[gmx-developers] Fw: About calculations of viscosity

[Berk Hess]

David van der Spoel wrote:

>
>>
>> I’m a new GMX user. I notice from either the user-manual or the 
>> references (Palmer, PRE 49 359, & B. Hess, JCP 116, 209, etc.) that 
>> there seem to be at least 3 methods available in Gromacs: Einstein 
>> relation, TCAF and periodic perturbation method. But I’m not very 
>> clear about how to set the calculations and how to get the according 
>> viscosity values.
>>
>> My understanding is this:
>>
>> For Einstein relation calculations, there is no special setting in 
>> the md.mdp file, and the viscosity value is gotten by the visco.xvg 
>> file generated by ‘g_energy –vis’;
>>
>> For TCAF, also there is no special setting in the md.mdp file, and 
>> the viscosity value is gotten by fitting the data of visco_k.xvg file 
>> generated by ‘g_tcaf’;
>>
>> For the periodic perturbation method, two NEMD items, a ‘accelerate’ 
>> value in X direction should be specified to all atoms and a 
>> ‘cos_acceleration’ value should be given, and the viscosity value is 
>> gotten from the ‘1/Viscosity(SI)’ item of the ‘g_energy’ results.
>>
>> I’m eager for any instructions about techniques on the calculations 
>> and the value ranges.
>>
>> Thanks.
>
>
> According to Hess, the periodic perturbation is more efficient than 
> other methods. In particular methods that depend on the pressure 
> autocorrelation tend to need very long simulation times to converge.
> We have used the NMED method by Hess (JCP 119 pp. 7308-7317, 2003) as 
> well and I will use it again. It is important to test different 
> accelerate values, and notice that the real viscosity is obtained in 
> the limit that the external perturbation goes to zero.
>
Note that I replied on the gmx-users list, as I thought that was the 
appropriate place.

Berk.