You are on page 1of 10

Journal of Crystal Growth 108 (1991) 637—646 637

North-Holland

New periodic morphologies observed during dendritic growth


of ammonium chloride crystals in thin layers
Eli Raz, S.G. Lipson
Physics Department, Technion, Haifa 32000, Israel

and

Eshel Ben-Jacob
School of Physics and Astronomy, Raymond and Beverly Sackler Faculty of Exact Sciences, Tel-A viv University,
Ramat Aviv 69978, Israel

Received 8 September 1990

We report novel morphologies observed during the growth of dendritic crystals from supersaturated solutions of NH
4CI and
NH4CI(CuSO4). These develop when the anisotropic crystal grows in a thin layer. The crystal shapes are distinctly periodic and
exhibit long-range order; their growth front appears to be marginally stable. Crystals of these forms grow as three-dimensional
structures in quasi-two-dimensional cells, and their dominant periodicity is determined quantitatively by the smallest dimension of
the cell. We argue that the observed phenomena result from an interplay between the geometrical constraint of the cell and the
kinetics of the phase transition at the interface. The observations are consistent with a model in which dendritic growth occurs at
temperatures below the equilibrium roughening temperature of the dominant crystal facet, although kinetic roughening of the facet
occurs.

1. Introduction pattern observed results from the interplay be-


tween these two factors. Introduction of a third
Dendritic growth of crystals is one of the most length scale, the thickness of the growth cell, re-
intensively studied systems which show sponta- suits in further pattern formation on an ap-
neous pattern formation during their evolution propriate scale, which is the subject of this paper.
[1—5],and an excellent experimental system for The results emphasise that the three-dimensional
this purpose is the crystallization of a salt from its reality of crystal growth must be recognized even
supersaturated solution [6—9].The cornerstone of in a quasi-two-dimensional environment [8,9].
the emerging picture is the recognition that growth Most recent studies focused on the role of
patterns can be determined by the combination of anisotropy in surface tension [1—4,11]and linear
external macroscopic thermodynamic forces and growth resistance [12,13] as controlling factors.
microscopic interfacial dynamics. Diffusion kinet- More recently, Raz et al. [9] measured directly the
ics determine the macroscopic approach towards surface kinetics in NH4C1 using an optical tech-
equilibrium, and tend to drive the system towards nique. Their results indicated that the boundary
“decorated” or irregular shapes via the Mullins— conditions at the interface are due to growth
Sekerka instability [10], whereas microscopic inter- kinetics and the relationship between interface
face dynamics introduce both characteristic length velocity and supersaturation is very non-linear.
scales and preferred growth directions. The final This suggests that growth below the thermody-

0022-0248/91/$03.50 © 1991 — Elsevier Science Publishers B.V. (North-Holland)


638 E. Raz et al. / New morphologies observed during dendritic growth of ammonium chloride

namic roughening temperature is involved. Under


such conditions it is also expected that there will
be a very strong anisotropy in the angular depen-
dence of the growth velocity; in fact this velocity
~ will be essentially zero except for growth in cer-
tain preferred directions. There has not yet been
any theoretical study predicting shapes of crystals
growing under the proposed dynamics; however it
is realistic to expect a sharp selection of growth
velocities and of directions and locations of side-
branches. The experimental results presented in
ref. [9] provide strong indication that this is in-
deed the case.
During the experiments described in ref. [91we
were frequently surprised to observe unusually
periodic dendritic shapes which we have not seen
described in the literature. These shapes include
dendrites with zig-zag stems and side-branches,
hexagonal cellular structures (according to the
origin of the term “dendrite”, these could hardly
~ be called dendritic!), and asymmetrical dendrites
with all the branches on one side. Most of these
structures show much more definite periodicity
,~ i~ than do conventional dendrites. Photographs of
some examples are shown in figs. 1—3. The
~4 surprising feature of these observations is the dy-
Fig. 1. A growing Stem gnes rise to alternating straight namical and structural stability which is evident.
sidebranches spaced at regular intervals on its two sides,. The It clearly originates in the geometrical constraint
stem has a zigzag structure. The side-branches do not branch .
of the cell together with the highly amsotropic
further. (Type Bi, d = 28 ~&m,with 0.05% impunty. The scale
bar is 200 jsm long.) growth velocity, and gives further credence to the
picture of growth at a temperature below an equi-
librium roughening transition.
A different class of periodic structures, which
- develop on an interface in a three-dimensional
system during its growth, has been investigated
- . quite intensively [15,16], but these are unrelated to
the ones which we describe here, which appear
• sequentially in time.

