No abstract

No abstract

No abstract

There has been a substantial effort made on the application of Nicollian–Goetzberger’s conductance technique to probe the Si-SiO2 interface traps on n-type substrates. However, it was reported that conductance measurement on the p-type substrate was impossible due to the strong surface potential fluctuations. By using metal-oxide-semiconductor (MOS) capacitors with thin (88–434 Å) oxides to damp out the fluctuations arising from the interface charge inhomogeneities, it is possible to carry out an accurate conductance measurement on as-oxidized p-MOS capacitors. A systematic dependence of the interface trap density on the oxide thickness and oxidation temperature is observed. The hole capture cross sections have no obvious dependence on the process conditions, but show an exponential dependence on the energy. Both the magnitude and bias dependence of the measured time-constant dispersion parameters are found to be much larger than those theoretically predicted. Results of numerical simulation show that the anomalous width of the conductance peaks observed for p-MOS structures is caused by the strong energy dependence of the hole capture cross sections, which has been overlooked in the conventional theory of MOS conductance.

Random telegraph signal (RTS) measurements have been used to study individual hot-camer-induced trap in nMOS-FET's. Single-trap flling and emptying can ,,cause 0.1% step noise in channel current. Trap location (3-10 A from interface), time constant ( -10 ms), and energy are found to be quite different from those of prestress (process-induced) traps. The type (acceptor or donor) of the traps can also be identified by RTS measurements; both the process and stress-induced traps with energies near the conduction band edge are found to be of the acceptor type for nMOSFET's and trap levels near the valence band edge are found to be of the donor type for pMOSFET's. I. HOT-CARRIER INDUCED RTS HE INFLUENCE of hot-carrier injection on the perfor-T mance of MOS transistors has been studied extensively during the last decade. The recent development of random telegraph signal (RTS) measurement offers a sensitive tool that may reveal some unexplored aspects of interface traps by studying the trapping-detrapping process of individual traps [11-[31. The deep-submicrometer MOSFET's used in this study were fabricated using a photoresist-ashing technique [4]. The oxide thickness is 8.6 nm and substrate doping density is 5 x 1017 ~m -~. The drain current fluctuations at room temperature were recorded by an HP3651A spectrum analyzer. Details of the measurement system can be found in [5]. Fig. l(a) shows the typical current fluctuation of an nMOSFET with We, = 0.5 pm and Le, = 0.35 pm. The striking twolevel current fluctuation is due to the filling and emptying of a single interface trap. It is known as random telegraph noise, which is observed only when the channel area is small enough to contain only one trap within kT's from the surface Fermi level. Fig. l@) shows the current noise after 10 min of channel hot-electron stressing at Vg = 2 V, V, = 4.5 V, and Zsub = 10 PA. New short pulses superimposing on the original RTS are observed. Fig. l(c) is a portion of Fig. l@) in Manuscript