Computational strategies for mapping equilibrium phase diagrams

Advances in Chemical Physics

Volumn 127, Issue , 2004, Pages 1-64

  Bruce, Alastair D ^a   Wilding, Nigel B ^b  

^a UNIVERSITY OF EDINBURGH   (United Kingdom)

^b University of Bath   (United Kingdom)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 78751654267     PISSN: 00652385     EISSN: None     Source Type: Book Series    
DOI: 10.1002/0471466603.ch1     Document Type: Review

References (162)

1. R. Evans, in Fundamentals of Inhomogeneous Fluids, D. Henderson, ed., Marcel Dekker, New York, 1992.
2. C. Caccamo, Phys. Rep. 274, 1 (1996)
3. R. A. Cowley, Adv. Phys. 12, 421 (1963)
4. H. N. V. Temperley, J. S. Rowlinson, and G. S. Rushbrooke, Physics of Simple Liquids, North-Holland, Amsterdam, 1968.
5. N. B. Wilding (unpublished)
6. In the arguments of exponentials we shall generally absorb factors of kT ≡ 1/β into effective energy functions.
7. K. Binder and D. W. Heermann, Monte Carlo Simulation in Statistical Physics, Springer, New York, 1998.
8. D. P. Landau and K. Binder, A Guide to Monte Carlo Simulations in Statistical Physics, Cambridge University Press, New York, 2000.
9. O. Narayan and A. P. Young, Phys. Rev. E 64, 021104 (2001)
10. N. Metropolis, A.W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, J. Chem. Phys. 21, 1087 (1953)
11. It would be more circumspect to say that the partition function cannot in general be naturally or usefully expressed as a canonical average since that average inevitably falls into the category for which canonical sampling is inadequate [13].
12. L. D. Fosdick, Methods Comput. Phys. 1, 245 (1963)
13. 0
14. Throughout this section we use the notation introduced in Eq. (7) in which the phase label is absorbed into an extended macrostate label ℓ. The path defined by the set of macrostates (ℓ) may (and sometimes will) extend from one phase to another.
15. We shall refer to a field l and a macrovariable M as conjugate if the l-dependence of the configurational energy is of the form ℰ({q}, c; δ = ℰ({q}, c)- δ M ({q})
16. It would be more conventional to refer to "densities," which are intensive. But we prefer to deal with their extensive counterparts, which we shall refer to as macrovariables. The preference reflects our focus on simulation studies which are necessarily conducted on systems of some given (finite) size.
17. Y. Iba, Int. J. Mod. Phys. C 12, 623 (2001)
18. C. J. Geyer and E. A. Thompson, J. R. Stat. Soc. B 54, 657 (1992)
19. K. Hukushima, H. Takayama, and K. Nemoto, Int. J. Mod. Phys. C. 3, 337 (1996)
20. E. Marinari, Optimized Monte Carlo Methods, in Advances in Computer Simulation, J. Kertesz and I. Kondor, eds., Springer-Verlag, New York, 1997
21. 2 ≡ α̃ c.
22. G. M. Torrie and J. P. Valleau, Chem. Phys. Lett. 28, 578 (1974)
23. G. M. Torrie and J. P. Valleau, J. Comp. Phys. 23, 187 (1977)
24. A. P. Lyubartsev, A. A. Martsinovski, S. V. Shevkunov, and P. N. Vorontsov-Velyaminov, J. Chem. Phys. 96, 1776 (1992)
