Post-doctorat - Diagramme de phases ab initio par inférence bayésienne - H/F
CEA · Essonne, Île-de-France, France
This listing has closed — view similar roles below.
- On site
- Internship
- €50,000 / year
- Essonne, Île-de-France, France
Job highlights
- Postdoctoral role in advanced materials simulation.
- Develop automated phase diagram sampling.
- Reduce computational costs significantly.
- Utilize cutting-edge machine learning potentials.
- Provide uncertainty estimates for results.
About the role
About the Role
We have recently demonstrated that AIMD simulations can be accelerated by a factor of 100 using an interatomic potential adjusted by machine learning (MLIP) [1-3]. This accelerated sampling of the equilibrium canonical distribution can be followed by free energy calculations via thermodynamic integration [4]. This 'on-the-fly' learning strategy ensures quasi-equivalent accuracy to AIMD calculations and is now available in the Python code MLACS, which we develop and use in production. Nevertheless, repeating such simulations tens or even hundreds of times to systematically sample a phase diagram remains extremely computationally demanding, or even unthinkable.
During this postdoctoral project, the candidate will develop an optimal and automated sampling strategy. The objective is twofold: to reduce the number of MLACS simulations required to construct a phase diagram, and to provide an estimate of the uncertainty associated with the obtained diagram. They will build upon recent developments in the literature [5, 6].
Key Responsibilities
- Develop an optimal and automated sampling strategy for phase diagram construction.
- Reduce the computational cost of phase diagram simulations.
- Provide uncertainty estimates for the generated phase diagrams.
- Collaborate with researchers on cutting-edge machine learning potentials for materials science.
Candidate Profile
- Ph.D. in a relevant scientific field (e.g., Physics, Materials Science, Chemistry, Computer Science).
- Strong background in computational materials science, statistical mechanics, and/or machine learning.
- Experience with programming, particularly in Python.
- Familiarity with ab initio methods and/or machine learning interatomic potentials is a plus.
- Excellent problem-solving skills and ability to work independently and collaboratively.
About CEA
CEA is committed to the integration of people with disabilities and offers accommodations for inclusion. A national security background check is conducted for all employees.
Key skills/competency
- Postdoctoral Researcher
- Bayesian Inference
- Phase Diagrams
- Machine Learning Potentials (MLIP)
- AIMD Simulations
- Thermodynamic Integration
- Python
- Computational Materials Science
- Statistical Mechanics
- Automated Sampling
Skills & topics
- Postdoctoral Researcher
- Phase Diagrams
- Bayesian Inference
- Machine Learning
- Computational Materials Science
- AIMD
- MLIP
- Python
- Thermodynamic Integration
- Statistical Mechanics
- Physics
- Materials Science
- Chemistry
- Computer Science
How to get hired
- Tailor your CV: Highlight relevant research in computational materials science, Bayesian inference, and machine learning potentials.
- Craft a strong cover letter: Emphasize your understanding of the project's goals and your specific skills in Python and simulation techniques.
- Prepare for technical questions: Be ready to discuss your experience with AIMD, MLIPs, and statistical mechanics during the interview.
- Showcase problem-solving skills: Demonstrate your ability to develop novel strategies for complex computational challenges.
- Research CEA: Understand their commitment to research and innovation in materials science.
Technical preparation
Behavioral questions
Frequently asked questions
- What is the main goal of this Postdoctoral Researcher position at CEA?
- The main goal of this Postdoctoral Researcher position at CEA is to develop an optimal and automated sampling strategy for constructing phase diagrams, reducing the number of simulations needed and providing uncertainty estimates.
- What are the key computational techniques used in this research?
- Key computational techniques include AIMD simulations accelerated by Machine Learning Interatomic Potentials (MLIP), on-the-fly learning strategies, and thermodynamic integration for free energy calculations. The project will also involve Bayesian inference for automated sampling.
- What programming skills are required for this role?
- Proficiency in Python is required for this role, as you will be working with and developing the MLACS Python code for simulations and analysis.
- What is the expected duration of this postdoctoral project at CEA?
- While not explicitly stated, postdoctoral positions are typically 1-3 years. Further details on the duration would be discussed during the application process.
- Does CEA offer support for integrating individuals with disabilities?
- Yes, CEA is committed to the integration of people with disabilities and offers accommodations and organizational possibilities for their inclusion.
- What is the significance of Bayesian inference in this research?
- Bayesian inference is crucial for developing an automated sampling strategy that provides an estimate of the uncertainty associated with the constructed phase diagram, leading to more robust and reliable results.
- What academic background is preferred for this Postdoctoral Researcher role?
- A Ph.D. in a relevant scientific field such as Physics, Materials Science, Chemistry, or Computer Science is preferred, with a strong background in computational materials science, statistical mechanics, or machine learning.