Optical excitations in Bi2Sr2CuO6 and Bi2Sr2CaCu2O8: Evidence for localized (excitonic) and delocalized charge-transfer gaps

The optical properties of Bi2Sr2CuO6 (2:2:0:1) and Bi2Sr2CaCu2O8 (2:2:1:2) have been investigated by transmission electron-energy-loss spectroscopy. At low energy, E(loss) less-than-or-equal-to 10.0 eV, common features are observed in the spectra of these two materials at E(loss) = 0.0, approximately 2.7, 3.6, and 4.6 eV. The effective number of charges associated with these excitations has been estimated using the optical sum rule, and from the ratio of effective charges in the two materials, the origin of this feature has been inferred based on the relative number of Cu-O2 and Bi-O layers per formula unit. The effective-charge ratio for the free carriers at zero energy loss, N2:2:1:2/N2:2:0:1 = 5.0, could not be used to determine whether the carriers were in the Cu-O2 or Bi-O planes because the oxygen doping in the two materials was not known. But at E(loss) = 2.7 and 3.6 eV, the effective-charge ratio is 2.6 indicating that these transitions are associated with the Cu-O2 planes since the ratio is close to 2/1. The effective-charge ratio is 0.9 for E(loss) = 4.6 eV suggesting that this excitation is localized in the Bi-O planes. If the two excitations in the Cu-O2 layers of 2:2:0:1 and 2:2:1:2 are identified with the delocalized and localized charge-transfer reactions [Mark S. Hybertsen, Michael Schluter, and Niels E. Christensen, Phys. Rev. B 39, 9028 (1989)], then the three-band Hubbard parameters E = E(p) - E(d) and U(pd) (E is the energy difference between the Cu 3d and O 2p levels and U(pd) is the Coulomb repulsion between two holes occupying adjacent Cu and O sites) are 1.8 and 0.9 eV, respectively, with an uncertainty of +/- 0.5 eV.