Canonical Ensamble with Nose Thermostat

Message

giacomo giorgi · #1 Post by **giacomo giorgi** » Tue Feb 21, 2012 3:56 am

Dear all,
I would like to run some MD simulations in order to anneal at first a Titanium hcp matrix with the Barendsen thermostat (SMASS=-1) and then once annealed I would like to simulate the canonical ensamble by using the Nose thermostat.

I have a doubt concerning the value I have to specify in this latter case for SMASS.

I mean, the manual reports that the optimal choice is that for which the induced temperature fluctuation show approximately the same frequencies as the typical 'phonon'-frequencies for the specific system.

I found some papers reporting the phonon dispersion curve of hcp Ti. Can I deduce the value of SMASS for Nose simulations from such curves? And if so, which the value to be considered? This is my INCAR

SYSTEM = Ti54_2d_b_Oh_scrambled

PREC = Low
EDIFF = 1E-05
ISPIN = 2
NELMDL = 4
NELMIN = 8

IWAVPR = 11

ISYM = 0
LREAL = A

ISMEAR = 2
SIGMA = 0.05

ISIF = 3

SMASS = <-----

?
TEBEG = 300.
TEEND= 300.

POTIM = 3.0

IALGO = 48

NSW = 2000

IBRION = 0

LWAVE = .FALSE.
LCHARG = .FALSE.

NGX= 31
NGY= 31
NGZ= 47

Thanks in advance,
Giacomo

#2 Post by **admin** » Tue Feb 21, 2012 1:26 pm

Nose-frequency (NF) in OUTCAR is in Hz (hertz). To convert it to cm-1 use the formula
NF*33.3567/2 pi (NF is Nose frequency in THz)

[ Edited Fri Feb 24 2012, 11:54AM ]
[ Edited Mon Feb 27 2012, 12:08PM ]

dodo · #3 Post by **dodo** » Fri Feb 24, 2012 2:10 pm

So, if I have a line in my OUTCAR

estimated Nose-frequenzy (Omega) = 0.64E+14 period in steps = 98.45 mass= 0.230E-26a.u.

is the frequency is .64*10^14 (radian-Hz)? =0.102Hz?

If so the formula above is slightly incorrect. It says frequency in cm-1 is (64*1000*33.3567/(2*pi)) = 339940.8917 cm-1

When it should be = (.64*10^14)/(2*pi*2.999*10^10) = 339.8160321 cm-1 ?

giacomo giorgi · #4 Post by **giacomo giorgi** » Mon Feb 27, 2012 2:59 am

Dear Admin,
thanks for the quick reply. Anyway, there is something I am not able to understand yet.
I mean if I do a single MD step with a trial mass (0.02) in the formula you suggested me

NF1/NF2 = SMASS2/SMASS1

I will get NF1 and SMASS1 (SMASS1=0.02) . I need to know SMASS2 (that I have to insert in my INCAR) . What about NF2? From where should I take it?

Sorry, maybe it is a silly question but I am starting right now doing MD calcs.

Very best,
Giacomo

giacomo giorgi · #5 Post by **giacomo giorgi** » Mon Feb 27, 2012 10:54 am

By the way,
the post that you replied has been changed. I dont see the equation you suggested me anymore..... (NF1/NF2 = SMASS2/SMASS1)...so...which value for SMASS in the Nose thermostat? SMASS=0? is it ok?
thanks once more,
G

#6 Post by **admin** » Mon Feb 27, 2012 12:03 pm

@dodo Taking NF in THz gives: 64*33.3567/2pi = 339.9 cm-1
@gg The formula for tuning the frequency is
NF1/NF2 = sqrt(SMASS2)/sqrt(SMASS1)

dodo · #7 Post by **dodo** » Mon Mar 05, 2012 2:01 pm

Thank you

martin wa · #8 Post by **martin wa** » Fri Oct 04, 2013 5:09 pm

Hello,

I hope my question is right here and contributing to this topic.

I am doing MD simulations on CoW systems and wondering about the SMASS value for the Nos? thermostat.

According to the formula from above I need a value of
SMASS = 270 to bring the NF down to ~5THz which is my phonon frequenzy.

I haven't run the calculations yet and can't say if I really get a good result or not.

I am just wondering if someone else needed to use such a high SMASS value since I cant find anything in the literature (most values I read about have been around SMASS = 2).

As I hope to understand it right, the SMASS is indicating the damping of the frequenzy, is there a physical or software internal limit for it?
Is there some information about reasonable values (since I am still a bit wondering about my high value)?

Regards
/Martin

#9 Post by **admin** » Mon Oct 07, 2013 3:36 pm

There is no formula defining SMASS (it is defined via the frequency). The formula above can be used to tune the value of SMASS to correspond to some desirable frequency. Use some value of SMASS, start vasp for a couple of seconds and find Nose frequency in OUTCAR. Then using NF1/NF2 = sqrt(SMASS2)/sqrt(SMASS1) calculate the SMASS value you need.

martin wa · #10 Post by **martin wa** » Tue Oct 08, 2013 2:18 pm

Hey.
Of course you are right. Its not the formula for defining SMASS.
For me it worked now this way to tune SMASS. The NF and the phonon frequency are similar. But I am still wondering about these high value of SMASS which is needed.

Thank you!