American Mineralogist, Volume 61, pages 448459, 1976

Crystal structureand crystal growth: I. The influenceof internal structureon morphology

Enrc Dowry
Department of Geological and Geophysical Sciences
Princeton Uniuersity, Princeton, New Jersey 08540

Abstract

A new model of the relationship between crystal structure and morphology has been
developed which uses the symmetry concepts of the Donnay-Harker method, as well as
bonding concepts developed by Buerger and Jackson. For each face, one computes the
templale fraction, which is the fraction of the total bond energy attaching a new layer to the
substrate.The layer is assumedto have thicknessd, the interplanar spacing. New layers will be
nucleated more rapidly on faces with higher template fraction, hence these faces will tend to
grow themselvesout of existence.The bonds defining the template fraction on each face are
those which cross the surface of least bonding, which is considered to be the exposed surface.
Prediction of the exposedsurfaceconfigurationis an important by-product of the new model.
Calcufations are made by computer methods,and in addition to template energy, lhe surface
or cleauage energy and other types of attachment energy may be routinely determined.
Although only approximate bond energieshave been used so far, examplessuggestthat the
new model is more effective than either the Donnay-Harker or Hartman-Perdok (PBC)
methods of rationalizing crystal morphology.

Introduction surface structure of crystals on an atomic level.

The use of external morphology as a clue to the
internal structure of crystals is no longer an active
branch of crystallography,exceptfor determiningthe
presenceor absenceof a center of symmetry (but see
the interesting paper by Schneer, 1970). However,
attempting to determine the influence of internal
structure on crystal morphology is still useful,for the
following reasons:( I ) Any model for predicting mor-
phology logically includes or relies on a model for
crystalgrowth. and thus suchpredictionscan provide
tests for crystal growth theories. (2) The observed
morphology of crystals has been shown by experi-
ment to be influencedby severallactors besidesinter-
nal structure, such as the composition and nature of
the growth medium and the rate of crystallization
( e . g .B u c k l e y , l 9 5 l ; K e r n , 1 9 5 5 ,1 9 6 8 ;K l e b e r , l g 5 7 ) .
The presenceof dislocations is also expectedto in-
fluence the growth rate and thus the development
of various forms. If the influenceof internal strucrure
could be accountedfor, it might be possibleto make
inferencesabout the external factors which were op-
erative during growth of particular geologicalspeci-
mens.(3) An important aspectof morphological and
crystal-growth models in general is the prediction of
448
Knowledge of the atomic structure of crystal faces,
and of the processeswhich take place on them, has
important applications in geology in the study of
sector zoning, the distribution of minor and trace
elements, and other problems involving crystal
growth. Such knowledge is also useful in electronics
and chemical technology, in such fields as semi-
conductors, catalysis, etc. There are experimental
techniqueswhich can give direct information on sur-
face structure (Estrup, 1975), but so far these have
been applied only to structurally simple substances.
This paper describesa new model of the influence
of internal structure on crystal growth and mor-
phology. The model uses the symmetry concepts
developedby Bravais and Donnay and Harker, but
also takes explicit account of the atomic arrangement
and bonding, by means of a quantity called the tem-
plate fraction, which is the fraction of the total bond
energy in a growth layer attaching that layer to the
substrate.Furthermore, the method involves predic-
tion of the atomic configuration on the surface of
each face;it is postulatedthat an exposedsurfacewill
usually be a surface of least bonding. In Part II
(Dowty, 1976) predictions of surface structure are
applied to the problem of sectorzoning in minerals.
 CRYSTAL STRUCTURE AND CRYSTAL GROI4/TH: I
Previous investigations
Most important ideas on the relation betweenin-
ternal structure and morphology have been reviewed
by Buckley (1951) and Hartman (1973).The two
methods which have been used most often to try to
predict or rationalize crystal morphology are the
reticular density method of Bravais, as extendedby
Donnay and Harker (referredto as the DH method)
and the periodic-bond-chain(PBC) method of Hart-
man and Perdok.
