In this work, we introduce a discretized-gate-capacitor (DGC) EPROM model and a post-processing quasi-transient (PPQT) method for efficient simulations of the programming and erasing of nonvolatile memory devices. The DGC model and PPQT method have been implemented in a 2D device simulator-UMDFET2[1] which solves a 2D Generalized Energy Transport (G-ET) model self-consistently, and includs impact ionization, hot-electron gate-injection, Fowler-Nordheim tunneling and Band-teBand tunneling. This integrated efficient method requires one order of magnitude less CPU time than conventional methods[2, 31 as used in commercial simulation software, and is very accuracy. This method has been used in the design and optimization of deepsubmicron Flash/EPROM devices.The DGC EPROM device model shown in Fig. 1 uses many small discretized gate capacitors Cf(i) defined by numerical mesh to take into account the interface potential variation V;.(i), so the floating gate voltage V,, can be accurately determined. This model also uses a "virtual control gate" and a control gate capacitor C., to replace the physical inter-poly dielectric structure. The 3-D gate coupling effect of the Poly 1 and Poly 2 overlap on the field oxide can thereby be effectively taken into account, and the treatment of the charge boundary condition is simplified. This DGC EPROM model combines the accuracy of the physical device model and the efficiency of the analytical device model. It requires much less CPI; time and has better analysis ability than that using a full physical device model. It also have the flexibility for post processing, shifting during the programming and erasing operations, the steady state solutions were first calculated to determine the initial threshold &(O), the interface potential vi(.), and the gate current 1, as a functions of Vj,, which 15 to 20 minutes CPU time on SUN/SPARC2. It then took less than 1 second of CPU time using PPQT method to obtain the &(t) shift curves shown in Fig. 3 for programming, and in Fig. 4 for erasing. The main advantage of the PPQT method is that it performs the steady state device simulation only once. The effects of different C,,, initial charge on floating gate, different control gate voltage V,, and arbitrary progra"ing/erasing time can be easily calculated within a second. The accuracy of deep-submicron device simulation has been verified by using 2D G-ET model [l], which is able IO accurately calculate the velocity overshoot, hot-electron energy, anci the impact ionization generation rate. The gate current was calculated by a modified Richarson's hot-electron injection formula, in which the Maxwellian exponent is replaced by the more realistic electron energy distributions based on Monte Carlo (MC) calculation. Fig. 5 shows the obtained excellent agreement between the calculation and the data from the indirect measurement method.The summary results shown in Table 1 demonstrate that the electrical characteristics of EPROM devices with L,jf=0.25 and 0.45 pm were predicted with better that 90% accura...
