Comment on Electronic structure of spin-(1)/(2) Heisenberg antiferromagnetic systems: Ba2Cu(PO4)(2) and Sr2Cu(PO4)(2)

Recently S.S. Salunke et al. [Phys. Rev. B 76, 085104 (2007)] reinvestigated the electronic and magnetic properties of the low-dimensional spin-1/2 materials Sr2Cu(PO4)(2) and Ba2Cu(PO4)(2). Based on a NMTO downfolding methodology their main result is a considerably reduced transfer term along the magnetic chains compared to an earlier study (M. D. Johannes et al., Phys. Rev. B 74, 174435 (2006)]. The discrepancy is assigned to the Nth-order muffin-tin orbital mapping procedure that is suggested to be more accurate than the tight-binding approach taken by Johannes et al. Here, we demonstrate that in contrast to the suggestion of Salunke et al., the discrepancy arises solely from the employment of the atomic-sphere approximation in the underlying band-structure calculation rather than from the mapping scheme used. By comparison of the bandwidths of Salunke et al. to those obtained using three different full-potential methods we find that the full-potential methods are all in nearly exact agreement with one another and yield an about 30% larger bandwidth compared to the results in Salunke et al.. In general, our results emphasize the need for a full-potential description especially for strongly anisotropic structures as a precondition for a subsequent accurate modeling. Furthermore, we comment on the exact diagonalization results given by Salunke et al..