Intrinsic and Extrinsic Threshold Voltage Shift Dependence on the Oxide Field Induced during Optically Assisted Electron Injection

Abstract

The effects of injecting carriers at different gate insulator fields on observed threshold voltage shifts (hVt), during optically assisted electron injection, to quantify charged and neutral defect densities in unirradiated and, for the first time, x-ray irradiated [2.4 Mrads (SiO2)] devices were examined. In unirradiated devices, where intrinisic fixed positive charge 16 2(FPC) densities are below measurement limits, AV~ accompanying an injection fluence of 1.25 • 10 e/cm [the fluence normally used to fill large cross section neutral electron traps (NETs)] in the range of gate insulator fields (EJ between 1.5 and 7 MV/cm was found to decrease from positive to negative values. This was interpreted as indicating a decrease in large NET cross sections with increasing gate insulator field above 1.5 MV/cm, accompanied by the net generation of FPCs, at Eo~ >-5 MV/cm, as evidenced by V~ dropping below its initial value. Based on two stage injections, the first injection for Eox= 0.5 to 7 MV/cm, followed by reinjection at Eox= 0.7 MV/cm, it was concluded also that for Er -> 5 MV/cm, large cross section NETs and fixed negative charges, in addition to the FPCs mentioned above, were generated, and that field alone does not result in defect generation. In irradiated devices, [2.4 Mrads (SiO~.)] examined in a similar manner, the high-field behavior was somewhat different. At an injection fluence Of 2.5 • 10 e/cm ~' , (the fluence normally used to fill FPCs) the accompanying hVt first decreased to around zero which was attributed to an FPC cross section decrease, and then increased slightly with increasing oxide fields, attributed to additional large angle scattering. At an additional injection fluence of 16 21.25 • 10 e/cm' (again, the fluence normally used to fill NETs), the accompanying AV. first increased, as expected due to the filling of both NETs and any unfilled FPCs, and then decreased at fox -> 2.5 MV/cm. The decrease in AVt at the higher fields is similar to that seen in the unirradiated devices, indicating again that the NET cross section decreased. However, in this case Vt never dropped below its initial value and AV, measured during the subsequent low field rcinjection for FPCs, showed no increase in the number of FPCs above that initially present, so there is no direct evidence of net FPC generation. In these devices using two stage injection sequences, it was determined also that large cross section NETs were not generated. The effects of the device being "on" or "off" during injection, depending on the applied substrate bias, were also examined using "continuous" injections. Evidence is presented which suggests strongly that to obtain the proper cross sections and/or densities of defects, the devices must be on.