GDR CNRS 3532 MODMAT
2012-2020


Les Webinaires du Jeudi

Animateur: Julien Lam (CEMES Toulouse)

Accéder à la réunion zoom

Meeting ID: 910 2522 6917

Passcode: FTp01u

Programme

8 Juillet 2021 à 14h

Modelling of the primary damage in Fe and W: impact of the interatomic potentials and multivariate multiple linear regression analysis of displacement cascades towards a cascade model.

Charlotte Becquart (UMET Lille)

The cohesive model, which describes the interactions between atoms, is the keystone of any simulations at the atomistic level. To model displacement cascades, it is not possible to use first principles calculations and one turns to empirical potentials. In this work, we compare the quality of a large number of empirical potentials for fundamental energy transfer quantities for tungsten and iron calculations, and then build a large database of displacement cascades using Molecular Dynamics. The potentials differ by their short-range part (called “soft” or “hard” potentials) as well as their long range (or equilibrium) part and we have analyzed the sensitivity of the defect distributions to the equilibrium part of the potential as well as its hardened part. We characterize the potentials using statics as well as dynamics simulations and find useful correlations among the properties, the impact of which I will discuss.

To describe the morphology of the individual displacement cascades we use 3 descriptors: the volume, the number of sub cascades and the sphericity, and for the primary damage, 7 descriptors: the total number of defects, the number of SIA and vacancy clusters and their full size distributions. Statistics studies reveal detailed features of the primary damage, particularly from the point of view of the defect production and the distributions of size of the vacancy and self interstitial defects (SIA) defects at the end of the cascades. A multivariate multiple linear regression analysis based on these 3+7 descriptors and the choice of the potential indicates that the combination of the volume and the sphericity is meaningful. This analysis also highlights several cascade properties and their link to the potentials.

Finally, this work demonstrates that the formation of vacancy clusters is different in Fe and in W.

A. De Backer, C. Becquart, P. Olsson, C. Domain, Modelling the primary damage in Fe and W: influence of the short-range interactions on the cascade properties: Part 2 – multivariate multiple linear regression analysis of displacement cascades, Journal of Nuclear Materials 549, 152887 (2021), [doi: 10.1016/j.jnucmat.2021.152887,

C. Becquart, A. De Backer, P. Olsson, C. Domain, Modelling the primary damage in Fe and W: Influence of the short range interactions on the cascade properties: Part 1 – Energy transfer, Journal of Nuclear Materials 152816 (2021), [doi: 10.1016/j.jnucmat.2021.152816,]

17 Juin 2021 à 14h

Stratégie multi-niveaux pour modéliser les défauts et leur formation dans les matériaux semi-conducteurs à l'échelle atomique : approches conventionnelles, développements nécessaires et nouveaux besoins.

Anne Hemeryck (LAAS Toulouse)

Lorsqu'il s'agit de modéliser les matériaux d'intérêt technologique pour la microélectronique et leurs propriétés, le défi est de taille, surtout si l'on considère les procédés actuels à basse température, les architectures 3D pour l'intégration de matériaux ultraminces et nanostructurés aux performances contrôlées. Les méthodes de continuum sur lesquelles repose généralement la simulation TCAD dans les environnements industriels sont intrinsèquement incapables de capturer les effets résultants à ces échelles, tels que la formation de défauts et les mécanismes de croissance. Dans cet exposé, nous verrons comment il est devenu nécessaire de mettre en œuvre une série d'outils atomistiques à différentes échelles de temps et de taille pour capturer tous les effets nécessaires à une prédiction fiable des événements à l’échelle atomique.

20 Mai 2021 à 14h

Machine Learning for materials science

Cosmin Marinica (DEN/SRMP-CEA Saclay)

I will present recent advances in atomistic material simulations by means of machine learning and data-driven approaches. Machine learning (ML) methods cannot fully replace traditional approaches in physics and/or materials science. The phase space in physics / materials science has a well defined structure and is too vast and complex to be described only by the inherent statistical correlations within the data points. To our knowledge, none of the current statistical methods alone, generically called Machine Learning and its subclass deep learning (DL), can provide a valuable alternative to the laws of physics. In order to provide reliable results in the field of physics, ML/DL should learn and be trained on logically coherent data provided by well established methods from the community of physics. Nevertheless, statistical methods trained on the physical data can be of great help when the traditional approaches are limited and/or their direct application is hindered by factors such as high computational cost.

