The size dependence of the mechanical properties and breaking behavior of metallic nanowires: A statistical description
Yunhong Liu,Jianwei Zhao
DOI: https://doi.org/10.1016/j.commatsci.2010.11.026
IF: 3.572
2011-01-01
Computational Materials Science
Abstract:Molecular dynamics simulations have been performed with an embedded-atom method to study the size dependence of the mechanical properties and the breaking behavior of copper nanowires aligned in the [1 0 0] direction. By statistical analysis with 300 samples for each cross-section, we have found that the average Young’s modulus and yield stress increase with increasing cross-sectional area when its side length exceeds 1.09 nm. However, for the systems smaller than this value, the average Young’s modulus presents an exceptional high value. The final breaking positions for each cross-section follow a Gaussian distribution with a most probable breaking position (MPBP). When the side length is greater than 2.20 nm, the MPBP locates at the center of the nanowire. As the side length is reduced, the MPBP shifts to the two ends, inferring a confinement of boundary on the mechanical shock-wave. Keywords Molecular dynamics Nanowire Nano-breaking Shock-wave 1 Introduction In the past decade, the studies on metallic nanowires are becoming important area because of their unique mechanical [1–3] , electrical [4,5] and optical properties [6] . Since nanowires are potentially applicable in nano-electromechanical systems [7] , electronic circuits [8] , and sensors [9,10] , well understanding the mechanical properties, such as Young’s modulus, yield stress and yield strain, is essential. However, when the wire size is reduced into a nanometer scale, the experimental and theoretical studies all revealed that the properties, such as the electrical properties [11,12] , thermodynamics propeties [13–15] , and the mechanical properties [16–18] , are sensitive to the size. Hasmy and Medina [19] illustrated that the 5 d metals played a reorientation from {1 0 0} to {1 1 1} orientation if the film thickness was less than the critical thickness. The deformation mechanism was found to be dependent on the material size as well. Using molecular dynamics simulations, Schiøtz and Jacobsen [20] found that the flow strength increased as the grain size was reduced from 50 to 15 nm, which accord with the Hall–Petch relationship. Once the grain size was smaller than 10 nm, the reverse Hall–Petch effect emerged. This maximum strength at a grain size of 10–15 nm was because of a shift in microscopic deformation mechanism from dislocation-mediated plasticity to grain boundary sliding. Although great advances have been made in the studies of size effect, many fundamental issues are still unsolved. For example, the final breaking position which is a vital factor related to the material utility is still unpredictable. Experimental measurements [21–24] of the mechanical properties have been carried out for various metallic nanowires. Theoretical simulations, especially, molecular dynamics (MD) simulations, have been also demonstrated to be powerful in this field [25–27] . Compared to the experimental measurements, theoretical simulations provide more structural details during the deformation, enabling us to analyze the mechanism from microscopic viewpoint. Present studies have demonstrated that, when the materials reach nanoscale, the stochastic motion of atoms becomes significant, and therefore, many qualities follow a statistical distribution. These characters are especially pronounced in the metallic or molecular junction conductance [28,29] , the breaking position [30] , and the formation probability of atomic chains [31] . Due to the high performance and low cost, MD simulations provide a feasible solution for analyzing the intrinsic properties of nanowires in a sustainable time with sufficient samples. Among the many atomic models, such as the effective medium theory [32] , tight binding model [33] , the Finnis–Sinclair [34] pair functional formulation and the Sutton–Chen [35] potential, the embedded-atom method (EAM) [36–38] is one of the most successful and versatile approaches because of the sound theoretical foundation – the density-functional theory and the simple analytical expression. To date, EAM has been applied to a variety of material systems, such as liquids, metals and alloys, semiconductors, ceramics, polymers, nanostructures, and composite materials. The structure [39,40] , mechanical properties [41,42] , and thermal properties [41,43] are widely studied. Following our previous work [44] that described the shock-wave propagation in one-dimension, we have further considered the possible dispersion of shock-wave in the two-dimension. Therefore, in this work we have performed the MD simulations on the size dependence of the mechanical properties and the breaking behavior of copper nanowires. Due to the size effect of the cross-section, the statistical most probable breaking position (MPBP) [30] has obvious shift inferring a dispersion of the shock-wave. We have also found an interesting behavior that the Young’s modulus averaged from 300 samples presented a distinct “V” shape with the increase of the nanowire size. This result might expand our understanding of traditional Hall–Petch relation due to the surface effect in such system [2,3,30,42,45,46] . 2 Simulation methods In this study, the initial geometric configurations of nanowires were generated from a rectangular face centered cubic lattice along [1 0 0] crystallographic direction, as shown in Fig. 1 . The nanowire length was kept constant as 16 lattices corresponding to 5.79 nm. The side length of cross-section varies from 0.72 to 4.34 nm. In order to investigate the size effect, free boundary is applied for x , y , and z direction. Since the tensile behavior of a nanowire is sensitive to the initial state, different but enough structural relaxations were carried out to gain the different stable initial states [47] . During the structural relaxations, all the atoms are free in x , y and z directions. Starting from the equilibrium state, the uniaxial stretching was performed along the z direction in opposite directions at constant strain rate of 1.92% ps −1 corresponding to 111 m/s by moving the two terminal atomic planes. Strain is defined as ε = ( l - l 0 ) / l 0 , where l is the current wire length and l 0 is the length right after the preliminary relaxation. The atoms in these regions had the same constant speed in the z direction, but they are free in x – y plane. The interaction of the atoms is described by embedded-atom method (EAM) [37,38] developed by Johnson. It has been proved that the EAM can provide a relevant description of the transition metals of either a face centered cubic or body centered cubic structure. The total energy is given by: (1) E = 1 2 ∑ ij V ( r ij ) + ∑ i F ( ρ i ) (2) ρ i = ∑ i ≠ j φ ( r ij ) where E , V , F , and ρ i are the total energy of the system, the pair potential, the embedded energy, and the electron cloud density, respectively. i and j are the number of the atoms, and r ij is the distance between them. φ ij is the contribution to the electron density at atom i due to atom j at a distance r ij . Concerning the special needs of nano-engineering, we have developed a software, NanoMD [48] , in which the Verlet leapfrog algorithm [49] was used for the integration of motion equations to obtain velocity and trajectories of atoms, with the time step of 1.6 fs which is small enough to ensure the validity. Nosé–Hoover thermostat [50–52] was applied so as to keep the system temperature constantly of 300 K. The reliability of the software as well as the algorithms has been validated not only by a large amount of theoretical simulations [30,31,53,54] , but also with the comparison to the experimental measurements [55] . Average stresses along z direction of all the atoms were recorded every step by virial scheme [56] . (3) σ α zz = 1 Ω α - m α v α z v α z + 1 2 ∑ β ≠ α ∂ ϕ ∂ r α β + ∂ F ∂ ρ α + ∂ F ∂ ρ β ∂ f ∂ r α β r α β z r α β z r α β where σ α zz is the zz component of the atomic stress tensor of atoms α , Ω α is its volume, m is the mass, and v α z is the velocity component in the z direction of atom α , φ , F , ρ , and f are parameters from EAM potential [38] . The first and second terms in the right side of the above equation represent the thermal and atomic interactions respectively. 