Periodic Reporting for period 1 - MUST (Comprehensive multiscale modelling of atomistic and electronic structure of radiation-induced defects in semiconductors)
Okres sprawozdawczy: 2023-03-01 do 2025-08-31
Particle irradiation can change the structure and properties of semiconductor materials, which can be either beneficial or harmful, depending on the situation. For example, irradiation, in particular with dopant species, can improve certain properties in silicon-based power devices, but in high-radiation environments, energetic particles can damage materials used in satellites or other critical technologies. Understanding how radiation affects these materials helps improve their performance and reliability.
The MUST project aims to improve how we predict the effects of radiation on semiconductors, with focus on wide band gap materials, which are strong candidates for next-generation devices. The main goal is to merge advanced physics models to make quantitative predictions about how these materials will behave in extreme environments, such as space or high-radiation areas. We do this by using powerful computers for advanced calculations, combined with cutting-edge multi-scale modeling techniques to accurately predict how radiation will affect these materials at the atomic level.
The project aims to push beyond current limits in radiation damage simulations, reaching a level of precision not yet achieved, to be able to predict the effects of irradiation based on a bottom-up approach from basic physics, rather than relying on empirical models. The methods developed and validated in the MUST project will be extendable to other compound semiconductor materials. This will significantly improve our ability to design radiation-resistant semiconductor devices, which are essential for technologies like satellite communications, space exploration, and secure global communications. In essence, this research will help design materials that can withstand extreme radiation conditions, opening up new possibilities for electronics in space and other high-risk environments, while making the development of these materials more cost-effective and efficient.
The MUST project aims to improve how we predict the effects of radiation on semiconductors, with focus on wide band gap materials, which are strong candidates for next-generation devices. The main goal is to merge advanced physics models to make quantitative predictions about how these materials will behave in extreme environments, such as space or high-radiation areas. We do this by using powerful computers for advanced calculations, combined with cutting-edge multi-scale modeling techniques to accurately predict how radiation will affect these materials at the atomic level.
The project aims to push beyond current limits in radiation damage simulations, reaching a level of precision not yet achieved, to be able to predict the effects of irradiation based on a bottom-up approach from basic physics, rather than relying on empirical models. The methods developed and validated in the MUST project will be extendable to other compound semiconductor materials. This will significantly improve our ability to design radiation-resistant semiconductor devices, which are essential for technologies like satellite communications, space exploration, and secure global communications. In essence, this research will help design materials that can withstand extreme radiation conditions, opening up new possibilities for electronics in space and other high-risk environments, while making the development of these materials more cost-effective and efficient.
The work so far has focused on developing two main tools needed to predict the atomic effects of energetic impacts in selected semiconductors. Firstly, the interaction of energetic projectiles with the electrons in the material has been modeled using quantum mechanical calculations that account for the dynamics of excited electrons. Based on these calculations, we are able to directly predict the slowing down of fast atomic projectiles due to interactions with electrons in the material. This effect is important in radiation damage predictions, since it dictates how much of the energy of incident particles is transferred to the electrons. This in turn has a direct impact on the atomistic damage that is formed, and is sensitive to the directions the projectiles travel through the material.
Based on these calculations, we have developed models that can transfer this behavior in a computationally efficient way to simulations involving millions of atoms, which is necessary to fully model the radiation damage events.
Secondly, we have developed two new interatomic potentials using a machine learning approach, with which we can model the interactions between atoms in germanium and silicon carbide. These models are superior to existing analytical models, and are able to predict the properties of the materials in different forms—such as solid, liquid, or disordered states—as well as how small defects form and move within them, and how atoms with high energies collide with other atoms. These are crucial details for predicting radiation damage formation.
We have further carried out a large number of simulations of the radiation damage formation in Si, using different existing models for the electronic and atomic interactions. These will be used to benchmark predictions with the new developed methods, and to gauge the impact of incorporating more rigorous physics in the models. This assessment is needed to balance the choice between high fidelity and high computational cost in modeling.
