POLEWCZYK, Franck; LAFOURCADE, Paul; COSTA, Jean-Pierre Da; VIGNOLES, Gerard; LEYSSALE, Jean-Marc

doi:10.1016/j.carbon.2023.118109

Voir/Ouvrir

LCTS_Carbon_2023_Polewczyk.pdf (3.103Mo)

Métadonnées

Afficher la notice complète

Licence d’utilisation du document

POLEWCZYK, Franck
Laboratoire Matière sous Conditions Extrêmes [LMCE]
Institut des Sciences Moléculaires [ISM]
DAM Île-de-France [DAM/DIF]

LAFOURCADE, Paul
Laboratoire Matière sous Conditions Extrêmes [LMCE]
Institut des Sciences Moléculaires [ISM]
DAM Île-de-France [DAM/DIF]

COSTA, Jean-Pierre Da
Laboratoire de l'intégration, du matériau au système [IMS]
Bordeaux Sciences Agro [Gradignan]

Langue

Article de revue

Ce document a été publié dans

Carbon. 2023-08-01, vol. 212, p. 118109

Résumé en anglais

Atomistic modeling of disordered yet textured carbons is notoriously difficult even though it can prove extremely helpful in rationalizing structure–property relationships for this class of materials. In this work we introduce a polygranular image-guided atomistic reconstruction method, which allows building models with fine-tuned values of the in-plane (L$_a$) and out-of-plane (L$_c$) coherence lengths, and of the orientation angle (OA). Applying a parametric study of grain size and orientation distribution, a database of 210 models is presented with parameters spanning domains characteristic of high and medium textured pyrolytic carbons: 1.5–8 nm, 2–5.5 nm and 25–110°, for L$_c$, L$_a$ and OA, respectively. A machine learning model based on a random forest regression shows that these three measurable properties can be accurately predicted from a limited set of microscopic information characterizing the distribution of local atomic environments in the models. Finally, the computed diffraction properties and high-resolution transmission electron microscopy images of a series of six models, that best match the properties of a set of well-characterized pyrocarbons, are extensively compared to experimental data, showing excellent agreement and drastically improving over former modeling studies on high textured pyrocarbons, in addition to providing the first atomistic model of a medium textured pyrocarbon.< Réduire

Mots clés en anglais

Pyrolytic carbon Structure Texture Modeling Machine learning