cond-mat0004011/part1
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%         The following is a LaTeX file        %%% 
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \documentstyle[12pt,epsfig]{article}             %%%
5: %%%%%%%%%%%%%%%%  Definitions  %%%%%%%%%%%%%%%%%%%%%
6: \textheight 225mm                               %%
7: \textwidth 160mm                                %%
8: \headheight 0mm                                 %%
9: \setlength{\oddsidemargin}{0mm}                 %% 
10: \topmargin -1cm                                 %%
11: %\hbadness=5000                                  %%
12: %\vbadness=8000                                  %%
13: %\addtolength{\voffset}{-1in}                    %%
14: %\addtolength{\hoffset}{-0.3in}                  %%
15: %\addtolength{\textheight}{1in}                  %%
16: %\addtolength{\textwidth}{0.6in}                 %%
17: %\newfont{\ffont}{msym10}                        %%
18: \newcommand{\beq}{\begin{equation}}             %%
19: \newcommand{\eeq}{\end{equation}}               %%
20: \newcommand{\bqry}{\begin{eqnarray}}            %%
21: \newcommand{\eqry}{\end{eqnarray}}              %%
22: \newcommand{\bqryn}{\begin{eqnarray*}}          %%
23: \newcommand{\eqryn}{\end{eqnarray*}}            %%
24: \newcommand{\NL}{\nonumber \\}                  %%
25: \newcommand{\preprint}[1]{\begin{table}[t]      %%
26:             \begin{flushright}                  %%
27:             \begin{large}{#1}\end{large}        %%
28:             \end{flushright}                    %%
29:             \end{table}}                        %%
30: \newcommand{\eref}[1]{(\ref{#1})}               %%
31: \newcommand{\DD}[2]{\frac{d^{#2}}{d#1^{#2}}}    %%
32: \newcommand{\PD}[2]                             %%
33:     {\frac{\partial^{#2}}{\partial #1^{#2}}}    %%
34: \newcommand{\cc}{\mbox{\ffont C}}               %%
35: \newcommand{\rr}{\mbox{\ffont R}}               %%
36: \newcommand{\kk}{\mbox{\ffont K}}               %%
37: \newcommand{\zz}{\mbox{\ffont Z}}               %%
38: \newcommand{\ii}{\mbox{\ffont I}}               %%
39: \newcommand{\nn}{\mbox{\ffont N}}               %%
40: \newcommand{\qq}{\mbox{\ffont Q}}               %%
41: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
42: %%%%%%%%%%%% numbering equations by sections %%%%%%%%
43: %% \catcode`\@=11 \@addtoreset{equation}{section}  %%
44: %% \renewcommand{\theequation}                     %%
45: %%          {\arabic{section}.\arabic{equation}}   %%
46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
47: \begin{document}
48: \preprint{LA-UR-99-5914}
49: \title{Melting as a String-Mediated Phase Transition}
50: \author{\\ Leonid Burakovsky\thanks{E-mail: BURAKOV@T5.LANL.GOV}, \
51: Dean L. Preston\thanks{E-mail: DEAN@LANL.GOV}, \
52: and Richard R. Silbar\thanks{E-mail: SILBAR@WHISTLESOFT.COM. Also at 
53: WhistleSoft, Inc., 168 Dos Brazos, Los Alamos, \hspace*{0.5cm} NM 87544, USA}
54:  \\  \\ 
55: %Theoretical Division, MS B283 \\  
56: Los Alamos National Laboratory \\ Los Alamos, NM 87545, USA}
57: \date{ }
58: \maketitle
59: \begin{abstract}
60: We present a theory of the melting of elemental solids as a 
61: dislocation-mediated phase transition. We model dislocations near melt as 
62: non-interacting closed strings on a lattice. In this framework we derive 
63: simple expressions for the melting temperature and latent heat of fusion 
64: that depend on the dislocation density at melt. We use experimental data 
65: for more than half the elements in the Periodic Table to determine the 
66: dislocation density from both relations. Melting temperatures yield a 
67: dislocation density of $(0.61\pm 0.20)\;\!b^{-2},$ in good agreement with 
68: the density obtained from latent heats, $(0.66\pm 0.11)\;\!b^{-2},$ where 
69: $b$ is the length of the smallest perfect-dislocation Burgers vector. 
70: Melting corresponds to the situation where, on average, half of the atoms 
71: are within a dislocation core.
72: \end{abstract}
73: \bigskip
74: {\it Key words:} melting, dislocation, phase transition, lattice 
75: 
76: PACS: 11.10.Lm, 11.27.+d, 61.72.Bb, 64.70.Dv, 64.90.+b 
77: \bigskip
78: 
79: \section{Introduction}
80: 
81: Nearly 50 years ago Shockley \cite{Shock} successfully accounted for the 
82: fluidity of a liquid by assuming a certain concentration of line defects in 
83: the liquid state. Bragg \cite{Bragg} had earlier estimated an upper bound on 
84: the core energy of a dislocation under the assumption that the core atomic 
85: configuration was like that of a liquid. Cotterill and Doyama \cite{CD} later 
86: confirmed the Bragg estimate. These early results implied that the liquid 
87: state is equivalent to a crystal saturated with dislocation cores. It was 
88: first suggested by Mott \cite{Mott} that the melting transition could be 
89: described in terms of dislocations. 
90: 
91: There are now compelling results from molecular dynamics \cite{MD} and 
92: Monte Carlo \cite{MC} simulations that imply that dislocations play a key role 
93: in three-dimensional melting, and moreover there is experimental evidence that 
94: linear defects are in fact generated near the melting transition \cite{Crawf}. 
95: 
96: Mizushima \cite{Miz} and Ookawa \cite{Ook} were the first to formulate a 
97: dislocation theory of melting. They based their theory on the fact that the 
98: self-energy of a dislocation decreases with dislocation density because of 
99: screening. In their theory melting is a first-order transition that occurs 
100: when the free energy of the crystal with a sufficiently high concentration 
101: of thermally-generated dislocations equals the free energy of the 
102: dislocation-free crystal. Their predicted melting temperatures agree with 
103: data for reasonable choices of the core energy. Many dislocation theories 
104: of melting then followed [10$\;\!$-13], most notably the exhaustive 
105: treatment of linear-defect-mediated melting by Kleinert and collaborators 
106: \cite{KleinII}. We refer the reader to several fine reviews of the literature 
107: for additional details and references on dislocation-mediated melting 
108: \cite{Cotterill,reviews}. 
109: 
110: Significant progress in our understanding of melting has been achieved by 
111: Kleinert \cite{KleinII} who pointed out that the melting process cannot proceed
112: through the mediation of dislocations alone. Dislocations are associated with 
113: the discrete translational symmetry of the crystal, so only this symmetry is 
114: lost when dislocations condense. But the rotational order of the solid is also 
115: lost as the solid converts into liquid, and for this to occur the defects 
116: associated with the rotational symmetry of the lattice, namely disclinations, 
117: must come into play. Kleinert assumes that the free energy of dislocations 
118: alone would lead to a second-order phase transition, and in addition shows that
119: a second-order proliferation of disclinations can occur in a background of 
120: dislocations above some critical density. The coupled dislocation-disclination 
121: system could, however, undergo a first-order transition, i.e., melting. 
122: 
123: Although disclinations must participate in the melting process, in this paper 
124: we consider only the dislocation degrees of freedom. Ideally, we would derive 
125: the precise form of the free energy of a dense ensemble of dislocations 
126: interacting via the full Blin potential \cite{KleinII} and subject to specific 
127: configurational constraints (Brownian, self-avoiding, etc.), but this problem 
128: has so far defied solution. Instead we develop an effective theory of melting 
129: based on perfectly screened, non-interacting dislocations. We employ the 
130: widely accepted $-\rho \ln \rho $ form $(\rho $ is dislocation density) for 
131: the self-energy density of dislocations \cite{Miz,YI,KV}, which results in 
132: a first-order phase transition. A dislocation in a dense ensemble of other 
133: dislocations is assumed to be a random loop, i.e., the possible configurations 
134: of a dislocation loop are closed random walks, and short-range steric 
135: interactions are neglected. Thus the partition function is evaluated in 
136: the independent-loop approximation. We obtain two new relations: a simple 
137: expression for the melting temperature (the melting relation) that explicitly 
138: takes into account the crystal structure, and another relation between melting 
139: temperature, latent heat of fusion, and critical density of dislocations. We 
140: carry out a comprehensive comparison of these relations with experimental data 
141: on over half of the Periodic Table. The melting relation is accurate to 17\% 
142: \cite{previous}. Dislocation densities as determined from the melting 
143: temperature and latent heat relations are $\rho =(0.61\pm 0.20)\;\!b^{-2}$ 
144: and $(0.66\pm 0.11)\;\!b^{-2},$ respectively, where $b$ is the length of the 
145: smallest perfect-dislocation Burgers vector. Both relations should also 
146: apply to alloys and compounds. 
