Creating mass-weighted covariance matrix

[fatih4534]

Hi sir, I would like to ask about creating mw covariance matrix. Do I need to align the frames before creating matrix? I am planning to compute configurational entropy for structure.

[drroe]

In cases where you want to remove global rotational and translational motions, you should definitely align the frames before calculating the mass-weighted covariance matrix. You can do so either by RMS-fitting to the first frame or to an average structure, or by aligning along principal axes; your best bet is usually aligning heavy atoms to an average structure. Hope this helps!

[fatih4534]

Hi Mr. Roe, thank you for your response. I actually cannot understand the eigenmode term. Should I use all eigenmodes, when I calculate the entropy?

[drroe]

Sorry this fell off my radar for a bit.

Should I use all eigenmodes, when I calculate the entropy?

So assuming you are referring to the thermodynamics analysis activated by the thermo keyword of the diagmatrix command, then yes you will want all eigenmodes. Note that however unless you have a fairly simple and harmonic system, and unless your covariance matrix is well-populated (i.e. you use many more frames to create the matrix than the columns in the matrix, which is 3 times the number of atoms), that the quasiharmonic entropy analysis done by thermo will be fairly inaccurate. The model used is a very simplistic one derived from the study of an ideal gas in the canonical ensemble (see D. A. McQuarrie, "Statistical thermodynamics", Harper & Row, New York, 1973, Chapters 5, 6, and 8) which tends to fall apart for larger and more dynamic/anharmonic systems, so it is recommended you calculate entropy in a different way - see e.g. publications like this from M. Gilson et al. for more details.

Hope this helps.