Voronoi 2
Voronoi 2
model
J. Neil Cape, John L. Finney, and Leslie V. Woodcock
Citation: The Journal of Chemical Physics 75, 2366 (1981); doi: 10.1063/1.442299
View online: http://dx.doi.org/10.1063/1.442299
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/75/5?ver=pdfcov
Published by the AIP Publishing
Softsphere model for the crystal–liquid interface: A molecular dynamics calculation of the surface stress
J. Chem. Phys. 73, 2420 (1980); 10.1063/1.440392
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
141.209.144.122 On: Tue, 25 Nov 2014 17:05:41
An analysis of crystallization by homogeneous nucleation
in a 4000-atom soft-sphere model
J. Neil CapeS)
Department 0/ Physical Chemistry. Lensfield Road. Cambridge. England
John L. Finney
Department o/Crystallography. Birkbeck College. University 0/ London. England
Leslie V. Woodcock b)
Physical Chemistry Laboratory. University 0/ Amsterdam. Nieuwe Achtergracht 127, Amsterdam.
The Netherlands
(Received 17 March 1981; accepted 27 April 1981)
2366 J. Chern. Phys. 75(5).1 Sept. 1981 0021·9606/81/172366·08$01.00 © 1981 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
141.209.144.122 On: Tue, 25 Nov 2014 17:05:41
Cape, Finney, and Woodcock: Crystallization by homogeneous nucleation 2367
accord with the predictions of classical homogeneous o 1000 2000 3000 Time Steps
nucleation theory.
FIG. 1. Variation of reduced pressure with time for the an-
II. SIMULATION METHOD nealing simulation; each time step is 0.005 mif IE.
4
T ABLE I. Initial and final properties of the annealed system.
2
Initial Final Equilibrium
Reduced density (p*) 1. 831 1.831 1. 831 o 1000 2000 3000 Time Steps
Reduced pressure ($) 156.0 152.1 137.3 FIG. 2. Mean-square displacement of atoms (beginning after
Structure amorphous crystallites fcc crystal 240 time steps) as a function of time for the annealing simula-
tion.
er than that of the Kauzmann point of zero excess en- TABLE II. Distribution offl , the number of faces per poly-
tropy as determined previously, 14 yet even after anneal- hedron, for the initial configuration.
ing over several thousand time steps the diffusion is
finite and measurable. il 11 12 13 14 15 16 17
Frequency 2 88 473 830 492 141 21
III. METHOD OF STRUCTURE DETERMINATION Percent of total 0.1 4.3 23.1 40.5 24.0 6.9 1.0
A. Voronoi polyhedra method
The identification of a crystal nucleus in a fluid of
effectively the same density is not a trivial problem. face-center-cubic structure would have index (0, 12,
It the nucleus eventually grows it may be possible to 0, 0, ... ) and a body-center-cubic structure would have
use the structure factor to characterize the growth, 3 index (0, 6, 0, 8, 0, ..• ), where the zeros continue in-
but this method is only practical when the nucleus is definitely. For simplicity, the index may be written
well defined. When the nucleus or nuclei are in the pro- neglecting the zeros to the right of the last nonzero di-
cess of formation a more detailed method is required, git, i. e., (0, 12) and (0, 6, 0, 8) for the examples
and the work described here follows the method of Tane- above. Typical distributions of the Voronoi polyhedra
mura etal.,7 which is based on the local topology about for a fluid, a dynamic fcc crystal, and a dynamic bcc
each atom. crystal are given in Fig. 3. For selected indices the
To each atom i is ascribed a "Voronoi polyhedron" distribution is markedly different for each of the three
(VP j) which bounds the region of space which is closer cases. As expected, there is no particular index typi-
to atom i than to any other atom. The faces of VP I lie cal of the fluid structure, whereas for the bcc crystal
within the planes which perpendicularly bisect the lines the proportion of (0, 6, 0, 8) polyhedra is very large.
joining atom i to neighboring atoms j. The volume de- Any others arise from vibrations about the lattice points
fined by VP I is a measure of the local atomic volume, which distort the local symmetry. The rhombic dode-
and the set of all the VP I fills the total volume uniquely. cahedra of a perfect fcc lattice are destroyed by any
The Voronoi polyhedra may be regarded as generalized kind of Vibration, and give rise to polyhedra with in-
Wigner-Seitz cells. If the total number of faces of VP I dices such as (0, 3, 6, 4), (0, 3, 6, 5), (0, 4, 4, 6),
is given by II> then/i represents a local geometrical co- (0, 4, 4, 7), etc. The types of Voronoi polyhedra
ordination number. The local configuration about atom which are found for a given configuration are indicative
i can be described by the number of edges of each of the of the short-range structure and the spatial distribution
faces of VP;, and may be indexed, following Finney, 5 by of related Voronoi polyhedra reflects the long-range
listing the number of faces having 3,4,5, .•. edges. The structure.
set of integers n 3, n4, n5, •.• , where I", n", =Ii describes For example, a nucleus with bcc symmetry is charac-
the detailed atomic arrangement around atom i. A perfect terized by Voronoi polyhedra of type (0, 6, 0, 8) or
(0, 4, 4, 6) which are contiguous to another polyhedron
of type (0, 6, 0, 8). A perfect fcc (or hcp) crystal gives
rise to a polyhedron of type (0, 12), but since this type
freq involves vertices at which four, rather than three, edges
% meet, it is transformed into a variety of different types
b.c.c. cry st a I
by the small distortions arising from thermal vibrations.
(=1.51 10 Although techniques existl1 for referring to the ideal fcc
polyhedron type, their application is not, at present, un-
o+---~~--~~--~~~ equivocal and cruder methods, as described in Sec. HID
below, were used to confirm the nucleus with fcc sym-
20
metry in the present work.
f.c.c. cry st a I
~*=1.50 10 B. Structure of the initial and annealed configurations
The atomic positions of the initial configuration were
O~~~~LL~~~-W~~
used to calculate the Voronoi polyhedra for the con-
20 figuration. Only those polyhedra which did not cross
fluid the periodic boundaries of the system were considered.
,"=1.12 10 This was to expedite the computational analysis since
the actual pOSition of the reference frame for the period-
ic boundary is entirely arbitrary.
Figure 4 shows the distribution of index types; a com-
parison with Fig. 3 shows no significant evidence of an
increase in the numbers of polyhedra associated with a
particular type of crystal structure. The distribution is
not identical to that of Fig. 3 because the reduced densi-
FIG. 3. Distribution of selected types of Voronoi polyhedra ties are different. The number of faces per polyhedron,
for the equilibrium fluid, fcc crystal, and bcc crystal. i. e., the distribution of Ii is given in Table II. The
15 20 Freq t g( r)
o 1 2 3
FIG. 6. Pair distribution function for the initial configuration.
FIG. 4. Distribution of selected types of Voronoi polyhedra The pair distribution function g(r) for the initial con-
for the initial configuration. figuration is shown in Fig. 6. This is typical of a highly
supercooled liquid or amorphous structure, with a split
second peak and with no signs of peaks or shoulders cor-
mean local coordination number so defined is 14.09, and responding to the lattice spacings of crystal structures.
the distribution of polyhedron volumes about the mean Taking into account the distribution of the polyhedral
is shown in Fig. 5. types, the absence of bcc nuclei and the absence of any
evidence of crystallinity in the pair distribution function,
Analysis of the initial configuration in terms of possi- it appears that the initial configuration is wholly amor-
ble bcc nuclei using the criteria described above gives phous in character.
only one group of atoms with more than ten members.
This group contains only three atoms of type (0, 6, 0, 8), Two further analyses in terms of the Voronoi poly-
and cannot be regarded as a potential crystal nucleus. hedra were performed after annealing. The first was
As both previous simulations have found bcc structures for the single configuration after 3180 time steps, and
occurring before fcc structures, even though the latter the second for the time-averaged atom positions over the
are the more stable crystalline forms for the bulk crys- period 3181 to 3230 time steps. The latter procedure
tal, no attempt was made to locate groups of atoms with reduces the effects of atom vibrations about fixed points.
fcc symmetry. The resultant polyhedron statistics were virtually iden-
tical, and the following results are based on the data
from the time-averaged configuration.
The distribution of polyhedron types, shown in Fig. 7,
freq is now quite different from that of the initial configura-
% tion (Fig. 4), and comparison with Fig. 3 suggests ele-
5 ments of both bcc and fcc structure. There is a marked
increase in the number of (0, 6, 0, 8) polyhedra, typical
4 of a bcc structure, and in the numbers of (0, 4, 4, 6) and
(0, 4, 4, 7) polyhedra, which may be related to fcc sym-
3 metry. The distribution of 1/ is given in Table III, and
the proportion of polyhedra with 16 or 17 faces is seen
2 to be smaller than for the initial configuration, although
the mean local coordination number is almost the same,
at 14. 05. The distribution of polyhedron volumes is
given in Fig. 8, and shows a smaller spread and a skew
O~--~~~~--------r-------~~~--------
0.9 1.0 1.1 VOlume/v towards smaller volumes when compared to the initial
configuration.
FIG. 5. Distribution of Voronoi polyhedra volumes about the
mean for the initial configuration. The pair distribution function (Fig. 9) is similar to
Ii index 5 10 15 20 Ireq%0
freq
12 0012 r.
13 ,,1102
5
0283
0364 I 4
0445
0526
~ 3441
0 2
3
14 0284 I
10365 I
0446 I
0608 I O~--------~~-------r------~~~-------
0·9 1·0
1264 II 1·1 VOlume/V
15 0285
~345 1 FIG. 8. Distribution of Voronoi polyhedra volumes about the
mean for the initial configuration.
0366 'I
0447 I If the whole system had a perfect bcc structure, then
0528
13461
l-J the edge of the cubic unit cell would have length 0.0794l,
where l = V l/3 • The direction cosines and lattice
all ot hers I constants for each of the three largest nuclei are given
in Table IV. The direction cosines of the cubic unit cell
FIG. 7. Distribution of selected types of Voronoi polyhedra are calculated as averages over the 12 edges of each
for the final configuration. unit cell. It is clear from the data in Table IV that the
nuclei are not correlated with each other in that the di-
rection cosines are different in each case. This sug-
that for the initial state. There is a slight shoulder on gests that the 4000-atom system is big enough to observe
the second peak at r= 1. 3 a(E:/kT)1/12, and the splitting homogeneous nucleation free from the effects of periodic
of the second peak is more pronounced, but the pair boundaries. The direction cosines in each case define
distribution function is seen to be relatively insensitive axes which are orthogonal to within the statistical un-
to the changes in detailed structure as shown by the dis- certainty, and the side length of the cubic unit cell is as
tribution of polyhedral types. expected. The relative positions of the atoms within
each of the nuclei is shown in Figs. 10 and 11, where
The analysiS in terms of contiguous polyhedra of type
the coordinates are plotted as a projection along the cube
(0, 6, 0, 8) gives the clearest indications of crystallinity
In all, seven groups each containing more than 20 atoms'
with polyhedra of types (0, 6, 0, 8) or (0,4, 4, 6) were
found, the largest of which had 41 members, including
19 of type (0, 6, 0, 8). The three largest groups are g(r)
described in detail in the next section, and groups with
fcc symmetry are discussed in Sec. IUD. 5
f, 11 12 13 14 15 16 17
Frequency 47 473 961 468 81 11 o 1 2
Percent of total 0.0 2.3 23.2 47.1 22.9 4.0 0.5
FIG. 9. Pair distribution function for the final configuration.
/ ' prOjection
TABLE VI. Predictions of classical nucleation tion can be derived from the forgoing analysis.
theory.
Previous simulations of crystallization have been per-
p* p nmu tl.C e (nm;a)!kT formed on systems with 500 atoms or fewer, and with
periodic boundary conditions, and only a single crystal-
1. 20 26.34 1503 93.6
1. 25 30.75 117 16.7 lite has been observed. It has not been clear to what
1. 30 35.75 23 5.5 extent this nucleation was induced by the periodic bound-
1.409 48.88 1. 93 1.0 ary although it has been suspected that it plays a domi-
1.445 54.16 1. 00 0.63 nant role for small systems. The present results show
1. 831 152 0.005 0.016 unequivocally that homogeneous nucleation is a natural
process in a large system and that it occurs independent
ly of the periodic boundary.
tl.C e (n) = YC-L A(n) + (J.1.e - J.1.L) n ,
These findings accord quantitatively with the predic-
where A is the surface area 41T(3V/41TN)2/3 n2/ 3 and tions of classical homogeneous nucleation theory. Ther,
YC-L is a value for the interfacial tension averaged over appears to be virtually no barrier to the occurrence of
all possible oreintations of the interface; J.1. is the chemi- the spontaneous nucleation observed in our computation~
cal potential in each phase. The approximation usually the only retarding factor seems to be the relaxation tim,
made is that the enthalpy of fusion is independent of tem- for cooperative translational diffusion. The height of th
perature, so that, from the Gibbs-Helmholtz relation, barrier at the density and temperature which it occurs
d(J.1.c - J.1.L)/ dT = tl.H,/ NT is predicted to be much less than kT by the classical
treatment.
and integration gives
J.1.e - J.1.L = (tl.H,/ N) In{T/T,) ,
In addition to the many bcc nucleites, a fcc nucleite
has been confirmed thus casting doubt on the "Ostwald
as J.1. e = J.1.L at T = T,. step rule,,3 that the first stage in phase transformation
This gives is to that phase of second lower stability and not to the
tl.Ce{n) A (n) tl.H, )
most stable. The most stable phase of the soft-sphere
--;;r- =---;;:r- YC-L + n NkT In(T/T, • crystal is the fcc.