THE EFFECT OF TEMPERATURE ON ELECTRON-CAPTURE CROSS-SECTIONS AND DENSITIES IN IRRADIATED [2.4 MRADS(SIO2)] IGFETS

The temperature dependence of electron trapping within the gate insula tor of x-ray irradiated [2.4 Mrads(SiO2)] insulated gate field effect devices, using low field optically assisted electron injection was stu died. Threshold voltage shifts (Delta V-t) were obtained at 100, 200, and 300 K via automatic ''continuous'' electron injection over the inj ection range, N-inj, from 10(11) to l0(17) e/cm(2), at a substrate bia s of -7 V and an insulator field of 1 MV/cm at controlled electron flu xes. Under these conditions, during the ''continuous'' injections, the devices were always ''on.'' In order to model the oxide defects using first-order trapping kinetics over the range of N-inj studied, it was necessary to invoke a three-trap model. At room temperature, the capt ure cross sections, sigma s, of the three traps are 1.04 +/- 0.1 x 10( -13), 6.5 +/- 0.9 x 10(-16), and 5.7 +/- 0.5 x 10(-15) cm(2). The smal lest of these traps had not been previously reported in studies coveri ng the injection range up to 1.25 x 10(16) e/cm(2). The largest trap i s coulombic in nature, and the two smaller traps are neutral, since an nealing of these latter is not accompanied by a threshold voltage chan ge. The 10(-13) trap is what we have previously termed fixed positive charge (FEC) and the 10-(16) trap is what we have previously termed th e large neutral electron trap (NET). The 10(-17) trap we term the NET( 2). Over the temperature range studied, nd new traps,other than those present at room temperature were detected. As the temperature is decre ased from 300 to 100 K, sigma and trap density, N-T, of the FPC did no t change significantly. On the other hand, both the sigma s and N(T)s of the NET and NET(2) were each found to increase, as the temperature was lowered from 300 to 100 K. The fact that both the derived values o f sigma and N-T of the NET and NET(2) change is probably an indication that at the higher temperatures, at least, a fraction of tile filled traps depopulate under the injection conditions employed. While sigma might be expected to increase as T is decreased, the value of N-T for each trap should not change. The values of sigma and N-T derived from an analysis of the 100 K data are, therefore, expected to be fundament ally more accurate than the values : obtained at 300K. The variations of transconductance, g(m), with both temperature and the number of inj ected electrons were also studied. During characterization, as FPCs ar e annihilated, g(m) increases, and as NETs and NET(2)s are ''labeled,' ' g(m) is observed to decrease,both showing, therefore, the effects of coulombic charge on g(m), which is similar to the effect of interface states on channel mobility.