This position is part of the Marie Skłodowska-Curie Actions Doctoral Network (MSCA DN) PHYMOL (https:// phymol.eu/):
Physics, Accuracy and Machine Learning: Towards the next-generation of Molecular Potentials.
In the PHYMOL collaboration we bring together leading experts in the fields of molecular simulations, quantum chemistry, crystal structure prediction, intermolecular modelling, spectroscopy, machine-learning, and nano-clusters, from 12 academic institutes and national laboratories, and 4 industrial entities, in an ambitious programme of research and training to develop a new generation of researchers in the field of molecular modelling. PHYMOL combines the most advanced physical understanding of molecular interactions with machine-learning in a symbiotic manner that will lead to a new generation of researchers capable of advancing solutions to problems of importance in healthcare, energy and the environment, as well as basic science. The private sector is an integral part of PHYMOL and participates in management, training and research. With this union of forces from academia, industry from the EU and the US, we seek to keep molecular simulation techniques at the fore-front of industry and science in the EU.
PHYMOL is both a research and training network. We need researchers equipped with a deeper physical understanding of molecular interactions as well as profound expertise in computer-aided learning: a single focus on just one of these skills will not get us to where we wish to be. Machine learning (ML) needs accurate data, and while we have some of the most accurate ab initio methods available (many developed by members of this consortium), we know very well their limitations and shortcomings. Therefore, in PHYMOL we have set up a research and training programme that spans all areas of importance in the field of intermolecular interactions: the development of methods, generation and testing of reference data, development of models using physical ideas and machine learning, and the applications of these models in high-accuracy spectroscopic calculations and large-scale molecular dynamics computer simulations of complex systems.
The PHYMOL doctoral network will see 11 doctoral candidates (DCs) recruited across the 10 participating institutions in the UK and EU.Your Role...
State-of-the-art: A method that can accurately and efficiently predict the stable phases of molecular crystals is highly desirable for use in the development of new pharmaceutical drugs 1. If a drug is manufactured in a metastable phase that later converts to a more stable phase, it can render the drug insoluble and ineffective. Several pharmaceuticals, such as ritovanir, an HIV treatment, have had to be recalled from the market due to not discovering the most stable polymorph in time. One of the main reasons crystal structure prediction is so challenging is that one of the dominant interactions holding molecular crystals together is van der Waals interactions, an inherently quantum mechanical phenomenon. Additionally, while van der Waals interaction between two individual molecules is weak, many-body effects cause the collective interaction between the molecules in a system, even ones that are hundreds of angstrom apart, to become appreciable.
The complex nature of van der Waals interactions and the size of pharmaceutically relevant molecular crystals has impeded the use of computational methods to predict crystal structure, even making density functional theory (DFT), the workhorse of modern computational chemistry, ineffective. Two developments are changing this state of affairs. One, is was recently demonstrated in our group that DFT can be supplemented with an approximate model of van der Waals interactions called the many-body dispersion (MBD) model and this can lead to quantitative accuracy in predicting molecular crystal structure 2. Two, enormous progress is being made in the realm of machine learned force fields such that they are now able to produce results of comparable accuracy to the quantum mechanical methods they are trained on at orders of magnitude reduced cost. We intend to capitalize on these new developments to improve the accuracy and efficiency of crystal structure prediction methods with a direct view on industrial applications.
Research goals: The PhD candidate will test machine learned force field methodologies trained on DFT+MBD and improve and integrate these methods into crystal structure prediction workflows, with an eye on properly balancing the complex interplay between different intra- and intermolecular interactions.
Day-to-day: The research will involve working on a linux-based high- performance computing cluster using python to run and analyse calculations and implement/extend methodology for different aspects of the crystal structure prediction workflow. Programming in various languages will become a main focus when integrating new methods into existing software packages. The TCP group has several experienced postdocs that will be able to assist the doctoral candidate in mastering the various aspects of this research. Additionally, this work will be co-supervised by our industrial partner in order to direct the research towards real world applications.Pre-requisites...
Good mathematical and programming skills, a good understanding of basic quantum mechanics, thermodynamics, and physical and chemical intuition.
Researchers can be of any nationality. They must comply with the rule for mobility. Researchers are normally required to undertake trans-national mobility (i.e. move from one country to another) when taking up their appointment. One general rule applies to the appointment of researchers in a network: At the time of recruitment by the host organisation, researchers must not have resided or carried out their main activity (work, studies, etc) in the country of their host organisation for more than 12 months in the 3 years immediately prior to the reference date. Short stays such as holidays and/or compulsory national service are not taken into account. As far as international European interest organisations or international organisations are concerned, this rule does not apply to the hosting of eligible researchers. However the appointed researcher must not have spent more than 12 months in the 3 years immediately prior to the reference deadline for submission of proposals or recruitment by the host organisation, depending on the action, in the same appointing organisation.In Short...
Applications should include:
Early application is highly encouraged, as the applications will be processed upon reception. Please apply ONLINE formally through the HR system. Applications by email will not be considered.
The University of Luxembourg embraces inclusion and diversity as key values. We are fully committed to removing any discriminatory barrier related to gender, and not only, in recruitment and career progression of our staff.References...
1 What is Crystal Structure Prediction? And why is it so difficult? - The Cambridge Crystallographic Data Centre (CCDC)
2 Johannes Hoja and Hsin-Yu Ko and Marcus A. Neumann and Roberto Car and Robert A. DiStasio and Alexandre Tkatchenko "Reliable and practical computational description of molecular crystal polymorphs" Sci. Adv., 5 , eaau3338 (2019) provides an excellent overview of the general problem and the method, DFT+MBD, that the doctoral candidate will use.
PI, co-supervisors: Alexandre Tkatchenko (main supervisor), Dahvyd Wing (co-supervisor), University of Luxembourg; Marcus A. Neumann (co- supervisor), Avant-garde Materials Simulation GmbH; Alston Misquita (mentor), University of London
Host institution: University of Luxembourg ranks 25th in the 2022 Times Higher Education Young University Ranking and 3rd worldwide in International Outlook. Avant-garde Materials Simulation Deutschland GmbH, in Freiburg, Germany, is a subsidiary of AMS SARL, founded by Dr. Marcus Neumann in 2002. It is privately owned and was created with funding and strong support from a major pharmaceutical company. The company's main goal is the development of software for crystal structure prediction. In particular, its software package, GRACE, is the most used crystal structure prediction software in the pharmaceutical industry.
Update: Description of PHYMOL, structure of the network, training, conferences, DC benefits.
Webpage: https: // www. tcpunilu.com/, https: // www. avmatsim.eu, PHYMOL