Conversion of Monte Carlo Steps to Real Time for Grain Growth Simulation

Abstract

Monte Carlo (MC) technique is becoming a very effective simulation method for prediction and analysis of the grain growth kinetics at mesoscopic level. It should be noted that MC models have no real time of physical systems due to the probabilistic nature of this simulation technique. This leads to difficulties when converting simulated time, the Monte Carlo steps [subscript]tMCS[/subscript] , to real time. The correspondence between Monte Carlo steps and real time should be proposed for comparing the kinetics of MC models with the experiments. In this work, the conversion of Monte Carlo steps to real time is attempted. The lattice sites spacing Δ and the temperature T cannot be ignored in the Monte Carlo simulation of grain growth. Real time will be associated with [subscript]tMCS[/subscript] , T, and Δ.

N. Maazi 1

Academic Editor:Xavier Leoncini

Département de Physique, Faculté des Sciences, Université 20 Août 1955-Skikda, route d'El Hadaiek, BP 26, Skikda, Algeria

Received 1 April 2017; Accepted 23 May 2017; 27 June 2017

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

Grain growth in polycrystals is achieved by decreasing the total number of grains as result of the small grains vanishing. As a grain grows, atoms just outside the boundary change the lattice arrangement from that of the neighboring shrinking grain to the growing grain. To model the microstructure evolution, there are some analytical theories that adequately describe grain growth kinetics [1-4]. Many efforts have been devoted to study real materials in the presence of particles and texture effects by using these theories [5-7]. Grain growth in a polycrystal is driven by the curvature of the grain boundaries. As a result, the grain boundary moves with a velocity that is proportional to its curvature. According to classical grain growth theory [3] based on the assumption that the driving pressure on a grain boundary arises only from its curvature, grain growth kinetics are represented by [figure omitted; refer to PDF] where R- is the mean grain size at time t, _R-0 is the initial mean grain size at t=0, and α is a constant.

In the limit where R>>_R0 , we get [figure omitted; refer to PDF] The simulation from experimental data demonstrates that the grain size increases with time in accordance with a power law ^tn , where n<=0.5 [8, 9].

With the progress of computational technology, significant progress has been made in quantitative understanding of grain growth by using computer simulation techniques, including Monte Carlo (MC) Potts model [10, 11], vertex model [12, 13], cellular automata [14], phase field approaches [15, 16], and molecular dynamics for nanocrystalline [17, 18]. Among these numerical methods, the MC Potts technique has become a very effective simulation method for prediction and analysis of microstructure evolution in polycrystalline materials at mesoscopic level. The MC Potts model is a discrete statistical simulation technique, which does not sample the properties in a deterministic way but stochastically. It treats the evolution of nonequilibrium discrete ensemble, which represents the microstructure. This model is based on the classical works of the Exxon group [10, 11]. The advantage of MC method relies on its simplicity and flexibility to implement it in 2 and 3 dimensions. One of the most serious problems of Monte Carlo simulation for grain growth is that the correspondence between Monte Carlo steps and real time is not well understood. The time progression of the sites position proceeds randomly. So MC procedure suffers from the inability to deal with the physical mechanisms characteristics of grain growth. Several works modeled grain growth by considering a linear relationship between simulated time _tMCS and real time. Ling et al. [19] have proposed the following relation between the real time and _tMCS : [figure omitted; refer to PDF] Other research groups [20, 21] have suggested the relation [figure omitted; refer to PDF] where _k1 is constant, Q is the activation energy, T is the temperature, and R is the gas constant.

The lattice sites spacing Δ plays an important role in the Monte Carlo simulation of grain growth. In the present study, the relation between real time and simulated time _tMCS is derived from the assumption that evolution of the mean matrix area S-, that is, the mean number of sites n- contents in this area, is an invariant between the simulation and the experimental. Therefore, n- was used to establish the relation between the real time and _tMCS . A correlation between the real time, _tMCS , Δ, and T is established.

2. Conversion of Monte Carlo Steps to Real Time

The variation of n- versus the time will be derived from the theory (real time) firstly and from Monte Carlo simulation (_tMCS ) secondly. If the microstructure contains N cells (lattice sites) (Figure 1), the total matrix area _ST is given by [figure omitted; refer to PDF] where _Sc is the cell area: [figure omitted; refer to PDF] Then (5) becomes [figure omitted; refer to PDF] The matrix average areas _S-0 and S-, which contain, respectively, _n-0 and n- sites, verify also relation (7): [figure omitted; refer to PDF] where S- and _S-0 are the matrix average areas at times t and _t0 =0, respectively.

