[Pw_forum] Negative values in density matrix.

[jorge.gallardo at cab.cnea.gov.ar]

Hello.
In order to obtain the electronic density matrix for an Aluminun
supercell, I run pw.x with the following SYSTEM section in the input file
( al.scf.in )

&SYSTEM
	ibrav = 6,
	celldm(1) = 50.0,
	celldm(3) = 0.1,
	nat = 1,
	ntyp = 1,
	ecutwfc = 30.0 ,
	ecutrho = 100.0,
	nr1 = 1200,
	nr2 = 1200,
	nr3 = 120,
	occupations = 'smearing' ,
	degauss = 0.06 ,
	smearing = 'marzari-vanderbilt' ,
/

I ran many times pw.x with differents values of nr1, nr2, nr3, ranging
from 300,300,30 to 1200,1200,120 (accordig to my desidered supercell
dimensions). And, as far as I know, these values determine the "FFT grid"
for output rho charge density.

The problem is, whatever the matrix grid values set in nr1, nr2, nr3, are,
the output file with electron density matrix contains points with NEGATIVE
electronic density values. These values are located in the supercell
sectors where the density is supposed to be very close to zero, and the
negative values are very small in modulus. Are these negative values, just
numerical errors of some kind (like rounding, chopping, etc) in pw.x? and
in such a case, what is the numerical precision, pw.x uses to calculate
the density? and  There is a way to minimize this error?

Thanks in advance.