2. Experimental technique
The experiments were carried out in a system
similar to that described in ref. [91. The dendrites
grew in thin parallel-sided cells and were observed
• by reflection in a microscope. In order to ensure
l-ig. 2. A growing stem gives rise apenodically to zigzag side-
branches. Each side-branch is like fig. 1. (Type B3, d = 28 gm, uniform temperature in the x, y plane, the base of
with 0.025% impurity. The scale bar is 200 urn long.) the cell was made from copper, covered by a thin
E• Raz et al. / New morphologies observed during dendritic growth of ammonium chloride 639

(0.1 mm) aluminium-coated glass mirror. By means and 170 ±2 tim. The top boundary of the cell was
of two wire spacers separated by about 25 mm, we a 1 mm thick microscope slide. We used flowing
could define cells with thicknesses d between 28 water to cool the cell base, and its temperature, in

~t

~ ~ ~
)•~••~
.~
~

~ ~_-c~p~-~

I ‘~‘~‘~/ I
ç~’ t’~~L
~ .•-c- - -
~ C

~ ~
Fig. 3. A hexagonal network of dendrites is set up (Type C). The scale bars are 200 pm long. (a) The hexagons may be regular (Cl,
d = 56 pm), but (b) are not necessarily so (Cl, d = 28 pm). (c) An overview shows the structure to have long range order, and to grow
with a well-defined growth front (C2, d = 28 pm). (d) In detail, a hexagonal unit may have subsidiary short and irregular
side-branches on its arms (CI, d = 120 pm). (e) Addition of 0.02% CuSO
4 produces the same structure, but the random
side-branching is supressed (Cl, d =120 pm).
640 E. Raz et al. / New morphologies observed during dendritic growth of ammonium chloride

differences between the results obtained in the two


8~ --~~io systems.
6 ~_______
0 \\\‘siwiw~s O.~ 7
L
_____I 3. Experimental results
— _________________ Depending on the thickness of the cell, the
____________________ degree of supersaturation and the orientation of
1 the initial nucleus, one of a variety of dendritic
Fig. 4. Schematic diagram of the growth cell used in the structures was observed. Under many conditions
present experiments; (1) microscope Stage; (2) tilting table; (3) we observed dendrites similar to those investigated
water-cooled chamber; (4), copper cell base; (5) thermocouple; by Dougherty et al. [7] and Chan et al. [6]. These
(6) thin mirror

