SIMSVerified Implant Models Released in ATHENA 4.3
ATHENA version 4.3 has been released containing significant developments in implantation and lithography simulation. The implantation improvements compliment the developments made in diffusion and RTA models released last year in ATHENA version 4.0. This combination of advanced implant and diffusion models provide a complete solution to simulation of deep submicron processes. The lithography developments allow Optical Proximity Correction (OPC) simulation for complex masks.
New SIMSVerified Dual Pearson (SVDP) Implant Models
The importance of accurate implant simulation has increased in recent years as thermal processing is reduced for subhalf micron processes. Modern diffusion models such as those released last year in ATHENA 4.0 require that the asimplanted profiles are highly accurate. This because the implant damage, the driving force behind RTA effects, has been consistently shown in research to be scaled to the doping profile though the UNIT.DAMAGE (or PLUS.ONE) model.
A new SIMS Verified Dual Pearson (SVDP) model has been implemented. This model is based on experiments and research conducted by Professor Al Tasch and coworkers in the University of Texas at Austin[1]. The model provides analytical implant tables that describe:
 dual pearson distributions to describe channeling for B, BF2, P and As
 Older models contained dual pearson distributions only for Boron. Channeling for P and As is significant at low tilt angles.
 tilt angle dependence of channeling profile
 The new implant tables contain separate implant moments from measured data for several tilt angles. Interpolation between tilt angles is done by the model. Previous model used a simple cosine relationship to describe the range as a function of tilt angle.
 dose dependent channeling
 The new tables have dose dependent moments that are especially apparent in the reduction of channeling in high dose source/drain implants when compared to lower doped LDD implants. (See Figure 1.)

Figure 1. Dose dependence of phosphorus implants along <100> channel in Si.

Curves  ATHENA simulations using the SVDP model,

Experimental points R.J. Schreutelkamp et.al., Nucl. Instr. Meth. B55, 615 (1991)


