Distribution of occupied near-surface band-gap states in a-Si:H.

Abstract

We investigate the distribution of occupied band-gap states in undoped, B-doped, and P-doped a-Si:H within the first \ensuremath{\sim}100 A\r{} of the surface using total-yield photoelectron spectroscopy in combination with the Kelvin probe. In clean, undoped a-Si:H the occupied density of states extracted from the measured yield spectrum consists of a linear valence-band edge, an exponential valence-band tail decreasing into the gap, and a broad band of deep defect states superposed on this tail. The deep-defect density of undoped a-Si:H measured by total yield is 2 orders of magnitude larger than that measured by photothermal deflection spectroscopy (PDS) on simultaneously prepared 1-\ensuremath{\mu}m-thick samples, from which we infer the existence of an excess density of deep defect states near the surface corresponding to an equivalent surface-state density of 3\ifmmode\times\else\texttimes\fi{}${10}^{11}$ ${\mathrm{cm}}^{\mathrm{\ensuremath{-}}2}$.The addition of diborane (phosphine) to the glow-discharge plasma decreases (increases) the density of excess occupied near-surface defect states in the gap and shifts the Fermi energy towards the valence- (conduction-) band edge. A 0.5-eV increase in the work function over that in undoped a-Si:H accompanies the \ensuremath{\sim}98% reduction of these excess occupied near-surface deep defect states by ${10}^{\mathrm{\ensuremath{-}}5}$ B doping, from which we infer an (0.4--0.5)-eV downward band bending at the clean, undoped a-Si:H surface. The inverse logarithmic slope of the exponential intrinsic valence-band tail observed in B-doped a-Si:H is 45 meV, in excellent agreement with that inferred from dispersive-transport, electron-spin-resonance (ESR), and traveling-wave data. Dividing the occupied density of states of ${10}^{\mathrm{\ensuremath{-}}3}$ P-doped a-Si:H by the Fermi-Dirac distribution function, we observe an exponential conduction-band tail extending over more than 3 orders of magnitude in the density of states. Its inverse logarithmic slope of \ensuremath{\sim}35 meV agrees with that inferred from ESR and traveling-wave data on P-doped a-Si:H, but is slightly larger than the slope inferred from dispersive-transport measurements on undoped a-Si:H. This discrepancy arises from the presence of a large concentration of P donors which affects the deeper tail-state distribution observed in heavily-P-doped a-Si:H. The average density of occupied states at the Fermi energy ${E}_{F}$ for the 30 doped samples we have studied is 8(\ifmmode\pm\else\textpm\fi{}2)\ifmmode\times\else\texttimes\fi{}${10}^{15}$ states/eV ${\mathrm{cm}}^{3}$, in good agreement with the thermodynamic minimum density of defect states permitted in undoped a-Si:H. This implies that the Fermi level lies at a minimum in the a-Si:H density of states for all doping levels.From this result we infer that the position of the ${D}^{+}$ defect level in B-doped a-Si:H lies more than 0.5 eV above the ${D}^{\mathrm{\ensuremath{-}}}$ level in P-doped a-Si:H; an arrangement in conflict with a fixed energy distribution of deep defect states and with the generally accepted positive value of the defect correlation energy in a-Si:H. We resolve this seeming discrepancy by postulating that the amorphous network responds to changes in ${E}_{F}$ by changing its defect structure so as to minimize the total energy of the system. This postulate leads to a variable energy distribution of deep defect states that depends only on the position of ${E}_{F}$ and the defect-formation energies; intimate pairing of dopants and defects, which has been suggested to account for this discrepancy, is not required. The relation between the surface and bulk distributions of localized gap states in a-Si:H is discussed.