25. E. Marinari and G. Parisi, Europhys. Lett. 19, 451 (1992
26. J. P. Valleau, Adv. Chem. Phys., 105 369 (1999)
27. M. Mezei, J. Comp. Phys., 68, 237 (1987)
28. B. A. Berg and T. Neuhaus, Phys. Lett. B 267, 249 (1991)
29. B. A. Berg and T. Neuhaus, Phys. Rev. Lett. 68, 9 (1992)
30. D. Frenkel and A. J. C. Ladd, J. Chem. Phys. 81, 3188 (1984
31. D. Frenkel and B. Smit, Understanding Molecular Simulation, Academic Press, New York, 1996.
32. J. P. Hansen and L. Verlet, Phys. Rev. 184, 151 (1969)
33. D. Frenkel, H. N. W. Lekkerkerker, and A. Stoobants, Nature 332, 822 (1988
34. W. G. Hoover and F. H. Ree J. Chem. Phys. 49, 3609 (1968)
35. H. Ogura, H. Matsuda, T. Ogawa, N. Ogita, and A. Ueda, Prog. Theor. Phys. 58, 419 (1977)
36. S. Pronk and D. Frenkel, J. Chem. Phys. 110, 4589 (1999)
37. P. G. Bolhuis, D. Frenkel, S.-C. Mau, and D. A. Huse, Nature 388, 235 (1997)
38. L. V. Woodcock, Nature 384, 141 (1997)
39. L. V. Woodcock, Nature 388, 236 (1997).
40. For hard spheres the Helmholtz free energy is purely entropic. Thus Table I shows the difference between the entropy per particle of the two phases, [Sfcc - Shcp]/N.
41. Q. Yan and J. J. de Pablo, J. Chem. Phys. 111, 9509 (1999)
42. Equation (10) shows that the logarithm of the probability ratio measures the free-energy difference, which is extensive.
43. In fact, HPT seems to require that all the points chosen for replicas lie suffiently close to coexistence that both phases have detectable signatures.
44. B. A. Berg, U. H. E. Hansmann, and T. Neuhaus, Z. Phys. B 90, 229 (1993)
45. B. Grossmann, M. L. Laursen, T. Trappenberg, and U. J. Wiese, Phys. Lett. B 293, 175 (1992
46. Reference 44 and N. B. Wilding (unpublished)
47. N. B. Wilding, Phys. Rev. E 52, 602 (1995
48. Some of the results from Ref. 44 are included in the LJ phase diagram shown in Fig. 4.
49. Operationally we may regard a phase as "metastable" and assign it a "free energy" (in relation to some other, also metastable) if the conditions for single-phase sampling [see Eq. (21) and accompanying discussion] are met.
50. The configurations unearthed by the ES algorithm used in Lattice Switch studies of hard spheres (Section IV.D.2) have some distinctive features [48].)
51. A. D. Bruce, A. N. Jackson, and G. J. Ackland, Phys. Rev. E 61, 906 (2000
52. See, for example, M. S. S. Challa and J. H. Hetherington, Phys. Rev. A 38, 6324 (1988)
53. When the interphase traverse proceeds by way of such two-phase configurations the free-energy barrier to be surmounted must scale with the area of the interface supporting two phases in equal measure. The probability of the ravine macrostates (and thus the probability of tunneling between the two phases at coexistence) must vanish as e-FI where the free-energy cost of the interface is of order FI ∼ δsLd-1, with δs the interface tension and L ∼ N1/d the edge of the ddimensional simulation cube.
54. See, for example, Ref. 41.
55. The method to be discussed in the next section belongs in this family; so does HPT, but less compellingly because of the limitations we have already touched on in Section IV.B.2.
56. See Ref. 29, p. 176.
57. The traverse might work acceptably well if the dynamics of the interface between the two phases is favorable: systems with martensitic phase transitions may fall into this category: Z. Nishiyama, Martensitic Transformations, Academic Press, New York, 1978. Note also that the special case in which the structural phase transition involves no change of symmetry can be handled within the standard multicanonical framework [55].
58. G. R. Smith and A. D. Bruce, Phys. Rev. E 53, 6530 (1996)
59. A. D. Bruce, N. B. Wilding, and G. J. Ackland, Phys. Rev. Lett. 79, 3002 (1997
60. A. D. Bruce, A. N. Jackson, G. J. Ackland, and N. B. Wilding, Phys. Rev. E 65, 36710 (2002)
61. N. B. Wilding and A. D. Bruce, Phys. Rev. Lett. 85, 5138 (2000
62. In general the partition function Eq. (2) should already incorporate a Jacobean reflecting the )transformation from cartesian coordinates to the generalized coordinates {q}. It may always be absorbed into the configurational energy and we have adopted that convention.