Figure 1: 2-Dimensional hexagonal lattice (cells) used in the calculation of the matrix area.

[figure omitted; refer to PDF]

With the circular grains hypothesis, the matrix average areas are given by [figure omitted; refer to PDF] Then (1) will be [figure omitted; refer to PDF] where t(s) is the real time in seconds.

By substituting (8) into (10), the real time will be [figure omitted; refer to PDF] Equation (11) gives the real time t(s) as function of n- and Δ. On the other hand, n- can be obtained from Monte Carlo simulation. When n- is plotted as a function of the simulated time _tMCS for grain growth simulation without particles consideration, one obtains by linear fitting a line with the equation [22] [figure omitted; refer to PDF] where "a" and "b" are obtained by regression analysis of the data generated from MC simulation.

Equation (11) will be [figure omitted; refer to PDF] Equation (13) gives the real time as function of the simulated time and the lattice sites spacing Δ.

3. Grain Growth Monte Carlo Algorithm

The grain growth Monte Carlo simulation is performed on a two-dimensional triangular lattice, where for each lattice site "i" a number _Si is assigned which corresponds to the grain orientation [10, 11]. The neighboring lattice sites are spaced of a distance Δ. A grain is defined by neighboring lattice sites with the same orientation, while neighboring sites with another orientation belong to a neighboring grain. A grain boundary lies between two adjacent lattice sites with different orientations. The energy of a lattice site "k" is given by [figure omitted; refer to PDF] where δ is the Kronecker delta function with δ(_Sk ,_Sm )=1 if _Sk =_Sm and 0 otherwise and _Jkm is a positive constant that represents the grain boundary (km) energy.

The grain growth MC algorithm is as follows:

(1) Randomly select a lattice site "j" (_Sj ).

(2) Calculate the site energy _Ej (see (14)).

(3) Assign to this site a new orientation (_Si ) among its near neighboring area "i."

(4) Calculate the new site energy _Ei (see (14)).

(5) Calculate the net energy change ΔE due to the reorientation [figure omitted; refer to PDF]

(6) Reorientation is accepted with the transition probability (TP): [figure omitted; refer to PDF]

where T is the simulation temperature and k is a constant.

The number of reorientation attempts N, that is, the number of lattice sites, is defined as one Monte Carlo step (MCS). Starting material used in this study is a real Fe-3%Si microstructure (400×400 µ m² ) obtained by orientation imaging microscopy (^OIMT ) (Figure 2). This microstructure corresponds to a hexagonal grid with Δ=2 µ m and N=46200 sites. The MC simulation is done for the case of isotropic grain boundaries (all energies and mobilities were set to unity) at T=0 K. The parameter α in (1) has been supposed to be equal to 1 for simplification.

Figure 2: Grain structure of the primary matrix analyzed by ^OIMT .

[figure omitted; refer to PDF]

4. Results and Discussion

From MC simulation results, Figure 3 depicts the MC time dependence of the matrix mean number of sites n-. One obtains by linear fitting a line with the equation [figure omitted; refer to PDF] Substituting (17) into (13) gives the relation between real time and _tMCS : [figure omitted; refer to PDF] Figure 4 illustrates the simulated time dependence of the real time for the matrix simulation.

Figure 3: Matrix mean sites number n- variation versus MC steps _tMCS .

[figure omitted; refer to PDF]

Figure 4: Variation of real time versus simulated time _tMCS .

[figure omitted; refer to PDF]

Equation (18) permits us to represent the MC simulation results as function of real time; for example, Figure 5 shows the variation of the matrix mean sites number n- according to the real time. By linear fitting, one obtains [figure omitted; refer to PDF]

Figure 5: Mean sites number n- variation with real time t (min).

