betaFeedback

WorksPhysics and Astronomy

Transferability of force fields for 2D silicon (silicene)

Abstract

An ability of various interatomic potentials to reproduce the properties of silicene (2D silicon) polymorphs were examined. Structural and mechanical properties of the flat (FS), low-buckled (LBS), trigonal dumbbell (TDS), honeycomb dumbbell (HDS) and large honeycomb dumbbell (LHDS) single-layer silicon (silicene) phases, were obtained using density functional theory (DFT) and molecular statics (MS) calculations with Tersoff, MEAM, Stillinger-Weber, EDIP, ReaxFF, COMB and machine-learning-based (ML-IAP) interatomic potentials. A quantitative systematic comparison and discussion of the results obtained are reported.

Introduction

We live in the "Silicon Age" due to the great importance of elemental silicon to the modern global economy, it is mainly about electronics. Silicon is one of the best investigated materials and the quality of its production is impressive. However, this applies to bulk silicon. The interest in the graphene has also led to interest in other non-carbon 2D materials [1, 2]. One such material is precisely the 2D silicon called silicene [3, 4]. Using first-principles methods, with current computer resources, we can model structures up to about a few hundred atoms. For larger systems, we need approximate methods, e.g. molecular dynamics/statics. For these methods, the quality of the interatomic potentials (IAPs) used is crucial. Because of the importance of silicon, but also its complexity, dozens of potentials have been proposed for it. In the very well-known NIST Interatomic Potentials Repository, there are 27 potentials for silicon (no other element has more potentials), the oldest one from 1985 and the latest from 2020 [5].

At least five 2D silicon polymorphs have been reported in the literature, that is, flat (FS), low-buckled (LBS) [6], trigonal dumbbell (TDS), honeycomb dumbbell (HDS) and large honeycomb dumbbell (LHDS) [7]. There are still doubts about their dynamic stability, e.g. for a flat phase the negative ZO phonon mode could be removed by the selection of an appropriate substrate, see [3, 8]. The ability of potentials for 3D silicon to reproduce 2D silicon is poorly studied. There are several papers where the quality of potentials for 3D silicon has been assessed, see e.g. [9, 10, 11] but not for silicene. The intention of this work is first to determine the structural and mechanical properties of 2D silicon using first-principles method and then to test the ability of different interatomic potentials to reproduce these properties.

Methods

Analyzing the available literature concerning all phases of single-layer (SL) Si, it is not feasible to find all structural, mechanical and phonon data obtained in one consistent way. The availability of experimental data is actually limited to the silicene grown on some support, a pristine free-standing SL sheet of silicene has not been discovered to date [4, 12], and therefore we must use ab initio calculations. Unfortunately, also ab initio calculations, most often DFT, differ in the calculation methodology, i.e., they use different functional bases, different pseudopotentials or exchange-correlation (XC) functionals, and parameters such as cohesive energy and elastic constants are poorly accessible. For this reason, structural and mechanical data, i.e., lattice parameters, average cohesive energy, average bond length, average height, 2D elastic constants as well as phonon data are determined here using a single consistent first-principles approach as described in the next § 2.1. These data were further considered as reference data and marked as ValueDFT\mathrm{Value}^{\mathrm{DFT}}. Then the same data were determined, as described in § 2.2 using the analyzed molecular potentials from Subsection 2.2.1 and are marked as Valuepotential\mathrm{Value}^{\mathrm{potential}}. Having both data, we can simply define mean absolute percentage error (MAPE):

MAPE=100%n∑t=1nValueDFT−ValuepotentialValueDFT(1)\mathrm{MAPE}=\frac{100\%}{n}\sum_{t=1}^{n}\frac{\mathrm{Value}^{\mathrm{DFT}}-\mathrm{Value}^{\mathrm{potential}}}{\mathrm{Value}^{\mathrm{DFT}}} \tag*{(1)}

that will allow us to quantify the potentials under examination.

For 2D materials, directional 2D Young’s moduli

E[10]2D=C11C22−C122C22andE[01]2D=C11C22−C122C11(2)E^{2\mathrm{D}}_{[10]}=\frac{C_{11}C_{22}-C_{12}^{2}}{C_{22}} \quad\text{and}\quad E^{2\mathrm{D}}_{[01]}=\frac{C_{11}C_{22}-C_{12}^{2}}{C_{11}} \tag*{(2)}

2D Poisson’s ratios

ν[10]2D=C12C22andν[01]2D=C12C11(3)\nu^{2\mathrm{D}}_{[10]}=\frac{C_{12}}{C_{22}} \quad\text{and}\quad \nu^{2\mathrm{D}}_{[01]}=\frac{C_{12}}{C_{11}} \tag*{(3)}

and the 2D shear modulus

G2D=C33(4)G^{2\mathrm{D}}=C_{33} \tag*{(4)}

are often used instead of elastic constants CijC_{ij}. Due to symmetry for hexagonal lattices, this reduces to one 2D Young’s modulus EE and a one 2D Poisson’s ratio ν\nu, see [13].

AB INITIO Calculations

The ab initio calculation methodology here is closely analogous to that used in [14], i.e., the density functional theory (DFT) [15, 16], ABINIT plane-wave approximation code [17, 18], local density approximation (LDA) [19, 20] as an exchange-correlation (XC) functional, optimized norm-conserving Vanderbilt pseudopotential [21] (ONCVPP). The cut-off energy and the electron configuration for Si used in the DFT calculations according to the pseudopotential and Gaussian smearing scheme with tsmear(Ha)=0.02t_{\mathrm{smear}}(\mathrm{Ha})=0.02 were used. To generate K-PointS grids kptrlenkptrlen was set to 43.0. Since 2D structures were analyzed in the zz-direction, a vacuum of 20 Å was applied. The initial data for the five structures analyzed were deduced from [6] and [7]. The structures were then carefully relaxed with full optimization of cell geometry and atomic coordinates [14].