Three data sets have been made publicly available: two sets of data for electronic energy losses, in silicon and in germanium, and a third data set containing primary radiation damage data from numerous cascade simulations in silicon. The primary damage data serves as an important starting point for models that simulate how radiation damage evolves over time, which will help predict the long-term effects of radiation on materials. Work is ongoing towards developing these methods for SiC.
Based on these calculations, we have developed models that can transfer this behavior in a computationally efficient way to simulations involving millions of atoms, which is necessary to fully model the radiation damage events.
Secondly, we have developed two new interatomic potentials using a machine learning approach, with which we can model the interactions between atoms in germanium and silicon carbide. These models are superior to existing analytical models, and are able to predict the properties of the materials in different forms—such as solid, liquid, or disordered states—as well as how small defects form and move within them, and how atoms with high energies collide with other atoms. These are crucial details for predicting radiation damage formation.
We have further carried out a large number of simulations of the radiation damage formation in Si, using different existing models for the electronic and atomic interactions. These will be used to benchmark predictions with the new developed methods, and to gauge the impact of incorporating more rigorous physics in the models. This assessment is needed to balance the choice between high fidelity and high computational cost in modeling.
Three data sets have been made publicly available: two sets of data for electronic energy losses, in silicon and in germanium, and a third data set containing primary radiation damage data from numerous cascade simulations in silicon. The primary damage data serves as an important starting point for models that simulate how radiation damage evolves over time, which will help predict the long-term effects of radiation on materials. Work is ongoing towards developing these methods for SiC.
Predictions of radiation damage formation from atomistic modeling hinge crucially on two factors; the interactions between atoms, and the interactions between atoms and the electrons of the material.
So far, the project has achieved major progress in understanding how energy is lost to electrons by atoms traveling at high velocities through silicon and germanium. When a high-energy particle strikes a material, it creates a chain reaction of atomic collisions—called a collision cascade—which can leave behind damage. A key part of predicting the damage formation process is accounting for how much of the energy of the atoms involved in the cascade is transferred to the material’s electrons. This energy transfer plays a key role in how much and what kind of damage occurs.
Previous simulation approaches have accounted for certain aspects of electronic energy losses in an ad hoc manner, or disregarded them completely. In the collision cascade, the energy loss process for each atom changes dynamically as the damage event progresses. With the models developed in MUST, these energy losses can now be captured accurately, with no free parameters, significantly improving the predictive power of simulations. For example, the new models accurately predict the amount of energy that is lost depending on the exact direction the particle travels, providing predictions in good agreement with experiments. The model parametrizations will be made available to the scientific community, and can be used directly in two existing open source atomistic simulation codes, one of which we have extended to include this functionality.
Additionally, we have developed two new advanced machine learning interatomic potentials suitable for radiation damage simulations, which offer much greater accuracy than existing interaction models, for germanium and silicon carbide. These will also soon be made available to the community.
These tools will allow predicting radiation damage with much higher fidelity than has previously been possible.
So far, the project has achieved major progress in understanding how energy is lost to electrons by atoms traveling at high velocities through silicon and germanium. When a high-energy particle strikes a material, it creates a chain reaction of atomic collisions—called a collision cascade—which can leave behind damage. A key part of predicting the damage formation process is accounting for how much of the energy of the atoms involved in the cascade is transferred to the material’s electrons. This energy transfer plays a key role in how much and what kind of damage occurs.
Previous simulation approaches have accounted for certain aspects of electronic energy losses in an ad hoc manner, or disregarded them completely. In the collision cascade, the energy loss process for each atom changes dynamically as the damage event progresses. With the models developed in MUST, these energy losses can now be captured accurately, with no free parameters, significantly improving the predictive power of simulations. For example, the new models accurately predict the amount of energy that is lost depending on the exact direction the particle travels, providing predictions in good agreement with experiments. The model parametrizations will be made available to the scientific community, and can be used directly in two existing open source atomistic simulation codes, one of which we have extended to include this functionality.
Additionally, we have developed two new advanced machine learning interatomic potentials suitable for radiation damage simulations, which offer much greater accuracy than existing interaction models, for germanium and silicon carbide. These will also soon be made available to the community.
These tools will allow predicting radiation damage with much higher fidelity than has previously been possible.