Centre Inter-universitaire de Recherche et d’Ingénierie des Matériaux - UMR CNRS 5085



Recherche de candidat(e) en postdoctorat

Development of a combined quantum chemical / machine learning based model to predict NMR parameters in large amorphous carbon structures

Project description

Porous carbons are an important class of materials used in many applications including energy storage, gas storage, water treatment, and catalysis. In all cases, the characterisation of the porous materials, still challenging from an experimental point of view, is an essential step in order to understand and optimise the performance of the systems. As a consequence, a number of theoretical works have aimed at generating atomistic structures which are very helpful to study structure-property relationships. Techniques used to generate these porous carbon structures include Hybrid reverse Monte Carlo1 and Quench Molecular Dynamics using empirical potentials2 or more recently machine-learning based interatomic potentials3. While these models are extremely useful to get insights into the microscopic mechanisms of adsorption, it is hard to assess precisely the quality of these structures. This is due in part to the challenge of obtaining quantitative experimental data (e.g. proportions of 5-/6-/7-membered rings) and the ambiguity of some results (i.e. structures with diffrent local structures and ring counts can have very similar pair distribution functions in agreement with the measured ones).

Recently, Nuclear Magnetic Resonance has been investigated as a promising technique to improve our understanding of such carbon structures. In particular, NMR has been proposed as a way to estimate pore size distributions4 and to assess local ordering 5. So far, the interpretation of NMR spectra of adsorbed molecules has mainly relied on the calculation of chemical shifts for small aromatic molecules (e.g. benzene, coronene). Over the years, a number of theorical models based on predicting ring currents in small aromatic molecules have been proposed to explain the magnetic properties observed experimentally6. These models give more or less satisfactory results but are overall successful, especially for planar molecules containing only 6-membered rings. Nowadays, for small molecules, it is also possible to use Density Functional Theory (DFT) as a straightforward way to calculate chemical shifts close to these carbon molecules. Nevertheless, the porous carbons used in common applications are extended solids, with defects, for which current models are not suitable.

The aim of this postdoctoral project is to develop a combined quantum chemical / machine learning based model to predict NMR parameters in large amorphous carbon structures and use it to improve our understanding of such disordered solids. The approach could be based on coupling a tight-binding model with tunable atomistic magnetic polarisabilities in order to predict NMR parameters. The model will be first implemented and tested for relatively small aromatic molecules containing different numbers 5-/6-/7-membered rings and will then be applied to periodic solids. The model should then allow us to better understand the in uence of defects on the electronic and magnetic properties of the carbon structure as well as discriminating between various proposed atomistic models. A fine interpretation of the NMR spectra of adsorbed molecules would also improve our description of the local structure and dynamics of these adsorbed species.

1/ J. C. Palmer et al., Characterization of porous solids VIII (2009)
2/ de Tomas et al., Carbon, 119, 1 (2017)
3/ V. L. Deringer and G. Csaanyi, Phys. Rev. B, 95, 094203 (2017)
4 /Xing et al., Carbon, 77, 1132 (2014)
5/ A. C. FORSE et al., Chem. Mater., 27, 6848 (2015)
6/ Gomes and Mallion, Chem. Rev., 101, 1349 (2001)

Applicant’s profile

The applicant should have a PhD in computational Materials Science, Chemistry or Physics. Skills in manipulating Density Functional Theory codes is required. Prior knowledge in machine-learning is a plus.


Please send a CV, a cover letter and the names and addresses (including e-mail address) of two referees to Céline Merlet (merlet chimie.ups-tlse.fr) by the 15th of June 2018.

Context and Funding

This postdoctoral position will be funded by an ERC starting grant which started on the 1st of July 2017. The title of this ERC is SuPERPORES : Structure - PErformance Relationships in PORous carbons for Energy Storage. This work will be done in collaboration with Albert Bartok from STFC (Science & Technology Facilities Council) and Chris Pickard from the University of Cambridge, the postdoctoral researcher will be expected to spend time in these institutions.

Duration, salary and location

2 years from October 2018,
2000 e/month,
CIRIMAT - Université Toulouse 3 - CNRS.


Céline Merlet, merlet chimie.ups-tlse.fr