Comparative electron spin resonance study of epi-Lu2O3/(111)Si and a-Lu2O3/(100)Si interfaces: Misfit point defects

Abstract

An electron spin resonance study has been carried out on heteroepitaxial Si/insulator structures obtained through growth of epi-Lu2O3 films on (111)Si (∼4.5% mismatch) by molecular-beam epitaxy, with special attention to the inherent quality as well as the thermal stability of interfaces, monitored through occurring paramagnetic point defects. This indicates the presence, in the as-grown state, of Pb defects (∼5×1011 cm−2) with the unpaired sp3 Si dangling bond along the [111] interface normal, the archetypical defect (trap) of the standard thermal (111)Si/SiO2 interface, directly revealing, and identified as the result of, imperfect epitaxy. The occurrence of Pb defects, a major system of electrically detrimental interface traps, is ascribed to lattice mismatch with related introduction of misfit dislocations. This interface nature appears to persist for annealing in vacuum up to a temperature Tan∼420 °C. Yet, in the range Tan∼420–550 °C, the interface starts to “degrade” to standard Si/SiO2 properties, as indicated by the gradually increasing Pb density and attendant appearance of the EX center, an SiO2-associated defect. At Tan∼700 °C, [Pb] has increased to about 1.3 times the value for standard thermal (111)Si/SiO2, to remain constant up to Tan∼1000 °C, indicative of an unaltered interface structure. Annealing at Tan>1000 °C results in disintegration altogether of the Si/SiO2-type interface. Passivation anneal in H2 (405 °C) alarmingly fails to deactivate the Pb system to the device grade (sub) 1010 cm−2 eV−1 level, which would disfavor c-Lu2O3 as a suitable future high-κ replacement for the a-SiO2 gate dielectric. Comparison of the thermal stability of the c-Lu2O3/(111)Si interface with that of molecular-beam deposited amorphous-Lu2O3/(100)Si shows the former to be superior, yet unlikely to meet technological thermal budget requirements. No Lu2O3-specific point defects could be observed.