Modeling phase separation in polymer-derived silicon carbonitride ceramics through extended machine learning molecular dynamics

Fabien Mortier, Sylvian Cadars, Olivier Masson, Mauro Boero, Guido Ori, Yun Wang, Samuel Bernard, Assil Bouzid ∗ · 2026

Read Modeling phase separation in polymer-derived silicon carbonitride ceramics through extended machine learning molecular dynamics by Fabien Mortier, Sylvian Cadars, Olivier Masson, Mauro Boero, Guido Ori, Yun Wang, Samuel Bernard, Assil Bouzid ∗ (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-20358.txt.

Introduction

Mortier and seven co-workers model phase separation in polymer-derived silicon carbonitride ceramics. They train a machine-learning interatomic potential on over 9000 varied setups, from amorphous models to hot states, surface models and crystal predictions. The potential reproduces pair distribution functions with high fidelity and explains the mix of ceramic and graphitic traits. Defective five- and seven-member carbon rings mediate the shift to stable six-member aromatic rings.

Catalog record

CREATORFabien Mortier, Sylvian Cadars, Olivier Masson, Mauro Boero, Guido Ori, Yun Wang, Samuel Bernard, Assil Bouzid ∗
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87739
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.20358

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Modeling phase separation in polymer-derived silicon carbonitride ceramics through extended machine learning molecular dynamics Fabien Mortier, Sylvian Cadars, Olivier Masson, Mauro Boero, Guido Ori, Yun Wang, Samuel Bernard, Assil Bouzid ∗ Abstract Polymer-derived ceramics combine the thermal stability of ceramics with the versatile properties of carbon domains, but modeling their atomic-scale evolution during processing remains elusive due to the limitations of traditional computational methods. To address this issue, here we develop and apply a machine learning interatomic potential for silicon carbonitride-based (Si-C-N-H) systems, trained on a diversified database of over 9000 configurations - including amorphous models, high-temperature states, surfaces, and crystal structure predictions - to capture the full complexity of these materials. This potential enables large-scale molecular dynamics simulations of 8000-atom systems revealing the atomic-scale evolution of the polymer-derived ceramic during thermal treatment. A key result of this work is the occurrence of a phase separation where carbon domains progressively nucleate from the amorphous SiCN matrix during thermal processing, forming distinct graphene-like sheets while preserving the integrity of the ceramic network. The resulting models reproduce the experimental atomic pair distribution functions with exceptional fidelity, validating our approach and providing microscopic explanations for the material unique combination of ceramic and graphitic properties. In this process, defective 5- and/or 7-member carbon rings, mediate the transformation to stable 6-member aromatic structures. These findings offer new atomic-scale insights into the thermal stability and structural transformation pathways of polymer-derived ceramics, while our methodology opens avenues for studying complex amorphous systems with first-principles accuracy at experimentally relevant scales. I Introduction Polymer-derived ceramics (PDCs) are advanced materials synthesized through pyrolysis of preceramic polymers such as polysilazanes, inorganic polysiloxanes or polycarbosilanes 16 ; 60 . This synthetic route, first pioneered by Fritz and Raabe in the 1950s and later expanded by Ainger, Herbert, and Chantrell in the 1960s, enables the production of ceramics with tailored compositions and microstructures at relatively low temperatures (800–1500 ∘ C) compared to conventional ceramic manufacturing 25 ; 2 ; 11 . PDCs can adopt complex shapes and are compatible with several shaping techniques such as additive manufacturing, fiber spinning, or surface coating. In addition, thanks to their chemical and structural diversity, their thermal, electrical, and mechanical properties can be tuned 17 ; 59 ; 12 ; 14 . Among the most studied PDC systems one can cite silicon carbide (SiC), silicon nitride ( \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{3}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{4}}} ), silicon oxycarbide (SiOC), and silicon carbonitride (SiCN). These materials exhibit distinct microstructures and functional properties that are tightly linked to the synthesis procedure 22 ; 5 . Among the PDCs, SiCN materials, when derived from polysilazane polymers pyrolysed at temperatures below 1400 ∘ C, feature a particularly interesting complex microstructure that arises from the nanoscale phase separation between an amorphous silicon-based network, typically composed of SiC x N 4-x tetrahedra, and a "free carbon" phase with a degree of organisation that depends on the pyrolysis temperature 71 ; 66 ; 70 . This phase separation is driven by the decomposition of organic groups during pyrolysis, leading to the precipitation of carbon-rich regions within the ceramic network. At temperatures below 1000 ∘ C, the free carbon phase consists of small, disordered sp 2 -hybridized C clusters, while at higher temperatures (above 1200 ∘ C), it progressively organizes into a turbostratic phase consisting of more-or-less extended stacked arrangements of potentially-defected graphitic layers, with various degrees of ordering up to graphite 66 ; 13 . Its presence is critical for boosting the overall electrical conductivity of the material, as sp 2 -hybridized carbon domains exhibit metallic-like behavior, whereas the SiCN matrix remains semiconducting 