We have used x-ray reflectivity and transmission electron microscopy to study the CaFi/SiCl 11) interface. The results are consistent with a reconstructed two-layer CaF interface which can be transformed to a diff'erent structure simply by increasing the thickness of the CaF2 overlayer. We are able to reconcile previous measurements of the interface structure and gain insight into the rich variety of phenomena that may be observed at heteroepitaxial interfaces.PACS numbers: 61.16. Bg, 68.35.Rh, 68.55.Jk CaF2/Si(lll) is a prototypical system for studying atomic and electronic structure at the ionic/covalent interface. As a consequence there has been a wide range of experiments aimed at determining the interface structure and a number of models have been proposed [1,2]. Despite this attention no consistent picture has emerged. Conflicting results have been attributed to differences in sample preparation and the varying sensitivity of experimental techniques [3][4][5]. The problem relates to more general issues concerning the differences between surface and interface structure. In particular, it is questionable whether the structure observed at monolayer coverages, which is accessible to many standard surface probes, is representative of the true interface between two materials [6]. The latter can only be studied with techniques that employ highly penetrating beams, such as x-ray diffraction and transmission electron microscopy (TEM). It is the structure of the buried interface that influences the electronic properties.In this Letter we present x-ray diffraction and TEM measurements of the CaFi/Sid 11) interface. The results are consistent with a reconstructed two-layer CaF interface between the top Si double layer and the "bulklike" CaF2 film. The interface structure is a metastable phase formed in the early stages of heteroepitaxy. By increasing the thickness of the CaF2 film it is possible to drive a transition to a single layer CaF interface. Our results are consistent with previous experiments and reconcile many of the differences in proposed interface models. The results indicate that the rich variety of phenomena observed at surfaces, e.g., reconstructions and phase transitions, may also be observed at buried interfaces.CaF2 was grown by molecular beam epitaxy (MBE) on well-oriented (miscut <0.5°) Si(lll) substrates in an ISA/Riber CBE 32P ultrahigh vacuum (UHV) system. The Shiraki etched Si substrates [7] were outgassed thoroughly and then heated to 880 °C to remove the protective oxide. Cooling to the growth temperature (720°C) routinely resulted in a sharp (7x7) reflection high-energy electron diffraction (RHEED) pattern. CaF2 was deposited by evaporation from an effusion cell at 1150°C. During deposition the pressure was typically ~1 xl0~^^ torr at a growth rate of ~ 1 CaF2 triple layer (TL) per 30 s. At 720 °C CaF2 films with a thickness of -7 TL's ( -22 A) were grown. This is below the critical thickness for strained layer growth (~12 TL's). The sample was then cooled to room temperature where, due t...