Donnay-Harker method
This method extendspreviouswork by Bravaisand
Friedel. According to the "Law of Bravais," the im-
portance of a face is proportional to the reticular
density,or concentrationof lattice points on the face
(including those due to centeringin nonprimitive lat-
tices). Ranking the facesin decreasingorder of their
reticular density is the same as ranking them in in-
creasingorder of their reticular area, or decreasing
order of their interplanar spacingd, as usedin X-ray
crystallography. Donnay and Harker (1937) pro-
posed modifications taking account of the space-
group symmetry, so that all equipoints, instead of
just lattice points, are considered.Thus, in the calcu-
lation of d, only those Miller indices are used which
satisfy the requirementsfor nonextinction of X-ray
reflections, as given in Volume I of the International
Tables for X-ray Crystallography. A physical inter-
pretation of the DH law was given by Buerger(1947;
see below). The implications of pseudosymmetry
have been discussedby Donnay and Donnay (1961).
The DH method is easy to carry out, and in its
purest form (neglecting pseudosymmetry) is com-
pletely objective.However, it can only be an approxi-
mation becauseit takes no direct account of atomic
arrangementor bonding. Also, it doesnot distinguish
betweengrowth morphology and cleavage,and it is
not applicableto the prediction ofsurface structure.
PBC method
The periodic-bond-chainmethod of Hartman and
P e r d o k( 1 9 5 5 a , b , cH; a r t m a n , 1 9 7 3 )e x t e n d st h e i d e a s
about strong bonding directionsdevelopedby Niggli
(1920;1926a)and Parker (1923, 1930),and also uses
the concept of attachment energy developed by Kos-
s e l ( 1 9 2 7 )a n d S t r a n s k i( 1 9 2 8 ) .
One first identifies infinite chains of atoms con-
nected by relatively strong bonds only, i.e. the peri-
odic bond chains. An interestingempirical approach
to the location of PBC's has been describedby Wolff
and Gualtieri (1962).The most important facesmor-
phologically are supposedto be the flat or F faces,
which contain two or more PBC's. Stepped or S
faces,containing only one PBC, occur lessoften, and
kinked or K faces,containing no PBC's, occur only
rarely or not at all. For the F faces,which are pre-
sumed to grow relatively slowly by means of two-
dimensional growth layers or steps,the location of
the PBC's definesthe surfaceof the growing face,and
one may identify a slice, whose thicknessis the min-
imum thicknessof the growth step. This thicknessis
d, the crystallographicrepeat distanceperpendicular
to the face,which is determinedby the DH rules.The
-F faces are ranked relative to each other by calcu-
lating the attachment energy across the interface be-
tween the establishedface and the newly-addedslice;
this is the averagebond energyholding atoms to the
surface. S and K faces are supposedto grow more
rapidly than .F faces becauseindividual atoms are
presumablyattachedto them by more than one bond;
thus they presentready-madenichesor kinks for the
building units to fit into, and supposedly do not
require the construction of growth stepsor slices.
The PBC method not only takesexplicit account of
atomic arrangementand bonding, but also predicts
surface structure. However, it is also an approxima-
tion, and has severaltheoretical and practical short-
comings: ( I ) The chains are supposed to have
stoichiometric composition, which in crystals with
even moderately complex structures leads to com-
plicated chains with various kinks and excrescences.
As the complexity of the chains increases,and the
directions of the bond vectors within them become
more variable, the relations which originally led to
the division of the faces into F, S, and K types are
o b s c u r e d ,a n d i t i s n o l o n g e r a m a t t e r o f i n d i v i d u a l
atoms fitting into locations with greater or lesser
numbers of bonds. (2) For even moderatelycomplex
structures,the PBC method can be extremelydifficult
to apply, requiring the consultation of projections
and structure models. lt is difficult to avoid sub-
jectivity and error in the classificationof a given face
as an F, S, or K face, as shown by the example of
barite (below). (3) If the PBC method is correct; it is
difficult to seewhy the DH method should be success-
ful in so many cases(severalexamplesare given be-
low). The PBC method does not systematicallycon-
sider the symmetry of the crystal, exceptto determine
slicethickness,and to predict the relativeimportance
of S facesin a single zone (Hartman, 1958).
