Deep-level transient spectroscopy was used to study polycrystalline material. Peaks were observed which corresponded to deep levels in samples where a p-n junction lay very close to the poly/monocrystalline interface. The sign of these peaks was the opposite to that which was expected for deep levels that were associated with point defects. This fact, and the unusually broad nature of the peaks, led to the deduction that the observed deep levels were due to grain boundaries. This was supported by the time dependence of the capture rate, and by an observed activation energy which was consistent with the value that had previously been reported for grain boundaries. The defect density which gave rise to the signal depended upon F implantation of the polysilicon and upon the chemical treatment which was given to the monosilicon before polysilicon deposition.

D.P.Parton, T.Markvart, P.Ashburn, J.C.Carter, L.Castañer: Materials Science Forum, 1995, 196-201 1903-8