[Pw_forum] error on phonon calculation (root not converged)

Nicola Marzari marzari at MIT.EDU
Mon Apr 10 19:45:08 CEST 2006


> k = 0.0000 0.0000 0.0000 ( 1045 PWs) bands (ev):
> 
> -45.3707 -22.1221 -21.9779 -21.9779 -21.9779 -7.0142 -5.9530 -5.9530
> -4.1784 -4.1784 -4.1784 7.3984 7.3984 7.3984 9.0084 9.0084
> 9.0084 9.9455 9.9455 9.9455
> 
> 2x2x1 super cell
> Total envergy: -1100.81820918
> 
> k = 0.0000 0.0000 0.0000 ( 4189 PWs) bands (ev):
> 
> -45.2796 -45.2775 -45.2775 -45.2757 -22.1971 -22.1971 -22.1306 -22.0751
> -21.9837 -21.9837 -21.8962 -21.8962 -21.8923 -21.8923 -21.8880 -21.8880
> -21.8847 -21.8808 -21.8808 -21.8782 -7.3110 -7.3110 -7.0254 -6.2803
> -6.1254 -6.1254 -5.9895 -5.9369 -5.9001 -5.9001 -5.5665 -5.5665
> -4.1921 -4.1844 -4.1844 -4.1328 -4.1328 -4.1254 -4.1254 -4.0581
> -3.6727 -3.6727 -3.4773 -3.4773 5.2683 5.5759 5.7750 5.7750
> 5.8965 7.1769 7.1769 7.3059 7.3059 7.3148 7.3148 7.3862
> 7.3862 7.4078 7.6449 7.6449 7.7312 7.7312 7.7554 7.7554
> 8.0881 8.0881 8.6134 8.6800 8.6800 8.9788 8.9973 8.9973
> 9.5426 9.5426 9.5604 9.5604 9.9266 9.9266 9.9563 10.1162
> 
> The total energies of them are very close. The largest eigenvalues at gamma
> differs from each other about 10.1162- 9.9455 = 0.1707. Is this could be
> reseason

no, probably not. In the 1x1x1 supercell you need to collect the
4 kpoints that are equivalent to gamma in the 2x2x1 supercell, and only
then check. but the discrepnacies are likely to be smaller that
0.17 eV (I am not sure on how the ewald terms are calculated in
pw - Paolo circulated a document a few days ago. Depending on that,
you might require a shift of .1 eV in the eigenvalues, between the two
supercells, to match).

Still, no solution in sight.

			nicola

---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu



More information about the users mailing list