User:Jeanmichel.sellier/Archimedes, the GNU package for semiconductor device simulations

Archimedes, the GNU package for semiconductor device simulations

Introduction

Archimedes is the GNU package for semiconductor device simulations that has been released for the first time on 2005 under GPL
license. It has been created by Jean Michel Sellier who is, since then, the leader of the project and the main developer. It is a Free software
and thus it can be copied, modified and redistributed under GPL license. This is the one of the big advantages of using Archimedes.

Archimedes belongs to the well-known family of TCAD software, i.e. tools utilized to assist the development of technologically relevant products.
In particular, this package assists engineers in designing and simulating submicron and mesoscopic semiconductor devices. According to the author,
in a future version Archimedes will also be able to simulate nanodevices, using the NEGF formalism. Today Archimedes is used in several big companies
for simulation and production purposes.

Archimedes is also useful for teaching purposes since everybody can access the sources, modify and test them. Today, it is used for teaching
purposes in several hundreds universities all around the world. Furthermore, a simplified version, developed for students, is available on nanoHUB.org.

The Ensemble Monte Carlo method is the method that Archimedes to simulate and predict the behaviour of a devices. Being the Monte Carlo
very stable and reliable, Archimedes can be used to know the characteritics of a device even before this last is built.

The Ethical Motivations, a New Paradigma in Science

Archimedes has been created after observing the situation of semiconductor simulations around the world.
One easily observes that the all codes developed for simulation are usually not free and/or proprietary codes.
That is a very bad situation, at least for accademic purposes, since it forces people to reinvent the wheel everytime
a piece of code is needed. This surely slows down the progress of Science (immagine you had to rediscover the Newtonian
laws every time you need them...).

The actual situation is that we have a huge amount of papers describing a lot of numerical methods for advanced simulations of
semiconductor devices, but nobody can access a single code on which to build new and even more advanced methods.

So, today, every university (and even every group in a university) has its own Monte Carlo simulator, its own NEGF simulator
and so on.. Would not it be better if we could avoid this incredibile duplication of efforts all around the world?

That is why Archimedes has been created...

Features

The main purpose of Archimedes is to simulate the transport of electrons and holes (light and heavy) in a semiconductor device.

Description of a device

A scripting language has been created and implemented with the aim to describe a semiconductor device in a fast and general manner.
The script is created by the user and Archimedes parses it to understand what has to be simulated and under which physical conditions.
For example, a Silicon diode script that can be parsed and interpreted by Archimedes is reported below:

# Silicon DIODE script
# To run it type:
# archimedes diode.input
MATERIAL SILICON
TRANSPORT ELECTRONS
FINALTIME 5.5e-12
TIMESTEP 0.0015e-12
XLENGTH 1.0e-6
YLENGTH 0.1e-6
XSPATIALSTEP 100
YSPATIALSTEP 25

# Definition of the doping concentration
# ======================================
DONORDENSITY 0. 0. 1.0e-6 0.1e-6 2.e21
DONORDENSITY 0. 0. 0.3e-6 0.1e-6 5.e23
DONORDENSITY 0.7e-6 0. 1.0e-6 0.1e-6 5.e23
ACCEPTORDENSITY 0. 0. 1.0e-6 0.1e-6 1.e20

# Definition of the various contacts
# ==================================
CONTACT LEFT 0.0 0.1e-6 OHMIC 0.0 5.e23
CONTACT RIGHT 0.0 0.1e-6 OHMIC 1.0 5.e23
CONTACT UP 0.0 1.0e-6 INSULATOR 0.0
CONTACT DOWN 0.0 1.0e-6 INSULATOR 0.0
NOQUANTUMEFFECTS
MAXIMINI
# SAVEEACHSTEP
LATTICETEMPERATURE 300.
STATISTICALWEIGHT 1500
MEDIA 500
OUTPUTFORMAT GNUPLOT
# end

Commands like MATERIAL and TRANSPORT specify the material and the model used in the device simulation.
For more info about the scripting language see the Archimedes manual.

Electron Transport Models

Archimedes implements several methods for the simulation of electron transport in semiconductor devices. Some of them are still work in progress,
somehow experimental, while others are completely implemented, tested and working models. The most reliable and most developed is the ensemble Monte Carlo method,
which has always been considered to be, both in the accademic and industry community, the most vigorous and trusted method for device simulation.
It has been shown to be reliable and predictive.

Monte Carlo Transport

For submicron devices, where semiclassic transport can be used succesfully, the Monte Carlo method is certainly the most reliable and trusted method.
The Monte Carlo method is based on the particle nature of electrons, when the quantum effects are so irrelevant that they can be neglected still getting
a good description of the physics involved in the device.

In this particular case, the main physical effects involved are:

• Free flight
• Phonons Scattering
• Band effects

All effects are taken into account in the current version of Archimedes.
For details see this book [1]

Wigner-Monte Carlo Transport

For mesoscopic devices, in which the transport can be still considered as semiclassic but where quantum effects start to be quite relevant, one has
to find a way to include quantum effects without loosing coherence with semiclassical transport. This is achieved using the so-called Wigner-Monte Carlo
method. An affinity, a mathematical artifact, is added to every particle in the simulation. This is still a work-in-progress in the actual version of Archimedes.<> For details see [2]

NEGF Transport

For very small devices, where the quantum effects dominate the scene, the NEGF formalism has shown to be the most suitable one.
This is still a work-in-progress in Archimedes but it will be released as soon as possible.
For details, refer to this book [3]

Materials

The materials taken into account in the simulations are:

• Silicon
• Germanium
• GaAs
• GaP
• GaSb
• AlAs
• AlP
• AlSb
• InAs
• InP
• InSb
• Stechiometric combinations of the materials abbove

i.e. the most important III-V materials for the semiconductor industry.

Examples

Some examples of results obtained by Archimedes are reported below. The two examples, a Silicon diode
and a Silicon MESFET, are well-known and well-understood devices that can be used as benchmarks to show
the reliability of Archimedes.

Simulation of a Silicon Diode

The script shown above has been run and the results are reported in this reference [4]

Simulation of a Silicon MESFET

A MESFET has been successfully simulated by Archimedes. The results are reported in this reference [5]

How to Support Archimedes

The development of Archimedes is a volunteer effort, and everyone can also contribute to its development.
For information about contributing to the Archimedes Projects and/or requests of enhancements and new features,
please contact the author at jeanmichel [dot] sellier [at] gmail [dot] com.

Future Developments

Future developments are certainly:

• Wigner-Monte Carlo methods
• NEGF formalism

The future of Archimedes strongly depends on the semiconductor community. The more people uses it, the more the feedbacks.
The more the feedbacks, the more new versions and so on..

References

Silicon Diode Simulation
Silicon MESFET Simulation
Jacoboni, Reggiani - Excellent paper on semiclassic Monte Carlo transport
Dollfus, Querlioz - Wigner-Monte Carlo method
Datta - Quantum transport

External links

• GNU archimedes website
• Jean Michel Sellier webpage
• GNU archimedes manual
• GPL