3 Results and discussion Fig. 2 a illustrates the typical stress–strain responses from the initial state to the final rupture at a constant temperature of 300 K for the copper nanowires with a square cross-section with the side lengths ranging from 0.72 to 4.34 nm. The point at the maximum stress is recognized as the yield point, and the stress and strain at this point are yield stress and yield strain, respectively. For all model nanowires, the stress–strain response is straight before the yield point, in accord with the Hooke’s law. In this region, before reaching the yield point, the nanowire experiences an elastic deformation and the whole crystal structure of the nanowire can keep well. After reaching the yield point, stress drops dramatically for all the copper nanowires, suggesting that the crystal structure experiences irreversible deformation caused by the mechanical stretching. During this process, multiple dislocations are generated and repeat to reduce the stress until the final rupture. In contrast to the thicker samples, the two thinnest ones with side lengths of 0.72 and 1.09 nm exhibit quite unique features. For example, they have more significant stress fluctuation, the steeper stress–strain slopes before the yield point, and much smaller yield strain. The size dependence of the stress fluctuation is directly related to the stochastic atomic motion. When the system is small, the stress averaged on each atom is great. Therefore, any small thermal vibration in the small system appears significant as represented by the stress fluctuation of curves a and b in Fig. 2 a. We draw a semi-logarithmic plot of the mean square error of the stress curves as a function of the cross-section ( Fig. 2 b). The plot involves two variation stages, a stage of precipitous decrease before 2.0 nm, and a stage of plateau hereafter. For smaller systems, due to the fewer nearest neighbor atoms and weaker binding, surface atoms are usually more active than the interior ones. Therefore, an abrupt decrease of the stress fluctuation is observed for these small systems. This effect is not proportional as indicated by the semi-logarithmic axis, corresponding to the nonlinear increase of the surface-to-volume ratio. The obvious size dependence of the mechanical properties of the nanowires can be seen from the yield points. Detailed explanation of this size effect is vital when designing a component in device since the yield point generally represents an upper limit to the load that can be applied. Fig. 2 c shows the yield strain averaged from 300 samples for each cross-section. Except the two smallest systems, all others keep almost constant at 0.11, though large statistical errors exit. From macroscopic mechanism, it is not difficult to understand that the yield strain is independent on the system size. This is also true even in the nanometer scale, while the nanowire is thicker than 1.5 nm in side length. However, for the models with 1.09 and 0.72 nm side length, an abrupt decrease is identified. Their average yield strains are 0.060 and 0.058, respectively, which are almost half of others. Such discontinuity in the yield strain infers the structural change for the smallest systems. We also studied the tension processes of the similar nanowires but along [1 1 1] and [1 1 0] directions [31] . When stretching along [1 1 1] direction, the averaged yield strain is 0.07. When stretching along [1 1 0], the averaged yield strain is reduced further to 0.05–0.06, which is very close to the values for the two thinnest samples. This comparison hints that after relaxation they might experience a crystal grain rotation due to the large surface-to-volume ratio ( see SI ). The yield stress