[figure omitted; refer to PDF]

4.1. Effect of Lattice Sites Spacing Δ on Real Time

The lattice sites spacing Δ plays an important role in the Monte Carlo simulation of grain growth. To see the influence of Δ on the real time, three cases will be considered for the second simulation: Δ=2 µ m (N=46200 sites and _n-0 =45.43), Δ=1.5 µ m (N=82236 sites and _n-0 =88.05), and Δ=1 µ m (N=184800 sites and _n-0 =190.94). Figure 6 shows the MC time dependence of the matrix mean number of the sites n- for different values of the space step Δ. One obtains by linear fitting a line with the equation [figure omitted; refer to PDF] Substituting (20) into (13) gives a relation between real time and simulated time _tMCS : [figure omitted; refer to PDF] The real time sensitivity to the space step Δ can be seen from plotting the variation of the real time versus _tMCS for different Δ values. At constant _tMCS , the real time increases with increasing Δ as shown in Figure 7. For example, Figure 8 shows the microstructure evolution after _tMCS =220 MCS for three cases: (a) t=3.10 min for Δ = 2 µ m, (b) t=1.92 min for Δ = 1.5 µ m, and (c) t=0.99 min for Δ = 1 µ m. It is clear that the grain growth is faster for the case where Δ = 2 µ m. The movement of grain boundaries controls grain growth process. Based on the equivalence of grain boundary migration and single-site switches in the Potts model, during one Monte Carlo step, the grain boundary between two adjacent grains is displaced over a distance Δ. This can explain why for constant MCS the growth of grains is faster for the case where Δ = 2 µ m than the case where Δ = 1 µ m. From (13), it is obvious that the real time at constant MCS becomes smaller with decreasing Δ.

Figure 6: Variation of n- versus _tMCS for different values of Δ.

[figure omitted; refer to PDF]

Figure 7: Variation of real time versus simulated time _tMCS for different values of Δ.

[figure omitted; refer to PDF]

Figure 8: Influence of Δ on microstructural evolution after _tMCS =220 MCS: (a) t = 3.10 min-Δ = 2 µ m, (b) t = 1.92 min-Δ = 1.5 µ m, and (c) t = 0.99 min-Δ = 1 µ m.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

(c) [figure omitted; refer to PDF]

4.2. Introduction of Temperature in the Monte Carlo Simulation Method

In addition to the problem of the real time conversion, the mechanism of the temperature influence on the calculated results using MC simulation for grain growth is not well understood. The influence of temperature can be introduced usually in the Monte Carlo simulation through the transition probability in reorientation attempts (see (16)). Based on isothermal experimental data and the grain growth regression analysis, Burke and Turnbull [1] deduced the following parabolic law for isothermal grain growth: [figure omitted; refer to PDF] where _k1 is a constant, t is the time, T is the temperature, R is the gas constant, and Q is the activation energy for grain growth. Both _k1 and Q are obtained from experimental data.

While using relation (22) instead of relation (1) in the previous calculations, (13) will be [figure omitted; refer to PDF] Equation (23) gives the real time as function of the simulated time _tMCS , the temperature T, and the lattice sites spacing Δ. The last simulation is done for the case where Δ = 1 µ m and _n-0 =190.94 with the transition probability in reorientation attempts: [figure omitted; refer to PDF] According to (23), the parameters that will be used in the calculation are as follows: Q=1500 cal·K/mol, _k1 =1, and R=2 cal/mol. Influence of the temperature T on real time can be seen from plotting the real time versus _tMCS for different T values. At constant _tMCS , the real time decreases when T increases as shown in Figure 9. Equation (23) permits us to introduce the influence of the temperature in the MC simulation instead of using the transition probability (see (16)). Figures 10 and 11 show the real time dependence of the square matrix mean-radius and the evolution of grain growth after 6 min by using MC simulation for different values of the temperature T.

Figure 9: Variation of the real time versus the simulated time _tMCS for various temperatures.

[figure omitted; refer to PDF]

Figure 10: Square matrix mean-radius variation versus real time as function of temperature.

[figure omitted; refer to PDF]

Figure 11: Simulated grain structure for 6 min at (a) T = 100°C, (b) T = 200°C, and (c) T = 400°C.

(a) [figure omitted; refer to PDF]

(b) [figure omitted; refer to PDF]

(c) [figure omitted; refer to PDF]

5. Conclusion

In addition to the problem of the correspondence between the simulated time _tMCS and the real time, the mechanism of the temperature's influence on the calculated results in the MC simulation for grain growth is not well understood. The lattice sites spacing Δ cannot be ignored in the Monte Carlo simulation. A new equation that gives the real time as function of simulated time _tMCS , the temperature T, and the lattice sites spacing Δ has been derived. The simulation results show that, for modeling grain growth, the relation between real time and _tMCS is achieved linearly. The influence of the temperature can be introduced in the Monte Carlo simulation through the proposed equation instead of using the transition probability.

References

[1] J. E. Burke, D. Turnbull, "Recrystallization and grain growth,", Progress in Metal Physics , vol. 3, no. C, pp. 220-IN15, 1952.

