If it's energy drift it wont reach steady state. The energy drift is from integrating with a discrete timestep. It's a general problem with MD. I think it might be exaggerated by the mechanism of fix_heat, also.
Quoting shiziyuan <[email protected]>:
Thanks a lot!
It really helps me.
Also make sure that your system is not translating in one direction (flying icecube). This will add kinetic energy to the whole system. Make sure your linear and angular momentum is zero to start with. Hope this helps.
When I am doing graphene/water system, I used fix heat and my system is not translating in one direction but there is energy drift due to long time calculation. But if I used fix langevin, the system was translating in one direction.
However, when I was doing copper/argon system, I used both method and the system was not translating in one direction.
I am so confused about this. Did I make any mistakes for my graphene/water system?
Can you explain more about “flying icecube”?
Thanks a lot.
Flying Icecube: This of it as a system which is not vibrating at all but translating with certain velocity in one direction, (or even rotating with certain angular velocity). As it has certain velocity, the simulation will show certain value of temperature of this system, although it should be zero as atoms are not vibrating at all.
With graphene, the transverse thermal conductivity of the system is very low (due to van der Waals interactions). So, simulating such systems can become quite tricky. What is the standard deviation associated with temperature of thermostats?