 ERIC DOWTY

Crystal growth models-the surface of least bonding
and the template fraction
When dealing with the growth of crystals with
complex structures, it is expedient to consider new
material as being added in the form of blocks, rather
than individual atoms; some modern kinetic models
of crystal growth use this approach (see review by
Bennemaand Gilmer, 1973).Except for the simplest
facesof the simplestcrystals,theseblocks must usu-
ally be regarded as a mathematical fiction. In this
paper, such blocks will be taken to be the smallest
units which can be repeatedby translationparallel to
a given face, to give a layer with the stoichiometric
composition of the crystal. The minimum height of
such a layer on a face is d, as computed by the DH
method, since this is the minimum repeat distance
perpendicularto the face. The two minimum repeat
distancesin the plane of the face will be the vectors
betweenlattice points, which also define the Bravais
reticular area. Thus a block has thicknessd and area
of attachment to the substrateequal to the Bravais
reticular area (neglecting any microscopic con-
volutions of the actual surface).The blocks thus de-
fined will have different shapeson different forms. If
the unit cell is primitive and no screw axes or glide
planesare present,the blocks will have the volume of
the unit cell. Otherwise,they will have volume l/n of
the unit cell when the DH rules call for using the
multiple indicesn(hkl) to determined.
It seemsthat the rapidity of growth on a given face
dependslargely on the rapidity with which the block
for that face attachesitself to the surface,which has
generally been considered to be dependent on the
energy of attachment. On slowly-growing faces the
block will usually fit into the "half-crystal" configu-
ration (Kossell, 1927) as in position I in Figure la.
Becausethree surfacesare available for attachment,
the attachment energy will be high and the block
rapidly and firmly attached at roughly the same rate
for all faces.However, if the growth is by two-dimen-
sional nucleation,the stageswhich will largely deter-
mine the relative growth rates of faces will be the
starting of new layers,as in position 2, Figure I a, and
secondarily the starting of new rows (position 3),
becausethesestagesare necessarilyslower (Buerger,
1947).Thus the critical factor would seem to be the
energy with which the block is bonded directly to the
face or substrate.The situation is somewhatdifferent
if spiral or other dislocationsare present,but again,
the attachmentenergy directly to the substrateseems
to be critical (Bennemaand Gilmer, 1973).An actual
two-dimensionalnucleus in most casesprobably has
a volume greater than an individual block, but it is a
reasonable hypotheses that the number of blocks
which go Io make up a nucleus will be qbout the same
on all faces.
Becausethe blocks are not, in general,actual dis-
crete units, some additional concepts are required.
First, we need to specify the particular level in the
crystal structure acrosswhich the attachment of new
material takes place.This is equivalentto specifying
the atomic structure of the exposed surface of the
face, if we neglect temporary surfacedistortion and
adsorption. It is a fundamental postulate of this pa-
per that such an exposed surface will be a surface of
least bonding,which is that surfaceparallel to a given
face acrosswhich the total potential bond energyis at
a minimum. Such a surfaceis not necessarilystrictly
planar but must possessthe translational symmetry
of the lattice.
Second,it might be askedwhy we consideronly the
energyof attachmentto the preexistingsurface,when
the entire block may have to bq built up from atoms
or molecules.The answer is that the rapidity with
which the blocks or nuclei are built up will dependon
the presenceof an appropriate template. The pre-

F t c . I G r o w t h o n d i f f e r e n t t y p e s o ff a c e s ( s c h e m a t i c ) . ( a ) O n t h e f l a t f a c e , t h e t e m p l a t e f r a c t i o n f o r t h e
main face is much smaller than that for the secondary faces a and b. The side face a has larger template
fraction than b, yielding rows approximately perpendicular to c. (b) On the stepped face, the template
fraction ofthe main face c is in between that ofa and b, leading to the starting ofnew rows before layers
are completed. (c) On the kinked face, the template fraction of the main face is larqer than either a or b.