The average cohesive energy EcE_{\mathrm{c}} (eV/atom) was computed as the difference in the total energy of a given relaxed earlier structure and its individual atoms placed in a cubic box of sufficient size. The theoretical ground state, T=0T=0 K, elastic constants, CijC_{ij}, of all the previously optimized structures were computed using the metric tensor formulation of strain in density functional perturbation theory (DFPT) [22]. The mechanical stability of the analyzed structures was verified by calculating the so-called Kelvin moduli [23, 24, 25]. To calculate the phonons, the density functional perturbation theory implemented in ABINIT [17, 18] was employed. The phonon dispersion curves along the path Γ[0,0,0]\Gamma[0,0,0]-M[1/2,0,0]M[1/2,0,0]-K[1/3,1/3,0]K[1/3,1/3,0]-Γ[0,0,0]\Gamma[0,0,0] [26] of the analyzed structures were then used to identify their dynamical stability [27], complementary to the mechanical stability.

Molecular Calculations

To perform molecular calculations the molecular statics (MS) method, T=0T=0 K, [28, 29, 30] was used by means of the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) [31] and analyzed by means of the Open Visualization Tool (OVITO) [32].

As for DFT calculations, the structures were here fully pre-relaxed with the conjugate gradient (CG) algorithm and then the elastic constants, CijC_{ij}, were calculated for them using the stress–strain method with the maximum strain magnitude set to 10−610^{-6} [30, 31]. In the zz-direction, a vacuum was set to 20 Å.

To measure performance of the interatomic potentials analyzed, the series of molecular dynamics (MD) simulations (200 atoms and 10000 timesteps, NVT ensemble) and LAMMPS’s built-in function timesteps/s were used. The results were then normalized relative to the longest run time.

Interatomic Potentials

The parameterizations of the potentials listed below were obtained from the NIST Interatomic Potentials Repository and/or from LAMMPS code sources.

  1. Tersoff1988 [33]: the original Tersoff potential for silicon (it is important to remember that this paper proposed a form of potential rather than a specific parametrization for silicon)

  1. Tersoff2007 [34]: the Tersoff potential fitted to the elastic constants of diamond silicon

  1. Tersoff2017 [35]: newer, better optimized the Tersoff potential for silicon

  1. MEAM2007 [36]: a semi-empirical interatomic potential for silicon based on the modified embedded atom method (MEAM) formalism

  1. MEAM2011 [37]: spline-based modified embedded-atom method (MEAM) potential for Si fitted for silicon interstitials

  1. SW1985 [38]: the Stillinger–Weber (SW) potential fitted to solid and liquid forms of Si

  1. SW2014 [39]: the Stillinger–Weber (SW) potential fitted to phonon dispersion curves of a single-layer Si sheet

  1. EDIP [40]: the environment-dependent interatomic potential (EDIP) fitted to various bulk phases and defect structures of Si

  1. ReaxFF [41]: the reactive force-field (ReaxFF) fitted to a training set of DFT data that pertain to Si/Ge/H bonding environments

  1. COMB [42]: the charge optimized many-body (COMB) potential fitted to a pure silicon and five polymorphs of silicon dioxide

  1. SNAP [43]: the machine-learning-based (ML-IAP) linear variant of spectral neighbor analysis potential (SNAP) fitted to total energies and interatomic forces in ground-state Si, strained structures and slab structures obtained from DFT calculations

  1. qSNAP [43]: the machine-learning-based (ML-IAP) quadratic variant of spectral neighbor analysis potential (qSNAP) fitted to total energies and interatomic forces in ground-state Si, strained structures and slab structures obtained from DFT calculations

  1. SO(3) [44]: the machine-learning-based (ML-IAP) variant of the SO(3) smooth power spectrum potential (SO(3)) fitted to the ground-state of the crystalline silicon structure, strained structures, slabs, vacancy, and liquid configurations from DFT simulations

  1. ACE [45]: the machine-learning-based (ML-IAP) variant of the atomic cluster expansion potential (ACE) fitted to a wide range of properties of 3D silicon determined from the DFT calculation

Results and Discussion

Structural and Mechanical Properties

Basic cells for the five silicene polymorphs, i.e, the flat (FS):(hP2hP2, P6/mmmP6/mmm, no.191), low-buckled (LBS):(hP2hP2, P3ˉm1P\bar{3}m1, no.164), trigonal dumbbell (TDS):(hP7hP7, P6ˉ2mP\bar{6}2m, no.189), honeycomb dumb-bell (HDS):(hP8hP8, P6/mmmP6/mmm, no.191) and large honeycomb dumbbell (LHDS):(hP10hP10, P6/mmmP6/mmm, no.191) are depicted in Figure 1 and additionally the crystallographic data for them are stored in Crystallographic Information Files (CIFs) in the Supporting Information.

Polymorphs of silicene

Figure 1. Polymorphs of silicene.

The results of first-principles calculations show that all the silicene phases have hexagonal symmetry. The symmetry characteristic of a structure determines the symmetry of its physical properties (Neumann’s Principle and Curie laws) [46, 47]. For 2D linear hyperelastic materials, there are four classes of symmetry [24] and hexagonal symmetry implies isotropy of the stiffness tensor, i.e., there are only two distinct elastic constants and they satisfy, in Voigt notation, such conditions that C11=C22C_{11}=C_{22}, C33=(C11−C12)/2C_{33}=(C_{11}-C_{12})/2.

Determined from DFT computations structural and mechanical characteristics, namely lattice parameters, average cohesive energy, average bond length, average height, 2D elastic constants, 2D Young’s modulus, Poisson’s ratio and 2D Kelvin moduli, of the five silicene polymorphs analyzed are gathered in Table 1. Since we are analyzing free-standing silicene here, which has not yet been observed in experiments, we compare the results of the calculations with those of other authors. We find that the lattice constants, average bond length, average height and cohesive energy agree at the DFT level of accuracy with other calculations. Mechanical properties of silicene are available in the literature only for the LBS phase and are limited to 2D Young’s modulus and Poisson’s ratio only. These quantities are also in reasonable agreement with the present results. It is immediately worth noting that all the calculated 2D Kelvin moduli for all silicene phases are positive, which results in mechanical stability [24].

PolymorphFSLBSTDSHDSLHDS
SourceThis workRefs.This workRefs.This workRefs.This workRefs.This workRefs.
aa3.8553.90a3.8283.87a, 3.83b6.4346.52b6.2976.38b7.3347.425b
bb3.8553.90a3.8283.87a, 3.83b6.4346.52b6.2976.38b7.3347.425b
−-4.5624.764a4.5774.784a, 5.16b4.6794.6794.769
†^{\dagger}2.2252.2492.25b2.3312.3992.357
hh0.00.0a0.421a0.45a, 0.44b2.7342.6352.683
C11C_{11}84.869.2100.5141.6104.5
C22C_{22}84.869.2100.5141.6104.5
C12C_{12}40.622.152.396.452.7
C33C_{33}22.123.661.8a24.122.625.9
EE65.462.273.376.077.9
0.480.320.31a0.520.680.50
125.491.3152.8238.0157.2
44.347.148.345.251.8
44.347.148.345.251.8

Table 1. Table 1: Structural and mechanical properties of flat (FS), low-buckled (LBS), trigonal dumbbell (TDS), honeycomb dumbbell (HDS) and large honeycomb dumbbell (LHDS) silicene phases from density functional theory (DFT) calculations: lattice parameters a,ba,b (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Young’s modulus EE (N/m), Poisson’s ratio and 2D Kelvin moduli (N/m).

The phonon spectra along the Γ−M−K−Γ\Gamma-\mathrm{M}-\mathrm{K}-\Gamma path for the five silicene polymorphs is depicted in Figure 2. The analysis of the computed curves shows that the phases TDS, LBS, HDS and LHDS are not only mechanically but also dynamically stable, i.e., all phonon modes have positive frequencies anywhere. The FS phase is mechanically stable, but can be dynamically unstable, i.e., optical ZO phonon mode has a negative frequency. Other authors have also observed similar FS phase behavior [4, 12], however, but since silicene is not a free-standing structure in nature the selection of a proper substrate may dampen the out-of-plane vibration mode and a flat silicene may be produced [3]. So it was decided that this phase was also included in the molecular calculations.

Image not extracted

Figure 2. Phonon dispersion and density of states (DOS) of the flat (FS), low-buckled (LBS), trigonal dumbbell (TDS), honeycomb dumbbell (HDS) and large honeycomb dumbbell (LHDS) single-layer silicon (silicene) phases. High symmetry points: Γ[0,0,0]\Gamma[0,0,0], M[1/2,0,0]\mathrm{M}[1/2,0,0], K[1/3,1/3,0]\mathrm{K}[1/3,1/3,0].

Performance of Interatomic Potentials

The computed with the use of molecular statics and the fourteen various interatomic potentials for silicon (Tersoff (×3\times3), MEAM (×2\times2), Stillinger-Weber (×2\times2), EDIP, ReaxFF, COMB and machine-learning-based (ML-IAP (×4\times4)), enumerated in § 2.2.1, twelve structural and mechanical properties, i.e., lattice parameters aa, bb, average cohesive energy, average bond length, average height hh, 2D elastic constants CijC_{ij}, 2D Kelvin moduli of the flat silicene (FS) phase are collected in Table 2, of the low-buckled silicene (LBS) in Table 3, of the trigonal dumbbell silicene (TDS) in Table 4, of the honeycomb dumbbell silicene (HDS) in Table 5 and of the large honeycomb dumbbell silicene (LHDS) in Table 6, respectively. The aforementioned results, for each of the five silicene phases, were then compared with those from the DFT calculations using the mean absolute percentage error (MAPE) defined in Equation (1). Let’s briefly analyze the results for each phase, and so for the FS phase the most accurate is the MEAM2011 potential, it has the lowest MAPE, see Table 2, for the LBS phase the Tersoff2107, see Table 3, for the TDS phase the ReaxFF, see Table 4, for the HDS the ReaxFF, see Table 5 and for the LHDS again the ReaxFF potential, see Table 6, respectively. Now let’s take a summary look. Seven of the fourteen potentials analyzed, namely Tersoff2007, Tersoff2017, SW1985, SW2014, ReaxFF, SNAP and ACE, are able to correctly reproduce the structural properties of the five polymorphs of silicene, see Table 3, Table 5, Table 6. Two potentials: ReaxFF and MEAM2011 give the best quantitative performance measured by the total mean absolute percentage error (MAPE), see Table 6. From the point of view of the cost of calculations in terms of relative performance measured as normalized timesteps/second in molecular dynamics (MD) simulation the EDIP and Stillinger-Weber potentials are the fastest, about 5 times faster than the MEAM and Tersoff potentials, about 100 times faster than ReaxFF and COMB, and up to 2000 times faster than the ML-IAP(ACE) potential, see Table 6. It is also worth noting, that the machine-learning-based (ML-IAP) interatomic potentials, according to the methodology used, are not superior to classical potentials in terms of performance (MAPE) and are instead even three orders of magnitude more computationally expensive, see Table 6.

MethodDFTTersoff 1988Tersoff 2007Tersoff 2017MEAM 2007MEAM 2011SW 1985SW 2014EDIPReaxFFCOMBML-IAP SNAPML-IAP qSNAPML-IAP SO(3)ML-IAP ACE
aa3.8554.0084.0194.0424.4573.9604.1043.8864.0183.9503.9904.1214.0194.0513.850
bb3.8554.0084.0194.0424.4573.9604.1043.8864.0183.9503.9904.1214.0194.0513.850
—4.5623.9263.8283.6873.2883.7933.1452.5644.0103.4083.9114.5754.4994.4070.962
2.2252.3152.3212.3332.5732.2882.3692.2432.3212.2822.3062.3812.3212.3402.222
hh0.00.00.00.00.00.00.00.00.00.00.00.00.00.00.0
C11C_{11}84.854.655.247.757.384.058.857.387.474.779.741.524.741.088.9
C22C_{22}84.854.655.247.757.384.058.857.387.474.779.741.524.741.088.9
C12C_{12}40.649.547.452.829.740.834.433.09.236.823.916.217.623.944.8
C33C_{33}22.12.63.9−2.613.821.612.212.239.119.027.912.63.58.622.0
125.4104.0102.6100.587.0124.893.290.396.6111.5103.657.642.364.9133.7
44.35.17.8−5.127.643.324.424.478.237.955.825.37.017.144.1
44.35.17.8−5.127.643.324.424.478.237.955.825.37.017.144.1
MAPEFS_{\mathrm{FS}}36.51534.61945.98728.1743.14726.09326.59532.82710.90415.80533.28448.28335.9469.757

Table 2. Structural and mechanical properties of flat silicene (FS) from molecular calculations: lattice parameters aa, bb (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Kelvin moduli (N/m), mean absolute percentage error (MAPE) (%).

