# [lammps-users] Temperate calculation

Dear lammps users,

I have a question about a system containing separate particles. When we have some particles (for example separate graphene sheets in water) is it correct to calculate the temperate of each particle separately? I have defined seperate group for each particle. The temperature of the whole system is 300k. But when I calculate the temperature of each graphene sheet, it is not 300. It is much bigger.

Hi Runak,
Usually in thermodynamics when we talk about temperature, it’s the temperature of the whole thermodynamic system. I don’t have any idea about your project, but can one talk about temperature of each particle! Can we?

Well, if you have graphene + water, it is probably best to use 2 thermostats (1 for graphene, 1 for water), rather than one single thermostat for the whole system. Have you tried that?

Simon Gravelle

Well, if you have graphene + water, it is probably best to use 2 thermostats (1 for graphene, 1 for water), rather than one single thermostat for the whole system. Have you tried that?

it is difficult to make any specific statements and suggestions on such a general question without knowing any details about the specific input, how and how well the system was equilibrated, and when and how the temperature is computed.

assuming there are no other manipulations of the system impacting atom velocities than thermostats, the temperature should be the same across the system (equipartitioning).
if it is not, that is often an indication that the system is not (yet) in equilibrium. unless you are using a dissipative thermostat during (part of) the equilibration, it can sometimes take a long time for a system to reach equilibrium, when multiple subsystems are initialized differently or are only weakly coupling. using multiple thermostats can mask that effect and even be one way to “guide” the system into reaching equilibrium. however, the true measure whether this has been achieved (and whether the simulation setup is otherwise consistent and suitable for studying an equilibrium system) would be to run the entire system without any thermostatting, after equilibrium is reached and then observe that equipartitioning is retained and energy is conserved sufficiently well for a long enough timespan.

Axel.

Thanks Andri. I mean the temperature of each graphene sheet.

Thanks Axel. I needed such a scientific advice. So this “removing ensemble” is just for testing? Or it is correct to not use ensemble for the rest of simulation. Can I study the adsorption of the particles ( the graphene sheets) in the absence of any ensemble after applying ensemble for enough long time?

First of all it is not correct to refer to a thermostat as an “ensemble”. I have been explaining this many times on this mailing list, please see in the archives why this is incorrect.
From what it seems based on your descriptions, you are likely not in any kind of the usual, well defined statistical mechanical ensembles (they usually require having a homogeneous bulk system and no other manipulations like fix momentum or fix setforce or fix wall/… or others). Whether you have a well defined statistical mechanical ensemble or not is independent from whether you have a meaningful simulation or not, or whether applying a thermostat and what kind and how and where is required or important or not.

Yes, what I was describing is an important test, but it may also be required for the “production” simulation of your system.

Again, it is not possible to provide a definite answer to such a question without knowing more details about what exactly the system and input is and what kind of system you want to represent and compare to. Either using a thermostat or not using a thermostat may be more adequate.

You need a thermostat when you simulate only a subset that is embedded into a buik system. For adsorption studies, especially with particles deposited from gas phase, you may want to apply the thermostat only to part of the system, i.e. the part that is in contact with the bulk (usually that part requires some kind of boundary to immobilize the system in space which represents being coupled to a 0K system which needs to be compensated for), but then time integrate the surface and the deposited particles without a thermostat applied (because they have either an explicit thermal coupling or no coupling at all).

This is much more a question about the science of your simulation than about LAMMPS, so my advice is to discuss this with people sufficiently familiar with your research and also with statistical mechanics to discuss the implications of different choices. There are no simple “do this, not that” kind of general rules in these cases.

Axel.

Thanks Axel. Is it possible to define ( velocity group create 300.0 4928459 rot yes mom yes dist uniform) separately each for one group?

Could you please describe for study the adsorption between two separate segments (two graphene sheets) how it is correct to define the npt ensemble and velocity command before and after adsorption? Is it correct to apply an npt ensemble and separate velocity commands and don’t remove them after after during the adsorption? I know that I should not apply other fixes like momentum and so on…

Thanks Axel. Is it possible to define ( velocity group create 300.0 4928459 rot yes mom yes dist uniform) separately each for one group?

It is certainly possible, technically speaking. Whether it is advisable depends on the specific simulated system. The major difference will be in how the keywords “rot yes” and “mom yes” are applied, i.e. to the individual groups or the whole system. You yourself are in the best position to decide whether that is what you want.

Could you please describe for study the adsorption between two separate segments (two graphene sheets) how it is correct to define the npt ensemble and velocity command before and after adsorption? Is it correct to apply an npt ensemble and separate velocity commands and don’t remove them after after during the adsorption? I know that I should not apply other fixes like momentum and so on…

As I already mentioned, this is a discussion that you need to have with the people that are familiar with your research and are advising you, not with me. I don’t have the time to learn what your research is about and I cannot give meaningful advice without. As I already wrote before, this is not a simple “do this, not that” kind of situation.
I also already reminded you that you need to correct how you use the term “ensemble”. From the limited information you have provided it sounds like a variable cell integrator like fix npt would not be a preferred choice at all, but that is just a guess.