63. The constraint is rarely made explict in MC studies: it is tacitly assumed that the conditions for "single-phase sampling" [Eq. (21) and accompanying discussion] are met.
64. C. H. Bennett, J. Comp. Phys. 22, 245 (1976)
65. R. W. Zwanzig, J. Chem. Phys 22, 1420 (1954
66. A. Rahman and G. Jacucci, Il Nuovo Cimento D 4, 357 (1984)
67. One sees this more clearly when one views this relationship as the result of integrating Eq. (55)
68. But a judicious choice of the representations chosen for the two phases, reflected in the transformation τ [Eq. (44)], can bring one closer to this attractive limit. See Refs. 70 and 71 and further remarks below.
69. We use the term "lattice" rather loosely. The set of vectors {R→ }α identifies the orthodox lattice convoluted with the orthodox basis.
70. This will frequently be true for solid-solid phase behavior too; in the systems studied with Lattice Switch methods to date, the effects have proved rather small.
71. M. C. Moody, J. R. Ray and A. Rahman, J. Chem. Phys. 84, 1795 (1985)
72. A. F. Voter, J. Chem. Phys. 82, 1890 (1985)
73. C. Jarzynski, Phys. Rev. E 65, 046122 (2002)
74. A. R. Acharya, A. D. Bruce, and G. J. Ackland, in preparation.
75. A. Z. Panagiotopoulos, Mol. Phys. 61, 813 (1987)
76. A. Z. Panagiotopoulos, J. Phys. Cond. Matt. 12, R25 (2000)
77. J. P Valleau, J. Chem. Phys. 108, 2962 (1998)
78. K. K. Mon and K. Binder, J. Chem. Phys. 96, 6989 (1992)
79. A. D. Bruce, Phys. Rev. E 55, 2315 (1997)
80. D. Möller and J. Fischer, Mol. Phys. 69, 463 (1990); ibid 75, 1461 (1992)
81. B. Widom, J. Chem. Phys. 39, 2808 (1963)
82. A. Lotfi, J. Vrabec, and J. Fischer, Mol. Phys. 76, 1319 (1992)
83. J. Vrabec and J. Fischer, Mol. Phys. 85, 781 (1995)
84. See, for example, H. Falk, Am. J. Phys. 38, 858 (1970)
85. C. Jarzynski, Phys. Rev. Lett. 78, 2690 (1997)
86. W. P. Reinhardt and J. E. Hunter III, J. Chem. Phys. 97, 1599 (1992)
87. M. A. Miller and W. P. Reinhardt, J. Chem. Phys. 113, 7035 (2000)
88. A. G. Khatchaturyan, Theory of Structural Phase Transformations in Solids, John Wiley & Sons, New York, 1983
89. A continuous space transformation like this can also be implemented in the lattice switch framework described in Section IV.D. By the same token, one can generally devise a continuous version of any LS transformation. Thus, for example, the obvious continuous counterpart of the fcc-hcp LS transformation depicted in Fig. 6 entails a gradual shear of the close-packed planes. Since the LS implementation remains always in the space of one or other of the crystalline structures, one might expect it to prove the safer choice. Where an explicit comparison of the two strategies has been made [87], this was indeed the conclusion.
90. S.-C. Mau and D. A. Huse, Phys. Rev. E 59, 4396 (1999)
91. We use {δ} to denote a set of physically different fields δ 2δ2 . . . . In contrast, [δ] (appearing in our discussion of paths) is shorthand for a set of different values δ(1) δ (1) . . . of some (particular) field δ. We use {M} and [M] in a similar way.
92. It reduces to the familiar form with the macrovariables M1 and M2 chosen as respectively the enthalpy and the volume whose conjugate fields are 1=T and P=T.
93. D. A. Kofke, J. Chem. Phys. 98, 4149(1993)
94. R. Agrawal and D. A. Kofke, Phys. Rev. Lett. 74, 122 (1995)
95. P. G. Bolhuis and D. A. Kofke, Phys. Rev. E 54, 634 (1996)
96. D. A. Kofke and P. G. Bolhuis, Phys. Rev. E 59, 618 (1999)
97. M. Lisal and V. Vacek, Molecular Simulation 18, 75 (1996)
98. F. A. Escobedo and J. J. de Pablo, J. Chem. Phys. 106, 2911 (1997)
99. Recall that knowing the weights that define a multicanonical distribution of some quantity M over some range is equivalent to knowing the true canonical M-distribution throughout the chosen range.