or the yield strength is relevant to the plastic instability of nanowire [57] . Hence, it is of great importance to characterize the yield stress. Fig. 2 d shows the average yield stress as a function of the nanowire size. Similar to the yield strain, the thinnest two nanowires exhibit a unique decrease with increasing the cross-section. However, when the cross-sectional side is wider than 1.5 nm, the yield stress experiences an increase till about 9.0 GPa. In the present study, the dependence of yield stress on cross-section shows a unique “V” shape relation, which looks opposite to the Hall–Petch relation, a widely used theory for interpreting the mechanical properties of polycrystals [58,59] . Due to the grain-boundary strengthening, it is observed in many polycrystalline materials that grain boundaries hinder the dislocation activity. Hall–Petch strengthening is especially obvious, once the grain size reaches about 10 nm [18,60,61] , beyond which the grain boundary sliding weakens the material strength. Therefore, the Hall–Petch relation presents a reversed “V” shape. We noted that all the models we simulated are narrower than 5.0 nm. It seems that it falls into only the left branch of the Hall–Petch relation. Recent molecular dynamics simulations also showed that for copper and nickel with grains smaller than 12 nm the flow stress is reduced as the grain size decreases [62–64] . Due to the limitation of the computation ability, a whole Hall–Petch relation can not be obtained at this stage. Therefore, we could not expect whether the yield stress will decrease while the side length exceeds 10 nm. However, our simulations demonstrated that when the nanowire is thinner than 1.09 nm, the yield stress behaves abnormal increase, which might extend the original description of the Hall–Petch relation. Young’s modulus, as an effective approach to evaluate the mechanical strength of the nanowire, is adopted with the force approach [65] . Fig. 2 e displays a typical stress–strain curve in the linear stage of the system with the cross-section of 2.53 nm × 2.53 nm. In spite of some fluctuations that are caused by the atomic motion and decrease with increasing the system width, the correlation coefficient of the linear fitting is as high as 99%. Fig. 2 f shows the Young’s modulus averaged from 300 samples for each cross-section. The average Young’s modulus increases from 85.2 ± 1.4 to 94.5 ± 0.3 GPa corresponding to the cross-section length from 1.44 to 4.34 nm. For the two thinnest nanowires, the average Young’s modulus displays abnormal high which are 164.0 ± 16.1 and 127.9 ± 9.7 GPa, corresponding to the cross-section length of 0.72 and 1.09 nm. Obviously, the surface effect might be responsible for the observation. From Fig. 2 , we observe an intriguing effect of surface on the mechanical properties of nanowires. For the larger systems, the effect is continuous. We obtain the nearly constant yield strain, a little decreased stress fluctuation, the increased yield stress and Young’s modulus. On the contrary, we observe an abrupt change when the nanowire becomes thinner than 1.09 nm. Therefore, further interpreting the surface effect on the nanowire structure is essential for the nanomaterial mechanics. Fig. 3 a shows the proportion of surface atoms as a function of the side length of the cross-section. For the smallest two systems, the proportion of surface atoms is 65% and 46%, respectively. In the nanowire bulk, each atom is pulled equally in every direction by neighboring atoms, resulting in a net force of zero. At the surface, the atoms are pulled inwards by others deeper inside the nanowire. Therefore, all of the atoms at the surface are subject to an inward force, which cannot be balanced by other atoms inside. The larger the ratio of the surface atoms, the greater the surface tension. Then, we expect that the great surface tension may lead to rearrangement of the atoms even the orientation rotation in nanowires. To see how significant of the surface effect, Fig. 3 b gives the cross-section