Multi-scale approaches in materials science face a traditional dichotomy in the choice of the atomistic force fields: robust, accurate and numerically expensive ab initio methods against less transferable but fast empirical methods. The ML methods propose a third avenue that allows control of the balance between the accuracy and numerical efficiency. Moreover, the ML-based vision of fundamental concepts in materials science, such as structural defects, can augment and revise traditional interpretations.

In metals, the interaction and transformation of crystal defect networks gives rise to an extraordinarily diverse range of defect morphologies [1]. Using the recently developed package MiLaDy (Machine Learning Dynamics) [2]: (i) we redefine the concept of defects in materials science [3]; (ii) we provide reliable force fields for complex defects such as interstitial, dislocation loops, dislocations; (iii) we are able to explore the atomistic free energy landscape of point defects in metals with ab initio accuracy up to the melting temperature [4], and, finally, (iv) we are able to propose surrogate models that bypass the traditional approaches [5]. We exemplify and discuss in the framework of experimental findings the case of energetic landscape of defects in body centered and face centered cubic metals.

[1] K. Arakawa , M.-C. Marinica et al. Nature Mat. 19, 508(2020) ; R. Alexander et al. Phys. Rev. B 94, 024103 (2016) [2] M.-C. Marinica, A. M. Goryaeva, T. D. Swinburne et al, MiLaDy - Machine Learning Dynamics, CEA Saclay, 2015-2021 ; A.M. Goryaeva, J.-B. Maillet, M.-C. Marinica. Comp. Mater. Sci. 166, 200 (2019) [3] A. M. Goryaeva et al. Nature Commun. 11, 4691 (2020) [4] C. Lapointe et al. (to be submitted). , T.D. Swinburne, M.-C. Marinica, Phys. Rev. Lett. 120, 135503 (2018) [5] C. Lapointe, T. D. Swinburne et al. , Phys. Rev. Materials 4, 063802 (2020); F. Bruneval et al. J. Chem. Theory Comput. 16, 4399 (2020)

15 Avril 2021 à 14h

Calculating vibrational free energies in solids: data-driven and analytical approaches

Thomas Swinburne (CINaM Marseille)

The free energy of atomic vibrations has a major influence on phase stability and defect thermodynamics, but even approximate methods of evaluation can be numerically challenging and inaccurate. I will highlight some recent contributions that calculate three different classes of vibrational free energy in three different ways. I will show a machine learning technique typically used to model energetics can accurately capture the harmonic formation and activation entropies of defects in crystals, nanoclusters and amorphous solids [1]. Going beyond the harmonic approximation, I will present recent developments of the PAFI code[2], an O(N) method to evaluate activation free energies in very large systems (>300000 atoms), recently applied to the migration of twin interfaces in Mg[3]. To finish, I will describe an analytic approach that evaluates the anharmonic vibrational free energy of FCC crystals using a “bond lattice” model, which has been shown to reproduce ab initio free energies to within an meV/atom [4].

[1] C Lapointe*, TDS*, L Thiry, S Mallat, L Proville, CS Becquart et M-C Marinica*, Phys. Rev. Mat. 2020, 10.1103/PhysRevMaterials.4.063802 + in prep. [2] TDS* et M-C Marinica, Phys. Rev. Lett. 2018, 10.1103/PhysRevLett.120.135503, https://github.com/tomswinburne/pafi [3] Y Sato, TDS, S Ogata, D Rodney*, Mat. Res. Lett. 2021, 10.1080/21663831.2021.1875079 [4] TDS*, J Janssen, M Todorova, G Simpson, P Plechac, M Luskin et J Neugebauer, Phys. Rev. B (Rap. Comm.) 2020, 10.1103/PhysRevB.102.100101

18 Mars 2021 à 14h

Simulations par dynamique moléculaire des propriétés mécaniques de nanoparticules

Laurent Pizzagalli (Institut P' Poitiers)

Les propriétés des matériaux sont fortement modifiées lorsque leurs dimensions sont réduites, principalement du fait de l'importance croîssante des surfaces. Les propriétés mécaniques n'échappent pas à ce constat, de nombreux travaux sur les nanopiliers ayant par exemple montré une résistance à la déformation bien supérieure à celle du matériau massif ou bien encore l'activation de mécanismes de plasticité originaux. Les nanoparticules constituent une classe importante de nanomatériaux pour laquelle nos connaissances sur ce sujet sont plus limitées. Cette carence tend toutefois à se résorber, en partie grâce à des études récentes basées sur des simulations numériques de type dynamique moléculaire. Quelques exemples choisis seront présentés au cours de cet exposé. On montrera ainsi comment la forme de la nanoparticule influe fortement sur la déformation plastique, ou bien encore comment un matériau "dur" peut se déformer avant un matériau "mou" dans le cas d'un architecture coeur/coquille. Enfin, nous avons récemment développé une approche permettant la compression par dynamique moléculaire ab initio. Son application à un fullerène C60 a permis de mettre en évidence les propriétés mécaniques tout à fait étonnantes de cette nanoparticule.

