[gmx-users] Re: g_velacc

[Vitaly Chaban]

Hi Ram,

You should divide the value of your integral to the square of mass of you protein. Be attentive with units and powers then.
I guess the final value will be essentially less than 0.99279 for floating protein.

rr> Here, though with the option -s we are calculting the momentum auto correlation function

It seems to me you calculate the momentum ACF in all the cases using g_velacc. Maybe Florian could correct me if I'm wrong here. :)

Vitaly



rr> Dear Vitaly Chaban,
rr> 
rr> Thanks for your kind sugestions. I did followed the way you mentioned for calcualting the diffusion constants. I like to have a better understanding of what I have done.
rr> 
rr> g_velacc:
rr> 
rr> g_velacc  -f   -s  -o  -aceflen 
rr> 
rr> Since, mine is a single protein, I have not defined any index file and I am calculating the g_velacc on backbone atoms.
rr> 
rr> as the manual says, -aceflen will define the number of frames to be taken into consideration i suppose.
rr> 
rr> Here, though with the option -s we are calculting the momentum auto correlation function, but still we are considering it as velocity auto correlation funciton. Is it alright or as the other user mentioned we need to devide the correlation value with square of the mass of the protein ?
rr> 
rr> g_analyze:
rr> 
rr> here, the numerical integration is done by trapezium rule. Ideally we need to carryout the integration from 0 to infinity but since our auto correlation function is calculated on a short period of time (which is close to t=0), the integration is evaluated only on this period i suppose. The output I got is the following:
rr> 
rr> 
rr>  Calculating the integral using the trapezium rule
rr> Integral 1     0.99279  +/-    0.00000
rr>                                       std. dev.    relative deviation of
rr>                        standard       ---------   cumulants from those of
rr> set      average       deviation      sqrt(n-1)   a Gaussian distribition
rr>                                                       cum. 3   cum. 4
rr> SS1   3.975160e-02   1.960813e-01   4.002493e-02       2.939    6.669
rr> 
rr> 
rr> is the 0.99279 is the integral value or any thing else ? 
rr> 

Yes, this is an integral value. But for diffusion constant divide it by <M*M>.


rr> Which value I can compare with the value obtained by g_msd. My g_msd value is 1.7*10^-6 cm**2/s which is reasonably good compared to the experimental value.
rr> 
rr> 
rr> Thanks and Regards,
rr> Ram.
rr> 
rr> 
rr> On Sun, Aug 31, 2008 at 2:18 PM, rams rams <rams.crux at gmail.com> wrote:
rr> 
rr> How to monitor the motion of center of mass of a protein as it is the case all the time to monitor this during the calculations of diffusion and correlation functions. How far the values will be different if we monitor the motion of backbone atoms rather than the center of mass motion.
rr> 
rr> I still dont have any idea how to get the diffusion constant using g_velacc. 
rr> 
rr> Ram. 
rr> 
rr> 
rr> 
rr> On Sun, Aug 31, 2008 at 4:28 AM, Vitaly Chaban <chaban at univer.kharkov.ua> wrote:
rr> 
>> No special reason, just mundane ones: computing the diffusion constant
>> through mean square displacement is easier in terms of convergence.
rr> 
rr> But it is not applicable in the anisotropic systems, for example in
rr> ones with spatial confinements present... :)




-- 
Vitaly V. Chaban
School of Chemistry
National University of Kharkiv
Svoboda sq.,4, Kharkiv 61077, Ukraine
email: chaban at univer.kharkov.ua
skype: vvchaban