Integration of New Physical Models in ATHENA

Introduction

As a commercial process simulator ATHENA has to combine together physical models from different research groups into a single program. ATHENA covers arbitrary customer requirements by a hierarchical set of models for the complete process flow in a fully integrated environment. Models developed by different research groups are linked together. Within ATHENA these models are subjected to rigorous conditions. They have to yield reliable results in all types of applications for arbitrary geometries. All models are combined with state of the art numerics and gridding features.

For years TCAD tools suffered from limited use through user hostile interfaces and lack of code stability. A comprehensive set of interactive tools makes ATHENA easy to use and provides common functionality in separate easy-to-maintain programs. User access to all model parameters together with the C-Interpreter environment, a common feature of all SILVACO products, guarantees a maximum extent of customization.

This article will focus on the latest deep sub-micron related developments in silicon such as defect controlled diffusion, implantation and induced defects.

 

Diffusion Models

The three original diffusion models (FERMI,TWO.DIM, FULL.CPL) are simple point defect damage models. The default, FERMI, assumes a constant level of point defects and therefore does not account for defect enhanced diffusion. As TWO.DIM solves the two-dimensional distribution of point defects and includes point defect generation during implant and oxidation, it is suitable for oxidation/ silicidation enhanced diffusion. The fully coupled model (FULL.CPL) takes into account coupling between point defects and individual dopants. However, it ignores reactions between defects and defect pairs. It can be used for simulation of transient diffusion phenomena, low temperature diffusion and co-diffusion of dopants (emitter push effect). Damage is described as a simple distribution of both types of defects and sealed to the implanted dopant with the "scaled +1" model.

The advanced diffusion models are made up of a) Daniel Mathiot's advanced diffusion models, CNET, b) Peter Griffin and Scott Crowder Advanced Diffusion Models, SU.MOD and c) extensions to control implant damage profiles.

a) CNET model is an extension to full.cpl, allowing better description at very high dopant concentrations.
The main additions are: percolation model, static clustering model for Boron and Arsenic, dopant / defect
pairs contribution in the total diffusivity, dopant / defect pairs involve in bimolecular point defect recombination.
b) SU.MOD extends FULL.CPLwith dopant/defect interaction with other defects and with interfaces. Moreover, it
allows introduction of <311>-cluster during implantation. The model suggests that the clusters dissolve in time,
injecting point defects as they disappear (S. Crowder, IEDM 95, p 427). The rate of interstitials released into silicon
is decaying exponentially with time, the time constant being a function of temperature. Optionally, dislocation loop
interstitial sinks can be introduced. This model is a first order approximation for dislocation loop interaction with
point defects. The recombination rate is proportional to the local non-equilibrium interstitial concentration.
c) The advanced diffusion models require more flexible control of implant damage generation. By setting max. and
min. concentration threshold values, regions related to the implanted profile were defined for <311> clusters and
dislocation loops. Similar to the scaled +1 model, clusters are distributed within these specified regions. Above the
max. cluster threshold, the material is assumed to be amorphous.
 

Implantation Models

The ion implantation hierarchy is made up of the 2 basic model types: analytic and Monte Carlo. Universal tilt and rotation capability for both analytic and Monte Carlo calculations are available. Damage is calculated with the scaled +1 model for interstitials and also extended defects or with physical damage calculation based on energy loss in collision cascade (MC).

Dual Pearson tables based on the measurements of Al Tasch (for example A.F. Tasch et al., J. Electrochem. Soc., 136, p.810, 1989) have been implemented for a large parameter space including screen oxides. This model allows accurate simulation of channeling during implant. For implant parameter in amorphous targets, single Pearson tables are used. For multilayer targets three different correction approaches are available: dose matching (default), range matching and maximal range scaling. In 2D cases, a parabolic approximation for the depth dependence of transversal standard deviation is implemented and matched against MC results.

 

Outlook

New models are continually being developed by researchers around the world. Silvaco will continue to incorporate the best-available models for all process steps into ATHENA. Work is currently being done to improve low energy implantation models, introduce a more advanced stress dependent oxidation model (ISEN IEDM 1996) and improve extended defect models.

 

 

Figure 1. New pre-deposition models accurately simulate high dose effects.

 

Figure 2. Improved default coefficients with the <311> cluster model
provide accurate modeling.

 

Figure 3. Integration of the latest implant research allows ATHENA to model vertical and horizontal implant profiles analytically.