18 Février 2021 à 14h

Etude de la conduction thermique dans les nanomatériaux par simulations de dynamique moléculaire d’approche à l’équilibre

Evelyne Martin (ICube Strasbourg)

Dans cet exposé seront présentées des simulations à l’échelle atomique du transport de chaleur dans des matériaux inorganiques, cristallins ou amorphes, dans des nanostructures, et au niveau de l’interface entre matériaux organiques et inorganiques. Ces études sont réalisées en dynamique moléculaire, classique pour les travaux les plus anciens, et ab initio plus récemment. La méthode AEMD (dynamique moléculaire d’approche à l’équilibre) développée pour le transport de chaleur sera d’abord présentée, tant en ce qui concerne son principe que l’analyse des résultats et les différentes informations auxquelles elle permet d’avoir accès, comme la conductivité thermique, les résistances d’interface et les libres parcours moyens des porteurs de chaleur. L’AEMD sera ensuite appliquée à divers nanomatériaux et nanostructures, ce qui permettra de comparer les comportements à petite échelle et d’identifier similitudes et différences.

21 Janvier 2021 à 14h

Embedded many-body perturbation theory for the electronic properties of organic systems

Xavier Blase (Institut Néel Grenoble)

Many-body perturbation theories, such as the GW and Bethe-Salpeter formalisms, have become a tool of choice in solid-state physics for studying the optoelectronic properties of crystals. Difficulties arise when attempting to explore with such techniques extended disordered systems : periodic boundary conditions cannot be used while the importance of long-range electrostatic and dielectric effects preclude the quantum chemistry approach of considering isolated molecules in the gas phase. We will present recent developments along the line of embedded, or QM/MM, formalisms allowing to perform accurate many-body calculations for the optoelectronic properties of organic systems immersed in complex electrostatic and dielectric environments. Applications to the study of the elusive doping mechanisms in organic semiconductors and to organic photovoltaics will be presented.

7 Janvier 2021 à 14h

Complex intermetallic compounds : original surface structures for unusual surface properties

Emilie Gaudry (Institut Jean Lamour / Mines Nancy)

Complex intermetallic compounds (CIMCs) with bulk cage-like structures are a class of ordered alloys made of highly symmetric polyhedra as building blocks. Representatives of this family includes quasicrystals and their approximants, as well as intermetallic clathrates and related compounds with large crystal cells. Their unique structures can lead to unusual surface properties – at least when compared to those of conventional alloys -- which make them attractive for a number of applications, like efficient coatings, templates for building new molecular nano-structures or as novel catalytic materials. The detailed knowledge of surface structures is a necessary step to understand and tune the surface structure-property relationships. In this talk I will show that Density Functional Theory calculations, possibly combined with experimental surface science techniques, can be used to determine the thermodynamic, atomic and electronic structures of the CIMC stable surfaces. I will discuss the influence of the intrinsic bulk properties of the compounds, on their surfaces structures and properties, using recent investigations achieved in the group.

26 Novembre 2020 à 14h

Deformation and flow of disordered solids - a statistical physics perspective.

Jean-Louis Barrat (LiPhy Grenoble)

Disordered or amorphous solids cover a wide class of systems, "soft" (colloidal, granular pastes) or "hard" (metallic glasses, oxide glasses), which are not ordered at the microscopic level. They therefore do not exhibit dislocations like crystalline solids. Their flow under sufficiently large stress is produced by the instability of localized zones, the “shear transformations”, which interact elastically with each other. I will present our current understanding, based on microscopic simulations and experiments, of the elementary mechanisms that govern the deformation and flow of these solids. I will show how these mechanisms can be incorporated into simple lattice models governed by long range elastic interactions, discuss the mean field analysis of these models as well as some fluctuation related issues such as the statistics of avalanche like events.