24th International Colloquium Tribology - January 2024 169 Tribochemical Reactions in the Degradation Process of Iron Nitride with Reactive Molecular Dynamics Simulation Development of Reactive Force Field Parameters for Elucidation of Tribochemical Reactions of Iron Nitride Mizuho Yokoi 1 , Masayuki Kawaura 1 , Shogo Fukushima 1 , Yuta Asano 1 , Yusuke Ootani 1 , Nobuki Ozawa 2, 1 , Momoji Kubo 1, 2* 1 Institute for Materials Research, Tohoku University, Sendai, Japan 2 New Industry Creation Hatchery Center, Tohoku University, Sendai, Japan * Corresponding author: E-mail: momoji@tohoku.ac.jp 1. Introduction Modern structural materials must exhibit long service lifespans and withstand harsh environmental conditions. Especially, components subjected to friction, such as gears, require exceptional wear resistance. Nitriding is a surface treatment technique that enhances wear resistance by forming a compound layer that consists of Fe 3 N, Fe 4 N, or a combination of both crystals. Previous studies have suggested that the compound layer is harder than steel, and thus more resistant to wear [1]. In addition, it has been reported that the Fe 4 N layer exhibited less wear than Fe 3 N [2]. As these previous studies show, the amount of wear has been thought to correlate with the hardness and crystal structure. However, the recent studies have shown that metal wear phenomena also depend on the chemical reactions between the sliding material and substance in diverse atmospheres. This dependence is evident from the observation of the reduced wear of steel in high-humidity environments and the increased wear of stainless steel in oxygenated environments [3, 4]. Therefore, to enhance the wear resistance of nitrided steels, understanding the chemical reaction dynamics between the surface compound layer and substances in diverse atmospheres is required. Especially, chemical reactions with water are important because they can strongly affect wear behaviors; water causes corrosive wear of steel in aqueous environments [5], whereas water reduces wear of steel in high humidity conditions [3]. However, the observation of chemical reactions at friction interface in an atomic scale is still challenging. Molecular dynamics (MD) simulations using the reactive force field (ReaxFF) play a crucial role in analyzing such chemical reaction dynamics [6]. ReaxFF determines the potential energy of a system based on variable bond-order and atomic charge depending on the atomic configuration. Therefore, ReaxFF can handle chemical reaction dynamics with less computational costs than first-principles MD calculations. ReaxFF has been widely used for studying the chemical reactions of various structural and catalytic materials [7, 8] as well as tribological materials [9]. Therefore, we aimed at the analysis of tribochemical reactions between iron nitride and water with ReaxFF for revealing the effects of tribochemical reactions on the wear. The simulation necessitates ReaxFF parameters for Fe, N, H, and O. While the parameters dealing with the chemical reactions between H/ O/ Fe, as well as those of H/ O/ N, have been previously published [10, 11], there exists no published parameter set of Fe and N suitable for iron nitride. This proceeding described the development of the ReaxFF parameters of the Fe/ N system to perform these tribochemical reaction simulations. 2. Parameterization Method The ReaxFF parameters of Fe/ N were developed so as to reproduce of the structure and energy obtained from the density functional theory (DFT) calculations. For this development, we performed DFT calculations for several iron nitride models. Bulk iron nitride and iron nitride crystal surfaces were used to develop the ReaxFF parameters. First, we need to develop ReaxFF parameters that can reproduce the stability of bulk iron nitride. Thus, volume energy curves were calculated for Fe 3 N and Fe 4 N crystals by changing the lattice constants a, b, and c uniformly using DFT for the reference to develop the ReaxFF parameter. Second, we need to develop ReaxFF parameters that can reproduce the stability of the iron nitride surfaces; Thus, the surface energy was calculated for Fe 3 N (001), (100), (110), and (111), as well as for Fe 4 N (100), (110) and (111) using DFT for the reference to develop the ReaxFF parameters. The surface energy E surface , was calculated from the bulk energies E bulk , slab model energies E slab , and the surface area of each orientation A using the following equation: In the DFT calculations, generalized gradient approximation type Perdew-Burke-Ernzerho (GGA-PBE) exchange-correlation functional and DNP basis set were used with effective core potentials. We employed DMol 3 code for DFT calculations [12, 13]. 