 CRYSTAL STRUCTURE AND CRYSTAL GROWTH: I
existing face or substrate will act as a template for a
new block, and the greater the fraction of the total
bonds in the block which are directedto the template,
the more rapidly the block will be assembled.The
template.fraction is therefore defined as thefraction of
the total bond energy in a growth layer or slice which
attaches that layer to the substrate.This is somewhat
similar in practiceto the anisotropy factor of Jackson
(1958)and others,although the anisotropy factor was
definedin a more limited context. If we consideronly
layers of thickness d, the template fraction is the
fraction of the total bond energy in a single block
holding that block to the substrate,acrossthe surface
of least bonding.
The larger the template fraction for a particular
face,the more rapid growth will be on that face,and
the ranking of the forms accordingto their increasing
template fractions is thus expected to approximate
their decreasingimportance as growth forms. It may
easily be seen, as explained by Buerger (1947), how
such a model reducesto that of Donnay and Harker;
if it is assumedthat the bond energy per unit area is
approximately the same for all faces, the template
fraction is directly proportional to the reticular area.
This model actually follows Buerger'sinterpretation
of the DH method in some respects,but by consid-
ering the templating properties of the substrate,
rather than the attachment energy of molecules,the
model is hopefully made applicableto ionic and other
nonmolecular structures. Although for simplicity
only layers of height d or a multiple of d have been
consideredin this paper, it may be that layers some-
times have smaller thickness.This seemsparticularly
likely when there exists a pseudospacingthat is a
submultiple of the true interplanar spacing. By con-
sidering such factors, it may be possibleto account
for the effectsof pseudosymmetryon morphology, as
discussed by Donnay and Donnay (1961) for ex-
ample.
Problems arise if some of the bonds are already
formed in the growth medium. This may not only
change the template fraction, which is properly
speaking only concerned with the bonds that are
formed during crystallization,but may also influence
the surface configuration, and thus the surface of
least bonding. In principle, such prestructuringmay
be dealt with, if the structureof the growth medium is
known, as discussedby Hartman and Perdok (1955b);
preformed bonds are simply ignored when com-
puting the template fraction, and they are not allowed
to be intersected by the surface of least bonding.
In practice, great difficulties may be encountered
451
if there is much prestructuring, and in the extreme
case of crystals composedsolely of large molecules,
the template fraction method would be completely
inappropriate;one might use in this casean approach
based more closely on Buerger's(1947) analysis.
Only the faces or forms near the head of the tem-
plate fraction list are expected to grow by the layer
mechanism,and thus only thesewill be identified on
mature crystals.These are the F facesof Burton and
Cabrera (1949) and Hartman and Perdok (1955), or
the "smooth" faces of Jackson (1958). The more-
rapidly-growing rough facesof Jackson or the ^Sand
K faces of Hartman and Perdok will either grow
themselvesout of existenceor be so poorly definedas
to escapeidentification as discrete faces. One could
redefine S and K facesstructurally on the basis of the
template fractions of possiblesub-faces(side facesof
growth stepsor units) versusthat of the main face, as
illustrated in Figure l. However, according to the
widely-acceptedmodel of Jackson (1958), the actual
mode of growth depends not only on the template
fraction (anisotropy factor in his equations)but also
the absolutevalue of bonding energy and the temper-
ature. Thus it is not possibleto establisha priori clt-
off values of template fraction for classifying faces as
F or smooth versus S, K or rough.
