Advanced Diffusion Models Released in ATHENA 4.0

ATHENA Version 4.0 contains the latest model developments from universities and research institutes worldwide. For process simulation of deep sub-micron devices, accurate and robust implantation and diffusion models are essential. ATHENA Version 4.0 includes a significant number of new implant and diffusion models for simulating high dose and RTA effects.

New Stanford Diffusion Models

  • 311 Cluster Model
  • Dislocation Loop Model
  • High Dose Model
  • Scaleable implant damage model

This set of three diffusion models developed at Stanford University together with a scaleable implant damage model developed at Silvaco allow users to model Transient Enhanced Diffusion processes such as RTA. The model extents the previous FULL.CPL model for point defects. It adds defect clusters that dissolve over time releasing interstitials and dislocation loops that act as point defect sinks. A model to derive cluster and loop concentrations and locations as a function of implant profile is also included.[1] [2] An example of RTA simulation using this model is shown in Figure 1.

Figure 1. RTA of Boron using the new Stanford diffusion models.

CNET Diffusion Models

  • Impurity-defect pairing statistics
  • Static clustering
  • Percolation
  • Correlated interstitial and vacancy mediated impurity diffusivities
  • Bimolecular recombination of defect through impurity states

This set of five diffusion models can be used as extensions to the current FULL.CPL model. These models were developed and implemented through a cooperative agreement with Daniel Mathoit of CNET. The models apply particularly at very high dopant concentrations. Figure 2 shows the comparison of experimental data for this model to 900C predisposition of phosphorus.[3] [4]

Ion Implant Enhancements

  • Three-Level Lateral Straggle Description
    - Simple scaling factor for lateral range (LAT.RATIO)
    - User control over lateral standard deviation
  • Full control of all moments for depth dependent lateral straggle

Figure 2. Comparison between experimental and simulated profiles at 900oC. The simulations are performed with the CNET model.

The reduction in post-implant diffusion for ULSI processes means that accurate models for lateral implant straggle are required to produce good agreements with MOSFET Leff. A hierarchical approach has been adopted in this release of ATHENA. The simplest level is a single scaling parameter (LAT.RATIO) which can be used to shift the lateral moments as a function of the default model (See Figure 3). At a more complex level the user has full control of depth or dose dependent lateral straggle though the MOMENTS statement.

  • Improved User Access to Implant Moments
    - allow users to add their own tables
    -allow lateral straggle definition in the tables

Figure 3. Variation in lateral straggle of phosphorus LDD implant.

New data for implant moments is continually becoming available from research institutions and universities. The USER_TABLE parameter of the MOMENTS statement will allow user to add new tables of implant moments to be used by ATHENA. The ATHENA executable and the tables of implant ranges are now decoupled. This will allow Silvaco to issue new and updated implant range tables asynchronously from ATHENA code releases. In the future new implant tables could be downloaded from the Silvaco WWW site.

  • Extended High Energy Implant Statistics (up to 8MeV into SiO2)
  • Allow printing of ion implant moments used
  • Ability to plot the Monte Carlo ion implant racks and secondary recoils in Tonyplot. (See Figure 4)
  • Ability to model silicon implants into silicon for pre-amorphisation of the substrate
  • Automatic extraction of spatial implant moments from Monte Carlo implant calculations

Updated Parameters for Interstitials

Updated model file values for Interstitial KSURF, THETA and KRAT to fall in line with FLOOPS [5]. This change will result in more accurate and quicker diffusion results using TWO.DIM or FULL.CPL models. Although results may vary from the previous release.

Adaptive Meshing

  • New Base Mesh Algorithm
  • New Global Smoothing Algorithm
  • Improved 1D To 2D Transition
  • Template Grid Rules Derived For Common Technologies

Figure 4. Tracks of implant ions and secondary recoils of silicon ions from a Monte Carlo implant.

Figure 5. MOS simulation in ATHENA using adaptive meshing algorithm.

A series of improvements have been made to the adaptive meshing routines in SSuprem4. Efficient adaption in 1D mode can now be performed. A transition to 2D mode can be made using an automated base mesh generator. For the most common technologies the meshing rules have been supplied as templates. Examples for MOSFETS, CCDs, BJTs, and large power structures are available.

Model Enhancements

  • Power Device diffusion model

For devices with extremely large geometry's (in the order 10um) some of the physics included in SSuprem4 is not required. In order to speed up simulation times of these large devices a new option METHOD POWER has been added. Results obtained using this method are identical for large devices with a 5x speed up.

Figure 6. Keyhole void formation during metal deposition into various sized contacts. Note how the position of the void rises for larger contacts.

  • Clustering Model Extended

A generalized SSUPREM4 clustering model is implemented to simulate impurity activation for both p-type and n-type impurities. The model can be selected by specifying parameter CLUSTER.S4 on the METHOD statement.

  • Extend Boron Solid Solubility Data Down To 700C
  • Oxide Threshold Model
  • Temperature Dependent Fractional Interstitialcy
  • Addition of Indium dopant

Void formation

An algorithm to allow formation of keyhole voids in deposited films has been added to Elite. Void boundary conditions are correctly handled so subsequent deposits do not fill the void. Void formation can be followed by simulation of viscous flow of the deposited material to reduce or eliminate the void.

Plasma Etch Model

  • Plasma ion flux based etch rates
  • Concentration dependent etch rates
  • Stress dependent etch rates

A Monte Carlo based plasma etching model has been added to Elite. It calculates the angular dependence of ions emitted by the dark spaces heath in RIE etchers. Etch rates over complex topography are calculated along with angle dependent sputtering efficiency. Shadowing effects are included.

Figure 7. Effect of plasma pressure on the etch profile of a trench.

Since Elite is a grid-based topography simulator it is also able to treat the etch rate dependence on physical quantities in the substrate.Etch rates as a function of doping or stress can be modeled. Faster and More Automated Lithography Simulation

  • Speed up of 10x in the calculation of aerial images
  • Speed up of 5x in the exposure calculation
  • Simpler CD extraction procedure
  • Tighter integration of OPTOLITH into VWF to allow characterization of lithography processes

The main area of Optolith enhancements has been in the numerics of the imaging and exposure calculation. For the imaging a conservative estimate of the speed-up is 10x the previous ATHENA version. This means that the imaging of very complex masks can be done in a reasonable CPU time. New routines have allowed Optolith to be used within the VWF automation tools for characterizing lithography processes. Figure 8 shows a contour plot of CD as a function of exposure and defocus. This is stored as a behavioral model in VWF to allow experimentation with second order parameters such as resist thickness or development rates.

Figure 8. Sensitivity of CD to lithography parameter can be analyzed using behavioral models created by VWF Automation Tools.


[1] "TED Modeling in ATHENA" Simulation Standard Aug 96.

[2] S. Crowder et al , IEDM 1995 p 427.

[3] "CNET Physical Diffusion Model in ATHENA" Simulation Standard Feb 96.

[4] D. Mathiot, Electrochemical Soc. Poc. 95-5 p 13.

[5] Journal Applied Physics, Park and Law. 72(8), p. 3431, Oct. 15, 1992.