67 ; 1 . Moreover, this free carbon phase acts as a diffusion barrier, and controlling its extent and degree of organization can prevent the crystallization of \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{3}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{4}}} or SiC domains, thereby improving the thermal stability of the final material 34 ; 49 ; 33 ; 20 . The functional properties of PDCs, particularly their electrical conductivity and catalytic activity, are closely tied to the organization of the free carbon phase. For instance, in metal-containing PDC such as SiCN-Ni or SiCN-Fe nanocomposites, the free carbon phase facilitates charge transfer between metal nanoparticles and the ceramic matrix, enhancing water splitting catalytic performance 50 ; 8 . However, if the carbon phase forms a dense or continuous network, it may also acts as a diffusion barrier, potentially hindering ionic transport 34 ; 49 ; 33 ; 20 . Thus, while electronic conductivity is improved, ionic diffusion could be compromised depending on the morphology of the carbon phase. Within this context, the atomic-scale mechanisms governing the formation, growth, and interaction of this phase with the surrounding matrix are of primary importance, and yet remain poorly understood. Experimental techniques such as transmission electron microscopy (TEM), Raman spectroscopy, and solid-state NMR provide valuable insights into the structure of PDCs, and specifically the free carbon phase, but each has inherent limitations. TEM can detect graphitic domains but struggles to reveal the fine details of how these domains form at the atomic level or how they interact with the surrounding ceramic matrix 66 . Raman spectroscopy effectively identifies sp 2 -hybridized carbon but cannot differentiate between small, isolated carbon clusters and larger, extended graphitic sheets 13 . Similarly, solid-state NMR offers detailed information about local C and Si atomic environments, but also lacks the longer-range information needed to fully map the phase separation within these complex materials 72 . These limitations highlight the necessity of atomic-scale computational models to provide a complementary picture of the complex PDC composition-structure relationship escaping experimental probes. Early computational studies of PDCs relied on empirical potentials, such as the Tersoff potential, or reverse Monte Carlo (RMC) methods. Dürr et al. 21 used RMC to model a \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{24}}}\text{C}{\vphantom{\text{X}}}_{\smash[t]{\text{43}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{33}}} composition, confirming phase separation between \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{3}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{4}}} -like tetrahedra and amorphous carbon. Matsunaga et al. 48 and Liao et al. 43 employed Tersoff potentials to study diffusion and mechanical properties, but these models lacked chemical reactivity and turned out to be insufficient to capture mixed SiC x N 4-x environments. As shown by Kroll, first-principles molecular dynamics (FPMD) on \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{40}}}\text{C}{\vphantom{\text{X}}}_{\smash[t]{\text{40}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{40}}} systems demonstrated the occurrence of phase separation and the formation of carbon clusters, but the small system sizes (120 atoms) and short simulation times (40 ps) limited insights into the long-term structural evolution 38 . In another work, Amkreutz et al. 3 used density functional-based tight binding (DFTB) to study the phase separation in \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{37}}}\text{C}{\vphantom{\text{X}}}_{\smash[t]{\text{32}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{31}}} systems, proposing that the free carbon phase could exist as either small clusters or extended layered structures. While insightful, these studies left several open questions about the dynamical evolution of the free carbon phase. In a recent FPMD-based study 51 we addressed some of these limitations by focusing on the \text{Si}{\vphantom{\text{X}}}_{\smash[t]{\text{32}}}\text{C}{\vphantom{\text{X}}}_{\smash[t]{\text{25}}}\text{N}{\vphantom{\text{X}}}_{\smash[t]{\text{24}}}\text{H}{\vphantom{\text{X}}}_{\smash[t]{\text{19}}} PDC derived from \text{\hskip 1.29167pt--\hskip 1.29167pt}\text{[}\text{Si}\text{(}\text{CH}{\vphantom{\text{X}}}_{\smash[t]{\text{3}}}\text{)}\text{(}\text{CH}\hbox to0.0pt{\raisebox{0.86108pt}{\text{\hskip 1.29167pt--\hskip 1.29167pt}}\hss}\raisebox{-0.86108pt}{\text{\hskip 1.29167pt--\hskip 1.29167pt}}\text{CH}{\vphantom{\text{X}}}_{\smash[t]{\text{2}}}\text{)}\text{\hskip 1.29167pt--\hskip 1.29167pt}\text{NH}\text{\hskip 1.29167pt--\hskip 1.29167pt}\text{]}\text{}{\vphantom{\text{X}}}_{\smash[t]{\text{0.2}}}\text{\hskip 1.29167pt--\hskip 1.29167pt}\text{[}\text{Si}\text{(}\text{H}\text{)}\text{(}\text{CH}{\vphantom{\text{X}}}_{\smash[t]{\text{3}}}\text{)}\text{\hskip 1.29167pt--\hskip 1.29167pt}\text{NH}\text{\hskip 1.29167pt--\hskip 1.29167pt}\text{]}\text{}{\vphantom{\text{X}}}_{\smash[t]{\text{0.8}}} – polyvinylsilazane. Using the ceramicNetworkBuilder algorithm (available in the Python Atomic Modeling Analyzes library 10 ), we generated several initial configurations differing in the organization and networking of the free carbon phase, including clusters, chains, and pre-inserted graphitic layers. After thermal annealing, the models were validated against experimental pair distribution functions (PDFs), revealing that models where carbon is essentially made of poorly spatially correlated layers, rather than

Read the complete text