The ABINIT Users Mailing List ( CLOSED )

["Jean-Yves Raty" <jyraty@ulg.ac.be>]

Title: Message

Dear Wu,

 

For the details on fluctuations quantification in such simulations, just check in classical textbooks.

 

1) In Allen and Tildesley, p51,

 

(note that I observe ~10% fluctuations also for simulations with ~60 atoms at 1000K...)        

 

2) Frenkel and Smith, p54

 

fluctuations in microscopic 'temperature' (average kinetic energy) scales as 1/ N**0.5

typical fluctuations 5-10% with systems of 100 - 1000 atoms.

 

So your results seem ok !

 

3) for the fictitious mass in the Nose thermostat, there is no absolute rule. You can just check the overall shape of your fluctuations (the default value

is apparently a good choice for Si, Ge...)

 

4) Your time for thermalization is large, BUT we sometimes need more than that when we deal with very 'viscous' liquids (or glass-forming)

 

Good luck with your simulation.

 

 

Jean-Yves Raty

----------------------------------------------------------------------------------------------------------------------------------------------------------
Charge de Recherches du FNRS
Universite de Liege                                                                      Tel :  + 32-4-3663747
Institut de Physique B5                                                                Fax : + 32-4-3662990
B4000 Sart-Tilman                                                                       Email : jyraty@ulg.ac.be
Belgique
 

-----Original Message-----
From: Wu Rongqin [mailto:g0203654@nus.edu.sg]
Sent: Tuesday, May 25, 2004 6:50 AM
To: forum@abinit.org
Subject: [abinit-forum] molecular dynamics

Dear all,

I use abinit to do molecular dynamics on a liquid system with 64 atoms at 1000k. after long time run, 15ps the temperature is still fluctuating, it should fluctuate as the system has finite degrees of freedom. But for such a system, what a fluctuating is acceptable? In my case, the maximum fluctuation is 150k. I use a nose thermostat.

As I use different nose thermostat frequency, the temperature fluctuation behavior is quite different. What is the difference between different nose frequencies?

I attach two temperature fluctuation with time, which one is better?

Regards

Rongqin