Please keep the list in the loop.
I don't think fix nvt will do better/worse than
any other thermostat at keeping fluctuations small.
Fluctuations are a function of the velocity of the
atom. So any thermostat that sets the velocities
to the desired temperature will have nearly the
same fluctiations, no?
Fix nvt (or any other thermostat) doesn't know anything
about (non)periodic BCs.
Steve
Hi,
How about setting the temperature at a low value eg. 2K , and then increase the temprature slowly (using Tstat, Tstop, Tdamp). i’m guessing the fluctuations of the atoms also depends on how fast heat goes in or out of the system i.e how quicly the sample is heated. Ofcourse its nonsence to think one can simulate a process that involved heating the sample in minutes our hours ( that would required alot of timesteps).
Oscar G.
Hi,
Also am … yl1007 wroteL "if ‘fix npt’ is used, then dimensions in x,y and z directions all will increase, but increase of x-dimension and z-dimension are very small, and increase of y-dimension is much smaller than increase of y-dimension " .
If you want dimensions to change evenly , then its better to use the “ISO” feature…, and yup usually when materials are heated they tend to “expand” (change volume), some materials experience isotropic expansion , other not.
Oscar G.
Ok let me try harder…
I guess that at low temperatures the atoms can be treathed as an harmonic oscillator i,e atoms the fluctuate around an equilibirum point. As you heat the sample, the atoms moves away from the equlibrium, if there is enough kinetic energy then the atoms moves out of the potential eall , it can be a liquid phase…
Thats all i can think …
Oscar G.
Hi Oscar G:
Thanks for your reply.
As wrote in my first email, the sample is non-periodic in y-direction, i.e. BCs psp, so ‘iso’ cann’t be used here.
Slowing down the speed of heating a sample may be a way to reduce atoms fluctuation, but will dividing heating process into several steps cause some other side effects?
Steve, as you said ,‘any thermostat will have nearly the same fluctuations’, for a new lammps-user, can i understand your conclusion like this:
fluctuations are related to atoms position, atoms position are related to atoms velocity, finally atoms velocity are closely related to temperature.
So for a sample, no matter which thermosta i used, once it is heated to a same temperature, then fluctuations will be nearly same.
But as wrote in my first email, for a sample with initial dimensions, Lx=38.518729 Ly=139.40627 Lz=34.457144,
when heated to 1400k by “fix npt”, the dimensions are Lx=39.288336, Ly=145.97492, Lz= 35.221666;
when heated to 1400k by “fix nvt”, the dimensions are Lx=38.518729, Ly=153.85074, Lz= 34.457144;
As we can see, for “fix nvt”, Ly is lager than that of “fix npt”.
Though these results just show changes of dimensions, which can not be used to completely reflect fluctuations ,
i think fluctuations are related to changes of dimensions, the larger changes of dimensions are, the larger the atoms fluctuations will be.
All i want to is, my simulational results seem to show that for an non-periodic sample, different thermostat may cause different fluctuations.
If this kind of phenomenon does exist, then attention should be put on it.
L.Yang
Hi L.Yang,
So for a sample, no matter which thermosta i used, once it is heated to a
same temperature, then fluctuations will be nearly same.
That is absolutely true. Same temperature = same thermal vibration.
But as wrote in my first email, for a sample with initial dimensions,
Lx=38.518729 Ly=139.40627 Lz=34.457144,when heated to 1400k by "fix npt", the dimensions are Lx=39.288336,
Ly=145.97492, Lz= 35.221666;when heated to 1400k by "fix nvt", the dimensions are Lx=38.518729,
Ly=153.85074, Lz= 34.457144;
I suggest we back up a little. You have a crystal bounded in a
simulation box with psp boundary conditions. With NVT, you are fixing
the p boundaries (x and z) and only the s boundary (y) is allowed to
relax. WIth NPT on the other hand, all psp boundaries (x, y and z)
are allowed to relax. Agree?
Okay, now you heat the system up to 1400K, and due to thermal
expansion you are almost guaranteed to see expansions in lattice
spacings and box dimensions (except for materials/potentials with
negative thermal expansion coefficient). For NVT ensemble with psp,
the crystal has no where to expand except for the y direction, but on
the contrary the crystal can, and will, expand in *all three*
directions with NPT. This is why you have no change in Lx and Lz with
NVT while Ly is doing all the expanding. With NPT, however, all three
dimensions expanded to relieve the thermal stress. Now can you see
why you obtained different box dimensions with different ensembles?
The crystal *had* to expand due to an increased temperature. It
expanded "uni-axially" with NVT and "volumetrically" with NPT. You
should get different expansions with these two conditions, and this is
why the final expanded volume are similar but not quite the same (1.1%
difference).
If this kind of phenomenon does exist, then attention should be put on it.
NVT and NPT fixes worked perfectly normal and absolutely reflected the
settings of your system. You should re-think which of the expansion,
uni-axially or volumetrically, you would want for your system.
Cheers,
Ray
Dear Ray:
I deeply appreciate your detailed and scientific explanation.
Now i clearly understand that fluctuations for 'fix npt'and 'fix nvt'actual are actually same.
Thanks again.
L.Yang