

Excitonic Effects In CoreExcitation Spectra Of Semiconductors
R. Buczko,^{1,2,3} G. Duscher,^{4} S. J. Pennycook^{1}, S.T. Pantelides^{2,1}
Full Article (PDF 168 KB)
Coreelectron excitation spectra are used widely for structural and chemical analysis of materials, but interpretation of the nearedge structure has been controversial because electronhole correlations may produce bound excitons below the band edge and also distort the continuum spectrum. For the important Si L_{2,3} edge, two mutually inconsistent interpretations have been used, in terms of effectivemass excitons and in terms of Bloch conductionband final states. Abinitio calculations show that neither interpretation is valid.
Calculations were performed using an allelectron energyband code (Fullpotential Linearized Augmented Plane Wave method). The resulting spectrum is shown in Fig. 1 using a threshold of 99.8 eV. The absolute magnitude of the absorption coefficient has not been adjusted. It is clear that spectra calculated without electronhole interactions cannot reproduce the sharp onset of the experimental spectrum. Core hole interactions were then included explicitly and the result showed good agreement with experiment (dotted line). We also used the Z + 1 approximation, replacing the core hole with an additional proton, which reproduced the shape but not the absolute magnitude of the edge. Finally we calculated the density of Bloch conductionband final states multiplied by a matrix element extracted from the same calculation that produced the dotted curve (dashdotted line). This predicts a cross section about an order of magnitude too small. Even adding effectivemass excitonic effects (longdashed line) cannot account for the observed spectrum.
The calculations show that the common belief that energy loss spectra measure the conduction band density of states is not correct. Accurate spectral shapes may be calculated through an allelectron code using the Z + 1 approximation or even a pseudopotential code provided matrix elements are calculated from an all electron code. Such calculations enable experimental spectra to be inverted to deduce local bonding configurations at nanostructures and interfaces.






Fig. 1. Experimental and theoretical Xray absorption spectra of the Si L_{2,3} edge. Thick curve: experimental; solid curve: full calculation including electronhole interactions with a selfconsistent hole orbital; dashed curve: calculation using the Z+1 model; dotted curve: calculation of excitation spectrum without electronhole interactions; dashdot curve: effectivemass (parabolic band) spectrum without electronhole interactions; longdash curve: effectivemass spectrum with hydrogenic excitonic effects 



 SSD ORNL
 Vanderbilt University, Nashville, TN.
 Polish Academy of Sciences, Warsaw, Poland
 North Carolina State University, Raleigh, NC