MethodDFTTersoff 1988Tersoff 2007Tersoff 2017MEAM 2007MEAM 2011SW 1985SW 2014EDIPReaxFFCOMBML-IAP SNAPML-IAP qSNAPML-IAP SO(3)ML-IAP ACE
aa3.8283.3093.8203.8704.1503.8373.8403.8124.0183.8433.9903.9163.8083.7433.702
bb3.8283.3093.8203.8704.1503.8373.8403.8124.0183.8433.9903.9163.8083.7433.702
—4.5773.9363.9363.7553.4043.8513.2522.5724.0103.4543.9114.6484.6244.6060.926
2.2492.3152.3122.3152.5342.2972.3512.2432.3212.3002.3062.3902.3662.3472.242
hh0.4211.3040.6900.3270.8200.6080.7840.4270.0‡0.0^{\ddagger}0.6100.0‡0.0^{\ddagger}0.7740.8590.9130.675
C11C_{11}69.20.00659.650.138.447.136.430.887.449.079.724.828.831.755.0
C22C_{22}69.20.00659.650.138.447.136.430.887.449.079.724.828.831.755.0
C12C_{12}22.10.0024.15.54.410.66.15.49.214.123.95.66.39.02.7
C33C_{33}23.60.00227.722.317.018.315.112.739.117.527.99.611.211.426.1
91.30.00863.755.742.757.742.636.296.663.1103.630.435.140.657.7
47.10.00455.544.634.036.530.325.578.234.955.819.222.422.752.3
47.10.00455.544.634.036.530.325.578.234.955.819.222.422.752.3
MAPELBS_{\mathrm{LBS}}79.46722.78119.23537.99123.56937.30635.91136.63022.97719.43345.30643.17642.06028.792

Table 3. Structural and mechanical properties of low-buckled silicene (LBS) from molecular calculations: lattice parameters aa, bb (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Kelvin moduli (N/m), mean absolute percentage error (MAPE) (%).

MethodDFTTersoff 1988Tersoff 2007Tersoff 2017MEAM 2007MEAM 2011SW 1985SW 2014EDIPReaxFFCOMBML-IAP SNAPML-IAP qSNAPML-IAP SO(3)ML-IAP ACE
aa6.4346.4806.4716.4757.1406.5116.6006.2916.7596.3806.5746.7776.7976.5186.448
bb6.4346.4806.4716.4757.1406.5116.6006.2916.7596.3806.5746.7776.7976.5186.448
—4.6794.2483.8653.8903.4273.8653.3222.5914.0753.5513.7234.7264.6934.6070.4606
2.3312.3622.3712.3622.5902.3762.4312.3152.4162.3442.3872.3982.4042.5052.429
hh2.7342.8703.1103.0563.2282.8563.2603.1002.6283.1152.9942.5182.5403.0582.649
C11C_{11}100.579.467.951.469.980.578.768.460.394.670.844.936.735.390.5
C22C_{22}100.579.467.951.469.980.578.768.460.394.670.844.936.735.390.5
C12C_{12}52.363.239.439.034.931.039.635.130.239.725.616.116.710.233.0
C33C_{33}24.18.114.36.217.524.819.616.715.127.422.614.410.012.528.8
152.8142.7107.390.5104.8111.5118.2103.590.5134.396.461.153.445.5123.5
48.316.228.512.435.049.539.133.330.154.945.228.820.125.057.6
48.316.228.512.435.049.539.133.330.154.945.228.820.125.057.6
MAPETDS_{\mathrm{TDS}}23.79022.99334.81523.82111.80917.07723.72325.53710.75516.88131.92938.10037.35719.317

Table 4. Structural and mechanical properties of trigonal dumbbell silicene (TDS) from molecular calculations: lattice parameters aa, bb (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Kelvin moduli (N/m), mean absolute percentage error (MAPE) (%).

Table 5: Structural and mechanical properties of honeycomb dumbbell silicene (HDS) from molecular calculations: lattice parameters aa, bb (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Kelvin moduli (N/m), mean absolute percentage error (MAPE) (%).

MethodDFTTersoff 1988Tersoff 2007Tersoff 2017MEAM 2007MEAM 2011SW 1985SW 2014EDIPReaxFFCOMBML-IAP SNAPML-IAP qSNAPML-IAP SO(3)ML-IAP ACE
aa6.2976.074†6.1336.1145.608†6.364†6.2726.0636.349†6.0486.3446.6446.510†5.030†6.250
bb6.2975.864†6.1336.1146.127†6.279†6.2726.0636.389†6.0486.3446.6446.344†5.818†6.250
−4.6794.3343.7383.8163.6634.0083.2432.4674.1053.3663.4814.7664.6994.8730.391
2.3992.3982.4252.4082.6412.4052.4952.3942.4892.3962.5112.4362.4572.6122.371
hh2.6352.5963.0503.0063.5002.9903.1893.0102.3652.9522.9892.4042.4382.1622.559
C11C_{11}141.689.359.464.056.278.257.435.678.075.414.341.930.876.569.6
C22C_{22}141.646.359.464.057.182.457.435.645.575.414.341.916.073.069.6
C12C_{12}96.420.424.631.317.37.125.314.946.331.4−37.416.213.8−10.210.6
C33C_{33}22.624.817.416.316.726.216.010.49.622.025.812.911.72.729.5
238.0102.184.095.376.687.882.750.6119.7106.851.758.141.284.980.1
45.257.734.832.744.572.932.020.713.643.9−23.1★25.724.364.759.0
45.225.534.832.7−1.2★52.332.020.79.243.951.725.723.45.559.0
MAPEHDS_{\mathrm{HDS}}28.36130.54029.93139.88030.39233.50345.40037.50022.74051.78037.68441.06245.57137.134

Table 5. † Potential does not reproduce the correct symmetry of the structure (a≠ba \ne b), ★ Negative Kelvin moduli indicating a lack of mechanical stability.

