Reactive force field (ReaxFF) molecular dynamics simulations have emerged as a powerful tool for studying chemical vapor deposition (CVD) processes, particularly for the synthesis of two-dimensional (2D) materials such as molybdenum disulfide (MoS2) and hexagonal boron nitride (hBN). These simulations bridge the gap between quantum mechanical methods and empirical force fields, enabling the study of complex chemical reactions and dynamic processes at larger scales and longer times than density functional theory (DFT) while retaining reasonable accuracy. The ReaxFF framework is especially suited for modeling CVD because it captures bond formation and breaking, charge transfer, and van der Waals interactions, all of which are critical for understanding precursor decomposition, surface diffusion, and growth mechanisms.

Parameterization of the ReaxFF potential for CVD simulations begins with high-quality DFT data, which provides reference energies, forces, and reaction barriers for key chemical species and intermediates involved in the deposition process. For MoS2 growth, this includes molybdenum and sulfur-containing precursors such as MoO3 and H2S, as well as intermediate species like MoSx clusters. Similarly, for hBN, precursors like ammonia borane or borazine are considered. The ReaxFF parameters are optimized to reproduce DFT-calculated reaction energetics, bond lengths, and angles, ensuring that the force field accurately describes precursor decomposition pathways and interactions with the substrate. Validation involves comparing simulated reaction pathways with known experimental or theoretical data, such as the activation energy for precursor dissociation or the stability of intermediate species.

In CVD simulations, ReaxFF captures the stepwise decomposition of precursors into reactive fragments that contribute to film growth. For example, in MoS2 synthesis, the reduction of MoO3 by H2S proceeds through intermediate oxysulfides, releasing water and forming Mo-S bonds. The simulations track these reactions in real-time, revealing how gas-phase chemistry influences the availability of growth species. Surface diffusion of these species is another critical factor, as it determines the likelihood of attachment at the edges of growing domains. ReaxFF simulations show that Mo and S adatoms exhibit distinct diffusion barriers on the substrate, with sulfur generally being more mobile. This difference in mobility affects the stoichiometry and morphology of the growing film.

Edge attachment kinetics play a crucial role in determining the size and shape of 2D domains. ReaxFF simulations reveal that MoS2 domains grow preferentially along zigzag edges due to lower energy barriers for adatom incorporation compared to armchair edges. For hBN, the growth is influenced by the competition between B and N attachment rates, which can lead to stoichiometric imbalances if not carefully controlled. The simulations also predict the formation of defects such as vacancies, antisites, and grain boundaries, which arise from kinetic limitations or non-ideal precursor ratios. For instance, sulfur vacancies in MoS2 are commonly observed in simulations when H2S concentration is insufficient, matching experimental observations.

One of the challenges in ReaxFF simulations of CVD is accurately modeling substrate interactions. The choice of substrate—commonly sapphire, silicon dioxide, or graphene—affects the adhesion and diffusion of growth species. Substrate interactions are often parameterized using DFT calculations of adsorption energies and diffusion barriers. However, simulating multilayer growth introduces additional complexity, as interlayer van der Waals forces and strain effects must be accounted for. ReaxFF can model these interactions but may require adjustments to the force field to prevent unrealistic stacking behaviors.

Experimental techniques such as Raman mapping and scanning tunneling microscopy (STM) provide critical validation for simulation predictions. Raman spectra of MoS2 and hBN films reveal information about layer thickness, strain, and defect density, which can be correlated with simulated domain structures. For example, the relative intensities of the E2g and A1g Raman modes in MoS2 indicate layer number and strain, matching trends observed in simulations of domain growth under varying conditions. STM characterization further confirms the presence of predicted defects and edge structures, providing atomic-scale validation of the simulation results.

ReaxFF simulations offer valuable insights for optimizing CVD processes. By identifying rate-limiting steps, such as precursor decomposition or edge attachment, simulations can guide the selection of temperature, pressure, and precursor ratios to achieve desired film quality. For instance, simulations suggest that higher H2S partial pressures reduce sulfur vacancies in MoS2, while controlled borazine flow rates improve hBN stoichiometry. Additionally, the simulations highlight the importance of substrate pretreatment and temperature gradients in promoting uniform growth.

Despite their advantages, ReaxFF simulations have limitations. The computational cost scales with system size, restricting simulations to domains on the order of tens of nanometers, whereas experimental films often cover micrometers. Furthermore, the force field’s accuracy depends heavily on the quality of the DFT training data, and certain exotic intermediates may not be well-represented. Future improvements may involve coupling ReaxFF with machine learning potentials to enhance accuracy and scalability.

In summary, ReaxFF molecular dynamics simulations provide a detailed atomic-scale understanding of CVD growth for 2D materials like MoS2 and hBN. By combining insights from simulations with experimental characterization, researchers can refine synthesis protocols to achieve larger, higher-quality films with controlled properties. The continued development of reactive force fields and computational methods will further enhance their predictive power for nanomaterials synthesis.