H a r t m a n a n d P e r d o k ( 1 9 5 5 a , b , cH
; artman, 1973)
also use a type of attachment energy to predict the
relative importance of F faces,but they compute it in
a different way. Instead of taking the total of bond
strengthscrossing the surface of least bonding on a
block basis,they averagethe energy on an atomic or
molecular basis, using only the atoms in the slice of
thickness d, and neglecting the atoms which do not
bond directly to the substrate.The reasonfor leaving
the nonbonding atoms out of the averaging is not
clear, sincetheselatter atoms are part of the structure
as well as the ones which are included in the averag-
ing. At any rate, many examples (some are given
below) have shown that the averagedper-atom or
per-molecule attachment energy does not give as
good an account of the morphology as the total per-
block template fraction.
Predicting cleavage
As assumedby others (e.g. Wells, 1946),a reason-
able hypothesisis that ifcleavagesare developedin a
crystal, they will be those planes across which the
bonding energy per unit area or the cleauageenergy is
at a minimum. This cleavageenergyis identicalto the
surfaceenergy as defined by Born (1923). Naturally,
the cleavagewill occur on a surfaceofleast bonding.
 ERIC DOWTY
The relative ranking of the faces on the basis of
cleavageenergy is in general not the same as the
ranking on the basis of template energy.Thus it may
be seenhow the cleavagesof a crystal may not always
be the most important growth faces. However, the
faces with lowest cleavage energy tend to be those
with low indices,and are usually near the head of the
morphological list.
The prediction of cleavageprovides an important
check on the method of calculation, because the
cleavagesof a crystal are generally well known, and
they may be presumed to depend on internal struc-
ture, but not on growth variables such as composi-
tion of the medium or rate of growth. It will be
assumedthat a cleavageis correctly predicted if the
plane in question has lower cleavageenergy than all
other planes which are not cleavages.
The equilibrium shapeof a crystal
Gibbs ( 1878;in his collectedworks. l96l ) defineda
specific surface free energy for the faces of crystals,
and proposedthat the crystalwould have equilibrium
shape when the total free energy is minimized, the
contribution of each face being its specificfree energy
m u l t i p l i e d b y i t s a r e a .C u r i e ( 1 8 8 5 )m a d e a n e q u i v a -
lent formulation, using the capillary constant instead
of free energy.Wulff (1901) assumedthat the growth
velocity of a face would be proportional to the cap-
illary constant. Born and Stern (1919) equated the
capillary constant with the surfaceenergy as defined
by Born (1923), which is the same as the cleavage
energy defined in this paper.
Most later work has tended to discount the effect
of surfacefree energy on growth forms of large crys-
tals for severalreasons(e.g. Buckley, 195l, Ch. 3,4).
Indeed, the surfaceor cleavageenergy as measured
by the methods of this paper is not at all a good
predictor of growth morphology. A few faces with
relatively low indices usually have low surface
energies, but otherwise the values tend to group
about some mean value of bonding density for the
crystal, and do not increasegreatly with increasing
indices.In some simple crystalswith high symmetry,
such as NaCl, it is possibleto identify only one or a
few low-index forms which constitutethe equilibrium
shape, by means of the Wulff construction (Wulff,
l90l). However, when the structure is complex and
the symmetry is low, it may be necessaryto consider
forms with rather high indices. The Gibbs-Curie
equilibrium shape of such crystals may be rather
complex, perhapsin somecasesapproachinga sphere
with only a few small faces.
It would be a misconception to assume that the
typical growth shape of a crystal is the thermody-
namic equilibrium shape,or that under ideal condi-
tions (very slow growth) the crystal tendstoward this
shape.The growth shape of crystals above a certain
size is evidently dominated by kinetic factors. It ap-
pears that large faces are developed during growth
becausethey can become at times distinctly out of
equilibrium with the growth medium, due to their
proportionately larger kinetic barriers. That is, with
respect to the faces which disappear, the larger faces
grow at a slower rate at a given degree of super-
saturation (although the relative growth rates of the
different faces may change with different degreesof
supersaturation).Thus in a thermodynamic senseit
might be more fitting to call the typical growth shape
of most crystalsthe "anti-equilibrium shape."