Table 6: Structural and mechanical properties of large honeycomb dumbbell silicene (LHDS) from molecular calculations: lattice parameters aa, bb (Å), average cohesive energy (eV/atom), average bond length (Å), average height hh (Å), 2D elastic constants CijC_{ij} (N/m), 2D Kelvin moduli (N/m), mean absolute percentage error (MAPE) (%), relative performance measured as normalized timesteps/second in molecular dynamics (MD) simulation.

MethodDFTTersoff 1988Tersoff 2007Tersoff 2017MEAM 2007MEAM 2011SW 1985SW 2014EDIPReaxFFCOMBML-IAP SNAPML-IAP qSNAPML-IAP SO(3)ML-IAP ACE
aa7.3347.000†7.2497.2367.900†7.3637.4037.0627.7057.1677.4227.6487.7417.4277.393
bb7.3346.978†7.2497.2367.560†7.3637.4037.0627.7057.1677.4227.6487.7417.4277.393
−4.7694.6983.9114.0043.5053.9113.3992.6424.1133.6734.6664.7844.7944.6984.689
2.3572.3812.3872.3692.6272.4072.4562.3452.4382.3702.4542.4362.4032.3852.515
hh2.6832.6923.1093.0503.0503.9113.2503.0102.6373.1122.9942.5382.5622.7002.712
C11C_{11}104.52.378.568.645.688.084.573.053.799.053.144.643.825.959.8
C22C_{22}104.57.378.568.656.288.084.573.053.799.053.144.643.825.959.8
C12C_{12}52.7−13.142.325.522.029.946.138.936.040.431.716.119.10.415.2
C33C_{33}25.99.018.121.614.029.919.217.18.829.310.714.212.312.722.3
157.218.5120.894.173.6117.9130.6111.989.7139.484.860.662.826.475.0
51.817.836.243.128.858.138.434.117.658.621.528.524.726.444.7
51.8−8.5★36.243.127.458.138.434.117.658.621.528.524.725.544.7
MAPELHDS_{\mathrm{LHDS}}55.69318.37320.28734.49913.63616.74123.85533.22810.80933.46633.21934.60340.92829.310
MAPE223.826129.306150.256164.36482.553130.721155.483165.72278.185137.365181.422205.224201.861124.310
timesteps/s387.2382.7355.2505.7416.9904.91753.82032.123.626.47.94.74.2

Table 6. † Potential does not reproduce the correct symmetry of the structure (a≠ba \ne b), ★ Negative Kelvin moduli indicating a lack of mechanical stability.

a^{a}Ref. [48], b^{b} Ref. [7]$.

†^{\dagger}An average bond lengths calculated using radial pair distribution function with a cut-off radius = 3.0 Å and a number of histogram bins = 1000 [32].

Conclusion

A systematic quantitative comparative study of various silicon interatomic potentials for reproducing the properties of five silicene (2D silicon) polymorphs was shown. In order to compare the fourteen potentials listed in § 2.2.1, the structural and mechanical properties of flat (FS), low-buckled (LBS), trigonal dumbbell (TDS), honeycomb dumbbell (HDS) and large honeycomb dumbbell (LHDS) silicene (Figure 1) obtained from the density functional theory (DFT) and molecular statics (MS) computations were used. The computational cost, the performance, of the analyzed potentials were also compared.

It can be stated that considering the performance and the cost of calculations, the classical potentials of Tersoff, SW and MEAM type seem to be the best choice here.

Although data for silicene polytypes were not used in the optimization of these potentials, they were able to reproduce their properties well. There is a consequence that they are based on physics, have a natural extrapolation ability, and not just interpolate data.

I hope that the findings done here will help other researchers in selecting the suitable potentials for their purposes and will be a hint to parameterize new potentials for silicene.

Supporting Information

Crystallographic Information Files (CIF) for polymorphs of silicene.

Supporting Information File 1:

File Name: Si_FS.cif

File Format: CIF

Title: flat silicene (FS)

Supporting Information File 2:

File Name: Si_LBS.cif

File Format: CIF

Title: low-buckled silicene (LBS)

Supporting Information File 3:

File Name: Si_TDS.cif

File Format: CIF

Title: trigonal dumbbell silicene (TDS)

Supporting Information File 4:

File Name: Si_HDS.cif

File Format: CIF

Title: honeycomb dumbbell silicene (HDS)

Supporting Information File 5:

File Name: Si_LHDS.cif

File Format: CIF

Title: large honeycomb dumbbell silicene (LHDS)

# Si_FS.cif -- in P1.1 -- non-magnetic

data_Si_FS_191

_audit_creation_date 222-8-4
_audit_creation_method "Bilbao Crystallographic Server"
_symmetry_Int_Tables_number 1
#_symmetry_space_group_name_H-M "P11’"
_cell_length_a 3.85481
_cell_length_b 3.85481
_cell_length_c 2.
_cell_angle_alpha 9.
_cell_angle_beta 9.
_cell_angle_gamma 12.
loop_
_symmetry_equiv_pos_site_id
_symmetry_equiv_pos_as_xyz
1 x,y,z
loop_
_atom_site_label
_atom_site_type_symbol
_atom_site_fract_x
_atom_site_fract_y
_atom_site_fract_z
_atom_site_occupancy
Si1 Si .33333 .66667 .5 1.
Si2 Si .66666 .33333 .5 1.
# Si_LBS.cif -- in P1.1 -- non-magnetic
data_Si_LBS_164
_audit_creation_date 222-8-4
_audit_creation_method "Bilbao Crystallographic Server"
_symmetry_Int_Tables_number 1
#_symmetry_space_group_name_H-M "P11’"
_cell_length_a 3.82781
_cell_length_b 3.82781
_cell_length_c 2.
_cell_angle_alpha 9.
_cell_angle_beta 9.
_cell_angle_gamma 12.