[2] J. von Neumann, C. Herring, Metal Interfaces , pp. 108-110, American Society for Metals, Cleveland, Ohio, USA, 1952.

[3] M. Hillert, "On the theory of normal and abnormal grain growth,", Acta Metallurgica , vol. 13, no. 3, pp. 227-238, 1965.

[4] D. A. Aboav, T. G. Langdon, "The distribution of grain diameters in polycrystalline magnesium oxide,", Metallography , vol. 1, no. 3-4, pp. 333-340, 1969.

[5] A. D. Rollett, W. W. Mullins, "On the growth of abnormal grains,", Scripta Materialia , vol. 36, no. 9, pp. 975-980, 1997.

[6] F. J. Humphreys, "A unified theory of recovery, recrystallization and grain growth, based on the stability and growth of cellular microstructures--I. The basic model,", Acta Materialia , vol. 45, no. 10, pp. 4231-4240, 1997.

[7] N. Maazi, "Modeling of the neighborhood topological effect on grain growth in presence of particles pinning,", Materials Science and Engineering A , vol. 528, no. 15, pp. 4994-5001, 2011.

[8] H. Cetinel, O. Kayacan, D. Ozaydin, "Investigation of nucleation and grain growth in 2-dimensional systems by using generalized Monte Carlo simulations,", Physica A. Statistical Mechanics and its Applications , vol. 392, no. 18, pp. 4121-4126, 2013.

[9] Y. Shi, D. Chen, Y. Lei, X. Li, "HAZ microstructure simulation in welding of a ultra fine grain steel,", Computational Materials Science , vol. 31, no. 3-4, pp. 379-388, 2004.

[10] D. J. Srolovitz, M. P. Anderson, P. S. Sahni, G. S. Grest, "Computer simulation of grain growth-II. Grain size distribution, topology, and local dynamics,", Acta Metallurgica , vol. 32, no. 5, pp. 793-802, 1984.

[11] M. P. Anderson, D. J. Srolovitz, G. S. Grest, P. S. Sahni, "Computer simulation of grain growth-I. Kinetics,", Acta Metallurgica , vol. 32, no. 5, pp. 783-791, 1984.

[12] K. Kawasaki, T. Nagai, K. Nakashima, "Vertex models for two-dimensional grain growth,", Philosophical Magazine B , vol. 60, no. 3, pp. 399-421, 1989.

[13] D. Weygand, Y. Brechet, J. Lepinoux, "A vertex dynamics simulation of grain growth in two dimensions,", Philosophical Magazine B , vol. 78, no. 4, pp. 329-352, 1998.

[14] H. W. Hesselbarth, I. R. Göbel, "Simulation of recrystallization by cellular automata,", Acta Metallurgica Et Materialia , vol. 39, no. 9, pp. 2135-2143, 1991.

[15] D. Fan, L.-Q. Chen, "Computer simulation of grain growth using a continuum field model,", Acta Materialia , vol. 45, no. 2, pp. 611-622, 1997.

[16] A. E. Lobkovsky, J. A. Warren, "Phase-field model of crystal grains,", Journal of Crystal Growth , vol. 225, no. 2-4, pp. 282-288, 2001.

[17] A. Rahman, "Correlations in the motion of atoms in liquid argon,", Physical Review Letters , vol. 136, article 405, 1964.

[18] D. M. Ceperley, "Microscopic simulations in physics,", Reviews of Modern Physics , vol. 71, no. 2, pp. S438-S443, 1999.

[19] S. Ling, M. P. Anderson, G. S. Grest, J. A. Glazier, "Comparison of soap froth and simulation of large-Q potts model,", Materials Science Forum , vol. 94-96, pp. 39-52, 1992.

[20] B. Radhakrishnan, T. Zacharia, "Monte Carlo simulation of grain boundary pinning in the weld heat-affected zone,", Metallurgical and Materials Transactions A , vol. 26, no. 8, pp. 2123-2130, 1995.

[21] J. Gao, R. G. Thompson, "Real time-temperature models for Monte Carlo simulations of normal grain growth,", Acta Materialia , vol. 44, no. 11, pp. 4565-4570, 1996.

[22] Q. Yu, S. K. Esche, "A Monte Carlo algorithm for single phase normal grain growth with improved accuracy and efficiency,", Computational Materials Science , vol. 27, no. 3, pp. 259-270, 2003.