Computation methods
Computerized methods have been developed for
locating the surface of least bonding, and for eval-
uating template fractions, cleavage energies and
Hartman-Perdok attachmentenergies.At presentthe
evaluation of these quantities is only approximate,
but the computing method is rapid, and provides an
essentially objective approach to predicting mor-
phology. The procedure is convenientlydivided into
three steps,and three separateFORTRAN computer
programs are currently used for the complete oper-
ation.l
Step I (Program MORPH)
The reticular area for each possible face is com-
puted according to the DH rules, and the faces are
ranked in increasing order of reticular area. This
gives the predicted DH morphological aspect(Don-
nay and Harker, 1937) for the crystal, and servesto
identify objectivelythe most likely faceson which to
carry out more detailed work.
Step 2 (Program BOND)
A complete and unique set of interatomic bonds
for the unit cell is identified, and a bond strength is
assignedto each one. The procedure for identifying
bonds and assigningstrengthsis basically that of a
normal bond-length computing routine. At present,
only bonds between first-nearest neighbor atoms,
cafled hereafterfirst-order bonds, are routinely used;
in the case of ionic structures, only cation-anion
bonds are used.It is not difficult. however.to include
' Copies of these programs are available from
the author
 CRYSTAL STRUCTURE AND CRYSTAL GROIATH: I
a limited number of higher-order bonds. A bond
table or matrix is constructed,with entriesfor specific
atoms, usually cations and anions. The number of
atoms used as entries for the table is at least as great
as that in the unit cell, and usually greater.If a bond
existsbetweentwo atoms, the correspondingelement
in the table is the bond energy;if no bond exists,that
element is zero.
The assignmentof bond strengths or energiesrs
critical in complex crystals,and severalmethodshave
been tried. For ionic crystals, one can use simple
point-chargecoulomb energies,as did Hartman and
Perdok (1955c) and others. However, use of formal
valencesas chargeshas been found to be unsatisfac-
tory for silicates,giving too little strengthto the Si-O
bonds for example. A different method is presently
used for complex oxides and silicatesalthough this
too has been found to be inadequatein some cases
(seePart II; Dowty, 1976).The lattice energy of the
simple oxide of each cation is calculatedfrom stand-
ard thermodynamic data (Kubaschewskiand Evans
1967;Moore, 1970)using the Born-Haber cycle, and
bond energiesare computed assumingthat all lattice
energyis in the first-orderbonds. It is then considered
that oxygen has the samechargein all the oxides(the
exact value is irrelevant, since oxygen is usually the
principal anion in the crystals of interest, and only
relative values are needed),and a relative "effective
valence" or charge for each cation is derived. Hydrox-
yls and halogens, where they occur, can be assumed
to have half the charge of oxygen. These chargesare
used in coulomb calculationsfor the crystal of inter-
est. This method of approximation hopefully takes
account of both a "normal" contribution from cov-
alent resonance,and those bond strength variations
in the crystal of interest which are reflected in the
interatomic distances.Some "effective valences"are
given in Table l.
Step 3 (Program PLANE)
This step identifiesthe surfaceof leastbonding and
measurestotal bond strensthsfor each rational face
Test-s I . Effective valencesof some ions
( n o r m a l i z e dt o S i ' + : 4 . 0 0 )
4+
Si 4.00 UA 1 <t
4+ )+
Ti' 3.r7 IE 1.50
?+ )+
UT a 1L
Mg-' 7.42
?+
+
IE 2.24 0.84
1+ ?
2.22 N 0.83
(b)
(o)
Flc. 2. Two dimensional illustration of the intersection of
several unit cells by a rational plane; (a) the real case; (b)
mathematically equivalent case of several planes intersecting a
singleunit cell.
selectedin Step l. The height of each atom from a
plane or planes parallel to the rational face is com-
puted, and the atoms are then ranked in order of this
height coordinate. Except for faces with the simple
indices(100), (010), (001),etc., a given rational plane
intersectsseveralunit cellsat differentlevels(Fig. 2a).