shapes and relative stress distributions in the x – y plane at equilibrium state. Due to the periodicity, only the atoms in three adjacent lattices in z direction are presented. The stress levels indicated by the color bar distributed in the range of 0–8.0 GPa. As expected, the stress of the outmost atoms is bigger than that of the inner ones. This result accords well with the above analysis that the surface atoms have lower neighbors and higher attractive force. This character is more pronounced for the small cross-sections. From Fig. 3 b, we can also observe that the atoms at the corner collapse inward for all the systems. It should be noted that the system of 1.09 nm × 1.09 nm in cross-section expands clearly by about 1.0 cell in the x – y plane totally. Similar phenomena can be also observed for the system of 0.72 nm × 0.72 nm in cross-section. This result indicates that the smallest two systems are unstable at the initial stretching stage because of the high surface tension. There is an empirical criterion that if a nanowire is longer than its circumference, it is unstable under high surface tension [66] . This is probably the reason why the smallest nanowires expand and possess small yield strain and large Young’s modules. After relaxation, the thinnest two nanowires partially rotated from [1 0 0] to [1 1 0] orientation, then, the new structure becomes more stable ( SI ). We also compared the nanowire length upon relaxation as shown in Fig. 3 c. A large contraction of about 26% occurs for the thinnest two nanowires along the z direction, just matching the cross-sectional expansion. On the contrary, while the nanowire cross-section is bigger than 1.44 nm, the nanowires exhibit slight elongation. The surface atoms play an important role during the stretching process, and would affect the mechanical properties such as the Young’s modulus. Fig. 3 d shows the relationship between the average Young’s modulus and the proportion of surface atoms as a small summary of the size effect. When the nanowire is thicker than 1.44 nm, the crystal structure is robust enough to impede the crystal rotation caused of the surface tension. Then, the average Young’s modulus increases almost linearly with decreasing the proportion of surface atoms. However, as the proportion of surface atoms reaches to 47.1%, the surface tension drives part of the nanowire rotated from [1 0 0] to [1 1 0] direction, making the average Young’s modulus unusually high. The statistical study on the materials breaking is of great significance due to the following reasons. The breaking is an ultimate failure of materials. If the breaking position is predictable, we can strengthen the nanowires near the breaking position to extend the operational lifetime. On the other hand, the nanowire can be a carrier body of mechanical shock-wave [20,44,46,67–69] When the nanowire is thin enough; the wave propagation can be confined by the nanowire boundary. Proving more evidence for the propagation of the mechanical shock-wave is important in the fundamental research. Fig. 4 shows the representative snapshots of copper nanowires with different length of cross-section from 0.72 to 4.34 nm at the final breaking moment. In most cases, the final breaking position occurs near one end of the nanowire when the cross-section is small. Then, the breaking position shifts gradually to the center part of the nanowire with the nanowire length increasing. From the snapshots of Fig. 3 , we can clearly observe the dislocation associated with disordered surface atoms for the thicker nanowires. On the contrary, the dislocations gradually disappear with the thinner nanowires. Especially, for the two thinnest nanowires, the whole structure exhibits amorphous state after deformation. In order to explain and eventually predict the final breaking position, we define it as ANR = ( N small / N total ) × 100% [30] , where N small is atom number in the smaller part after breaking, and N total is total atom number of the nanowire. Fig. 5 a shows the statistical histograms