Insights into the primary radiation damage of silicon by a machine learning interatomic potential

Hamedani, A.; Byggmästar, Jesper; Djurabekova, Flyura; Alahyarizadeh, Gh.; Ghaderi, Reza; Minuchehr, A.; Nordlund, K.

doi:10.1080/21663831.2020.1771451

“…The machine learning potential is also found to behave similarly as the SW potential for the annealing process in terms of N int . 14 This corroborates the More atoms are treated with the higher-level method in QM/MM than in MM/MM according to Fig. 9(c), indicating a larger disordered region in the QM/MM simulation.…”

Section: Sdac Qm/mm Simulation Of the Dd Generation In Bulk Siliconsupporting

confidence: 83%

“…5,7,18,28,[93][94][95] Different properties are used in literature to identify the atoms in the disordered region, such as the atomic potential energies, 4,7 the bond angles, 5 the ring structures, 15,28 the Lindemann spheres, 9,30 and the time-average of atomic positions. 10 The Wigner-Seitz defect analysis 84 has been widely used as well, 6,11,14,21 even though it is less suitable for amorphous structures. I use the atomic potential energy for simplicity, and atoms with potential energies greater than the crystalline average by 0.2 eV 4,7 are identified as disordered atoms.…”

Section: Comparison Of Dynamicsmentioning

confidence: 99%

“…6,17,[20][21][22]43 The deficiencies of these methods in simulating the DD can be seen in the QM studies of model systems, [20][21][22][23] which reveal effects and processes in the DD that were missing from BCA and classical MD simulations. A more clear demonstration of the deficiencies is recently provided by Hamedani, et al, 14 in which a DFT-trained machine-learning potential is shown to yield highly different DD results from those of commonly used classical potentials.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Speed-Dependent Adaptive Partitioning QM/MM for Displacement Damage Simulations

Yang

¹

2020

Preprint

0

View full text Add to dashboard Cite

Solids receive displacement damages (DD) when interacting with energetic particles, which may happen during the fabrication of semiconductor devices, in harsh environments and in certain analysis techniques. Simulations of the DD generation are usually carried out with classical molecular dynamics (MD), but classical MD does not account for all the effects in DD, as demonstrated by ab initio calculations of model systems in literature. A fully ab initio simulation of the DD generation is impractical due to the large number of atoms involved. In my previous paper [Phys. Chem. Chem. Phys. 22, 19307 (2020)], I developed an adaptive-center (AC) method for adaptive-partitioning (AP) quantum mechanics/molecular mechanics (QM/MM) simulations, allowing the active region centers and the QM/MM partition to be determined on-the-fly for energy-conserving AP-QM/MM methods. I demonstrated that the AC-AP-QM/MM is applicable to the simulation of the DD generation, so that the active regions can be treated with an ab initio method. The AC method was unable to identify the fast-moving recoil ions in the DD generation as active region centers, however, and the accuracy is negatively affected by the rapid change in QM/MM partition of the system. In this paper, I extend the AC method and develop a speed-dependent adaptive-center (SDAC) method for proper AP-QM/MM simulations of DD. The SDAC method is applicable to general problems with speed-dependent active regions, and is compatible with all existing energy-conserving partition-by-distance AP-QM/MM methods. The artifact due to the speed-dependent potential energy surface can be made small by choosing proper criteria. I demonstrate the SDAC method by simulations of the DD generation in bulk Silicon.

show abstract

“…Together with the previously evidenced high quality of the deposition simulations, i.e., the good agreement with experimental observables observed in initial work [24], this suggests that the carbon GAP is indeed able to capture the deposition process correctly. In this context, we mention the recently demonstrated usefulness of GAP simulations for radiation damage in elemental tungsten and silicon, where the impact of (very) highly energetic ions must be correctly described as well [41][42][43].…”

Section: A Gaussian Approximation Potential (Gap) Modeling Of Amorphmentioning

confidence: 99%

Machine learning driven simulated deposition of carbon films: From low-density to diamondlike amorphous carbon