3. Results and Discussion First, we performed DFT and ReaxFF calculations for volume-energy curves of Fe 3 N and Fe 4 N. Figure 1 illustrates the volume-energy curves obtained by DFT as well as ReaxFF 170 24th International Colloquium Tribology - January 2024 Tribochemical Reactions in the Degradation Process of Iron Nitride with Reactive Molecular Dynamics Simulation with the developed parameters. The crystal structures of Fe 3 N and Fe 4 N are shown in the inset. The origin of each volume-energy curve is the minimum energy value for each curve. As shown in Figure 1, the lattice volume that gives the minimum energy obtained from the DFT calculations and those obtained using ReaxFF are almost identical for both crystals. This indicates that the developed ReaxFF parameters reproduce the lattice constants of both crystals. Moreover, since the curve shapes obtained from the DFT and ReaxFF are in good agreement, the developed ReaxFF parameters are expected to reproduce the expansion and compressibility coefficients of the Fe 3 N and Fe 4 N. Figure 1 Volume-energy curves of (a) Fe 3 N and (b) Fe 4 N crystal obtained by DFT as well as ReaxFF with the developed parameters. Second, to reproduce the energetic stability of Fe 3 N and Fe 4 N surfaces, which is important for describing tribochemical reactions, we performed the DFT and ReaxFF calculations for slab models of Fe 3 N and Fe 4 N. Figure 2 shows slab models of Fe 3 N (001), (100), (110) and (111), as well as for Fe 4 N (100), (110) and (111) planes. Figure 2 Slab models of (a) (001), (b) (100), (c) (110), and (d) (111) planes of Fe 3 N and (e) (100), (f) (110), and (g) (111) of Fe 4 N. Figure 3 presents the surface energy values obtained by DFT as well as ReaxFF with developed parameters. The DFT calculation results indicate that the surface energy of Fe 4 N is generally lower than that of Fe 3 N, except for the Fe 3 N (111) surface. Furthermore, the surface energy of Fe 3 N exhibited greater variation across surface orientations compared to the surface energy of Fe 4 N. These surface energy trends are also well reproduced in the developed ReaxFF parameters. The errors in the surface energy obtained by ReaxFF are within 10% for all planes. Therefore, ReaxFF with the developed parameters adequately reproduces the surface stability, although the ReaxFF with the developed parameters slightly overestimates the surface energies. Figure 3 Surface energies of Fe 3 N and Fe 4 N crystal obtained by DFT as well as ReaxFF with the developed parameters. From these comparisons, it is evident that the newly developed ReaxFF parameters for Fe/ N effectively represent the bulk and surface states of iron nitride. The developed ReaxFF parameters are listed in the Appendix. 4. Conclusion To investigate the wear and degradation mechanisms of iron nitride, we developed the ReaxFF parameters of the Fe/ N system. The ReaxFF parameters can reproduce volume-energy curves and surface energies of iron nitrides. Therefore, we can utilize these ReaxFF parameters for sliding simulations to elucidate the tribochemical reactions of iron nitrides. References [1] S. Karaoğlu, Mat. Charact., 49, 349 (2002). [2] T. Peng, et al., Surf. Coat. Technol., 403, 126403 (2020). [3] G. Bregliozzi, et al., Tribol. Lett., 17, 697 (2004). [4] M. Esteves, et al., Tribol. Int., 88, 56 (2015). [5] J. Jiang, et al., Wear, 261, 964 (2006). [6] A. C. T. van Duin, et al., J. Phys. Chem. A, 105, 9396 (2001). [7] Y. Ootani, et al., J. Phys. Chem. C, 126, 2728 (2022). [8] L. Chen, et al., Appl. Surf. Sci., 563, 150097 (2021). [9] Y. Wang, et al., Sci. Adv., 5, eaax9301 (2019). [10] M. Aryanpour, et al., J. Phys. Chem. A, 114, 6298 (2019). [11] Y. Wang, et al., J. Phys. Chem. C, 124, 10007 (2020). [12] B. Delley, J. Chem. Phys., 92, 508 (1990). [13] B. Delley, J. Chem. Phys., 113, 7756 (2000).