fitted with Gaussian function over 300 independent samples for each size system. The ANR distribution is distinct symmetrical for all the systems. The peak center of the curve is related to the most probable breaking position abbreviated as MPBP, which varies regularly with the increasing nanowire size. Because the statistical breaking position is symmetrical, for example, the most probable breaking position for the system with the side length of 1.09 nm is located at 25% and 75%; only one portion is given and presented in Fig. 5 b, which gives the detailed results of the MPBP as a function of the cross-section for all the systems. The most probable breaking position shifts from 25% to 50% when the length of cross-section side increases from 1.09 to 2.5 nm and the MPBP almost fixes at 50.5% even the cross-section increases to 4.34 nm. These results indicate that the 2.5 nm thickness might be the limitation of the confinement of the mechanical shock-wave. Although many factors may be responsible for the wave confinement, we would like to explain this observation from a viewpoint of surface effect. At the equilibrium state, the interatomic force restricts atoms vibrate in a certain region and the crystalline structure keeps ordered. When the external applied momentum coming from the stretching is introduced, the atoms of the nanowire would have sufficient energy to break the surrounding bonding. Due to the availability of open bond compared to atoms situated within the interior bulk, the binding of surface atoms might break firstly. Then, the dislocations or defects are generated at surface. Güseren et al. [14] have demonstrated that surface premelting for thin nanowires which have improved proportion of surface atoms. Therefore, for the thinner nanowires, the deformation is controlled by the “surface melting” because of the dominated surface atoms, which can up to 65%. For the thicker nanowires, with the increasing cross-section and decreasing surface atoms, the deformation mechanism receives contribution mainly by the deformation, minor part being caused by “surface melting”, resulting in the shift of MPBP. 4 Conclusion In conclusion, we have studied the size and surface atoms influence on the mechanical properties and breaking behavior of copper nanowire by performing the molecular dynamics simulations. When the side length of the cross-section is 1.09 nm or less, the surface effect becomes dominant, which makes the crystal grain reorientation from [1 0 0] to [1 1 0] and, therefore, results in small yield strain and high Young’s modulus. For other bigger nanowires, the size effect is dominant, showing a continuous variation in the mechanical properties. Due to the boundary confinement, the most probable breaking position in the statistical histogram increases with increasing the nanowire cross-section, and reaches to the nanowire center. Acknowledgements This project was supported by the National Natural Science Foundation of China (Nos. 20873063, 20821063, and 51071084) and National Basic Research Program of China (973 Program, Nos. 2007CB936302 and 2010CB732400). Appendix A Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.commatsci.2010.11.026 . Appendix A Supplementary data Supplementary data 1 A) The amplitude-frequency relationships of the [1 0 0] orientation nanowires; and B) The amplitude-frequency relationships of the [1 1 0] orientation nanowires. . References [1] H. Mehrez S. Ciraci Phys. Rev. B 56 1997 12632 [2] P.S. Branício J.P. Rino Phys. Rev. B 62 2000 16950 [3] H. Ikeda Y. Qi T. Çagin K. Samwer W.L. Johnson W.A. Goddard Phys. Rev. Lett. 82 1999 2900 [4] S. Dubois L. Piraux J.M. George K. Ounadjela J.L. Duvail A. Fert Phys. Rev. B 60 1999 477 [5] T. Miyake S. Saito Phys. Rev. B 65 2002 165419 [6] A. Christ T. Zentgraf J. Kuhl S.G. Tikhodeev N.A. Gippius H. Giessen Phys. Rev. B 70 2004 125113 [7] S. Buttgenbach S. Butefisch M. Leester-Schadel A. Wogersien Microsyst. Technol. 