Instead of considering severalunit cells, however, a
mathematically equivalent procedure is to consider
severalequidistant planes cutting the same unit cell
(Fig. 2b), which is in practice much simpler. The
number of planesis at leastas large as the sum of the
indices. The height coordinate of each atom must
then be determinedwith respectto each plane, and a
given atom appearsas many times in the final sorted
list as there are separateplanes.Fortunately, it is not
necessaryto take account ofcell parametersor actual
interatomic distancesin this step; the atomic loca-
tions may be specifiedby their fractional coordinates
only, and the cell may be consideredto be a cube with
edge length one.
ln order to find the surfaceofleast bonding, the set
of parallel planesis in effectcausedto move through
the structure atom by atom, and the total bond
strength which crossesthe entire set is computed at
each step. They are moved through a distanceequal
to their separation,which would be the interplanar
spacingd if computed in the actual unit cell. This is
done in practice by considering successivelythe
atoms in the master sorted list, then going through
the bond table and identifying all the atoms which lie
above the given atom and are bonded to it or any
other atoms with the same or smaller height coordi-
nates. The location of the set of planes which gives
the minimum total bond energy is specifiedin terms
of the atoms above and below the planes.
The template fraction, the cleavageenergy and the
Hartman-Perdok (HP) attachment energy are rou-
tinely computed on the same pass through the unit
cell. Minimizing the HP attachment energy some-
454
times givesa different location for the surfaceof least
bonding than the template fraction and cleavageen-
ergy (which necessarilygive the samelocation). There
is an ambiguity in computing the template fraction in
some cases.If we consider that actual growth layers
can have thickness greater than d, the number of
bonds attaching the layer to the substrate may be
greater than if we consider the layers to have thick-
ness d only; this is illustrated in Figure 3. A given
bond in the unit cell may be consideredonly once in
computing the template fraction for a slice of thick-
nessd (Fig. 3a), but may be counted severaltimes for
a layer of indefinitethickness(Fig. 3b). In view of the
fact that some growth layers have actually been ob-
served by light optics, the template fractions for the
examples given in this paper have been computed
assuming a layer of indefinite thickness. The diffe-
rence only appearswhen d is smaller than the bond
lengths, and thus high-index faces are usually the
only ones affected.It still seemsreasonablethat, bar-
ring pseudosymmetry,the actual thicknessof layers
parallel to different faces will be at least roughly
proportional to d. The cleavageenergy is necessarily
computed using a// possiblebonds as in Figure 3b.
Limitations of the method
Two important limitations of this method should
be mentioned at the outset. First, the bounding surface
is restrictedto a plane. This limitation is lessserious
than might be supposed,sincethe atom locationsare
specifiedin terms of their centers,and a plane divid-
ing atoms by means of their centersmay correspond
to a quite undulatory surface with respect to the
atoms as a whole. However, errors can be expected,
especially in dealing with molecules or compound
ions. The PBC method does not in principle have this
limitation; but note that in the example of barite
(below) the resultsare identical.Some other cases,in
which the approximation fails, are discussedin Part
(o)
Ftc. 3. Two-dimensional illustration of computation of template, attachment or surface energies.(a) A
layer of thickness d only is considered. (b) A layer of indefinite thickness is considered.A single reticular
unit is shown by the heavy line. In case(a) one bond crossesthis area and in case(b) four bonds cross the
area.
ERIC DOWTY
II (Dowty, 1976). lt has been found necessaryto
consult crystal structure projections in some casesto
check for non-planarity of the surfaceof least bond-
ing. A fourth FORTRAN computer program
(PROJ) has been written to plot to scale the projec-
tion of the atoms and bonds determined in Step 2.
The projection direction can"be made parallel to the
zone axis defined by any two faces.
The secondlimitation is that only first-orderbonds
are taken into account, whereasin ionic crystalselec-
trostatic interactionsmay be expectedto operateover
fairly great distances;lattice energycalculationsgen-
erally do not converge unless many unit cells are
included in the sums, This probably will not lead to
mistakesin the location of the plane of leastbonding
for a given face very often, but will certainly cause
errors in the quantitative evaluation of relative tem-
plate flractionsand attachment and surface energies
for different faces.
