[gmx-developers] rmsf units in g_anaeig

[Berk Hess]

Ran Friedman wrote:
> Hi All,
>
> I was looking at the gmx_anaeig code to see how it calculates the rmsf
> for each vector.
> At function rmsf, I find:
>
> (GMX 4.04, lines 713-714):
>       for(i=0; i<natoms; i++)
>           y[g][i] = sqrt(eigval[eignr[v]]*norm2(eigvec[v][i]))/sqrtm[i];
>
> The units of the eigenvalues are kJ mol-1 nm-2 u-1
>   
How did you get to this unit?
g_covar generates eigenvalues with the unit nm^2 u, if you use mass
weighting.
If you don't use mass weighting, all the mass units disappear.

Or are you using a matrix from a normal mode analysis?
In that case I think it is not useful to look at rmsf, you could simply
look at the vector components.

Berk
> The eigenvectors are normalized by u-1/2 so:
>     norm2(eigenvec[v][i])
> should give units u-1
>
> the units of sqrtm seem to be u-1/2.
>
> but the rmsf units should be nm
>
> What am I missing?
>
> Thanks,
> Ran.
>
>