147: 
148: In Section 2 we discuss the statistical mechanics of dislocation loops on a 
149: lattice. These results are used to derive the melting relation in Section 3, 
150: the free energy density in Section 4, and the formula for the latent heat of 
151: fusion in Section 5. The values for the critical dislocation density extracted 
152: from both the melting relation and the formula for the latent heat of fusion   
153: are checked in Section 6 with the formula for volume change at melt. Our 
154: concluding remarks appear in Section 7. 
155: 
156: \section{Statistical mechanics of dislocation loops on a lattice}
157: 
158: The energy per unit length, $\sigma ,$ of a dislocation 
159: %(which we will treat as a thick string) 
160: can be very large. However, this energy can always be 
161: compensated at sufficiently high temperatures by the large entropy of 
162: line-like structures, as will be seen in what follows. 
163: 
164: In a Bravais lattice with coordination number $z$ we consider the graph 
165: $\Gamma ,$ the edges of which are all nearest-neighbor links. The set 
166: of $z$ links from any lattice site is identical to the set of shortest 
167: perfect-dislocation Burgers vectors, of length $b.$ We now evaluate the 
168: partition function for a single Brownian loop on $\Gamma .$
169: 
170: The line tension, $\sigma ,$ is assumed to be independent of its length, 
171: $L,$ as is the case for a dislocation in a dense complex network. (In a 
172: dilute network, interactions between distant segments of a dislocation lead 
173: to a logarithmic dependence of $\sigma $ on $L.)$ The number of configurations 
174: of a string of length $L$ is $(z')^{L/b},$ where $z'$ is the number of 
175: possible directions that a line segment can take from a given lattice site. 
176: If backtracking is not allowed, $z'=z-1.$ For a simple cubic lattice in $D$ 
177: dimensions $z=2D.$
178:  
179: Hence, the partition function for a single closed dislocation 
180: (in 3 dimensions) is 
181: \beq
182: Z_1=\sum _Lp(L,V)(z')^{L/b}e^{-\beta \sigma L}=
183: \sum _Lp(L,V)e^{-\beta \sigma _{{\rm eff}}L},\;\;\;\sigma _{{\rm eff}}\equiv 
184: \sigma \left( 1-\frac{T\ln z'}{\sigma b}\right) ,
185: \eeq
186: where $\beta \equiv 1/k_BT,$ $p(L,V)$ is the sum of probabilities over all 
187: lattice sites that a dislocation of length $L$ will close, $V$ is the volume 
188: of the system, and $\sigma _{{\rm eff}}$ is the {\it effective} energy cost 
189: to create unit length of a string at temperature $T.$ 
190: 
191: In order to calculate $p(L,V),$ let $p({\bf r}',{\bf r};L/b)$ be the 
192: probability density for a dislocation of $L/b=n$ steps to start at ${\bf r}$ 
193: and end at ${\bf r}'.$ In the limit $n\rightarrow \infty ,$ $b\rightarrow 0,$ 
194: and $L={\rm const,}$ $p({\bf r}',{\bf r};L/b)$ satisfies the diffusion 
195: equation \cite{Wie} 
196: \beq
197: \frac{\partial p}{\partial n}=\frac{b^2}{z'}\nabla ^2p,
198: \eeq
199: the solution of which is the heat-kernel expansion
200: \beq
201: p({\bf r}',{\bf r};L/b)\equiv p({\bf r}'-{\bf r};L/b)=\sum _{{\bf k}}f_{\bf k}
202: ({\bf r})f^\ast _{\bf k}({\bf r}')e^{-E_{{\bf k}}L/b},
203: \eeq
204: where the $f_{{\bf k}}({\bf r})$ are eigenfunctions of the Laplacian,
205: \beq
206: -\frac{b^2}{z'}\nabla ^2f_{{\bf k}}=E_{{\bf k}}f_{{\bf k}},
207: \eeq
208: which we take as normalized according to
209: \beq
210: \int d^3{\bf r}\;\!|f_{{\bf k}}({\bf r})|^2=1.
211: \eeq
212: When $V\rightarrow \infty ,$ we have
213: \beq
214: f_{{\bf k}}({\bf r})=\frac{1}{\sqrt{V}}\;\!e^{i{\bf k}\cdot {\bf r}},\;\;\;
215: E_{{\bf k}}=\frac{b^2k^2}{z'},
216: \eeq
217: where $0\leq k=|{\bf k}|\leq \infty .$ It then follows from Eq.\ (3), in 
218: which we replace a sum by an integral, $\sum _{\bf k}\rightarrow V/(2\pi )^3
219: \int d^3{\bf k},$ that
220: \beq
221: p({\bf r}'-{\bf r};L/b)=\int \frac{d^3{\bf k}}{(2\pi )^3}\;\!e^{i{\bf k}\cdot 
222: ({\bf r}-{\bf r}')-bLk^2/z'}=\left( \frac{z'}{4\pi bL}\right) ^{3/2}
223: e^{-z'({\bf r}'-{\bf r})^2/4bL}.
224: \eeq
225: The normalization (5) of the eigenfunctions thus imparts unit  
226: normalization to the probability density:
227: \beq
228: \int d^3{\bf u}\;p({\bf u};L/b)=1.
229: \eeq
230: The partition function for a dislocation loop $({\bf r}'={\bf r})$ is 
231: therefore 
232: \beq
233: Z_1=\sum _L\frac{b}{L}\int d^3{\bf r}\;p({\bf 0};L/b)e^{-\beta \sigma _{{
234: \rm eff}}L}=\left( \frac{z'}{4\pi }\right) ^{3/2}\frac{V}{b^3}\sum _L\left( 
235: \frac{L}{b}\right) ^{-5/2}\!e^{-\beta \sigma _{{\rm eff}}L}\equiv \sum _LN(L)
236: e^{-\beta \sigma L},
237: \eeq
238: where the factor $b/L$ removes the overcounting due to the degeneracy in 
239: the number of starting points on the loop. Here, $N(L)$ is the number of 
240: configurations of a loop of length $L.$ The exponent 5/2 becomes $1+D/2$ in 
241: $D$ dimensions \cite{GSW}. 
242: 
243: Real dislocations are not necessarily Brownian loops. In fact, they are 
244: expected to be self-avoiding and/or neighbor-avoiding loops, so they do not 
245: penetrate each other's core. Eq.\ (9) can then be extended to non-Brownian 
246: or open dislocations by means of an effective exponent $q+1\neq 5/2$ and 
247: normalization constant $A(q,z')$ \cite{Copeland}, as follows:
248: \beq
249: Z_1=A(q,z')\;\!\frac{V}{b^3}\sum _L\left( \frac{L}{b}\right) ^{-q-1}\!\!
250: e^{-\beta \sigma _{{\rm eff}}L}.
251: \eeq
252: Here, $q=-1$ for non-interacting (Brownian) open dislocations \cite{Copeland} 
253: and $q\approx 7/4$ for self-avoiding dislocations at low densities in 3 
254: dimensions \cite{Copeland}. In the string literature, the value $q=0$ has also 
255: been quoted. A general argument based on modular invariance \cite{BV} shows 
256: that for non-interacting closed strings $q=0$ for sufficiently high energy 
257: on any compact target space. The same value of $q$ was also obtained in a 
258: discrete model for strings \cite{SS}, and as a static solution to the 
259: string Boltzmann equation \cite{LT}. In principle, $q$ may even be a 
260: function of temperature. Although we may expect $3/2\leq q\leq 7/4$ 
261: \cite{Copeland2}, our main conclusions do not depend on the precise value 
262: of $q.$ The normalization constant, $A(q,z'),$ can be calculated analytically 
263: for Brownian loops in any dimension $(q=D/2),$ analogously to the calculation 
264: of $A(3/2,z')=(z'/4\pi )^{3/2}$ in Eq.\ (9), and numerically in other cases. 
265: 
266: The average length of a loop is 
267: \beq
268: \langle L\rangle =\frac{\sum _LLN(L)e^{-\beta \sigma L}}{\sum _LN(L)e^{-\beta 
269: \sigma L}}=\frac{\xi (T)}{\bar{\xi }(T)}\;\!b, 
270: \eeq
271: where we define 
272: \beq
273: \xi (T)\equiv A(q,z')\sum _{L/b}\left( \frac{L}{b}\right) ^{-q}e^{-\beta 
274: \sigma _{{\rm eff}}L}, 
275: \eeq
276: and 
277: \beq
278: \bar{\xi }(T)\equiv A(q,z')\sum _{L/b}\left( \frac{L}{b}\right) ^{-q-1}
279: e^{-\beta \sigma _{{\rm eff}}L}=\frac{b^3}{V}\;\!Z_1. 
280: \eeq
281: 
282: The grand canonical partition function for an ensemble of non-interacting 
283: indistinguishable loops is given by 
284: \beq
285: Z=Z(T,V,\mu )=\sum _{N=1}^\infty \frac{Z_1^N}{N!}\;\!e^{\mu N/k_BT}=
286: \exp \;\!\{ \exp \left( \frac{\mu }{k_BT}\right) Z_1\}, 
287: \eeq
288: where $\mu $ is the chemical potential. The free energy of the ensemble is
289: \beq
290: F=-k_BT\ln Z=-k_BTe^{\mu /k_BT}Z_1.
291: \eeq
292: The average number of loops in the ensemble is
293: \beq
294: \bar{N}=-\left( \frac{\partial F}{\partial \mu }\right) _{T,V}\!\!\!=
295: e^{\mu /K_BT}Z_1. 
296: \eeq
297: 
298: Since $Z=\sum _{L_i}N(L_i)e^{-\beta (\sigma L_i-\mu )},$ where $N(L_i)$ is 
299: the number of dislocation configurations of total length $L_i,$ the average 
300: total dislocation length in the ensemble is
301: $$\bar{L}=\frac{1}{Z}\;\!\sum _{L_i}L_iN(L_i)e^{-\beta (\sigma L_i-\mu )}=
302: -\frac{\partial \ln Z}{\partial (\beta \sigma )}=-e^{\mu /k_BT}\frac{\partial 
303: Z_1}{\partial (\beta \sigma )}$$
304: \beq
305: =\bar{N}\;\!\frac{\sum _LLN(L)e^{-\beta \sigma L}}{\sum _LN(L)e^{-\beta \sigma
306: L}}=\bar{N}\langle L\rangle ,
307: \eeq
308: i.e., the average total dislocation length is equal to the average number of 
309: loops times the average loop length. 
310: 
311: The dislocation density, $\rho ,$ is the average total length per 
312: unit volume. It then follows from (12),(13) and (17) that  
313: \beq
314: b^2\rho (T)=\frac{\bar{N}\langle L\rangle }{V}\;\!b^2=\frac{\bar{N}}{V}
315: \;\!\frac{\xi (T)}{\bar{\xi }(T)}\;\!b^3=e^{\mu /k_BT}\xi (T). 
316: \eeq
317: 
318: \section{New melting relation}
319: 
320: The effective line tension, $\sigma _{{\rm eff}},$ [see Eq.\ (1)] vanishes 
321: at the critical temperature $k_BT_{cr}\equiv \sigma b/\ln z'.$ Consequently, 
322: dislocations proliferate as $T_{cr}$ is approached from below. At temperatures 
323: above $T_{cr},$ the divergence of $Z_1$ signals the breakdown of the 
324: underlying theory, and the system enters a new phase. Hence, the temperature 
325: $T_{cr}$ corresponds to a phase transition, in which dislocations are 
326: copiously produced in the solid. We therefore equate the melting temperature, 
327: $T_m,$ to $T_{cr}.$ 
328: 
329: The line tension, i.e., the dislocation self-energy per unit length, is 
330: assumed to be that of a dislocation in a complex array, or tangle, of other 
331: dislocations. In that case the stress field of a given dislocation beyond 
332: $\simeq R/2,$ where $R$ is the mean interdislocation separation, is largely 
333: cancelled out by the stress fields of the other dislocations in the complex 
334: array \cite{HL,Sar}. The line tension is then the sum of the core energy 
335: plus the elastic energy inside a cylinder of radius $R/2$ \cite{HL}: 
336: \beq 
337: \sigma =\kappa \frac{Gb^2}{4\pi }\ln \left( \frac{\alpha }{b}\frac{R}{2}
338: \right) =\kappa \frac{Gb^2}{8\pi }\ln \left( \frac{\alpha ^2}{4b^2\rho }
339: \right) . 
340: \eeq 
341: Here, $\kappa $ is 1 for a screw dislocation and $(1-\nu )^{-1}\approx 3/2$ 
342: for an edge dislocation, $\nu $ being the Poisson ratio. Also, $G$ is the 
343: shear modulus, $b$ is the Burgers vector magnitude, and $\alpha $ is 
344: a constant of order unity. In the second half of this equation we have taken 
345: distance $R$ to be approximately equal to $1/\sqrt{\rho },$ where $\rho $ is 
346: the dislocation density defined in Eq.\ (18). An expression of the form (19) 
347: with $R=\rho ^{-1/2}$ for the dislocation self-energy was originally proposed 
348: by Mizushima \cite{Miz}, later put on a sound theoretical basis by Yamamoto 
349: and Izuyama \cite{YI}, and was recently employed by Kierfeld and Vinokur 
350: \cite{KV} to model dislocation-mediated phase transitions of vortex-line 
351: lattices in high-$T_c$ superconductors.  
352: 
353: The constant $\alpha $ accounts for the nonlinear elastic effects in the 
354: dislocation core. Hirth and Lothe \cite{HL} compare dislocation energies in 
355: the Peierls-Nabarro (discrete) and Volterra (continuum) dislocation models and
356: find
357: \beq
358: \frac{1}{\alpha }=\frac{d}{eb}\left( \frac{\sin ^2\beta }{e^\gamma (1-\nu )}+
359: \cos ^2\beta \right) ,
360: \eeq
361: where $\gamma =(1-2\nu )/4(1-\nu )\approx 1/8,$ $d$ is the interplanar 
362: spacing, and $\beta $ is the angle between the Burgers and sense vectors of 
363: the dislocation. In a face-centered cubic (fcc) crystal, the smallest 
364: perfect-dislocation Burgers vectors are $\frac{1}{2}\langle 110\rangle a,$ and 
365: the primary glide planes are $\{111\}$ with $d=a/\sqrt{3},$ where $a$ is the 
366: lattice constant. Experimental evidence (ref. \cite{HL}, Table 9-2, p.\ 275) 
367: suggests that the predominant high-temperature glide system in body-centered 
368: cubic (bcc) lattices is $\{110\},$ which has $d=a/\sqrt{2}.$ The smallest bcc 
369: perfect-dislocation Burgers vectors are $\frac{1}{2}\langle 111\rangle a.$ 
370: Thus, in both cases $d/b=\sqrt{2/3}.$ Averaging over $\beta ,$ 
371: %$(\langle \sin ^2\beta \rangle =\langle \cos ^2\beta \rangle =\frac{1}{2}),$ 
372: we find $\alpha \approx 2.9$ for {\it both} fcc and bcc lattices. Atomistic 
373: calculations of core energies in ionic crystals (ref. \cite{HL}, p.\ 232) 
374: indicate that $\alpha \approx 3.$ In metals, no such calculations have been 
375: performed. We use $\alpha =2.9$ for all elements. 
376: 
377: We have also assumed that no backtracking is allowed for dislocations, 
378: $z'=z-1,$ since each backtracking would result in a divergence in the linear 
379: elastic interaction energy between the overlapping segments. The coordination 
380: numbers for the elements considered in our analysis below are $z=6$ for a 
381: simple cubic (sc) lattice, $z=8$ for bcc and body-centered tetragonal (bct) 
382: lattices, and $z=12$ for fcc, hexagonal close-packed (hcp), and double hcp 
383: (dhcp) lattices. Replacing $$b^3\equiv \lambda v_{WS},$$ where $v_{WS}$ is 
384: the volume of the Wigner-Seitz cell of the crystal lattice and $\lambda $ 
385: is a geometric constant, we finally obtain our formula for the melting 
386: temperature of the elements:
387: \beq
388: T_m=\frac{\lambda Gv_{WS}}{4\pi \delta \ln (z-1)},\;\;\;
389: \delta ^{-1}\equiv \kappa \ln \left( \frac{1.45}{b\sqrt{\rho (T_m)}}\right) .  
390: \eeq   
391: 
392: In ref.\ \cite{previous} we evaluated $Gv_{WS}/4\pi T_m\ln (z-1)$ for 51 
393: elements and found $\delta /\lambda $ to be $1.01\pm 0.17,$ where the error 
394: is the root-mean-square deviation. These $\delta /\lambda $ are summarized in 
395: Fig. 1.
396: 
397: 
398: %\hskip 1.9in    
399: % \epsfysize=3in
400: \begin{center}
401: \vspace{1.5cm} 
402: %\parbox{6in}{abs } 
403: \epsfig{file=delta_Z2_1.eps,width=15cm,angle=0}
404: \end{center}
405: Fig.\ 1. Values of $\delta /\lambda =Gv_{WS}/4\pi T_m\ln (z-1)$ from 
406: experimental data for 51 elements.
407:  \\
408: 
409: We now assume that the dislocation ensemble is dominated by perfect 
410: dislocations with the smallest possible Burgers vectors, since dislocation 
411: energy is proportional to $b^2.$ 
412: %it is energetically preferable for a dislocation with Burgers vector 
413: %${\bf b}$ to split into two dislocations with Burgers vectors ${\bf b}_1$ and 
414: %${\bf b}_2,$ so that ${\bf b}_1+{\bf b}_2={\bf b},$ $b_1^2+b_2^2<b^2,$ and 
415: %thereby reduce its energy.) 
416: For a bcc crystal $b=a\sqrt{3}/2,$ $v_{WS}=a^3/2,$ and $b=a/\sqrt{2},$ $v_{WS}
417: =a^3/4$ for a fcc crystal. Then, $b^3\approx 1.30\;\!v_{WS}$ and $1.41\;\!v_{
418: WS},$ respectively. For a hcp lattice, $\lambda =(4/\sqrt{3})\;\!(c/a)^{-1},$ 
419: so that for an ideal hcp crystal $(c/a=\sqrt{8/3})$ one would have $\lambda =
420: \sqrt{2}.$ As estimated for two hcp metals \cite{PP}, $b^3\approx 1.42\;\!v_{
421: WS}$ for Mg and $1.24\;\!v_{WS}$ for Zn. Hence, we take $\lambda =1.33\pm 0.09
422: \approx 4/3.$ This embraces all of the values quoted above. 
423: %Note that for a simple cubic lattice $\lambda =1.$
424: 
425: In an ensemble of loops there are roughly equal amounts of edge and screw 
426: dislocation in the crystal, so we have $1/\kappa =(1-\nu /2)\pm \nu/2\simeq 
427: 5/6\pm 1/6.$ Therefore, as follows from (21),
428: \beq
429: \ln \left( \frac{2.1}{b^2\rho (T_m)}\right) =\frac{2\;\!(5/6\pm 1/6)}{
430: (1.33\pm 0.09)(1.01\pm 0.17)}=1.24\pm 0.33.
431: \eeq
432: Hence, 
433: \beq
434: \rho (T_m)=(0.61\pm 0.20)\;\!b^{-2}.
435: \eeq 
436: 
437: It follows from Eqs.\ (21)-(23), with $\kappa \lambda =1.6\pm 0.3,$ that 
438: to $\sim 20$\% accuracy 
439: \beq
440: T_m=\frac{Gv_{WS}}{4\pi \ln (z-1)}. 
441: \eeq
442: We regard Eq.\ (24) as a new dislocation melting law. 
443: %It compares with the elemental data better than the well-known Lindemann 
444: %relation (which was derived from a mechanical model of melting).
445: 
446: \section{Free energy of the dislocation ensemble}
447: 
448: To calculate the free energy of a dislocation ensemble, Eq.\ (15), let 
449: us rewrite Eqs.\ (10) and (12), using Eq.\ (19) with $R=1/\sqrt{\rho },$ 
450: and replace the sums (which start with $L=4b,$ the smallest loop length) 
451: by the corresponding integrals:
452: \beq 
453: Z_1=\frac{V}{b^3}\;\!A(q,z')\int _4^\infty dx\;\!x^{-q-1}\left[ \frac{4b^2\rho 
454: }{\alpha ^2}\;\!(z')^{1/c}\right] ^{cx}\!, 
455: \eeq
456: \beq 
457: \xi (T)=A(q,z')\int _4^\infty dx\;\!x^{-q}\left[ \frac{4b^2\rho }{\alpha ^2}
458: \;\!(z')^{1/c}\right] ^{cx}\!,\;\;\;c\equiv \frac{\kappa Gb^3}{8\pi k_BT}.  
459: \eeq   
460: Here, $(4b^2\rho /\alpha ^2)\;\!(z')^{1/c}=\exp \;\!\{-8\pi \sigma 
461: _{{\rm eff}}/\kappa Gb^2\}\leq 1,$ since $\sigma _{{\rm eff}}\geq 0.$ 
462: Integrating Eq.\ (25) by parts we find
463: $$Z_1=\frac{V}{qb^3}\left( \!\frac{\kappa Gb^3}{8\pi k_BT}\ln \left[ \frac{
464: 4b^2\rho }{\alpha ^2}\;\!(z')^{1/c}\right] \!\cdot \xi (T)+\frac{A(q,z')}{4^q}
465: \left[ \frac{4b^2\rho }{\alpha ^2}\;\!(z')^{1/c}\right] ^{4c}\right) $$
466: \beq
467: =\frac{V}{qb^3}\left[ -\frac{\sigma _{{\rm eff}}b}{k_BT}\;\!\xi (T)+\frac{
468: A(q,z')}{4^q}\;\!e^{-4\sigma _{{\rm eff}}b/k_BT}\right] .
469: \eeq
470: Hence,
471: \beq
472: F=-k_BTe^{\mu /k_BT}Z_1=
473: %\frac{V}{qb^3}\left( \sigma _{{\rm eff}}\rho \;\!b^3-\frac{A(q,z')}{4^q}
474: %k_BTe^{\mu /k_BT}\left[ \frac{b^2\rho }{2.1}\;\!(z')^{8\pi k_BT/\kappa 
475: %Gb^3}\right] ^{\kappa Gb^3/2\pi k_BT}\right) \!, 
476: \frac{V}{qb^3}\left( \sigma _{{\rm eff}}\rho \;\!b^3-\frac{A(q,z')e^{
477: \mu /k_BT}}{4^q}k_BTe^{-4\sigma _{{\rm eff}}b/k_BT}\right) ,
478: \eeq
479: where we have replaced $(\kappa Gb^3/8\pi )\;\!\ln(\alpha ^2/4b^2\rho )-k_BT
480: \ln z'$ by $b\sigma _{{\rm eff}},$ in view of (1),(19), and used Eq.\ (18). 
481: 
482: The second term on the right-hand side of Eq.\ (28) takes its largest value at 
483: $T=T_m,$ where $\sigma _{{\rm eff}}=0.$ To estimate its contribution to the 
484: free energy, consider the case of Cu discussed in more detail below. In this 
485: case, to estimate $A(q,z')\exp\{\mu (T_m)/k_BT_m\}/4^q,$ we use Eqs.\ 
486: (12),(18), and replace the sum by an integral: 
487: $$\frac{A(q,z')e^{\mu (T_m)/k_BT_m}}{4^q}=\frac{A(q,z')}{4^q}\;\!\frac{b^2\rho 
488: (T_m)}{\xi (T_m)}=\frac{b^2\rho (T_m)}{4^q\int _4^\infty dx\;\!x^{-q}}=\frac{
489: b^2\rho (T_m)\;\!(q-1)}{4}.$$ 
490: As discussed in Section 2, the value of $q$ may be expected between 
491: 3/2 (Brownian loops) and $\approx 7/4$ (self-avoiding loops). 
492: With $b^2\rho (T_m)$ given in Eq.\ (23), we therefore obtain  
493: $A(q,z')\exp\{\mu (T_m)/k_BT_m\}/4^q=0.095\pm 0.037\approx 0.1.$ 
494: 
495: Hence, the contribution of the second term to $qF/V$ would be $\simeq -0.7$ 
496: meV $\stackrel{\circ }{{\rm A}}$$^{-3}.$ As seen in Fig.\ 2, this contribution 
497: is negligibly small. In fact, the second zero of $F$ for $T=T_m$ and $A(q,z')
498: \exp\{\mu (T_m)/k_BT_m\}/4^q=0.1$ occurs at $b^2\rho =0.61,$ which is within 
499: $\sim 5$\% of the value of 0.64 (the second zero of $F$ at $T=T_m$ with $F/V$ 
500: given in (29)), and within uncertainties in the values of $b^2\rho (T_m)$ in 
501: Eq.\ (23).  
502: 
503: Thus, we have derived the dislocation free energy density, and it is given 
504: approximately by 
505: \beq
506: \frac{qF(\rho )}{V}\simeq \sigma _{{\rm eff}}\;\!\rho =-\left( \frac{\kappa 
507: G}{8\pi }\ln \left( \frac{4b^2\rho  }{\alpha ^2}\right) +\frac{k_BT}{b^3}
508: \ln (z-1)\right) b^2\rho .
509: \eeq
510: This form for the free energy density was previously suggested but not 
511: derived by Cotterill \cite{Cotterill}. It was later put on a firm theoretical 
512: basis by Yamamoto and Izuyama \cite{YI}. It also is a fundamental ingredient 
513: in the recently developed theory of dislocation-mediated phase transitions 
514: of vortex-line lattices in high-$T_c$ superconductors \cite{KV}.  
515: 
516: In Fig.\ 2 we plot $qF(\rho )/V$ from (29) for Cu for three different 
517: temperatures: $T<T_m,$ $T=T_m$ and $T>T_m.$ We take $\kappa =6/5,$ $\alpha =
518: 2.9,$ $G=47.7$ GPa \cite{GS}, $T_m=1356$ K, and $b=2.55$ $\stackrel{\circ }{
519: {\rm A}}$ \cite{previous}. A first-order phase transition, that is melting, 
520: takes place when the second zero of $F(\rho )$ occurs at the critical 
521: dislocation density, $\rho (T_m).$ This is a transition from a perfect 
522: crystalline solid to a highly dislocated {\it solid,} not a liquid. In fact, 
523: our theory describes dislocations, which do not exist in liquids. If a 
524: dislocation is viewed as a disclination dipole \cite{dipole}, the dislocated 
525: solid may in turn undergo a Kosterlitz-Thouless-like transition \cite{KT} to 
526: a phase of free disclinations, i.e., a liquid. This dislocated solid may then 
527: be viewed as the three-dimensional analog of an intermediate hexatic phase, 
528: between a solid and a liquid, in the Halperin-Nelson theory of two-dimensional 
529: melting \cite{HN}. The clarification of this point needs further 
530: investigation, to be undertaken elsewhere. Patashinskii {\it et al.} 
531: \cite{Pat} also identified melting as a transition from a perfect crystalline 
532: solid to a highly dislocated solid, and Nelson and Toner \cite{NT} found 
533: residual bond-orientational order in a three-dimensional solid with an 
534: equilibrium concentration of unbound dislocation loops, which is analogous 
535: to that in the two-dimensional hexatic phase. 
536: 
537: Note that it is not possible to increase the dislocation density progressively
538: from zero to $\rho (T_m)$ at a temperature lower than $T_m$ (e.g., by 
539: deformation) because of the high energy barrier at the maximum of $qF(\rho )
540: /V.$ Hence, the dislocation density, as a function of temperature, is
541: \beq
542: \rho (T)=\left[ 
543: \begin{array}{cc}
544: 0, & T<T_m, \\
545: \rho (T_m), & T=T_m.
546: \end{array}
547: \right.
548: \eeq
549: In fact, it can be shown that (30) is the only physical solution of (12) 
550: written as a ``gap'' equation: 
551: $$e^{-\mu /k_BT}b^2\rho (T)=A(q,z')\sum _{n=4}^\infty \frac{(z')^n}{n^q}\left( 
552: \frac{4b^2\rho (T)}{\alpha ^2}\right) ^{\kappa Gb^3n/8\pi k_BT}\!\!\!.$$ 
553: %In view of (18),(30), $\xi (T)$ behaves similarly to (30), i.e., 
554: %\beq
555: %\xi (T)=\left[ 
556: %\begin{array}{cc}
557: %0, & T<T_m, \\
558: %\xi (T_m), & T=T_m,
559: %\end{array}
560: %\right. \;\;\;\xi (T_m)=A(q,z')\sum _{L/b}(L/b)^{-q}. 
561: %\eeq
562: %
563: 
564: 
565: %\hskip 1.9in    
566: % \epsfysize=3in
567: \begin{center}
568: \vspace{1.5cm} 
569: %\parbox{6in}{abs } 
570: \epsfig{file=FforCu.eps,width=15cm,angle=0}
571: \end{center}
572: \vspace*{-3.7cm}
573: %\hspace*{3.5cm} 
574: Fig.\ 2. $qF(\rho )/V$ for Cu at three different temperatures, in units 
575: of meV $\stackrel{\circ }{{\rm A}}$$^{-3}.$ The vertical line denotes 
576: the critical dislocation density value of $0.64\;\!b^{-2}.$
577:  \\
578: 
579: \section{Latent heat of fusion}
580: 
581: For the ensemble of strings on a lattice considered in Section 2, 
582: the internal energy and pressure are%\footnote{If the fugacity $e^{\mu /k_BT}$ 
583: %is not fixed, the formula for $U$ takes on the form $U=-\partial \ln Z/
584: %\partial \beta +\mu \bar{N},$ and leads, via (16), to the same expression for 
585: %$U$ as in Eq.\ (31).} 
586: \beq 
587: U=-\left( \frac{\partial \ln Z}{\partial \beta }\right) _{e^{\mu /k_BT},V}
588: \!\!\!=-\frac{\bar{N}}{Z_1}\;\!\frac{\partial Z_1}{\partial \beta }=
589: e^{\mu /k_BT}\;\!\frac{V\sigma }{b^2}\;\!\xi (T), 
590: \eeq
591: \beq
592: P=k_BT\left( \frac{\partial \ln Z}{\partial V}\right) _{T,\mu }=k_BT\frac{
593: \bar{N}}{Z_1}\;\!\frac{\partial Z_1}{\partial V}=e^{\mu /k_BT}\;\!\frac{
594: k_BT}{b^3}\;\!\bar{\xi }(T). 
595: \eeq
596: Hence, the enthalpy is 
597: \beq
598: H=U+PV=\frac{V}{b^3}\;\!e^{\mu /k_BT}\left[ \sigma b\;\!\xi (T)+k_BT\;\bar{
599: \xi }(T)\right] .
600: \eeq
601: The latent heat of fusion is the enthalpy difference:
602: \beq
603: L_m\equiv H(T_m)-H(0).
604: \eeq
605: In our case, $H(0)=0,$ which follows directly from (30)-(33) and Eq.\ (18). 
606: Using $\sigma _{{\rm eff}}(T_m)=0$ and the melting condition 
607: $k_BT_m=\sigma b/\ln (z-1),$ we obtain 
608: \beq
609: L_m=\frac{V}{b^3}\;\!e^{\mu (T_m)/k_BT_m}k_BT_m\left[ \ln (z-1)\xi (T_m)+
610: \bar{\xi }(T_m)\right] .
611: \eeq
612: To obtain the latent heat per mole, the quantity tabulated in the literature, 
613: one has to multiply the expression (35) by the ratio of the number of atoms 
614: per mole, $N_A,$ to the total number of atoms in the volume $V,$ which is 
615: equal to $V/v_{WS}.$ Replacing $N_Ak_B$ by the gas constant $R,$ and using 
616: $b^3=\lambda v_{WS},$ we obtain
617: \beq
618: L_m=\frac{e^{\mu (T_m)/k_BT_m}}{\lambda }\;\!\xi (T_m)R\;\!T_m\ln (z-1)
619: \left[ 1+\frac{1}{\ln (z-1)}\;\!\frac{\bar{\xi }(T_m)}{\xi (T_m)}\right] .
620: \eeq
621: 
622: To estimate the ratio $\bar{\xi }(T_m)/\xi (T_m),$ we replace 
623: the sums in Eqs.\ (12),(13) by the corresponding integrals:
624: \beq
625: \frac{\bar{\xi }(T_m)}{\xi (T_m)}=\frac{\int _{L/b=4}^\infty d(L/b)\;\!
626: (L/b)^{-q-1}}{\int _{L/b=4}^\infty d(L/b)\;\!(L/b)^{-q}}=\frac{q-1}{4q}.
627: \eeq
628: With $3/2\leq q\leq 7/4,$ as discussed in Section 2, 
629: $0.083\leq (q-1)/4q\leq 0.107,$ i.e., 
630: \beq
631: \frac{\bar{\xi }(T_m)}{\xi (T_m)}=0.095\pm 0.012.
632: \eeq
633: Therefore, the contribution of the second bracketed term on the right-hand 
634: side of Eq.\ (36) (corresponding to the work contribution to enthalpy) is 
635: $0.04-0.06$ $(6\leq z\leq 12).$ We expect, therefore, that neglecting the 
636: second bracketed term on the right-hand side of Eq.\ (36) will introduce an 
637: error not larger than $\sim 6$\%. Hence, with accuracy of $\sim 94$\% 
638: we have the following formula for the latent heats of the elements:
639: \beq
640: L_m=\frac{1}{\lambda }\;\!b^2\rho (T_m)R\;\!T_m\ln (z-1), 
641: \eeq
642: where we have replaced $e^{\mu (T_m)/k_BT_m}\xi (T_m)$ by $b^2\rho (T_m),$ in 
643: view of (18). The proportionality of latent heat of fusion to the critical 
644: concentration of defects (multiplied by the core energy) has been noted 
645: previously by Cotterill \cite{Cotterill}.
646: 
647: In Fig.\ 3 we plot the values of $b^2\rho (T_m)$ extracted from the 
648: experimental data on latent heats for 75 elements. For this analysis, the 
649: values of both $T_m$ and $L_m$ are mostly taken from \cite{Gschn}. For Be, Hf,
650: Sc, Sr, Y, the lanthanides Dy, Ce, Er, Gd, Ho, La, Nd, Sm, Tb, Yb, and the 
651: actinides Am, Cm, Th, we disregard their high-$T$ bcc phases which exist 
652: only in the very vicinity of melting. (The intermediate hcp$\rightarrow $fcc 
653: phase transition for Yb, dhcp$\rightarrow $ fcc for Am, Ce and La, and 
654: fcc$\rightarrow $hcp for Sr, as well as hcp$\rightarrow $fcc for Co, do not 
655: change coordination number.) The crystal structure chosen for the evaluation 
656: of Ca, Co, Mn, N, Np, O, Sm, Ti, Tl, U and Zr corresponds to the phase from 
657: which melting occurs. The data on both $T_m$ and $L_m$ for H, N, O, Pa and Rn 
658: are taken from \cite{Emsley}. The data on both $T_m$ and $L_m$ for Am and Cm, 
659: and on $L_m$ for Ar, Kr, Ne and Xe are taken from \cite{Tonkov}. The data on 
660: $L_m$ for the lanthanides are taken from \cite{Samsonov}. The following values 
661: of $\lambda $ are used: 1 for sc, 1.3 for bcc, 1.41 for fcc, 1.24 for Zn, 1.42 
662: for Mg, and 1.33 for all other elements. 
663: 
664: 
665: %\hskip 1.9in    
666: % \epsfysize=3in
667: \begin{center}
668: \vspace{2cm} 
669: %\parbox{6in}{abs } 
670: \epsfig{file=latent_Z2.eps,width=15cm,angle=0}
671: \end{center}
672: Fig.\ 3. Critical dislocation density as extracted from the 
673: experimental data on latent heat of fusion for 75 elements.
674:  \\
675: 
676: For all these elements we find
677: \beq
678: \rho (T_m)=(0.66\pm 0.11)\;\!b^{-2}, 
679: \eeq
680: where the error is the root-mean-square deviation. This value is in good 
681: agreement with that obtained from the melting temperatures alone, Eq.\ (23). 
682: 
683: Note that the possible inaccuracy in the value of $\lambda $ for the hcp, 
684: dhcp and bct elements used in this analysis, on the order of $\sim 7$\%, may 
685: slightly increase uncertainty in the value of $\rho (T_m)$ in Eqs.\ (23) and 
686: (40). 
687: 
688: We do not have a reasonable explanation for the anomalously high values of 
689: $\rho (T_m)$ for the noble gases. If the noble gases 
690: are excluded from the analysis, then $\rho (T_m)$ turns out to be 
691: $(0.63\pm 0.06)\;\!b^{-2}$ for the remaining 70 elements. Note also that 
692: for deuterium (D), which is not included in Fig.\ 3, with the data on 
693: $T_m$ and $L_m$ from \cite{Tonkov}, we obtain $b^2\rho (T_m)=0.70.$
694: 
695: The uncertainty-weighted average of the values of $\rho (T_m)$ given in 
696: Eqs.\ (23) and (40) is
697: \beq
698: \rho (T_m)=(0.64\pm 0.14)\;\!b^{-2}, 
699: \eeq
700: which we take as our result for the critical dislocation density at melt. 
701: 
702: \section{Volume change at melt} 
703: 
704: As an independent consistency check on the relations (23) and (40), we 
705: determine the critical dislocation density using the formula \cite{PP,Nin}
706: \beq
707: \varepsilon \equiv \frac{\triangle V}{V}=\frac{\lambda }{2\pi }\;\!\frac{G}{B}
708: \;\!\left( \gamma _G-\frac{1}{3}\right) b^2\rho (T_m),
709: \eeq
710: where $\triangle V$ is the difference between the liquid and solid specific 
711: volumes at melt, $G$ and $B$ are the shear and bulk moduli, respectively, and 
712: $\gamma _G$ is the Gr\"{u}neisen constant. Here, $\varepsilon $ is identified 
713: with the dilation of the lattice as the reaction of the crystal to the sudden 
714: proliferation of dislocations.
715: In Table 1 we show the values of $b^2\rho (T_m)$ calculated for 32 elements 
716: for which we could find zero-pressure data on $\gamma _G$ and $\varepsilon .$ 
717: The experimental values of $\varepsilon $ are mostly taken from ref.\ 
718: \cite{Tonkov}, and those of $G,$ $B$ and $\gamma _G$ from \cite{GS}. For 
719: Ar, Kr, Ne and Xe, the values of $G$ and $B$ are taken from \cite{KL}, 
720: and those of $\gamma _G$ from \cite{Vor}. 
721: %The values of $\lambda /2\pi $ are 0.22 for fcc, 
722: %and 0.21 for all other elements used in this analysis.
723: %
724: 
725: %\begin{table}[t]
726: \begin{center}
727: %\hspace*{-1cm}
728: %\vspace*{-1cm}
729: {\footnotesize
730: \begin{tabular}{|c||c|c|c|c||c||c|}
731: \hline %-----------------------------------------------------------------------
732: element & $B,$ GPa & $G,$ GPa & $\gamma _G$ & $\varepsilon $ & $b^2\rho (T_m)$ 
733: from Eq. (42) & $b^2\rho (T_m)$ from Eq. (39)    \\
734: \hline %-----------------------------------------------------------------------
735:  Ag & 103  & 29.8 & 2.40 & 0.052 & 0.39 & 0.67       \\ 
736:  Al & 76.0 & 26.1 & 2.19 & 0.064 & 0.44 & 0.81       \\ 
737:  Ar & 1.83 & 0.75 & 2.59 & 0.144 & 0.69 & 1.00       \\ 
738:  Au & 173  & 28.0 & 2.99 & 0.055 & 0.57 & 0.65       \\ 
739:  Be & 111  &  151 & 1.11 & 0.115 & 0.51 & 0.63       \\ 
740:  Ca & 16.7 & 7.4  & 1.15 & 0.048 & 0.64 & 0.63       \\ 
741:  Cs & 2.01 & 0.65 & 1.41 & 0.026 & 0.36 & 0.56       \\ 
742:  Cu & 137  & 47.7 & 2.02 & 0.046 & 0.35 & 0.68       \\ 
743:  Eu & 17.0 & 7.53 & 1.39 & 0.048 & 0.50 & 0.67       \\ 
744:  Gd & 37.8 & 21.6 & 0.63 & 0.021 & 0.59 & 0.65       \\ 
745:  Ho & 40.8 & 26.3 & 1.18 & 0.075 & 0.65 & 0.66       \\ 
746:  In & 42.0 & 4.78 & 2.43 & 0.025 & 0.49 & 0.62       \\ 
747:  K  &  3.3 & 0.9  & 1.29 & 0.025 & 0.46 & 0.56       \\ 
748:  Kr & 2.04 & 0.85 & 2.64 & 0.151 & 0.70 & 1.00       \\ 
749:  Li & 12.1 & 3.85 & 0.92 & 0.016 & 0.41 & 0.53       \\ 
750:  Lu & 47.6 & 27.2 & 1.06 & 0.036 & 0.41 & 0.67       \\ 
751:  Na & 6.74 & 1.98 & 1.19 & 0.027 & 0.52 & 0.56       \\ 
752:  Nb & 171  & 37.6 & 1.77 & 0.029 & 0.44 & 0.59       \\ 
753:  Nd & 32.9 & 17.4 & 0.57 & 0.009 & 0.34 & 0.56       \\ 
754:  Ne & 0.88 & 0.40 & 2.79 & 0.156 & 0.62 & 0.98       \\ 
755:  Ni & 183  & 85.8 & 1.93 & 0.063 & 0.37 & 0.72       \\ 
756:  Pb & 44.7 & 8.6  & 2.74 & 0.037 & 0.36 & 0.56       \\ 
757:  Pd & 193  & 48.0 & 2.56 & 0.059 & 0.47 & 0.66       \\ 
758:  Pt & 283  & 63.7 & 2.87 & 0.066 & 0.51 & 0.68       \\ 
759:  Rb & 2.3  & 0.63 & 0.99 & 0.026 & 0.70 & 0.60       \\ 
760:  Ta & 193  & 69.0 & 1.74 & 0.052 & 0.50 & 0.59       \\ 
761:  Tb & 38.7 & 22.1 & 0.74 & 0.032 & 0.65 & 0.67       \\ 
762:  Tl & 35.7 & 5.4  & 2.10 & 0.033 & 0.60 & 0.60       \\ 
763:  Tm & 46.2 & 29.1 & 1.43 & 0.069 & 0.47 & 0.68       \\ 
764:   W & 310  & 160  & 1.67 & 0.090 & 0.63 & 0.77       \\ 
765:  Xe & 2.1  & 1.0  & 2.56 & 0.130 & 0.54 & 1.01       \\ 
766:  Yb & 14.9 & 8.06 & 1.04 & 0.036 & 0.44 & 0.56       \\ 
767: \hline %-----------------------------------------------------------------------
768: \end{tabular}
769: }
770: %\caption{ {\it Tentatively:} All the elements for which the values of 
771: %$Gv/(T_m\ln (z-1))$ are concentrated around 20.}
772: \end{center}
773: %\end{table} 
774: \vspace*{0.1cm}
775: Table 1. Values of $b^2\rho (T_m)$ from experimental data on volume change at 
776: melt for 32 elements. For comparison, we also show values of $b^2\rho (T_m)$ 
777: extracted for the same elements from the data on latent heats.
778:  \\
779:  
780: For all 32 elements in Table 1 we find 
781: \beq
782: \rho (T_m)=(0.51\pm 0.11)\;\!b^{-2},
783: \eeq
784: where the error is the root-mean-square deviation. This is somewhat lower 
785: than but still in agreement with both Eqs.\ (23) and (40) taking into account 
786: uncertainties associated with the three values. 
787: 
788: For comparison, we show in the last column of Table 1 the values of $b^2\rho (
789: T_m)$ extracted for the same elements from the data on $L_m.$ It is seen that 
790: the agreement between two sets of the values of $b^2\rho (T_m)$ is reasonably 
791: good, except for Ag, Al, Cs, Cu, Ni, Pb and Xe, for which the difference in 
792: both values of $b^2\rho (T_m)$ is on the order of $\sim 50-60$\%, Lu, Ne and 
793: Nd for which the difference is $\sim 45$\%, and Ar, Kr, Pd and Tm, for which 
794: it is $\sim 35$\%. For all other elements, the difference does not exceed 
795: $\sim 30$\%. 
796: %Most probably, the source of disagreement is due to inaccuracies 
797: %in experimental data, especially those on $\gamma _G.$ 
798: 
799: Note that the contribution of the volume change at melt, $\varepsilon ,$ 
800: to the latent heat of fusion is proportional to $\varepsilon ^2\ll 1$ 
801: \cite{PP,Nin}, and is therefore negligibly small compared to the right-hand 
802: side of Eq.\ (39). 
803: 
804: \section{Concluding remarks}
805: 
806: Our theory of dislocation-mediated melting was developed in the approximation 
807: that dislocations are non-interacting. This approximation is good only in the 
808: vicinity of melt where the dislocation density is very high and the otherwise 
809: long-range interactions are sufficiently screened \cite{Sar}. The statistical 
810: mechanics of non-interacting dislocations on a lattice yields simple, accurate 
811: relations between the dislocation density at melt and both the melting 
812: temperature and latent heat of fusion, despite the indeterminacy of the 
813: parameter $q$ that takes into account the possible non-Brownian nature of 
814: the dislocation network. The values of $\rho (T_m),$ as determined from an 
815: extensive analysis of $T_m$ and $L_m$ data, are remarkably consistent: 
816: $(0.61\pm 0.20)\;\!b^{-2}$ and $(0.66\pm 0.11)\;\!b^{-2},$ respectively. The 
817: uncertainty-weighted average of these values is $\rho (T_m)=(0.64\pm 0.14)
818: \;\!b^{-2},$ which we take as our result for the dislocation density at melt. 
819: Poirier and Price \cite{PP} analyzed 14 elements and found $\rho (T_m)v_{
820: WS}/b=0.48\pm 0.12.$ Using $v_{WS}=b^3/\lambda $ with $\lambda \approx 4/3,$ 
821: their result corresponds to $\rho (T_m)=(0.64\pm 0.16)\;\!b^{-2},$ which is 
822: in excellent agreement with ours. Kierfeld and Vinokur \cite{KV} modelled  
823: dislocation-mediated phase transitions of a vortex-line lattice and found 
824: $\rho (T_m)\approx 0.6\;\!b^{-2}.$ Vachaspati's \cite{Vach} study of 
825: topological defect formation gave $a^2\rho (T_m) \approx 0.88$ for a simple 
826: cubic lattice. This translates into $\rho (T_m)\approx 0.66\;\!b^{-2}$ for 
827: bcc lattices $(a=2/\sqrt{3}\;b)$ and $\rho (T_m)\approx 0.44\;\!b^{-2}$ for 
828: fcc lattices $(a=\sqrt{2}\;\!b),$ which are consistent with our result. In 
829: agreement with Vachaspati, Kibble \cite{Kibble} found $a^2\rho (T_m)\approx 
830: 0.89$ for a simple cubic lattice. 
831: 
832: Although our main results do not depend on the precise value of $q,$ there is 
833: a particular value of $q$ at which the relations (29) and (39) become exact: 
834: $q=1.$ In this limit, as seen in (12), $\xi (T_m)\rightarrow \infty ,$ so that 
835: Eq.\ (39) becomes exact in view of (36). Requiring finite internal energy in 
836: this limit leads, via (31), to $\exp\{\mu (T_m)/k_BT_m\}\rightarrow 0$ 
837: $(\mu (T_m)\rightarrow -\infty ),$ and therefore, Eq.\ (29) becomes exact, 
838: since the second term on the right-hand side of (28) disappears. In fact, 
839: the study of cosmological networks of string loops in 3 dimensions by 
840: Magueijo, Sandvik and Steer \cite{MSS} results in a scale-invariant loop 
841: distribution of the form of Eqs.\ (9),(10) with $q+1<5/2:$ \cite{Steer} $1.9
842: <q+1<2.1,$ or \cite{Mag} $q+1=2.03$ (plus error bars), and so in this study 
843: $q\approx 1.$ Thus, it is quite possible that linear defects which correspond 
844: to two apparently distinct physical phenomena, namely cosmic strings and 
845: crystal dislocations, are of a very similar statistical-mechanical nature. 
846: 
847: The average total dislocation length per Wigner-Seitz cell at melt is $\rho 
848: (T_m)v_{WS}=b\rho (T_m)/\lambda \approx b/2,$ since $\lambda \approx 4/3.$ 
849: Since a Wigner-Seitz cell contains $z$ links, each of length $b/2,$ it follows 
850: that, on average, one of $z$ links in each Wigner-Seitz cell is covered by a 
851: dislocation. Since each such a link is shared between two atoms, on average, 
852: half of the atoms are within a dislocation core at melt. 
853: 
854: If we use $\rho (T_m)=0.64\;\!b^{-2},$ then to $\sim 20$\% accuracy the 
855: melting temperatures and latent heats are given by 
856: \beq
857: k_BT_m=\frac{Gv_{WS}}{4\pi \ln (z-1)}, 
858: \eeq
859: \beq
860: L_m=\frac{\ln (z-1)}{2}\;\!RT_m. 
861: \eeq
862: The accuracy of these relations depends critically on the factor of $\ln (z-1
863: ),$ which is characteristic of a theory based on line-like degrees of freedom.
864: 
865: \section*{Acknowledgements}
866: We thank T. Goldman for valuable discussions during the preparation of this 
867: work. One of us (L.B.) wishes to thank J. Magueijo and D.A. Steer for very 
868: useful correspondence. 
869: 
870: 
871: \bigskip
872: \bigskip
873: \begin{thebibliography}{9}
874: %\bibitem{Ons} L. Onsager, Nuovo Cim. Suppl. {\bf 6} (1949) 249, Ann. N.Y. 
875: %Acad. Sci. {\bf 51} (1949) 627 
876: \bibitem{Shock} W. Shockley, in {\it L'Etat Solide,} Proceedings of 
877: Neuvienne-Consail de Physique, Brussels, 1952, Ed. R. Stoops, (Solvay Institut
878: de Physique, Brussels, Belgium) 
879: \bibitem{Bragg} W.L. Bragg, in {\it Proc. Symp. on Internal Stresses,} 
880: (Institute of Metals, London, 1947), p.\ 221
881: \bibitem{CD} R.M.J. Cotterill and M. Doyama, Phys. Rev. {\bf 145} (1966) 465 
882: \bibitem{Mott} C. Mott, Proc. Roy. Soc. A {\bf 215} (1952) 1
883: \bibitem{MD} R.M.J. Cotterill, Phys. Rev. Lett. {\bf 42} (1979) 1541 
884: \bibitem{MC} W. Janke, Int. J. Theor. Phys. {\bf 29} (1990) 1251 
885: \bibitem{Crawf} R.K. Crawford, Bul. Am. Phys. Soc. {\bf 24} (1979) 385 \\
886: R.M.J. Cotterill and J.K. Kristensen, Phil. Mag. {\bf 36} (1977) 453 
887: \bibitem{Miz} S. Mizushima, J. Phys. Soc. Japan {\bf 15} (1960) 70 
888: \bibitem{Ook} A. Ookawa, J. Phys. Soc. Japan {\bf 15} (1960) 2191 
889: \bibitem{Siol} M. Siol, Z. Phys. {\bf 164} (1961) 93 \\ 
890: D. Kuhlmann-Wilsdorf, Phys. Rev. {\bf 140} (1965) A1595 \\ 
891: S.F. Edwards and M. Warner, Phil. Mag. A {\bf 40} (1979) 257 \\ 
892: S.P. Obukhov, Zh. Eksp. Teor. Fiz. {\bf 83} (1982) 1978 
893: \bibitem{Pat} A.Z. Patashinskii and B.I. Shumilo, Zh. Eksp. Teor. Fiz. 
894: {\bf 89} (1985) 315 \\ 
895: A.Z. Patashinskii and L.D. Son, Zh. Eksp. Teor. Fiz. {\bf 103} (1993) 1087 
896: \bibitem{Copeland2} E. Copeland, D. Haws, S. Holbraad and R. Rivers, 
897: Physica A {\bf 179} (1991) 507  
898: \bibitem{PP} J.P. Poirier and G.D. Price, Phys. Earth Planet. Inter. {\bf 69}
899: (1992) 153 
900: \bibitem{KleinII} H. Kleinert, {\it Gauge Fields in Condensed Matter,} (World
901: Scientific, Singapore, 1989), Vol.\ II, and references therein 
902: \bibitem{Cotterill} R.M.J. Cotterill, J. Cryst. Growth {\bf 48} (1980) 582 
903: \bibitem{reviews} B.I. Halperin, Statistical mechanics of topological defects, 
904: in {\it Physics of Defects,} Eds. R. Balian, M. Kleman and J.P. Poirier, 
905: (North Holland, Amsterdam, 1981) \\ 
906: J.P. Poirier, Geophys. J. Roy. Astr. Soc. {\bf 85} (1986) 315 
907: \bibitem{YI} T. Yamamoto and T. Izuyama, J. Phys. Soc. Japan {\bf 57} (1988) 
908: 3742 
909: \bibitem{KV} J. Kierfeld and V. Vinokur, Dislocations and the critical 
910: endpoint of the melting line of vortex line lattices, cond-mat/9909190 
911: \bibitem{previous} L. Burakovsky and D.L. Preston, Los Alamos preprint 
912: LA-UR-99-4171 [cond-mat/0003494], to appear in Solid State Comm. 
913: \bibitem{Wie} F.W. Wiegel, {\it Introduction to Path-Integral Methods in 
914: Physics and Polymer Science,} (World Scientific, Singapore, 1986)
915: \bibitem{GSW} M.B. Green, J.H. Schwarz and E. Witten, {\it Superstring 
916: Theory,} (Cambridge University Press, Cambridge, 1987), Vol.\ 1 
917: \bibitem{Copeland} E. Copeland, D. Haws, S. Holbraad and R. Rivers, 
918: Physica A {\bf 158} (1989) 460, Nucl. Phys. B {\bf 319} (1989) 687  
919: \bibitem{BV} R. Brandenberger and C. Vafa, Nucl. Phys. B {\bf 316} (1989) 391 
920: \bibitem{SS} P. Salomonson and B.S. Skagerstam, Nucl. Phys. B {\bf 268} (1986) 
921: 349, Physica A {\bf 158} (1989) 499  
922: \bibitem{LT} D.A. Lowe and L. Thorlacius, Phys. Rev. D {\bf 51} (1995) 665 
923: %\bibitem{Copeland2} E. Copeland, D. Haws, S. Holbraad and R. Rivers, 
924: %Physica A {\bf 179} (1991) 507  
925: \bibitem{HL} J.P. Hirth and J. Lothe, {\it Theory of Dislocations,} 2nd ed.,
926: (Krieger Publishing, Malabar, FL, 1992)
927: \bibitem{Sar} G.F. Sarafanov, Phys. Solid State {\bf 39} (1997) 1403 
928: \bibitem{GS} M.W. Guinan and D.J. Steinberg, J. Phys. Chem. Solids {\bf 35}
929: (1974) 1501
930: \bibitem{dipole} N.D. Mermin, Rev. Mod. Phys. {\bf 51} (1979) 591 \\ 
931: N.K. Gilra, Cryst. Lattice Defects {\bf 8} (1979) 59 \\
932: K.J. Strandburg, Rev. Mod. Phys. {\bf 60} (1988) 161
933: \bibitem{KT} J.M. Kosterlitz and D.J. Thouless, J. Phys. C {\bf 6} (1973) 1181,
934: in {\it Progress in Low Temperature Physics,} Vol. VII-B, Ed. D.F. Brewer, 
935: (North-Holland, Amsterdam, 1978), p. 373. The $3D$ Kosterlitz-Thouless-like 
936: transition is discussed in C. Giannessi, J. Phys. Cond. Mat. {\bf 3} (1991) 
937: 1649, and N.K. Kultanov and Yu.E.~Lozovik, Solid State Comm. {\bf 88} (1993) 
938: 645, Physica Scripta {\bf 56} (1997) 129  
939: \bibitem{HN} D.R. Nelson and B.I. Halperin, Phys. Rev. Lett. {\bf 41} (1978) 
940: 121, Phys. Rev. B {\bf 19} (1979) 2456 
941: \bibitem{NT} D.R. Nelson and J. Toner, Phys. Rev. B {\bf 24} (1981) 363 
942: \bibitem{Gschn} K.A. Gschneidner, Jr., in {\it Solid State Physics, Advances 
943: in Research and Applications,} Eds. F. Seitz and D. Turnbull, (Academic Press,
944: New York, 1965), Vol. 16, p.~275
945: \bibitem{Emsley} J. Emsley, {\it The Elements,} (Clarendon Press, Oxford, 1989)
946: \bibitem{Tonkov} E. Yu. Tonkov, {\it High Pressure Phase Transformations,} 
947: (Gordon and Breach, Philadelphia, 1992)
948: \bibitem{Samsonov} {\it Handbook of the Physicochemical Properties of the 
949: Elements,} Ed. G.V. Samsonov, (IFI/Plenum, New York, 1968)
950: \bibitem{Nin} T. Ninomiya, J. Phys. Soc. Japan {\bf 44} (1978) 263
951: \bibitem{KL} P. Korpiun and E. L\"{u}scher, in {\it Rare Gas Solids,} Eds. 
952: M.L. Klein and J.A. Venables, (Academic Press, London, 1977), Vol. II, p. 741
953: \bibitem{Vor} V.S. Vorob'ev, High Temp. {\bf 34} (1996) 197
954: \bibitem{Vach} T. Vachaspati, Phys. Rev. D {\bf 44} (1991) 3723 
955: \bibitem{Kibble} T.W.B. Kibble, Phys. Lett. B {\bf 166} (1986) 311 
956: \bibitem{MSS} J. Magueijo, H. Sandvik and D.A. Steer, Phys. Rev. D {\bf 60} 
957: (1999) 103514 
958: \bibitem{Steer} D.A. Steer, private communication 
959: \bibitem{Mag} J. Magueijo, private communication  
960: \end{thebibliography}
961: \end{document}
962: \end
963: