Dear All, I believe this is a basis problem. See below from the archive.
Marty http://www.mail-archive.com/[email protected]/msg02951.html On 14/11/2004, at 6:42 AM, Javier Junquera wrote: Dear Rainer: When you are working with a non-orthogonal basis set, as is the case in Siesta, neither the PDOS nor the Mulliken population analysis are positive definite magnitudes. The reason is due to the fact that the PDOS is defined as: g_mu(eps) = sum_nu rho_mu,nu S_nu,mu delta(eps-eps_i) rho_mu,nu = sum_i C_mu,i C*_nu,i (i stands for the bands, S stands for the overlap matrix, and C stands for the coefficients of the wave function) and the non-diagonal elements of the density matrix might be negative. However, the term in the diagonal IS positive definite, and it is usually larger than the rest of the terms. Therefore, when the PDOS is negative the value should be small in absolute value. Hope this helps, Javier
