Experimental and Simulation Analysis of Program/Retention Transients in Silicon Nitride-Based NVM Cells

Abstract

A new characterization technique and an improved model for charge injection and transport through ONO gate stacks are used to investigate the program/retention sequence of silicon nitride-based (SONOS/TANOS) nonvolatile memories. The model accounts for drift-diffusion transport in the conduction band of silicon nitride (SiN). A priori assumptions on the spatial distribution of the charge at the beginning of the program/retention operations are not needed. We show that the carrier transport in the SiN layer impacts the spatial distribution of the trapped charge and, consequently, several aspects of program and retention transients. A few model improvements allow us to reconcile the apparent discrepancy between the values of silicon nitride trap energies extracted from program and retention experiments, thus reducing the number of model parameters

Explanation of the Charge Trapping Properties of Silicon Nitride Storage Layers for NVMs—Part II: Atomistic and Electrical Modeling

“…In this respect, it is worth noting that the obtained N T SidB value (2.5 · 10 19 cm −3 ) is comparable with the N T 1Si−H concentration, in agreement with the similar Gibbs free energies calculated for the two defects considering the fabrication conditions in the Si-rich recipe [3]. All the other model parameters, and in particular those describing the traps, have been taken as in [46], except for E T = 1.4 eV, as previously explained. Fig.…”

“…As a first step to verify if the insight obtained by means of atomistic simulations is indeed useful to explain the dependence of cell performance on SiN stoichiometry, we plugged the calculated trap parameters for std SiN in the electrical model described in detail in [46] and then simulated high temperature retention experiments. The model solves the electrostatics self-consistently with the in and out electron tunneling fluxes, the drift-diffusion transport in the SiN conduction band, and the Shockley-Read-Hall generation/recombination (considering Poole-Frenkel emission).…”

“…Therefore, in this condition, the retention curves are very sensitive to the energy level of the filled traps. All the main model parameters have been taken from [46], except for the trap energy (in retention conditions), which has been adjusted to the value E T = 1.4 eV, in agreement with the G 0 W 0 calculations. Fig.…”

“…However, the net charge responsible for the threshold voltage shift is only Q N = −q(n T − N T SidB ). By using this approach, it is possible to simulate with the same model equations described in [46] a cell featuring a SiN layer with any mixture of 1Si-H and Si dB defects (with a density of N T 1Si−H and N T SidB , respectively). In fact, as summarized in Table II, while considering a unique trapped electron concentration (n T ), we simply assume that the initial (corresponding to the neutral state) n T value is N T SidB .…”

Impact of Total Ionizing Dose on the Data Retention of a 65 nm SONOS-Based NOR Flash

“…Other researchers have shown already that the direct tunneling mechanism is only minor temperature dependent and the thermal emission component is very sensitive to temperature and the energy distribution of the trapped charge. A sharply peaked Gaussian energy profile with an average trap energy of 0.9 eV-1.6 eV has been demonstrated to fit experimental data retention data well [7], [8].…”