Mitsuru Yamaji

A MONTE CARLO SIMULATION OF ANISOTROPIC ELECTRON TRANSPORT IN SILICON INCLUDING FULL BAND STRUCTURE AND ANISOTROPIC IMPACT-IONIZATION MODEL

The physics of electron transport in bulk silicon is investigated by using a newly developed Monte Carlo simulator which improves the state‐of‐the‐art treatment of hot carrier transport. (1) The full band structure of the semiconductor was computed by using an empirical‐pseudopotential method. (2) A phonon dispersion curve was obtained from an adiabatic bond‐charge model. (3) Electron‐phonon scattering was computed by using a rigid pseudo‐ion model. The calculated scattering rate is consistent with the full band structure and the phonon dispersion curve of silicon, thus leaving no adjustable parameters such as deformation potential coefficients. (4) The impact‐ionization rate was calculated by using Fermi’s golden rule directly from the full band structure. We took into account the dielectric function depending on both wave vector and transition energy in the numerical calculation of the rate. The impact‐ionization rate obtained in the present study strongly depends on both wave vector and band index of t...

Impact ionization model for full band Monte Carlo simulation

The impact ionization rate in silicon is numerically derived from wave functions and energy band structure based on an empirical pseudopotential method. The calculated impact ionization rate is well fitted to an analytical formula with a power exponent of 4.6, indicating soft threshold of impact ionization rate, which originates from the complexity of the Si band structure. The calculated impact ionization rate shows strong anisotropy at low electron energy (e<3 eV), while it becomes isotropic at higher energy. Numerical calculation also reveals that the average energy of secondary generated carriers depends linearly on the primary electron energy at the moment of their generation. A full band Monte Carlo simulation using the newly derived impact ionization rate demonstrates that calculated quantum yield and ionization coefficient agree well with reported experimental data.

Moment expansion approach to calculate impact ionization rate in submicron silicon devices

A method to calculate the impact ionization rate in submicron silicon devices is developed using both an average energy and an average square energy of electrons. The method consists of an impact ionization model formulated with the average energy and conservation equations for the average square energy in the framework of an energy transport model. Parameters for the transport equations are extracted in such a way that calculated moments based on these equations match Monte Carlo simulation results. The impact ionization generation rate in an n+nn+ structure calculated with this method agrees well with the results obtained from Monte Carlo simulation. The new method is also applied to a submicron n‐MOSFET. The calculated distribution of the generation rate is found to be quite different from the results based on a conventional method.

Monte Carlo study of impact ionization phenomena in small geometry MOSFET's

We developed a multi-valley Monte Carlo simulator in which realistic physical parameters based on ab-initio calculations are implemented. A nonlocal impact ionization coefficient in exponentially increasing field is extracted using the Monte Carlo simulator. On the basis of the new nonlocal impact ionization coefficient, an analytical substrate current expression for n-MOSFETs is derived. The new substrate current expression clarifies the reason why a reported theoretical characteristic length used in a pseudo two-dimensional MOSFET model differs from empirically derived ones. The nonlocal impact ionization coefficient implemented in a device simulator demonstrates that the new coefficient can predict substrate current correctly in the framework of the drift diffusion model.<<ETX>>

Impact Ionization Model Using Average Energy and Average Square Energy of Distribution Function

A new impact ionization model is developed using both the average energy and the average square energy of electrons. The accuracy of the model in an inhomogeneous electric field is verified through comparison with results calculated with Monte Carlo simulation. Our model predicts the generation rate even in an inhomogeneous field more accurately than local field, non-local field, and local energy models.

Dopant Profile Design Methodology for 65 nm Generation via Inverse Modeling

Abstract We have applied inverse modeling technique to analysis and design of scaling for 65 nm generation MOSFETs. Our model has well described trade-off relation between drive current and tolerance to short channel effect. Using the model, we have succeed to clarify specific problems of the present device and to quantitatively design a future targeted device.

Novel Impact Ionization Model Using Second- and Fourth-Order Moments of Distribution Function for Generalized Moment Conservation Equations

Device degradation caused by hot carriers has been main concern from the reliability point of view. Because secondary-generated carriers created by impact ionization (I.I.) have great influence on the degradation of gate oxide, accurate modeling of I.I. is necessary. We propose an I.I. model which is formulated using second- and fourth-order moments of distribution function for precise description of I.I. in inhomogeneous electric field. A set of moment conservation equations for carrier transport is also presented to perform practical device simulation with the I.I. model.

Impact Ionization Model Using Second- and Fourth-Order Moments of Distribution Function

This paper describes an impact ionization model suitable for calculation of an impact ionization rate in inhomogeneous electric field. The model is formulated using second- and fourth-order moments of an electron energy distribution function. A set of model equations for carrier transport in semiconductor devices is also presented to perform practical device simulation with the impact ionization model. The calculation result with the new models agrees to Monte Carlo simulation result.