> Dear Sir, > > *your structure is not exactly cubic* (look at your atoms position). >> I wonder why you use atomic coordinate which looks like >> the result of relaxation along with lattice parameters >> which look like given by hand. Relax either both, or none. >> > > 1) i relaxed both atomic coordinates and lattice parameters using variable > cell option and got > > outcoor: Relaxed atomic coordinates (scaled): > 0.00005993 0.00006110 0.00005744 1 Mg 1 > 0.48553111 0.48553523 0.48553182 2 O 2 > > outcell: Unit cell vectors (Ang): > -0.000022 2.044803 2.044814 > 2.044808 -0.000026 2.044818 > 2.044803 2.044802 -0.000021 > so i used lattice constant=4.089606~4.09 Ang.
Dear Sonu, Your coordinates are either not accurately enough converged, or they get some numbers with respect to your "old" lattice constant, and then you freeze these coordinates and declare them as scaled with respect to your new (slightly relaxed) lattice constant. The difference (0.4855 vs.0.5000) is not that large, but I do not understand why do you want to keep it. Perfect MgO has no internal parameters, so your atoms must sit exactly in the middle. > 2) when i did study of energy v/s volume, i got minima at 4.10 Ang, with > out > above relaxation. Fine; take this value of lattice constant if you like it, and place the atoms at (0 0 0) and (1/2 1/2 1/2). > should i proceed this way, keeping care of symmetry or should i see the > final results(freqs) to decide, which one to use (relaxation or no > relaxation)? I don't know what you want. If you intentionally lower the symmetry by moving your atom out of symmetry position then do not wonder that your phonon modes are not any more triple degenerated... In fact I am not sure that your "effect" is completely due to this; other sources of error might be k-mesh and mesh-cutoff, which are generated based on your primitive cell which is not cubic but "trigonal" (around the [111] diagonal). > 3) can we use macroscopic polarisation analysis for small band gap > materials? sorry, if my question is stupid. I think, any band gap will do, which survives throughout the path as you displace the atoms to calculate the polarization. (Is your MgO a small gap material?) Best regards Andrei Postnikov