loop_
_symmetry_equiv_pos_site_id
_symmetry_equiv_pos_as_xyz
1 x,y,z

loop_
_atom_site_label
_atom_site_type_symbol
_atom_site_fract_x
_atom_site_fract_y
_atom_site_fract_z
_atom_site_occupancy
Si1 Si .33333 .66667 .48915 1.
Si2 Si .66666 .33333 .5185 1.

# Si_TDS.cif -- in P1.1 -- non-magnetic
data_Si_TDS_189
_audit_creation_date 222-8-4
_audit_creation_method "Bilbao Crystallographic Server"
_symmetry_Int_Tables_number 1
#_symmetry_space_group_name_H-M "P11’"
_cell_length_a 6.43381
_cell_length_b 6.43381
_cell_length_c 2.
_cell_angle_alpha 9.
_cell_angle_beta 9.
_cell_angle_gamma 12.
loop_
_symmetry_equiv_pos_site_id
_symmetry_equiv_pos_as_xyz
1 x,y,z
loop_
_atom_site_label
_atom_site_type_symbol
_atom_site_fract_x
_atom_site_fract_y
_atom_site_fract_z
_atom_site_occupancy
Si1 Si .33333 .66667 .5 1.
Si2 Si .66666 .33333 .5 1.
Si3 Si .29982 . .5 1.
Si4 Si . .29982 .5 1.
Si5 Si .718 .718 .5 1.
Si6 Si . . .4322 1.
Si7 Si . . .56798 1.

# Si_HDS.cif -- in P1.1 -- non-magnetic

data_Si_HDS_189

_audit_creation_date 222-8-4

_audit_creation_method "Bilbao Crystallographic Server"

_symmetry_Int_Tables_number 1

#_symmetry_space_group_name_H-M "P11’"

_cell_length_a 6.29732

_cell_length_b 6.29732

_cell_length_c 2.

_cell_angle_alpha 9.

_cell_angle_beta 9.

_cell_angle_gamma 12.

loop_

_symmetry_equiv_pos_site_id

_symmetry_equiv_pos_as_xyz

1 x,y,z

loop_

_atom_site_label

_atom_site_type_symbol

_atom_site_fract_x
_atom_site_fract_y
_atom_site_fract_z
_atom_site_occupancy
Si1 Si . . .5 1.
Si2 Si .35934 . .5 1.
Si3 Si . .35934 .5 1.
Si4 Si .6466 .6466 .5 1.
Si5 Si .33333 .66667 .56577 1.
Si6 Si .66666 .33333 .43423 1.
Si7 Si .33334 .66667 .43423 1.
Si8 Si .66667 .33334 .56577 1.

# Si_LHDS.cif -- in P1.1 -- non-magnetic

data_Si_LHDS_191

_audit_creation_date 222-8-4
_audit_creation_method "Bilbao Crystallographic Server"
_symmetry_Int_Tables_number 1
#_symmetry_space_group_name_H-M "P11’"
_cell_length_a 7.33381
_cell_length_b 7.33381
_cell_length_c 2.
_cell_angle_alpha 9.
_cell_angle_beta 9.
_cell_angle_gamma 12.

loop_
_symmetry_equiv_pos_site_id
_symmetry_equiv_pos_as_xyz
1 x,y,z
loop_
_atom_site_label
_atom_site_type_symbol
_atom_site_fract_x
_atom_site_fract_y
_atom_site_fract_z
_atom_site_occupancy
Si1 Si .18192 .36384 .5 1.
Si2 Si .8188 .18192 .5 1.
Si3 Si .63616 .8188 .5 1.
Si4 Si .8188 .63616 .5 1.
Si5 Si .18192 .8188 .5 1.
Si6 Si .36384 .18192 .5 1.
Si7 Si .33333 .66667 .43292 1.
Si8 Si .66666 .33333 .43292 1.
Si9 Si .66666 .33333 .5678 1.
Si1 Si .33333 .66666 .5678 1.

Acknowledgements

Additional assistance was granted through the computing cluster GRAFEN at Biocentrum Ochota, the Interdisciplinary Centre for Mathematical and Computational Modelling of Warsaw University (ICM UW) and Poznań Supercomputing and Networking Center (PSNC).

Funding

This work was supported by the National Science Centre (NCN – Poland) Research Project: No. 2021/43/B/ST8/03207.

