Stress calculations

[fgygi]

The calculation of the stress tensor using version 1.44.0 is inaccurate when very small unit cells are used (e.g. cells containing only one atom). This happens when the unit cell dimensions are comparable to the width of the gaussian pseudocharge distributions (parameter rcps, initialized to 1.5 (a.u.) in AtomSet.C.
This bug is corrected in 1.45.0.

[gpanchap]

I am trying to understand the different 'sigma' terms in the output file, and its relation to the thermodynamic 'pressure'. I guess the sigma_eks_** comes from the Kohn-Sham total energy and the eks_kin_** is the contribution from the 'instantaneous' temperature (temp_ion). So that the instantaneous pressure is (1/3) Tr[sigma_eks_*+sigma_kin_*], (note there is no negative sign in here as one would otherwise expect. ). I am lost at what sigma_** could be though, as it is not the sum of eks and kin terms. Thanks.

[fgygi]

There are several possible sign conventions that can be adopted to describe stress and pressure. Qbox uses the following convention (which happens to agree with that adopted in the Quantum Espresso package): The sign of the stress <sigma_xx> is such that a positive value of <sigma_xx> indicates a tendency of the cell to expand. The contribution to the stress from the Kohn-Sham energy is sigma_eks = - dEks/dV (or rather the appropriate tensor equivalent expression). The kinetic contribution to the stress is coming from the instantaneous ionic temperature and is positive, i.e. always makes the cell expand. The externally applied stress is defined by the Qbox variable ext_stress. Giving a positive value to ext_stress will tend to compress the cell. The total stress is <sigma_xx> = sigma_eks + sigma_kin - sigma_ext. With this convention, the total stress is what drives the cell dynamics (e.g. if cell_dyn == SD): if sigma > 0, the cell will expand, if sigma < 0, the cell will shrink. Note that in the absence of external stress, sigma is the sum of sigma_eks and sigma_kin.