Miguel

¹

,

Csänyi

²

,

Laurila

³

et al. 2020

View full text Add to dashboard Cite

Amorphous carbon (a-C) materials have diverse interesting and useful properties, but the understanding of their atomic-scale structures is still incomplete. Here, we report on extensive atomistic simulations of the deposition and growth of a-C films, describing interatomic interactions using a machine learning (ML) based Gaussian Approximation Potential (GAP) model. We expand widely on our initial work [Phys. Rev. Lett. 120, 166101 (2018)] by now considering a broad range of incident ion energies, thus modeling samples that span the entire range from low-density (sp 2-rich) to high-density (sp 3-rich, "diamond-like") amorphous forms of carbon. Two different mechanisms are observed in these simulations, depending on the impact energy: low-energy impacts induce sp-and sp 2-dominated growth directly around the impact site, whereas high-energy impacts induce peening. Furthermore, we propose and apply a scheme for computing the anisotropic elastic properties of the a-C films. Our work provides fundamental insight into this intriguing class of disordered solids, as well as a conceptual and methodological blueprint for simulating the atomic-scale deposition of other materials with ML-driven molecular dynamics.

show abstract

“…I develop the SDAC method by extending the AC method to allow speed-dependent criterion, so as to carry out QM/MM simulations of the DD generation in solids. Such simulations are carried out in the NVE ensemble with temperature controlling layers, [4][5][6][7][8][9][10][11][12][13][14][15] so I only discuss NVE simulations in the following. A speed-dependent ξ results in a PES that depends on both the atomic positions and the speeds, and the Newton's second law EOM no longer applies.…”

Section: The Sdac Equation Of Motionmentioning

confidence: 99%

Speed-Dependent Adaptive Partitioning QM/MM for Displacement Damage Simulations

Yang¹

2020

Preprint

0

View full text Add to dashboard Cite

Solids receive displacement damages (DD) when interacting with energetic particles, which may happen during the fabrication of semiconductor devices, in harsh environments and in certain analysis techniques. Simulations of the DD generation are usually carried out with classical molecular dynamics (MD), but classical MD does not account for all the effects in DD, as demonstrated by ab initio calculations of model systems in literature. A fully ab initio simulation of the DD generation is impractical due to the large number of atoms involved. In my previous paper [Phys. Chem. Chem. Phys. 22, 19307 (2020)], I developed an adaptive-center (AC) method for adaptive-partitioning (AP) quantum mechanics/molecular mechanics (QM/MM) simulations, allowing the active region centers and the QM/MM partition to be determined on-the-fly for energy-conserving AP-QM/MM methods. I demonstrated that the AC-AP-QM/MM is applicable to the simulation of the DD generation, so that the active regions can be treated with an ab initio method. The AC method was unable to identify the fast-moving recoil ions in the DD generation as active region centers, however, and the accuracy is negatively affected by the rapid change in QM/MM partition of the system. In this paper, I extend the AC method and develop a speed-dependent adaptive-center (SDAC) method for proper AP-QM/MM simulations of DD. The SDAC method is applicable to general problems with speed-dependent active regions, and is compatible with all existing energy-conserving partition-by-distance AP-QM/MM methods. The artifact due to the speed-dependent potential energy surface can be made small by choosing proper criteria. I demonstrate the SDAC method by simulations of the DD generation in bulk Silicon.

show abstract

Insights into the primary radiation damage of silicon by a machine learning interatomic potential

Cited by 19 publications

References 64 publications

Speed-Dependent Adaptive Partitioning QM/MM for Displacement Damage Simulations

Speed-Dependent Adaptive Partitioning QM/MM for Displacement Damage Simulations

Machine learning driven simulated deposition of carbon films: From low-density to diamondlike amorphous carbon

Speed-Dependent Adaptive Partitioning QM/MM for Displacement Damage Simulations

Contact Info

Product

Resources

About