References

  1. [1]Akinwande, D.; Brennan, C. J.; Bunch, J. S.; Egberts, P.; Felts, J. R.; Gao, H.; Huang, R.; Kim, J.-S.; Li, T.; Li, Y.; Liechti, K. M.; Lu, N.; Park, H. S.; Reed, E. J.; Wang, P.; Yakob-son, B. I.; Zhang, T.; Zhang, Y.-W.; Zhou, Y.; Zhu, Y. Extreme Mechanics Letters 2017, 13, 42–77. doi:10.1016/j.eml.2017.01.008.DOI
  2. [2]Lei, Y.; Zhang, T.; Lin, Y.-C.; Granzier-Nakajima, T.; Bepete, G.; Kowalczyk, D. A.; Lin, Z.; Zhou, D.; Schranghamer, T. F.; Dodda, A.; Sebastian, A.; Chen, Y.; Liu, Y.; Pour-tois, G.; Kempa, T. J.; Schuler, B.; Edmonds, M. T.; Quek, S. Y.; Wurstbauer, U.; Wu, S. M.; Glavin, N. R.; Das, S.; Dash, S. P.; Redwing, J. M.; Robinson, J. A.; Terrones, M. ACS Nanoscience Au 2022, 2 (6), 450–485. doi:10.1021/acsnanoscienceau.2c00017.DOI
  3. [3]Roome, N. J.; Carey, J. D. ACS Applied Materials & Interfaces 2014, 6 (10), 7743–7750. doi:10.1021/am501022x.DOI
  4. [4]Spencer, M.; Morishita, T. Silicene: Structure, Properties and Applications; Springer Cham, 2016; Vol. 235. doi:10.1007/978-3-319-28344-9.DOI
  5. [5]Hale, L. M.; Trautt, Z. T.; Becker, C. A. Modelling and Simulation in Materials Science and Engineering 2018, 26 (5), 055003. doi:10.1088/1361-651x/aabc05. https://www.ctcms.nist.gov/potentials/system/Si/
  6. [6]Cahangirov, S.; Topsakal, M.; Aktürk, E.; Şahin, H.; Ciraci, S. Phys. Rev. Lett. 2009, 102, 236804. doi:10.1103/PhysRevLett.102.236804.DOI
  7. [7]Cahangirov, S.; Özçelik, V. O.; Xian, L.; Avila, J.; Cho, S.; Asensio, M. C.; Ciraci, S.; Ru-bio, A. Phys. Rev. B 2014, 90, 035448. doi:10.1103/PhysRevB.90.035448.DOI
  8. [8]Lee, C.-C.; Fleurence, A.; Friedlein, R.; Yamada-Takamura, Y.; Ozaki, T. Phys. Rev. B 2013, 88, 165404. doi:10.1103/PhysRevB.88.165404.DOI
  9. [9]Balamane, H.; Halicioglu, T.; Tiller, W. A. Phys. Rev. B 1992, 46, 2250–2279. doi:10.1103/PhysRevB.46.2250.DOI
  10. [10]Lysogorskiy, Y.; Hammerschmidt, T.; Janssen, J.; Neugebauer, J.; Drautz, R. Mod-elling and Simulation in Materials Science and Engineering 2019, 27 (2), 025007. doi:10.1088/1361-651x/aafd13.
  11. [11]Michelin, J. V.; Gonçalves, L. G.; Rino, J. P. Journal of Molecular Liquids 2019, 285, 488–499. doi:/10.1016/j.molliq.2019.04.076.DOI
  12. [12]Cahangirov, S.; Sahin, H.; Le Lay, G.; Rubio, A. Introduction to the Physics of Silicene and other 2D Materials; Springer Cham, 2017; Vol. 930. doi:10.1007/978-3-319-46572-2.DOI
  13. [13]Andrew, R. C.; Mapasha, R. E.; Ukpong, A. M.; Chetty, N. Phys. Rev. B 2012, 85, 125428. doi:10.1103/PhysRevB.85.125428.DOI
  14. [14]Maździarz, M. Materials 2021, 14 (3), . doi:10.3390/ma14030519.DOI
  15. [15]Hohenberg, P.; Kohn, W. Phys. Rev. 1964, 136, B864–B871. doi:10.1103/PhysRev.136.B864.DOI
  16. [16]Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140, A1133–A1138. doi:10.1103/PhysRev.140.A1133.DOI
  17. [17]Gonze, X.; Jollet, F.; Araujo, F. A.; Adams, D.; Amadon, B.; Applencourt, T.; Audouze, C.; Beuken, J.-M.; Bieder, J.; Bokhanchuk, A.; Bousquet, E.; Bruneval, F.; Caliste, D.; Côté, M.; Dahm, F.; Pieve, F. D.; Delaveau, M.; Gennaro, M. D.; Dorado, B.; Espejo, C.; Geneste, G.; Genovese, L.; Gerossier, A.; Giantomassi, M.; Gillet, Y.; Hamann, D.; He, L.; Jomard, G.; Janssen, J. L.; Roux, S. L.; Levitt, A.; Lherbier, A.; Liu, F.; Lukačevič, I.; Martin, A.; Mar-tins, C.; Oliveira, M.; Poncé, S.; Pouillon, Y.; Rangel, T.; Rignanese, G.-M.; Romero, A.; Rousseau, B.; Rubel, O.; Shukri, A.; Stankovski, M.; Torrent, M.; Setten, M. V.; Troeye, B. V.; Verstraete, M.; Waroquiers, D.; Wiktor, J.; Xu, B.; Zhou, A.; Zwanziger, J. Computer Physics Communications 2016, 205, 106–131. doi:10.1016/j.cpc.2016.04.003.DOI
  18. [18]Gonze, X.; Amadon, B.; Antonius, G.; Arnardi, F.; Baguet, L.; Beuken, J.-M.; Bieder, J.; Bottin, F.; Bouchet, J.; Bousquet, E.; Brouwer, N.; Bruneval, F.; Brunin, G.; Cavignac, T.; Charraud, J.-B.; Chen, W.; Côté, M.; Cottenier, S.; Denier, J.; Geneste, G.; Ghosez, P.; Gi-antomassi, M.; Gillet, Y.; Gingras, O.; Hamann, D. R.; Hautier, G.; He, X.; Helbig, N.; Holzwarth, N.; Jia, Y.; Jollet, F.; Lafargue-Dit-Hauret, W.; Lejaeghere, K.; Marques, M. A.; Martin, A.; Martins, C.; Miranda, H. P.; Naccarato, F.; Persson, K.; Petretto, G.; Planes, V.; Pouillon, Y.; Prokhorenko, S.; Ricci, F.; Rignanese, G.-M.; Romero, A. H.; Schmitt, M. M.; Torrent, M.; van Setten, M. J.; Troeye, B. V.; Verstraete, M. J.; Zérah, G.; Zwanziger, J. W. Computer Physics Communications 2020, 248, 107042. doi:10.1016/j.cpc.2019.107042.DOI