 Screen oxide dependent channeling
 The new tables contain separate implant moments from measured data for several screen oxide thicknesses for Boron. Interpolation between tilt oxide thickness is done by the model. The parameter S.OXIDE is used to set the screen oxide thickness. Automation of S.OXIDE can be done by measuring the oxide prior to implant and using $substition of the variables.
The SVDP implant model is used over a range of dose, energy, tilt,rotation and screen oxide thickness determined by the original research[1]. Outside this range the program uses implant moments from the previous model tables. The parameter PRINT.MOM can be used on the implant statement to get explicit data on the implant moments used and their source table.
Modeling of Lateral Implant Range
The FULL.LAT model for accurate simulation of 2D implant profiles has been improved. It can now be used together with the SVDP model. It also includes lateral distributions based on symmetrical Pearson functions.
The MOMENTS statement functionality has been extended to support the SVDP and improved 2D implant models.
Figure 2. Simulated 2D profiles for 200 keV
B implant into amorphous Si are shown in Figure 2a .
The solid contours are the FULL.LAT analytical model, dashed contours are
Monte Carlo simulation.
Implantation takes place at the point (0,0) in the vertical direction downwards.
Figure 2b is the same as Figure 2a, with the standard analytical model is
used showing poor agreement.
Implantation into MultiLayer Targets
Two additional models for implants in multi layered structures are added. They are based on the range (RP.SCALE) or maximum range (MAX.SCALE) scaling. The dose matching model (DOSE.MATCH) remains a default. Differences between these models are important when the range of the implant is close to the thickness of a screen layer. (See Figure 3.) The parameter SCALE.MOM invokes moments scaling algorithm which can be used with any of above models to scale the default implant moments.
Figure 3. 100 keV phosphorus implant into multilayer
structure.
Solid curve  analytical dose matching method (default);
Dashed curve analytical maximal range scaling method
(MAX.SCALE); Squares  Monte Carlo simulation.
Improvements in MonteCarlo Implantation
Amorphous Monte Carlo Implant Model has been improved. It now uses much more accurate model for electronic stopping which substantially increased the accuracy of MC profiles for intermediate (few hundreds KeV) and high (few MeV) implant energies. In addition nonphysical spikes in MC profiles near the surface are eliminated because a more sophisticated freepathlength estimation algorithm is used now. (See Figure 4.)
New parameter IMPCT.POINT allows simulation of point source Monte Carlo implant to produce a 2D distribution. Extraction of spatial moments using parameter PRINT.MOM is also supported in this mode.
Figure 4. 400 keV Phosphorus Implant into Si.
Solid curve
 Amorphous Monte Carlo Simulation using version 4.3.0.R Dashed line
 the same using version 4.0.0.R Squares  experiment (single crystal silicon,
7 degrees tilt) Amorphous MC model matches well to the surface profile
and peak. The crystalline MC model is required to match channeling tail
Diffusion Simulation Enhancements
In the <311> cluster model[3] used for RTA simulations (CLUSTER.DAM) the decrease of interstitial concentration in <311> clusters during interstitial release is now included. This allows correct simulation of RTA cycles which include several independent temperature steps. (See Figure 5).
Figure 5. Comparison of default diffusion model
parameters
to measured data. Coefficients in Version 4 based
on latest research provide best fit.
The default model file "athenamod" has been updated. It includes better set of parameters for interstitial diffusion and the <311> cluster model. Unlike the FULL.CPL model the CLUSTER.DAM model can predict the transient enhanced diffusion effect over a range of time and temperature with the same default conditions. An example is shown in Figure 6 and 7.
Figure 6. Transient enhanced diffusion of arsenic
at 850oC.
The 311 clusters dissolve over approx. 60 seconds
releasing interstitials to fuel the enhanced diffusion.
Figure 7. Temperature dependent RTA effects
can be modeled using default diffusion coefficients in ATHENA 4.3.
For each of four temperatures (750oC > 900oC) simulations are run for
a critical TED time documented in [2]
and for 10 times the critical time. Results show how long the TED effect
lasts at low temperatures.
In order to support backward compatibility the model files for two previous versions 4.0 and 3.0 are also available by using commands athena modfile 96 and athena modfile 95 correspondingly. This allows users with concerns about backwards compatibility to check the new parameter set against the old easily.
A prototype model file is also available using athena modfile prototype. This model file contains parameters collected from recent published research. The effect of these newest parameter sets on typical simulation is still under evaluation. However used parameters from this file can be as a guideline for calibration purposes.
Oxidation Simulation Enhancements
The parameter FLIP.FACTOR has been added to the METHOD statement. It allows to control mesh behavior and improve convergence of oxidation simulation. This parameter varies from zero to one. Changes from the default are recommended only for those cases when oxidation simulation experiences convergence problems.
FLASH Enhancements
Cinterpreter capability is implemented for all impurities recognized by FLASH. Parameters DIF.COEF, SEG.CALC, and ACT.CALC in the IMPURITY statement specify the files which contain Cinterpreter functions for diffusivity, segregation,and activation of specified impurity in selected material.
Optolith Enhancements
For this release, there are two new features for Optolith. One is the capability for performing optical proximity correction (OPC). This OPC capability is done in conjunction with MASKVIEWS. Details of this feature have been described in the previous issue of Simulation Standard [3] and will not be re)peated here.
The other new feature is the pattern on the mask is no longer approximated by upright rectangles. The partitioning of the mask pattern can now be done by using rectangular, triangular and circular shapes with arbitrary orientation. With this flexibility, it will greatly enhance the accuracy and computational speed for the nonManhattan geometry, which occurs quite often in the actual layout pattern. The aerial image for a 45 ( wire joint and a 0.5 mm circular contact opening is displayed in Figure 8. The LAYOUT command has been revised to accommodate these changes, for example,
layout x.circle=0.0 z.circle=0.0 radius=0.25 layout x.lo=1.0 z.lo=0.75 x.hi=1.3 z.hi=0.75 rot.angle=0 layout x.tri=1.3 z.tri=0.25 width=0.3 height=0.3 rot.angle=90
The first LAYOUT command describes the 0.5 mm circle at the center. The second LAYOUT command specifies the rectangle at the lowerright corner. The last LAYOUT command draws the triangle adjacent to the previous rectangle. The angle of rotation is specified in degree and is measured with respect to the xaxis.
Figure 8. Images from nonrectangular mask
shapes can now be simulated in Optolith.
References
[1] Simulation Standard Dec 96 "Advanced Analytical Implantation in ATHENA"
[2] Simulation Standard Aug 96 "TED Modeling in ATHENA"
[3] Simulation Standard Feb 97 "Performing Optical Proximity Correction in ATHENA"