# [Pw_forum] Negative values in density matrix.

jorge.gallardo at cab.cnea.gov.ar jorge.gallardo at cab.cnea.gov.ar
Tue Oct 9 19:12:24 CEST 2012

> Dear Jorge,
>  I do not understand what you are trying to do. In particular, nr1, nr2
> and
> nr2 have nothing to do with defining a supercell: you should not change
> them except in very special cases.
>
> The fact that you get some negative rho  may or may not be normal
> depending
> on a lot of factors you did not specify. Especially the geometry of your
> crystal and the magnitude of said negative rho.
>

Thank you for you response.
I'm trying to find a numerical approximation for the electronic density as
a function of its coordinates rho(x,y,z) (Yes, I know QE wasn't made just
for this, but is the better tool I found) for an aluminum orthogonal
supercell :

ibrav = 6,
celldm(1) = 50.0,
celldm(3) = 0.1,

Of course, nr1, nr2 and nr3 have nothing to do with the supercell
dimensions. But as far as i know, they do with the matrix density pw.x
uses. And according to the responses when I queried before, these
parameters change the dimensions of the electronic density matrix pw.x
uses for its calculations.

As a control, integrating numerically that matrix:   int rho(x,y,z) dx dy
dz,  in the space surrounding the atom, I found a number that approaches
the charge of 3 electrons  which is the valence charge of an Aluminum one.
In addition, by plotting this matrix I get a map density (as other methods
in QE do).

But I need the matrix rho(x,y,z) for other calculations (related with
plasmon excitations) QE doesn't perform. The only problem I found is to
deal with the negative density values (very, very close to zero) which *I
suspect*, but not sure, are just numerical errors (because negative
density has no meaning in this context).

thank you