  19. [19]Bloch, F. Zeitschrift für Physik 1929, 57, 545–555. doi:10.1007/BF01340281.DOI
  20. [20]Perdew, J. P.; Wang, Y. Phys. Rev. B 1992, 45, 13244–13249. doi:10.1103/PhysRevB.45.13244.DOI
  21. [21]Hamann, D. R. Phys. Rev. B 2013, 88, 085117. doi:10.1103/PhysRevB.88.085117.DOI
  22. [22]Hamann, D. R.; Wu, X.; Rabe, K. M.; Vanderbilt, D. Phys. Rev. B 2005, 71, 035117. doi:10.1103/PhysRevB.71.035117.DOI
  23. [23]Maździarz, M.; Mościcki, T. Materials 2020, 13 (13), 3022. doi:10.3390/ma13133022.DOI
  24. [24]Maździarz, M. 2D Materials 2019, 6 (4), 048001. doi:10.1088/2053-1583/ab2ef3.DOI
  25. [25]Maździarz, M.; Psiuk, R.; Krawczyńska, A.; Lewandowska, M.; Mościcki, T. Archives of Civil and Mechanical Engineering 2022, 22 (4), 193. doi:10.1007/s43452-022-00513-6.DOI
  26. [26]Hinuma, Y.; Pizzi, G.; Kumagai, Y.; Oba, F.; Tanaka, I. Computational Materials Science 2017, 128, 140–184. doi:10.1016/j.commatsci.2016.10.015.DOI
  27. [27]Grimvall, G.; Magyari-Köpe, B.; Ozolinš, V.; Persson, K. A. Rev. Mod. Phys. 2012, 84, 945–986. doi:10.1103/RevModPhys.84.945.DOI
  28. [28]Tadmor, E. B.; Miller, R. E. Modeling Materials: Continuum, Atomistic and Multiscale Tech-niques; Cambridge University Press, 2011; doi:10.1017/CBO9781139003582.DOI
  29. [29]Maździarz, M.; Young, T. D.; Jurczak, G. Archives of Mechanics 2011, 63 (5-6), 533. http://am.ippt.pan.pl/am/article/view/v63p533
  30. [30]Maździarz, M.; Gajewski, M. Computer Modeling in Engineering & Sciences 2015, 105 (2), 123–150. doi:10.3970/cmes.2015.105.123.DOI
  31. [31]Plimpton, S. Journal of Computational Physics 1995, 117 (1), 1 –19. doi:10.1006/jcph.1995.1039.DOI
  32. [32]Stukowski, A. Modelling Simul. Mater. Sci. Eng. 2010, 18, 015012. doi:10.1088/0965-0393/18/1/015012.DOI
  33. [33]Tersoff, J. Phys. Rev. B 1988, 37, 6991–7000. doi:10.1103/PhysRevB.37.6991.DOI
  34. [34]Kumagai, T.; Izumi, S.; Hara, S.; Sakai, S. Computational Materials Science 2007, 39 (2), 457–464. doi:10.1016/j.commatsci.2006.07.013.DOI
  35. [35]Pun, G. P. P.; Mishin, Y. Phys. Rev. B 2017, 95, 224103. doi:10.1103/PhysRevB.95.224103.DOI
  36. [36]Lee, B.-J. Calphad 2007, 31 (1), 95–104. doi:10.1016/j.calphad.2006.10.002.DOI
  37. [37]Du, Y. A.; Lenosky, T. J.; Hennig, R. G.; Goedecker, S.; Wilkins, J. W. physica status solidi (b) 2011, 248 (9), 2050–2055. doi:10.1002/pssb.201147137.DOI
  38. [38]Stillinger, F. H.; Weber, T. A. Phys. Rev. B 1985, 31, 5262–5271. doi:10.1103/PhysRevB.31.5262.DOI
  39. [39]Zhang, X.; Xie, H.; Hu, M.; Bao, H.; Yue, S.; Qin, G.; Su, G. Phys. Rev. B 2014, 89, 054310. doi:10.1103/PhysRevB.89.054310.DOI
  40. [40]Justo, J. a. F.; Bazant, M. Z.; Kaxiras, E.; Bulatov, V. V.; Yip, S. Phys. Rev. B 1998, 58, 2539–2550. doi:10.1103/PhysRevB.58.2539.DOI
  41. [41]Psofogiannakis, G.; van Duin, A. C. Surface Science 2016, 646, 253–260. doi:10.1016/j.susc.2015.08.019. Surface science for heterogeneous catalysis, a special issue in Honour of Richard Lambert
  42. [42]Yu, J.; Sinnott, S. B.; Phillpot, S. R. Phys. Rev. B 2007, 75, 085311. doi:10.1103/PhysRevB.75.085311.DOI
  43. [43]Zuo, Y.; Chen, C.; Li, X.; Deng, Z.; Chen, Y.; Behler, J.; Csányi, G.; Shapeev, A. V.; Thompson, A. P.; Wood, M. A.; Ong, S. P. The Journal of Physical Chemistry A 2020, 124 (4), 731–745. doi:10.1021/acs.jpca.9b08723.DOI
  44. [44]Yanxon, H.; Zagaceta, D.; Wood, B. C.; Zhu, Q. The Journal of Chemical Physics 2020, 153 (5), 054118. doi:10.1063/5.0014677.DOI
  45. [45]Lysogorskiy, Y.; Oord, C. v. d.; Bochkarev, A.; Menon, S.; Rinaldi, M.; Hammerschmidt, T.; Mrovec, M.; Thompson, A.; Csányi, G.; Ortner, C.; Drautz, R. npj Computational Materials 2021, 7 (1), 97. doi:10.1038/s41524-021-00559-9.DOI
  46. [46]Nye, J.; Nye, P. Physical Properties of Crystals: Their Representation by Tensors and Matrices; Oxford science publications; Clarendon Press, 1985.
  47. [47]Maździarz, M.; Gajewski, M. Computer Modeling in Engineering & Sciences 2015, 105 (2), 123–150. doi:10.3970/cmes.2015.105.123.
  48. [48]Yoo, S.; Lee, B.; Kang, K. Nanotechnology 2021, 32 (29), 295702. doi:10.1088/1361-6528/abf26d.DOI

Paper details

vlnc:W_my8xmd2s
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents