Você está na página 1de 9

Cryst. Res. Technol. 46, No. 12, 1241 – 1249 (2011) / DOI 10.1002/crat.

201100310

Comparison of thermodynamic and kinetic models of single-layer


crystal–mother-phase interface

M. Izdebski* and M. Włodarska


Institute of Physics, Technical University of Łódź, ul. Wólczańska 219, 90-924 Łódź, Poland

Received 11 July 2011, revised 8 September 2011, accepted 21 September 2011


Published online 7 October 2011

Key words Jackson model, solid-fluid interface, Monte Carlo simulations, Bragg-Williams approximation,
roughening.

Simple analytical kinetic model of single-layer crystal–mother-phase interface is proposed, which provides
results that can be compared directly with thermodynamic Jackson model. Both models are based on zeroth
order approximation known from lattice-gas models, in which solid and fluid growth units are treated as
mixed randomly in the interface layer. It is shown that the kinetic and thermodynamic approaches can lead to
very similar predictions about growth mechanism. The parameters characterising growth conditions, obtained
from these models, are significantly different from those obtained from Monte Carlo simulations applied to
study stable states of single-layer interface. Monte Carlo calculations describe crystal growth in detail and
their results can predict characteristic parameters for real experiment. Observed differences seem to be
strongly influenced by the use of zeroth order approximation.
© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction
Thermodynamic and kinetic models are nowadays two corner stones of crystal growth physics [1-6].
Thermodynamic models consider the minimum of free energy which determines the conditions of crystal
growth. These models can be classified with respect to the constraints they put on the thickness of phase
interface: Jackson model (single-layer interface) [1,7], Mutaftschiev model (two-layer interface) [8] and
Temkin model (no limitation of the interface width) [9,10]. Multi-layer Temkin model leads to a complex
expression describing the free energy, which can only be studied with numerical methods, whereas calculations
in Jackson model are much simpler. Hence, Jackson model is still frequently used as a didactic tool, which
helps to understand the basic concepts. However, approximations used in either model (by Jackson,
Mutaftschiev or Temkin) may result in conclusions which are not always valid. Among particularly rough
simplifications we must mention Bragg-Williams approximation used to calculate the total average number of
solid-fluid bonds between closest-neighbouring growth units, and Boltzmann statistical expression for entropy,
in which all possible configurations of solid and fluid units are treated as equally probable.
Kinetic models study sequences of elementary processes described by mean frequencies at which they
occur. These models are often the basis for Monte Carlo computer simulations. Direct comparison of kinetic
and thermodynamic models is difficult because predictions given by models from both groups could be caused
by the following fundamental differences whose effects are not easy to separate:
1 General approach – thermodynamic or kinetic.
2 Description of the state of crystal–mother-phase interface – in thermodynamic models the state of layers in
the interface is characterised typically by average fractions of solid growth units and the distribution of
solid and fluid units within a layer is assumed to be completely random. This zeroth order approximation
does not apply in Monte Carlo simulations, where full information about the location of each unit in the
lattice is available.
____________________

* Corresponding author: e-mail: izdebski@p.lodz.pl

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


1242 M. Izdebski and M. Włodarska: Models of single-layer crystal-mother-phase interface

3 Different way of analyzing the results – traditionally, a study of stable states in thermodynamic models vs.
a study of growth rate and surface structure in Monte Carlo simulations.
The zeroth order approximation is still a standard assumption in Jackson and Temkin thermodynamic
models, and their later modifications which are still often discussed in the literature. However, research on
these models has been focusing on other problems so far, whereas distortions being a consequence of that
assumption seem to have been neglected or underestimated in discussions shown in other papers [11-15].
The aim of this paper is to compare predictions of such models, which differ in only one selected feature
with reduced other fundamental differences. The use of popular models known from the literature is not
sufficient to achieve this goal. Therefore, in this work we also propose a new approach combining selected
features of known models.
The content of the work is as follows: Section 2 shows key assumptions of the Jackson model and discusses
conclusions resulting from some of the approximations used. In section 3 we propose a new concept of a
simple analytical kinetic model using zeroth order approximation. Our approach focuses on stable states
resulting from the analysis of a function describing the kinetics of the interface, which is analogous to the study
of free energy minima in thermodynamic models. The proposed model is easily comparable to the
thermodynamic Jackson model with reduced differences 2) and 3) above. The model is easy to understand, so
it may also have a didactic value. In Section 4 we present the results of Monte Carlo simulations applied to
study the stable states of single layer crystal–mother-phase interface. The algorithm of the simulation is not
new, but its application to the study of stable states seems to be a new idea. In Section 5 the results of our
simple kinetic model are compared with Jackson model as well as with Monte Carlo simulations. The
comparison with Jackson model allows to conclude that kinetic and thermodynamic approaches can lead to
very similar outcome when other significant differences between them have been eliminated. Comparing
results of the simple kinetic model using zeroth order approximation with results of Monte Carlo simulations
with fully described configuration of growth units in the interface allowed to determine the effect of the
approximation alone, without overlapping effects of other differences 1) and 3) between the models. It is
shown that this approximation considerably changes predictions related to the conditions in which particular
growth mechanisms occur. Removal of zeroth order approximation appears to be much more important for
improving the predictions than allowing multi layer interface.

2 Thermodynamic Jackson model


Jackson model of the crystal–mother-phase interface was originally published in [7]. In this work we refer to
the presentation of the model given by Bennema in [1].
The conditions of crystal growth in Jackson model are described by Temkin coefficient α and another
parameter β, related to kinetic roughening. The Temkin coefficient is defined as

α= , (1)
kT
where z = 4 is the number of nearest neighbour sites around each unit cell in the plane of the interface for the
assumed cubic or tetragonal crystal, k is the Boltzmann constant, T is temperature, and Φ is the change in
bonds energy when a solid cell in a pure solid crystal and a fluid cell in a pure fluid area are exchanged
1
Φ = Φ sf − (Φ ss + Φ ff ) . (2)
2
The symbols Φsf, Φss and Φff denote negative energies related to solid-fluid, solid-solid and fluid-fluid bonds,
respectively. The second parameter, β, describes the driving force of crystallization:
μ − μs
β= f = ln(1 + σ ) , (3)
kT
where μf and μs are standard chemical potentials of fluid and solid cells, respectively, and σ is relative
supersaturation of the mother-phase.

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org


Cryst. Res. Technol. 46, No. 12 (2011) 1243

The Jackson model considers a single-layer solid-fluid interface composed of 100% pure solid crystal and
100% pure fluid area as the thermodynamic reference state. The free energy of this interface layer, composed
of Ns solid and Nf fluid cells, is given by
F r = µs Ns + µf N f . (4)
When solid and fluid cells are mixed, the free energy can be expressed by
F = F r + ΔU mix − T ΔSmix , (5)
r
where the chemical potentials and F are assumed to stay unchanged during mixing, ΔUmix is the energy of
mixing and ΔSmix is the entropy of mixing. The energy of mixing is related to the formation of solid-fluid bonds
in place of broken solid-solid and fluid-fluid bonds:
ΔUmix = kTα Nsf / z, (6)
where Nsf is the average number of solid-fluid bonds between closest-neighbouring units. It is common for the
thermodynamic models proposed by Jackson, Mutaftschiev and Temkin to assume completely random
distribution of solid cells in the boundary layers, as in lattice-gas models. This zeroth order (or Bragg-
Williams) approximation leads to the following estimation of the number of solid-fluid bonds:
Nsf = N z xs (1 – xs), (7)
where N = Ns + Nf is the total number of cells in a single layer and xs = Ns/N is the fraction of solid cells in the
interface layer.
The entropy of mixing is derived from Boltzmann statistical expression. In the form employed in the
model, this expression is only valid if all possible states are equally probable, which means reusing of zeroth
order simplification:
ΔSmix = k ln(g), (8)
where
N!
g= (9)
Ns ! N f !
is the number of configurations of solid and fluid cells.
After substituting eqs. (3), (4), (6-9) into (5) and making some transformations the following formula can
be derived [1]:
F
f ( xs ) = = α xs (1 − xs ) + xs ln xs + (1 − xs )ln(1 − xs ) − β xs . (10)
NkT

Fig. 1 The function f(xs) defined by eq. (10) plotted for Fig. 2 The fraction xs of solid cells corresponding to the
selected values of α (2 and 3), and β equal to 0 and 0.1. minimum of the free energy F given by formula (10). For
β ≠ 0 no minimum exists in a certain range of the fraction xs.

The analysis of occurrences of free energy minima allows to determine boundary parameters at which barriers
preventing free crystallization disappear. In particular, the function f(xs) given by eq. (10) has one minimum for

www.crt-journal.org © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


1244 M. Izdebski and M. Włodarska: Models of single-layer crystal-mother-phase interface

α < 2 and two minima for α > 2 in the case of thermodynamic equilibrium (β = 0). With increasing values of β,
the fluid phase becomes less stable as depth of its respective energy minimum decreases. Exemplary curves for
α equal to and greater than 2 are shown in figure 1. The curves, which are symmetric for β = 0, become
inclined in the case of supersaturation.
The relation between the values of the fraction xs corresponding to the minima of function f(xs) of eq. (10)
and the parameter α is shown in figure 2 for several values of parameter β. One can easily see in the plots that
α = 2 is the threshold value for β = 0, whereas at higher supersaturation (e.g. β = 0.05 and 0.4) there is only
one minimum at xs > 0.5 up to a certain critical value αcrit > 2 above which the curve splits into two branches.
It appears that curves visible in figure 2, which show stable values of xs for various α and β, can be easily
compared with other models. One can also read threshold values αcrit and corresponding values of xs from the
plots, and the same quantities or their analogues can also be found in the kinetic approach. It can be shown that
for β > 0 the minimum of the function f(xs) for xs < 0.5 disappears suddenly for α = αcrit without reaching the
value of xs for the second minimum at xs > 0.5. The left branches of the graph in figure 2 are drawn up to the
disappearance of the left minimum.
When the value of β exceeds the threshold βcrit [1]
⎛ 1− 1− 2 / α ⎞
β crit = ± ⎜⎜ α 1 − 2 / α + ln ⎟ (11)
⎝ 1 + 1 − 2 / α ⎟⎠
only one minimum of free energy remains, for xs close to 1, which means that only the solid phase is
thermodynamically stable. This result can be directly derived from function (10) by zeroing out the first and
second derivatives. In this case there is no thermodynamic barrier for crystallization and the surface is
roughened due to the positive driving force of crystallization. The plot of the dependency given by formula
(11) is shown further in the paper, along with the discussion of kinetic theory (Fig. 4).
The boundary values of parameters α and β shown above, which result from Jackson theory, may differ
from real thresholds due to the fact that derivation of function f(xs) is consequently based on the zeroth order
approach i.e. assumed completely random distribution of solid units on the entire surface, with probability xs.
This assumption was used both to derive the number of solid-fluid bonds in eq. (7) and to fulfil the condition of
equal probability of configurations in Boltzmann formula (8). This problem was raised in earlier papers [10,16]
concluding that both the energy of mixing ΔUmix and the entropy of mixing ΔSmix are significantly
overestimated in Jackson model. An improved estimation of the number of bonds Nsf, based on the results of
Monte-Carlo simulations was also proposed in [16]. The new estimation agrees with Bragg-Williams
approximation (7) only for α = 0 and decreases considerably with growth of α. Deviations from the results of
thermodynamic theories (Jackson, Temkin) can also be seen in other papers describing crystal growth with the
use of Monte-Carlo simulations [17-19].

3 The zeroth order kinetic approach


In classic Monte-Carlo simulations, an array containing full information about the location of each solid block
on the crystal face is used to describe the surface. Such an approach is fundamentally different from typical
description of the surface in thermodynamic models. In order to investigate the influence of thermodynamic
and kinetic approaches in similar conditions, we have considered a simple kinetic model inspired by Monte-
Carlo simulations, in which information about the configuration of solid and fluid blocks is reduced to a single
fraction xs of solid cells in the layer. The distribution of neighbours around the site under consideration is
assumed to be completely random as in thermodynamic zeroth order Jackson and Temkin models.
In our model, we assumed that among all N sites available on the surface of crystal face, annihilations are
possible only in N xs places filled with solid blocks and creations are possible only in N(1 - xs) locations that
are not filled by solid blocks. The probability of finding a solid block at any selected location is xs, regardless
of the type of blocks in the neighbourhood. The probability pi of finding exactly i solid neighbours in four
closest sites around a particular location is thus modelled by Bernoulli trial:

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org


Cryst. Res. Technol. 46, No. 12 (2011) 1245

⎛ 4⎞
pi = ⎜ ⎟ xsi (1 − xs )4−i . (12)
⎝i⎠
The number of blocks attached to the surface is equal to the difference between the number of creations and
number of annihilations. Hence, we get that the change dxs in the fraction of solid cells during time dt can be
calculated as
⎡ 4 4 4 4 ⎤
⎛ ⎞ ⎛ ⎞
dxs = ⎢(1 − xs ) ∑ ⎜ ⎟ xsi (1 − xs )4−i ki+ − xs ∑ ⎜ ⎟ xsi (1 − xs )4−i ki− ⎥ dt , (13)
⎣⎢ i =0 ⎝ ⎠ i =0 ⎝ ⎠ ⎦⎥
i i

where ki+ and ki− are average frequencies of creation and annihilation, respectively, in a single location that
has i solid closest neighbours. The results presented in this paper relate to a crystal growing from solution.
According to the formulae derived by Binsbergen, the frequencies of creation ki+ and annihilation ki− are
[17,20]:
⎡ α β⎤
ki+ = f t ⋅ exp ⎢ − ( 2 − i ) + ⎥ , (14)
⎣ 4 2⎦
⎡α β⎤
ki− = f t ⋅ exp ⎢ ( 2 − i ) − ⎥ . (15)
⎣4 2⎦
The frequency ft is
kT ⎛ ΔGη ⎞
ft = exp ⎜ − ⎟, (16)
h ⎝ kT ⎠
where ΔGη is the free energy of activation of viscous flow or rotational diffusion. Surface diffusion, which is
often taken into account in Monte Carlo simulations, is ignored in this simplified model, because it cannot
affect in any way the value of xs.

Fig. 3 The stable values of the fraction xs obtained Fig. 4 Curves separating the areas of kinetic
from equation (13) for a very long time t and selected roughening and layer growth obtained on the basis of
values of β. Jackson model and zeroth order kinetic model.

Equation (13) was solved for two types of initial conditions: xs(0) = 0 (completely fluid interface) and xs(0) = 1
(completely solid interface). After a very long time t the fraction xs always tends to a certain stable value. We
obtained the same stable value of xs for both considered initial conditions, or two different values, depending
on the chosen values of α and β parameters. The results of calculations are shown in figure 3. The transitions
between one and two stable values of xs show a very clear analogy to the transitions between one and two
minima of free energy predicted by Jackson model (Fig. 2). The resulting critical value αcrit = 2.04 for this
transition in thermodynamic equilibrium state (β = 0) is very similar to the one resulting from the Jackson

www.crt-journal.org © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


1246 M. Izdebski and M. Włodarska: Models of single-layer crystal-mother-phase interface

model. Both models can be directly compared on one picture (Fig. 4) where curves separating the areas of
kinetic roughening and layer growth are shown. The curve for Jackson model according to equation (11)
corresponds to transition between one and two minima of the free energy F as a function of the fraction xs. The
curve for kinetic model separates the areas of one and two stable values of xs obtained numerically from
equation (13) for a very long time t. For a given value of β we started calculations from the assumed interval of
α ∈ [1, 6], which corresponds to the existence of a single stable value of xs for αmin = 1 and two stable values
for αmax = 6. This interval has been repeatedly bisected to find αcrit inside the interval. Such calculations were
performed for many selected values of β in the range from 0 to 1.2.

4 Monte Carlo kinetic simulations


In Monte Carlo simulations, both the sequence of elementary processes and coordinates at which they occur
are determined by comparing random numbers with some number intervals. In the simulation algorithm
applied in older works the lengths of the intervals are constant throughout the simulation. Thus, adapting the
distribution of generated events to changing distribution of solid and fluid growth units requires additional
decisions whether the selected event should be performed or abandoned. Implementation of the algorithm with
constant intervals is particularly easy, but its efficiency decreases rapidly with increasing α. For this reason, we
decided to use a different algorithm in this work, similar to a fast algorithm previously described in detail in
[19]. The only change is that the phase boundary has been reduced to a single layer. This means that all
creations are prohibited at these coordinates (x, y), where solid blocks already exist inside the boundary layer.
Analogously, annihilations are not allowed at those sites where solid blocks are not attached. According to the
solid-on-solid assumption all layers below the boundary layer are completely filled with solid blocks, and all
layers above it are fluid.

Fig. 5 The stable values of the fraction xs of solid cells in Fig. 6 Curve separating the areas of one and two
the solid-fluid interface obtained employing single-layer stable values of xs obtained employing single-layer
Monte Carlo simulations. Monte Carlo simulations.

The formulas describing frequencies of creation and annihilation for growth from solution are given by eqs.
(14) and (15). The fast algorithm used in our Monte Carlo simulations provides results equivalent to those
obtained using the classical algorithm known from early works of Gilmer and Bennema [17,18], but it is more
efficient and enables calculations of explicit time in the crystal–mother-phase system instead of only rough
measure of time available in the classical algorithm via number of events. High efficiency of the algorithm was
achieved by introducing an array containing frequencies of occurrence for all combinations of coordinates and
individual elementary phenomena potentially possible in these places. At each repetition of the algorithm main
loop one element of this array is randomly selected with probability proportional to the frequency stored in it.
Each selected element determines uniquely an event for execution without any further decision. After every

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org


Cryst. Res. Technol. 46, No. 12 (2011) 1247

event some elements of the array situated near the site of the last event must be updated. The need for efficient
search and update of the elements of this huge array makes this algorithm not easy to implement. Some details
on the optimal implementation of the algorithm can be found in [19].
The simulations were performed for two types of initial conditions: xs(0) = 0 (completely fluid interface)
and xs(0) = 1 (completely solid interface). After a sufficiently long time the fraction of solid blocks xs in the
interface tends to a stable value and is subject to only minor fluctuations. We found that there may exist two
different stable values of xs or only one value common to various initial conditions, depending on the α and β
parameters (Fig. 5). The transition between one and two stable values of xs is quite sharp, which enabled us to
plot the transition curve in α and β coordinates shown in figure 6.
The results of previous Monte Carlo simulations for multi-layer phase boundary showed that surface
diffusion significantly accelerates the growth rate of crystal faces [17,19]. However, we found no clear
influence of this phenomenon on the number of stable values of xs for single layer model (results presented in
figures 5 and 6). Average path lengths of surface diffusion investigated in our simulations ranged from 0 to 3a,
where a is the lattice spacing. Implementation of surface diffusion in the simulation algorithm was described
previously in [19].

5 Discussion
In an attempt to compare the predictions of Monte Carlo simulations of crystal face growth and thermodynamic
models of crystal–mother-phase interface proposed by Jackson, Mutaftschiev and Temkin we encounter the
problem of overlapping effects of several fundamental differences listed earlier in section 1. In order to
separate the impact of kinetic and thermodynamic approaches, a simple kinetic model has been proposed in
section 3, which was purposely conformed to thermodynamic Jackson model: in both models the state of the
interface is described only by the average fraction xs of solid growth units in a single-layer interface, the
distribution of units within a boundary layer is assumed to be completely random (zeroth order approximation),
and results are given in analogous form. The results presented in figures 2 and 3 show that in the case of
thermodynamic equilibrium state (β = 0) both models predict the transition between one and two stable states
of the interface layer for very similar values αcrit = 2.00 for Jackson model and 2.04 for the kinetic model. If
there is a driving force for crystallization (β > 0) then the value of αcrit increases similarly in both models with
increasing β (Fig. 4). Corresponding to a stable interface for α > αcrit, the following features of the behaviour of
xs values may be noted:
1 In Jackson model an abrupt transition occurs between two distinct values of xs for α > αcrit and one value of
xs for α < αcrit (Fig. 2).
2 The kinetic model predicts that two stable values of xs come close to each other for α decreasing towards
αcrit (Fig. 4). However, the curve α(xs) is so close to a horizontal line for α slightly above αcrit that
distinguishing between discontinuous and continuous transition does not appear to have significant physical
meaning.
Thus, we see that thermodynamic Jackson model and analogous kinetic model lead to very similar predictions
when other fundamental differences between the models (mentioned in Introduction) have been eliminated.
The aim of the classic Monte Carlo simulations is usually to investigate the growth rate of crystal faces and
to monitor the structure of their surfaces. Moreover, Surface Roughening Coefficient, being an average number
of vertical walls adjacent to a location, has been recently proposed to describe the roughness of the surface of
crystal face [21]. In the case of phase boundary reduced to a single layer this coefficient becomes equivalent to
the ratio of solid-fluid bonds to all cells in the interface layer: Nsf / N. The results presented in all these ways,
however, do not show rapid qualitative change in the function of simulation parameters and it is difficult to
accurately locate the transition between various growth mechanisms in the coordinates of α and β. Although
single-layer Monte Carlo simulations proposed in section 4 are not suitable for investigating crystal growth
rate, they may be convenient to study the stable states of crystal–mother-phase interface. After a sufficiently
long time the fraction xs tends to a stable value and the number of possible stable values obtained for various
www.crt-journal.org © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1248 M. Izdebski and M. Włodarska: Models of single-layer crystal-mother-phase interface

initial conditions xs(0) = 0 and xs(0) = 1 depends on the α and β parameters. The transition between one and
two stable values of xs is quite sharp (Fig. 5) and the boundary curve may also be drawn in the coordinates of α
and β (Fig. 6). The observed differences in the results of single-layer Monte Carlo simulations and the kinetic
zeroth order model described in section 3 is merely a consequence of different level of accuracy in the
description of the state of the crystal–mother-phase interface. Full information about the long range ordering of
solid and fluid growth units is available only in Monte Carlo simulations. The distribution of units in the
simplified kinetic model is assumed to be completely random (zeroth order approximation) and the state of
interface is described only by averaged fraction xs of solid units in the boundary layer. Thus, we can observe
the impact of zeroth order approach separated from the effects of other differences between the models.
Comparing the results shown in figure 3 with figure 5 and the results in figure 4 with figure 6 we can see that
the shapes of the obtained curves are similar, but the plots differ greatly in characteristic values, namely:
1 the value of αcrit for thermodynamic equilibrium (β = 0) is about 2.04 for the zero-order model and 3.4 for
Monte Carlo simulations,
2 according to both models αcrit increases with increasing β, but the zeroth order model predicts significantly
slow increase than single-layer Monte Carlo simulations.
The results of Monte Carlo simulations presented in our previous work [16] showed that the estimation of Nsf
given by eq. (7) is valid only for α close to zero and the value of Nsf should decrease quickly with an increase
in α, although no such dependence is included in eq. (7). Moreover, any form of regular arrangement of solid
and fluid growth units within the interface leads to different probabilities of various configurations and the
entropy estimated by the formula (8) becomes overvalued. The combination of overvalued expressions for the
energy and entropy of mixing with the term "–β xs", which seems to be most accurate in eq. (10), results in too
low value of αcrit for β = 0 and too weak dependency of αcrit on β.
Multilayer models may provide more results, e.g., by studying interface thickness, which does not make
sense for the interface reduced to a single layer. Also, the proposed simple kinetic model may be extended to
multilayer interface and compared with thermodynamic Temkin model. However, when the curve separating
various mechanisms of crystal growth is of interest, the differences between prediction of single-layer Jackson
and multi-layer Temkin thermodynamic models are not significant [1]. In addition, we found that limiting the
interface to a single layer does not significantly affect the predictions of the Monte Carlo simulations, either. It
can be explained by the minor role played by vertical bonds in thermal and kinetic roughening, which allowed
other authors (e.g. [2]) to use 2D models as well. Thus, continuation of work on single-layer models seems to
be still promising. In particular, it might be interesting to further develop the single-layer model for the case of
ordered monolayer adsorption of foreign molecules on a crystal surface.

6 Conclusions
Our results can be summarized as follows:
1 Comparison of the results derived from thermodynamic and kinetic models known from the literature is
difficult because of overlapped effects of several major differences between models. We have proposed a
simple kinetic model of single-layer crystal–mother-phase interface, designed as a close analogy to
thermodynamic Jackson model. Their comparison shows that the kinetic and thermodynamic approaches
lead to very similar predictions about the growth mechanism of crystal face when we remove other
fundamental differences between them.
2 The simple kinetic model proposed in this work is derived from Monte Carlo simulations in which the
description of the configuration of solid and fluid units inside the layer has been reduced to zeroth order
approximation, as is typically assumed in thermodynamic models. The results arising from this simplified
model were compared with the results of Monte Carlo simulations used to study long-term stable states of
phase interface limited to a thickness of one layer. The results obtained show that the zeroth order
approximation is very rough and strongly influences the characteristic parameters describing the growth
mechanisms, which is very important for predicting conditions of a real experiment.

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org


Cryst. Res. Technol. 46, No. 12 (2011) 1249

3 The above findings suggest a direction for further research on thermodynamic models of crystal–
mother-phase interface. The most widely discussed thermodynamic models (especially single-layer Jackson
model and multi-layer Temkin model) are based on zeroth order approximation, which is used to estimate
both the energy ΔUmix and the entropy ΔSmix of mixing of solid and fluid cells. In the light of the results
presented in this study, the effects of this very coarse approximation are even stronger than the effect of
limiting the interface in Jackson model to a single layer. In this situation, it seems essential to remove the
zeroth order approximation in order to bring these thermodynamic models closer to reality.

References
[1] P. Bennema, KONA Powder Part. J. 10, 25 (1992).
[2] P. Bennema, in: D. T. J. Hurle (Ed.), Handbook of Crystal Growth, Vol. 1a (North-Holland, Amsterdam, 1993), p.
477.
[3] Y. Shiohara and E. A. Goodilin, Single Crystal growth for science and technology, in: Handbook on the physics
and chemistry of rare earths, edited by K. A. Gschneidner, Jr., L. Eyring, M. B. Maple (Elsevier Science,
Amsterdam, 2000), vol. 30, chapter 189.
[4] I. V. Markov, Crystal Growth for Beginners: Fundamentals of Nucleation, Crystal Growth, and Epitaxy, 2nd ed.
(World Scientific, Singapore, 2003), Sec. 1.6.3.1, p. 61 and 1.6.3.2, p. 64.
[5] K. A. Jackson, J. Cryst. Growth 264, 519 (2004).
[6] J. J. Derby, in: Proceedings of the 14th International Summer School on Crystal Growth, AIP Conf. Proc. 1270
(Melville, New York, 2010), pp. 221-244.
[7] K. A. Jackson, Liquid Metals and Solidification, (Am. Soc. for Metals, Cleveland, 1958), p. 174.
[8] B. Mutaftchiev, Adsorption et Croissance Cristalline (Centre Nationale de la Recherche Scientifique, Paris, 1965),
p. 231.
[9] D. E. Temkin, Crystallization Processes (Consultants Bureau, New York, 1966), p. 15.
[10] P. Bennema and G. H. Gilmer, in: Crystal Growth: an Introduction, ed. P. Hartman, (North–Holland Publ., 1973),
p. 263.
[11] C. Van Leeuwen, P. Bennema, and D. J. van Dijk, Acta Metall. 22, 687 (1974).
[12] Y. Ishibashi, J. Phys. Soc. Jpn. 55, 2099 (1986).
[13] D. Y. Li, J. Mater. Sci. Technol. 13, 457 (1997).
[14] A. Mori and I. L. Maksimov, J. Cryst. Growth 200, 297 (1999).
[15] T. Ivas, Continous Temkin theory of interface (arXiv:0810.3509v1 [cond-mat.mtrl-sci], 2008), retrieved from:
http://arxiv.org/abs/0810.3509.
[16] M. Izdebski, Scientific Bulletin Physics / Technical University of Łódź 31, 25 (2010).
[17] G. H. Gilmer and P. Bennema, J. Appl. Phys. 43, 1347 (1972).
[18] G. H. Gilmer and P. Bennema, J. Cryst. Growth 13/14, 148 (1972).
[19] M. Rak, M. Izdebski, and A. Brozi, Comp. Phys. Commun. 138, 250 (2001).
[20] F. L. Binsbergen, Kolloid Z. Z. Polym. 237, 289 (1970).
[21] M. Rak, A. Brozi, and J. Żmija, J. Achiev. Mater. Manufac. Eng. 41, 91 (2010).

www.crt-journal.org © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Você também pode gostar