回复:Re: 回复:Re: Can LAMMPS do_this?

Sorry, I forgot to tell you that the snapshot is not the real size of the spindle, because the real spindle contains so many atoms that my PC couldn’t compute it,so I built a smaller model to simulate the process, if the small model can make it, then I will build the real size.The real spindle is like the snapshot below. Do you think LAMMPS can do this?
附件1.png

发件人:Ray Shan <rshan@…1795…>
收件人:“nihaohao360@…414…” <nihaohao360@…414…>
抄送人:lammps-users [email protected]
主题:Re: 回复:Re: [lammps-users] Can LAMMPS do_this?
日期:2016年10月29日 02点39分

From your snapshot the system looks like a nano-scale system with just several thousand atoms, which is inconsistent with your description below. Which is it?

Ray

附件1.jpg

Sorry, I forgot to tell you that the snapshot is not the real size of the
spindle, because the real spindle contains so many atoms that my PC couldn't
compute it,so I built a smaller model to simulate the process, if the small
model can make it, then I will build the real size.The real spindle is like
the snapshot below. Do you think LAMMPS can do this?

you are asking the wrong question here. it is not about "can LAMMPS do
this", but rather "can *you* do this?".
you seem to have your length and timescales completely wrong, and that
can be easily estimated from existing benchmark data with a simple
back-of-the envelope calculation:

a system of the dimension you are describing would have to have so
many atoms, that you would have to have exclusive access to a
supercomputer much larger than anything currently available. just look
up the dimensions of your system (which is in angstroms) and scale it
up to the intended dimensions, assuming the number of atoms scales
linearly with the volume (that isn't very accurate, but good enough
for this). LAMMPS can theoretically handle systems that large, but
none of the LAMMPS developers ever had a chance to test it.

but that isn't where the story ends. regardless of size of the system,
you have to simulate it at the same time step (which is determined by
the shape of the potential, the mass of the atoms and the largest per
atom velocity). at the same angular velocity, that would mean for a
larger system many more time steps, as some atoms would be moving much
faster for a larger system and thus the timestep would have to be much
smaller.

in conclusion, unless you find a suitable mesoscale model, there won't
be much of a chance that you can complete this calculation with a
reasonable amount of compute resources and time, even if the software
is technically able to handle such a system.

axel.

If you are only interested in microvibrations in what is essentially a macroscopic system, there is no good science reason for attempting to use particle-based simulation. Any off-the-shelf PDE-based solid/fluid mechanics code would be far more appropriate.

Aidan