100. For a review of FSS see Finite Size Scaling and Numerical Simulation of Statistical Systems, V. Privman, ed., World Scientific Publishing, Singapore, 1990.
101. C. Borgs and R. Kotecký, J. Stat. Phys. 61, 79 (1990)
102. C. Borgs, R. Koteck, and S. Miracle-Sol, J. Stat. Phys. 62, 529 (1991)
103. C. Borgs and W. Janke, Phys. Rev. Lett. 68, 1738 (1992)
104. As an immediate corollary, the field values that satisfy the equal-areas criterion will identify the coexistence curve of two such ensembles to within exponentially small corrections [95].
105. J. M. Polson, E. Trizac, S. Pronk, and D. Frenkel, J. Chem. Phys. 112, 5339 (2000)
106. J. J. Binney, N. J. Dowrick, A. J. Fisher, and M. E. J. Newman, The Theory of Critical Phenomena: An Introduction to the Renormalization Group, Oxford University Press, New York, 1992.
107. See, for example, A. L. Sengers, R. Hocken, and J. V. Sengers, Phys. Today 30, 42 (1977)
108. A. D. Bruce and N. B.Wilding, Phys. Rev. Lett. 68, 193 (1992);
109. N. B.Wilding and A. D. Bruce, J. Phys: Condens. Matter 4, 3087 (1992)
110. The treatment of scaling-field mixing in Ref. 100 has recently been corrected: M. E. Fisher and G Orkoulas, Phys. Rev. Lett. 85, 696 (2000)
111. G. Orkoulas, M. E. Fisher and C. Ü stün, J. Chem. Phys 113, 7530 (2000)
112. In a number of recent papers the ideas outlined here have been developed to yield a framework in which the liquid-vapor coexistence curve can be tracked with remarkable precision deep into the critical region: E. Luijten, M. E. Fisher and A. Z. Panagiotopoulos, Phys. Rev. Lett. 88, 185701 (2002)
