The interpretation of the soft-x-ray absorption spectra of alkali halides has been very controversial. Many of the observed spectra have been interpreted by several authors in terms of transitions to the Bloch states of the conduction bands, allowing the possibility of excitonic enhancement of peaks but excluding the possibility that excitons are formed below the band edge. Other interpretations have ranged from the identification of excitons below the band edges with relatively small binding energies (∼1 eV), excitons within the band continuum associated with critical points, and free-atom-like transitions shifted in energy by crystal fields. This controversy can be conclusively resolved by determining the position of the threshold for band transitions on the energy axis. Any structure appearing below this threshold can then unambiguously be identified as pure excitons, i.e., bound states of the electron to the core hole. Structure above this threshold must be carefully checked against theories predicting interband spectra. In this paper we present the theory governing soft-x-ray excitations and use x-ray photoelectron spectroscopy (XPS) and optical-gap data to determine band thresholds for a total of 39 soft-x-ray absorption spectra. In almost all cases we present a detailed analysis of the observed structure. Our main conclusions are: (a) In the case of excitations from the alkali-ion core levels, the observed spectra are almost entirely excitonic in nature. Electron-hole interaction is overwhelmingly large since the core hole in the alkali ion is not effectively screened by the valence-electron cloud, which lies almost completely on the halogen ions: This interaction mixes band states over the range of a rydberg and resulting binding energies of excitons are very large. No critical-point association for the observed excitons can thus be supported. (b) In the case of excitations from the halogen-ion core levels, electron-hole interaction is weaker due to effective screening by the valence electrons. Excitons are still found in all cases and binding energies are again larger than hitherto accepted. Critical-point analysis of excitons is again doubtful but transitions to band states may be responsible for some of the observed structure. (c) Transitions to states above the conduction-band edge do not reflect the structure present in the density of states of the conduction bands; this is also a consequence of the strong electron-hole interactions. We believe that the the present analysis gives a conclusive interpretation of the major features of each spectrum which is not always in agreement with either the latest publication on the given spectrum or, sometimes, any of the interpretations proposed in the past.

• Received 24 June 1974

DOI:https://doi.org/10.1103/PhysRevB.11.2391