(7) spacer; (8) sample;
.
(9) cover•glass (10) dry dendrites grow mainly parallel to the cell
nitrogen flow; (11) microscope objective.
boundaries (the base and the glass cover slide) and
branch in the [100], [110] or [111] directions, de-
pending on the degree of supercooling. The pre-
the range 1 ±0.2 to 15 ±0.2°C was measured sent investigations concentrated on the region of
with a copper—constantan thermocouple. This high supercooling, at growth temperatures be-
method of cooling from the base introduced a tween I and 8°C.Then growth occurs only in the
vertical temperature difference of less than 0.1°C, [111] directions, as was confirmed by measuring
because of the insulation afforded by the upper the angles between the various branches, which
boundary slide. A schematic diagram of the cell is are 70.530 and 109.470. Several new dendritic
shown in fig. 4. structures were observed which we attribute to
We prepared solutions containing 27—29% boundary effects and the strong anisotropy of the
NH4C1 by weight in distilled water from analyti- growth. These were as follows:
cal grade material and mixed them well at temper- (A) Dendrites which grew with aperiodic
atures above their saturation. We also prepared branches on one side only. (By the word
solutions in which 0.02—0.1% CuSO4 was added; “aperiodic”, we mean that no clearly defined pen-
the effect of this impurity will be discussed below. odicity was present. No detailed analysis is im-
For each experiment we placed a drop of the plied.)
thoroughly mixed solution on the base of the cell (B) Dendrites in which the stem zigzags in a
and covered it while maintaining the temperature periodic manner:
above saturation. The drop would spread to a (B1) Side-branches are produced every time the
diameter of about 12 mm. We then cooled the cell stem changes direction. The branches then grow
to its working temperature, in order to obtain long and straight, with no further branching (fig.
supercooling in the range 10—25°C. These figures 1). This type of growth was observed in layers
are equivalent to dimensionless supercoolings of thinner than about 50 ~.tm.
0.08—0.22 [17]. Higher degrees of supersaturation (B2) The pattern of BI grows, but the straight
generally resulted in spontaneous nucleation. We side-branches grow secondary branches which
initiated nucleation by gently tapping the cell, and zigzag like the original stem, usually in antiphase
recorded the growing crystals on video tape; the with it.
record included the temperature (mY output from (B3) A straight stem, which aperiodically pro-
the thermocouple) and the time. We also built a duces side-branches which zig-zag. This is really
transparent cell with a sapphire base, whose pur- like B2, starting at the second stage (fig. 2).
pose was to allow direct photography of some (C) Growth in a hexagonal honeycomb pattern
examples of the results on 35 mm film to illustrate (fig. 3).
this paper. This cell was cooled by a thermoelec- (Cl) The honeycomb is created in thick cells by
tric (Peltier) device. There were no observable the interlacing of several zigzags which grow
E. Raz et a!. / New morphologies observed during dendritic growth of ammonium chloride 641

parallel to one another with equal spacings. The later publication. However, the morphologies A, B
zigzags are synchronized in time, and clearly all and Cl were observed both with the pure and
the stems originated from the same seed. The impure media; one of the striking effects of the
aspect ratio of the hexagon changes from experi- impurity, which served to emphasize the periodic
ment to experiment. The arms of the hexagons are behaviour, was to reduce or even remove com-
often decorated with small aperiodic sidebranches pletely the aperiodic aspects of the growth. Mor-
which rarely develop (figs. 3a and 3b). phology C2 was observed only in pure NH4CI.
(C2) The honeycomb grows out from a single The figure legends indicate the addition of impuri-
nucleus which branches and rebranches in a highly ties. Compare in particular figs. 3d and 3e.
periodic and synchronized manner, giving a two-
dimensional lattice with surprisingly long-range
order. Occasional defects occur, but the lattice 4. Discussion
heals around them. The growth front observed as
the crystal spreads out is smooth and does not Our explanation for the observed structures is
develop instabilities. This form of growth occurs based mainly on two ideas. The first is a hypothe-
in thin cells (d < 50 ~tm) and requires very uni- sis, which was presented in detail in ref. [9], that a
form temperature. The hexagons are regular and nucleation process is involved in the creation of
their edges do not develop subsidiary side-branches side-branches. At temperatures below the equi-
(fig. 3c). librium roughening transition temperature of a
(D) An aperiodic version of C2 appears at low given crystal surface, growth nuclei on that surface
supercooling, when the solution is not uniform, have to be created statistically. It follows that
This gives rise to a form of densely branched growth of the crystal is very slow until a certain
“spherulite” and is similar to structures which (barrier) supersaturation is reached. Above this
have been observed in the Ge—Si system [18] (fig. level, growth in that direction of the anisotropic
5). crystal becomes very fast (its rate might approach
The addition of small amounts of CuSO4 as an asymptotically a maximum value which is inde-
impurity had great effect on the way in which the pendent of the exact degree of supersaturation).
crystals grow. The minimum amount to have an The concentration gradient at the interface does
effect was about 0.02%, while above 0.2% the not, according to this picture, affect the growth
growth mode changed completely. We shall dis- rate alone; it does so consistently with the rate at
cuss the effect of impurities in more detail in a which supersaturation is achieved and the degree
at which it should be maintained for a given
velocity [11]. Thus the ingredients of an instability
~ ~ ~ ~ ~‘1 of the Mullins—Sekerka type are present, since of
~ course the usual requirements of conservation of
~ ~ matter must still be incorporated into the rate
~ equations. Once growth has been initiated, we
~ 4f ~ suggested that the surface becomes dynamically
I~ rough,
growth inresemble
that interface
those fluctuations
of a rough created
surfacebyeven
the

.~
~ /
the supersaturanon fails below the barrier
level, and only ceases at a lower critical value.
Then the relaxation oscillations induced by this
diode-like behaviour provide a selection mecha-
Fig. 5. Growth of densely branched crystals from a non-uni- nism which is responsible for the quasi-periodic
form fluid. (Type D. The scale bar is 200 pm long.) structure of the dendrites.
642 E. Raz et a!. / New morphologies observed during dendritic growth of ammonium chloride

The second constraint is that the boundary (fast enough growth), curious growth patterns can
conditions in the fluid have to be satisfied at the emerge. Suppose that a dendrite stem grows to-
surfaces of the cell. During crystal growth from wards the cell boundary at an oblique angle. The
solution both latent heat and excess solvent are concentration field ahead of the tip is locally
ejected by the advancing interface. These must diluted by solvent which is rejected by the growing
diffuse away through the fluid. There is a small crystal. When the tip gets close enough to the
component of convection involved too since the boundary, the dilution will also be affected by the
densities of the solid and fluid are unequal [191 field around the image and the growth eventually
but this will be ignored since it does not affect the stops for lack of solute. It should be noticed that
explanation significantly. Thermal convection does the crystal does not have to touch the boundary
not occur because the cell is cooled from below, for this to happen, but only to approach it closely
The heat and solvent diffusion obey Fick’s laws, enough for the image field to have a decisive
subject to approriate boundary conditions at the effect. But if once started, the growth can continue
cell edges. These conditions are: (a) the upper and in only a small degree of supersaturation, the
lower surfaces of the cell are isotherms and (b) distance from the boundary at which it eventually
there is no solvent flow across the boundaries, ceases may be quite small compared with the
Mathematically, on the boundaries T TB = = diffusion length.
constant and v7,,C = 0. Using the principle of Meanwhile, the solvent diffuses away and the
images as in electrostatics [20], it immediately solute concentration at the interface increases,
follows that a dendrite growing near a planar cell fastest where the concentration gradient is highest.
boundary sees, concentration-wise, an identical Let us assume that the preferred growth directions
mirror-image dendrite growing the other side of are the [Ill] group. A new side-branch will emerge
the boundary. Temperature-wise, it sees a negative at a location which has a [111] surface orientation
image dendrite (i.e. for a point on the dendrite and where the supersaturation first reaches the
whose temperature is T, that of the equivalent critical level. It seems that a specific and well-dc-
point on the image is TB T). If we ignore cou-
— fined mechanism (i.e. not noise-generated) must
pling between the temperature and concentration be involved at this stage in order to generate the
fields, two general possibilities exist. If the tern- long-range order observed. It is clear that this will
perature field is dominant in determining the be on the side remote from the boundary, and
growth, then the dendrite and its image “attract”, growth will restart on that side. Since initiation of
because the temperature gradient between them growth is competitive, and occurs in the first
increases as they get closer, and the growth is equivalent direction for which the supersaturation
diverted towards the boundary [21]. If, on the exceeds the barrier, it follows that sidebranches
other hand, the concentration field dominates the will never start in a backward direction when an
growth, the dendrite and its image “repel” and the equivalent forward direction exists, or between
dendrite avoids growth too close to the boundary. two close sterns. Smoothness of the boundaries is
The latter represents the case of NH4C1 in which also relevant since the local orientation of the
the temperature field is very weak. In the normal surface may be a deciding factor in the nucleation
case of growth in an infinite medium, side- competition. Honjo et al. [22] have shown that
branches on two parallel-growing dendrites avoid irregularities of a boundary result in disordered
one another for the same reason. dendrites.
Now, if the crystal growth were isotropic, or This process leads to several possible mor-
approximately so, and if gravity is ignored, it phologies, depending on the direction of growth
would follow that the dendrite would choose to and the orientations with lowest nucleation barrier
grow in the central plane of the cell so as to at the temperature in question. We recall that
remain as far from its images as possible. How- growth in the [111] directions is preferred at the
ever, if the crystal growth is very anisotropic and supersaturations used in our experiments. It is
the diffusion length is less than the cell thickness easy to see by calculating angles that the formal
E. Raz et a!. / New morphologies observed during dendritic growth of ammonium chloride 643

rule for forward branching is that only one of the

two would indicate backward branching, were it


to occur).
A first example shows how zigzag crystals arise.
We consider a situation in which the (111) direc-
indices
tion is parallel
of the orientation
to the boundary
changes
planes,
sign but
(change
the cell
of
thickness is now several times greater than the
diffusion length (fig. 6). The cell surface is normal
to (112). A growing stem approaches the upper
boundary in the direction (111). When its ap-
proach is close enough, it stops growing, and
renucleation starts in the two equivalent orienta-
tions (111) and (ill). The latter dips down to- Fig. 7. A model showing the honeycomb growth in three
wards the lower boundary and on reaching its dimensions. The model was photographed with the sun over-
vicinity, it once again branches, this time to (111) head; W is the wire model and S is its shadow, representing the
and (111), from which point the process repeats projection on the boundary plane.
itself. Now the two branches (111) and (iii) re-
main in the vicinity of the boundaries, or very
slowly recede from them. They each have only one 3. We can illustrate the geometry of this case by a
possible-direction of branching on a side remote model (fig. 7). The boundaries are planes normal
from the boundary. In the former case, this is to (111), or close to this direction [23]. Growth is
(111), but this branch is a dead end since it preferred along the [111] axes again. Suppose that
quickly reaches the lower boundary, where it could a seed starts to grow close to the upper surface in
only branch back to (111), thus forming a branch the downward direction (ill). When it reaches the
parallel and too close to the original (111) to proximity of the base, it will branch into two
nucleate. Thus the original (111) does not appear equivalent directions which nucleate together,
to branch at all (fig. 1) although short branches of (111) and (111). The first of these, on reaching the
this sort would be obscured from vision, proximity of the upper surface, branches into (111)
A second scenario which has been observed and (ill) and the second into (111) and (111). So
gives rise to the honeycomb structure seen in fig. the process continues. However, it should be not-
iced that two opposite directions have been pro-
duced, (111) and (111), and from fig. 3 it can be

~Al,,i,il
(1 1 1) (1,1,1) (1,1,1)
c•i,i,i>
B (1,1,1)A (111) seen
coincide
close.
course,
is
gives
cells
that
Annot
fairly
bilities,
larger that
This
rise also
regular
than at
small
towayin
aalthoughadiffusion
interesting
the
closing
growth space,
later
angle
defects
theand stage
aexactly.
featuresolength
honeycomb
between
front,
thewhich
does
size of
not that
on these
result
ofthe the
network
the cells
two
boundary
macroscopic
seem
thehoneycomb
atin
tothe appear
branches
ishoneycomb
set
develop
individual
the tips up.
scale,
growthto
planes
Of
insta-
(fig.
cells is
3c).
is
This feature was also observed in dense branching
Fig. 6. Geometry of growth giving rise to the zigzag crystal
shown in fig. 1. The diagram shows the projection on the plane morphologies [18] and occurred in the present
of the cell: The points marked A are on the base, and B at the experiments when crystals grew in a more or less
top of the layer. radial manner from a fluctuation in a non-uniform
644 E. Raz et a!. / New morphologies observed during dendritic growth of ammonium chloride

initial fluid (fig. 5). A possible explanation for this asymptotically independent of the degree of su-
phenomenon is that the growth velocity of a crystal persaturation, but at present we do not have mdc-
after the nucleation barrier has been reached is pendent experimental evidence to confirm this.
The observation could also follow from the non-
I I I linear relationship between the velocity and the
diffusion field, which might allow only a discrete
150 — - set of stable velocities.

5. Dimensions of the cells

If we assume that the branches emerge at two


parallel planes, and that the growth directions are
known, it is easy to find a connection between the
50 - - dimensions of the periodic structure and the sep-
aration between the planes d. For the case where
+ the growth directions are [1111,we find the follow-
ing relationship between d and the projections l~,
12 and 13 of three sides of a hexagonal cell:
50 100 150
(1)
d(~m) . ~ (c,+c1±1)2=4/3,
Fig. 8. Illustrating the relationship between the dimensions of
the honeycomb units and the thickness of the experimental
cell. The ordinate shows the value of d~aicobtained by invert- where
ing eq. (1), and the abscissa the diameter of the spacers d. The
line has unit slope. The data include zigzag structures, for ~2 = d2,/(d2 + 12) (2)
which one 1 is infinite.

____
B ............~ 250

100- • -100

B ~ -50

A ....

0— ~ I I I I —0
0 1 2 3
t (sec)
Fig. 9. Positions of the tips of the two branches of zigzag growth as a function of time. Note the hesitation after one branch has
stopped and before the second one starts. One cycle is shown.
E. Raz et a!. / New morphologies observed during dendritic growth of ammonium chloride 645

and c4 c1. By way of contrast, for the case where saturation which occur after the crystal growth has
growth is along [100] ceased for a while (see fig. 9). This results in
3 clearer pictures of the periodic structures. Indeed,
= 1. (3) the unrestricted growth of this mixture often gives
i=1 long dendrites with no side-branches and either
cusped or quasi-paraboloidal tips, depending on
To a first approximation we can take d as being the degree of supersaturation [24]. Maurer et al.
the physical distance between the boundary planes, [25] have seen similar dendrite tips in pure NH4Br
although we should expect the value of d ap- and also attributed them to dynamic roughening.
propriate to eqs. (1)—(3) to be somewhat less by an
amount of the order of the diffusion length (which
determines the distance over which the dendrite
“feels” the image). These formulae are invariant 7. Conclusions
with respect to the orientation of the boundary
planes, and can be applied to cells of any aspect The experiments described in this paper show
ratio or to zigzag growth, which corresponds to the existence of a new family of well-defined pen-
cellular growth with one infinite dimension (i.e., odic morphological structures for an anisotropic
one c, is zero). The results of measurements on crystal growing in a thin, quasi-two-dimensional
many examples of cellular and zigzag structures cell. The morphologies show long range order and
are shown in fig. 8 in which we display the value their growth fronts are marginally stable. It would
of dcaIc, derived from the observed structures by seem very unlikely that a model for dendritic
inverting eq. (1), as a function of d, the distance growth involving noise could explain the long
between the boundaries (i.e. the thickness of the range order [71,and we attribute the very de-
spacers). It will be seen that the points are con- terministic character of the growth to a nucleation
sistent with a line of unit slope for [111] growth model for the development of dendrites which was
(i.e., calculated and measured values agree). recently proposed [9]. This model applies only to
Looked at in detail, we can follow the growth crystals growing when the facet in the preferred
of a single branch crossing the medium. Fig. 9 growth direction is below its equilibrium roughen-
shows the position of the tip of growth as a ing temperature, which explains the strong ani-
function of time for two successive edges in zigzag sotropy observed as well as the apparently limited
growth. First one tip stops growing as it reaches temperature range in which the structures have
the boundary. Then after a while, the branch been seen. The addition of CuSO4 impurities ap-
nucleates, reaches a maximum velocity and then parently inhibits nucleation and therefore empha-
stops at the far boundary. Of significance is the sizes the periodic patterns. Our understanding of
hesitation [8,9] (in even greater detail, the branch the structural and dynamic stability of this new
thickens in a rather isotropic manner during this meta-ordering is still incomplete.
period) before the new branch nucleates.

Acknowledgements
6. Effect of impurities
We acknowledge the technical assistance of Y.
Finally we should remark on the significance of Moses and S. Hoida, and stimulating conversa-
the CuSO4 impurity. Although we have not yet tions with J.S. Langer, S. Fishman, E. Polturak
carried out any systematic experiments on this and T. Viscek. We are grateful to the US—Israel
system, we surmise that the impurity increases the Binational Science Foundation, the German--Israel
nucleation barrier. This has the result of suppress- Foundation, the Fund for Basic Research admin-
ing the creation of sidebranches which only ap- istered by the Israel Academy for the Sciences for
pear under the more extreme conditions of super- support of this research.
646 E. Raz et a!. / New morpho!ogies observed during dendritic growth of aminonium ch!oride

References and footnotes [17] The dimensionless unit of supercooling is I — C~


5/c, where
ceq is the saturated concentration of the solution at the

[1] J.S. Langer, Rev. Mod. Phys. 52 (1980) I~ temperature in question and c is the actual concentration
J.S. Langer, in: Chance and Matter, Proc. Les Houches of the solution.
Summer School, Session XLVI, 1986, Eds. J. Souletie, .1. [181 E. Ben-Jacob, G. Deutscher, P. Garik, ND. Goldenfeld
Vannimenus and R. Stora (North-Holland, Amsterdam, and Y. Lereah. Phys. Rev. Letters 57 (1986) 1903.
1987). [19] GB. McFadden and SR. Coriell, J. Crystal Growth 74
[2] D.A. Kessler. J. Koplik and H. Levine, Advan. Phys. 37 (1986) 507.
(1989) 255. [20] The method of images is usually presented as a method of
[31J.S. Langer, Science 243 (1989) 1150. solving Laplace’s equation. However, it only requires
[4] E. Ben-Jacob and P. Garik, Nature 343 (1990) 523. uniqueness of the solution and definition of static
[5] ME. Glicksman, Ri. Schaefer and J.D. Ayers, Met. Trans. boundary conditions 4~= constant or V.’, = 0 (let us say,
A7 (1976) 1747. on the cell surfaces). The non-linear equations involving
[6] S-K. Chan, H-H. Reimer and M. Kahlweit, J. Crystal the diffusion of solvent and heat and crystal growth are
Growth 32 (1976) 303 deterministic and appear to us to satisfy these criteria.
[7] A. Dougherty, PD. Kaplan and J.P. Gollub, Phys. Rev. [21] Although we have done no experiments in this direction,
Letters 58 (1987) 1652. it is of interest to speculate on the effect that boundaries
[8] H. Honjo, S. Ohta and Y. Sawada, Phys. Rev. Letters 55 would have on growth in a thermally dominated system
(1985) 841. (ice, for example, at temperatures below roughening, which
[9] E. Raz, S.G. Lipson and E. Polturak. Phys. Rev. A40 is probably around 0°C). We should expect growth to
(1989) 1088. occur in the preferred growth direction which remains
[10] W.W. Mullins and R.F. Sekerka, J. AppI. Phys. 35 (1964) closest to the boundary, which would indicate a single
444. growth direction where this is parallel to the boundary, or
[11] M. Ben Amar and Y. Pomeau, Europhys. Letters 2 (1986) a non-periodic network if two such directions are parallel
307. to it. But if no growth direction is close to lying in the
[12] SR. Coriell and R.F. Sekerka, J. Crystal Growth 34 boundary plane. a three -dimensional network hugging the
(1976) 157; boundary will be Set up. and will presumably have a
R. Bowley, B. Caroli, C. Caroli, F. Graner, P. Nozières periodic nature, although we do not see at present what
and B. Roulet, J. Physique 50 (1989) 1377. determines its periodicity.
[131 E. Ben-Jacob, ND. Goldenfeld, J.S. Langer and G. Schhn, [22] H. Honjo, S. Ohta and M. Matsushita, .1. Phys. Soc. Japan
Phys. Rev. A29 (1984) 330. ~ (1986) 2487.
[14] E. Ben-Jacob, P. Garik, T. Mueller and D. Grier, Phys. [23] In this case, Orientations with zero or one negative index
Rev. A38 (1988) 1370. are upward-going.
[15] J. Bechhoefer and A. Libchaber, Phys. Rev. B35 (1987) [24] E. Raz, S.G. Lipson and F. Ben-Jacob. to be published.
1393. [25] J. Maurer, P. Bouissou, B. Perrin and P. Tabeling.
[16] L.R. Morris and W.C. Winegard, J. Crystal Growth 5 Europhys. Letters 8 (1989) 67.
(1969) 361.

You might also like