS and its effect on the results.
this is a question that you could answer yourself by making a sufficient set of rigorous tests.
- Is there any chance to see the same implementation for Cauchy stress in “compute stress/atom” soon?
not likely, unless you would sit down and program it yourself or convince somebody with sufficient time, knowledge, and interest to do it.
- I apologize if I am asking a cliche question. Still, I couldn’t find on LAMMPS forum that if I want to shear a material by freezing atoms in two blocks on the top and bottom of it using “fix steforce 0.0 0.0 0.0” commend what should be the thickness of each block. Some people say two times of cutoff radius of the potential; some say freezing atoms up to the cutoff radius of the potential will suffice. I appreciate it if someone can elaborate on this for me if there is any physical or implementation reasoning behind it.
no comment.
- Is it rigorous to use the “fix balance” commend every 100 steps like this “fix 1 all balance 100 1.0 shift xyz 10 1.0 weight time 0.8”
what is a good choice of settings for load balancing depends strongly on the individual system setup and geometry. my rule of the thumb: don’t use a feature unless you need it and thus avoid overhead and unexpected complications.
that means, that often load imbalances can be avoided by simply using a smart choice for the processors command and then no extra operations are needed. beyond that, one can often achieve significant improvements by just using the balance command once between runs. fix balance should only be needed for systems that undergo significant changes in the distribution of particles per subdomain over time. one also has to be aware of the impact of the different communication style settings brick vs tiled.
axel.