113. Y. C. Kim, M. E. Fisher and E Luitjen, cond-mat/0304032.
114. M. R. Stapleton, D. J. Tildesley, and N. Quirke, J. Chem. Phys. 92, 4456 (1990)
115. T. Kristof and J. Liszi, Mol. Phys. 99, 167 (2001)
116. P. V. Pant and D. N. Theodorou, Macromolecules 28, 7224 (1995)
117. N. B. Wilding and P. Sollich, J. Chem. Phys. 116, 7116 (2002)
118. L. Gomez, A. Dobry, and H. T. Diep, Phys. Rev. B 63, 224103 (2001)
119. I. A. Rudnev, V. A. Kashurnikov, M. E. Gracheva, and O. A. Nikitenko, Physica C 332, 383 (2000)
120. C. H. Bennett and B. J. Alder, J. Chem. Phys. 54, 4796 (1971)
121. S. Pronk and D. Frenkel, J. Phys. Chem. B 105, 6722 (2001)
122. D. Alfe, G. A. De Wijs, G. Kresse, and M. J. Gillan Int. J. Quant. Chem. 77, 871 (2000)
123. Ecclesiastes 1:9. [The Holy Bible, King James Version, American Bible Society, New York, 1999).]
124. G. J. Ackland, J. Phys. Condens. Matter 14, 2975 (2002)
125. F. Wang and D. P. Landau, Phys. Rev. Lett. 86, 2050 (2001)
126. Phys. Rev. E, 64 056101 (2001)
127. F. H. Zong, D. M. Ceperley, Phys. Rev. E 58, 5123 (1998)
128. C. Rickwardt, P. Nielaba, M. H. Muser, and K. Binder, Phys. Rev. B 63, 045204 (2001)
129. M. Dijkstra, R. van Roij, and R. Evans, Phys. Rev. Lett. 82, 117 (1999)
130. D. Frenkel, G. C. A. M. Mooij, B. Smit, J. Phys. Condens. Matter 4, 3053 (1992)
131. S. Consta, N. B. Wilding, D. Frenkel, and Z. Alexandrowicz, J. Chem. Phys. 110, 3220 (1999)
132. M. E. Fisher and A. Z. Panagiotopoulos, Phys. Rev. Lett. 88, 185701 (2002)
133. W. Kob, C. Brangian, T. Stuhn, and R. Yamamoto, in Computer Simulation Studies in Condensed Matter Physics XIII, D. P. Landau, S. P. Lewis, and H. B. Schttler, eds., Springer, Berlin, 2000.
134. M. Kolesik, M. A. Novotny, and P. A. Rikvold, cond-mat 0207405.
135. In this section we refer to E rather than its dimensionless counterpart ℰ ≡ βE since it seems advisable to make the measure of temperature β = 1/kT explicit.
136. J. Lee, Phys. Rev. Lett. 71, 211 (1993)
137. Erratum: Phys. Rev. Lett. 71, 2353 (1993)
138. Different authors have made different choices of notation and adopted different sign conventions at this point. We adopt those of Ref. 122.
139. G. R. Smith and A. D. Bruce, J. Phys. A 28, 6623 (1995)
140. The constant k merely serves to assign a convenient lower (or upper) bound to the weights; the value chosen is absorbed in the normalization of the sampling distribution.
141. B. A. Berg, Int. J. Mod. Phys. C 4, 249 (1993)
142. Fields Inst. Commun. 26, 1 (2000)
143. The WL scheme is terminated when the histogram is " sufficiently" flat and the weightmodification factor f is "sufficiently" close to unity.
144. Q. Yan, R. Faller, and J. J. de Pablo, J. Chem. Phys. 116, 8745 (2002)
145. M. Fitzgerald, R. R. Picard and R. N. Silver, Europhys. Lett 46, 282 (1999)
146. One may count the macrostate-transitions that are actually implemented, as in Ref. 55 or (retaining somewhat more information) store the acceptance probabilities of all attempted transitions, as in Ref. 126b.
147. J. Stat. Phys. 98, 321 (2000)
148. There is a caveat here: The traverse through the interesting region needs to be slow enough to allow a rough local equilibrium in each macrostate to be established; this condition is satisfied relatively poorly at the outset (where the MC trajectory heads rapidly for the equilibrium states); but it )gets better the closer the sampling distribution comes to the multicanonical limit; one requires no more than this.
149. P. M. C. de Oliveira, T. J. P. Penna and H. J. Herrmann, Braz. J. Phys. 26, 677 (1996);
150. P. M. C. de Oliveira, T. J. P. Penna, and H. J. Herrmann, ibid Eur. Phys. J. B 1, 205 (1998)
151. P. M. C. de Oliveira, Eur. Phys. J. B 6, 111 (1998)
152. J. S. Wang and L. W. Lee, Comput. Phys. Commun. 127, 131 (2000)
153. J. S. Wang, T. K. Tay, and R. H. Swendsen, Phys. Rev. Lett. 82, 476 (1999)
154. J. S. Wang and R. H. Swendsen, cond-mat/0104418.
155. We subsume into the definitions of the macrovariables any parameters that do not have the credentials ("linearity") of the fields flg.
156. The unspecified constant in Eq. (B3) is prescribed by normalization.)
157. A. M. Ferrenberg and R. H. Swendsen, Phys. Rev. Lett. 61, 2635 (1989)
158. A. M. Ferrenberg and R. H. Swendsen, Phys. Rev. Lett. 63, 1195 (1989)
159. R. H. Swendsen, Physica A 194, 53 (1993)
160. The name seems to reflect no more than the fact that the resulting estimates of PDFs are as always based on (normalized) histograms accumulated in the sampling process.)
161. The observable Q is then a delta function that picks out prescribed values for the macrovariables in the setfMg.
162. To do so, one must determine a set of weights that moderate the relative contributions of each histogram. There is more than a passing similarity here to the expanded ensemble framework discussed in Section III.C.