7 2001 165 [8] N.A. Melosh A. Boukai F. Diana B. Gerardot A. Badolato P.M. Petroff J.R. Heath Science 300 2003 112 [9] Q. Wan Q.H. Li Y.J. Chen T.H. Wang X.L. He J.P. Li C.L. Lin Appl. Phys. Lett. 84 2004 3654 [10] H.Q. Liu J. Kameoka D.A. Czaplewski H.G. Craighead Nano Lett. 4 2004 671 [11] Q. Huang C.M. Lilley M. Bode R. Divan J. Appl. Phys. 104 2008 023709 [12] C.Y. Nam P. Jaroenapibal D. Tham D.E. Luzzi S. Evoy J.E. Fischer Nano Lett. 6 2006 153 [13] P. Buffat J.P. Borel Phys. Rev. A 13 1976 2287 [14] O. Güseren F. Ercolessi E. Tosatti Phys. Rev. B 51 1995 7377 [15] S.L. Lai J.Y. Guo V. Petrova G. Ramanath L.H. Allen Phys. Rev. Lett. 77 1996 99 [16] Y.J. Wen Y. Zhang Z.Z. Zhu Phys. Rev. B 76 2007 [17] C.Q. Sun Prog. Solid State Chem. 35 2007 1 [18] M.A. Meyers A. Mishra D.J. Benson Prog. Mater. Sci. 51 2006 427 [19] A. Hasmy E. Medina Phys. Rev. Lett. 88 2002 096103 [20] J. Schiøtz K.W. Jacobsen Science 301 2003 1357 [21] T. Kizuka Phys. Rev. B 57 1998 11158 [22] J. Chen L. Lu K. Lu Scr. Mater. 54 2006 1913 [23] C.M. Lieber Mrs Bull. 28 2003 486 [24] U. Landman W.D. Luedtke N.A. Burnham R.J. Colton Science 248 1990 454 [25] H.S. Park J.A. Zimmerman Phys. Rev. B 72 2005 054106 [26] P. Walsh W. Li R.K. Kalia A. Nakano P. Vashishta S. Saini Appl. Phys. Lett. 78 2001 3328 [27] P.A.T. Olsson S. Melin C. Persson Phys. Rev. B 76 2007 [28] B. Xu N.J. Tao Science 301 2003 1221 [29] Pramod Reddy Sung-Yeon Jang R.A. Segalman A. Majumdar Science 301 2007 1568 [30] D. Wang J. Zhao S. Hu X. Yin S. Liang Y. Liu S. Deng Nano Lett. 7 2007 1208 [31] Y. Liu F. Wang J. Zhao L. Jiang M. Kiguchib K. Murakoshib Phys. Chem. Chem. Phys. 11 2009 6514 [32] J.K. Nørskov Phys. Rev. B 26 1982 2875 [33] F. Cleri V. Rosato Phys. Rev. B 48 1993 22 [34] M.W. Finnis J.E. Sinclair Philos. Mag. A-Phys. Condens. Matter Struct. Defect Mech. Prop. 50 1984 45 [35] A.P. Sutton J. Chen Philos. Mag. Lett. 61 1990 139 [36] M.S. Daw M.I. Baskes Phys. Rev. B 29 1984 6443 [37] R.A. Johnson Phys. Rev. B 39 1989 12554 [38] R.A. Johnson Phys. Rev. B 37 1988 3924 [39] H.S. Nam N.M. Hwang B.D. Yu J.K. Yoon Phys. Rev. Lett. 89 2002 275502 [40] Y. Mishin M.J. Mehl D.A. Papaconstantopoulos A.F. Voter J.D. Kress Phys. Rev. B 63 2001 224106 [41] T. Cağin G. Dereli M. Uludog˘an M. Tomak Phys. Rev. B 59 1999 3468 [42] S.J.A. Koh H.P. Lee C. Lu Q.H. Cheng Phys. Rev. B 72 2005 085414 [43] J.B. Sturgeon B.B. Laird Phys. Rev. B 62 2000 14720 [44] Y. Liu, J. Zhao, F. Wang, Phys. Rev. B 80 (2009) (AIP Code: 030935PRB, inpress). [45] Y.-H. Wen Z.-Z. Zhu R.-Z. Zhu Comp. Mater. Sci. 41 2008 553 [46] A.S.J. Koh H.P. Lee Nano Lett. 6 2006 2260 [47] G. Rubio N. Agraït S. Vieira Phys. Rev. Lett. 76 1996 2302 [48] J.W. Zhao X. Yin S. Liang Y.H. Liu D.X. Wang S.Y. Deng J. Hou Chem. Res. Chin. Univ. 24 2008 367 [49] L. Verlet Phys. Rev. 159 1967 98 [50] D.J. Evans B.L. Holian J. Chem. Phys. 83 1985 4069 [51] W.G. Hoover Phys. Rev. A 31 1985 1695 [52] S. Nosé J. Chem. Phys. 81 1984 511 [53] J. Zhao F. Wang L. Jiang X. Yin Y. Liu Acta. Phys.-C 25 2009 0001 [54] L. Jiang Y. Xing J. Zhao Chin. J. In. C 25 2009 176 [55] J. Zhao K. Murakoshi X. Yin M. Kiguchi Y. Guo Nan Wang S. Liang H. Liu J. Phys. Chem. C 112 2008 20088 [56] H.A. Wu Eur. J. Mec. A 25 2006 370 [57] L. Philippe Z. Wang I. Peyrot A.W. Hassel J. Michler Acta Mater. 57 2009 4032 [58] E.O. Hall Proc. Phys. Soc. B64 1951 747 [59] N.J. Petch J. Iron Steel Inst. 174 1953 25 [60] P.G. Sanders J.A. Eastman J.R. Weertman Acta Mater. 45 1997 4019 [61] U. Erb Nanostruct. Mater. 6 1995 533 [62] J. Schiøtz F.D. Di Tolla K.W. Jacobsen Nature 391 1998 561 [63] H. Van Swygenhoven M. Spaczer A. Caro D. Farkas Phys. Rev. B 60 1999 22 [64] H.V. Swygenhoven Science 296 2002 66 [65] H. Rafii-Tabar Phys. Rep. 390 2004 235 [66] T.R. Powers R.E. Goldstein Phys. Rev. Lett. 78 1997 2555 [67] E.J. Reed M. SoljaCic R. Gee J.D. Joannopoulos Phys. Rev. B 75 2007 174302 [68] E.M. Bringa J.U. Cazamias P. Erhart J. Stolken N. Tanushev B.D. Wirth R.E. Rudd M.J. Caturla J. Appl. Phys. 96 2004 3793 [69] E.M. Bringa A. Caro Y. Wang M. Victoria J.M. McNaney B.A. Remington R.F. Smith B.R. Torralva H. Van Swygenhoven Science 309 2005 1838