The PBC method also uses a first-order bonding
approximation for locatingthe exposedsurfaceof the
crystal on F faces.Once this is found, however,there
is sometimes a convenient way of approximating
complete electrostaticpotentials,using the chains as
units, rather than individual atoms (see Hartman,
1956).Unfortunately, this method is applicablewith
easeonly to very simple chains,and quickly becomes
impracticableas the structural complexity increases.
For the presentmethod, an atom-by-atom direct sum
of electrostaticinteractionsor some similar approxi-
mation would probably be most suitable. Such a
computation is not in principle difficult, using ma-
chine methods, but would require a rather different
procedure from that described above, and consid-
erably more computing time. In effect,one complete
lattice energy sum would have to be carried out for
each face, even if the location of the surfaceof least
bonding were initially determined by the first-order
approximation.
(b)
 CRYSTAL STRUCTURE AND CRYSTAL GROWTH: I 455

Examples Tnglp 2. Predicted morphology of NaCl

Structural data for the following examples have Forn ,2

been obtained from Wyckoff ( 1968)where not other-
wise noted. It is naturally almost impossibleto sepa- r) 111 3 0 . 5 0( 2 ) I.73 3.0
rate completelythe influenceof externalfactors from 2) 200 4 0.33(r) 1.00 1.0
3) 220 8 0.67(3) 1.41 2.0
that of internal structure on the morphology of indi- 4) 311 11 0.83(4) 1.51 2.0
vidual specimens.In the absenceof data from care- s) 133 19 1 . 1 7( 6 ) 1.61 2.5

fully controlled laboratory growth, it is still possible 6) 420 20 1.00(s) 1.34 2.O
to eliminate much of the influenceof external factors 7) 422 24 1 . 3 3( 8 ) 1.63 2.0
8) 51r 27 r . L 7( 1 ) 1.35 3.0
on the morphology of minerals by using the per- e) 531 35 1.s0(10) L.52 2.5
sistenceuuluesof Niggli (1920).Niggli (1926a,b)and r0) 244 36 L . 6 7( r 2 ) L.67 3.0
Parker (1930) tabulate severalpersistencevalue stud-
-
i e s ,a n d N i g g l i ( 1 9 2 6 b )g i v e ss u b j e c t i v ee v a l u a t i o n so f Forn Ranked in Donnay-Harker order, with the
, indices as used by this method.
the morphology of all important minerals.The crys- S' - Square of the relative reticular area.
t a l d r a w i n g so f G o l d s c h m i d t( 1 9 2 3 )m a y a l s o b e u s e d T - Template fraction, with ranking in parentheses
CL - Cleavage energy, divided by that of the first'
to evaluate morphology qualitatively, and to derive ranked fom.
persistencevalues if necessary. HP' - Hartman-Perdok attachment energy, as average
number of first-order Na-Cl bonds per atom.

NqCl structure
B u c k l e y ( 1 9 5 1 ) s u m m a r i z e st h e r e s u l t s o f m a n y the template fractions and cleavageenergiesin Table
experiments on NaCl itself. The forms disappear 2 are approximations, since they consider only first-
(1919) found that the
from a growing spherical seed in the order {ll0}, order bonds. Born and Stern
surfaceenergy of {l l0} is 2.706 times that of {100}on
{ 2 1 0 } ,{ l l 1 } , a n d t h e c u b e { 1 0 0 }i s n o r m a l l y t h e o n l y
a calculation.This
form remaining after long growth. These results ap- the basis of completeelectrostatic
differs considerably from the value 1/2 obtained
ply to a number of methods of growth, and do not
first-order approximation; complete elec-
seemto be reflecting the effectsof adsorption, accord- from the
i n g t o B u c k l e y . N i g g l i ' s ( 1 9 2 6 b )r a n k i n g . o f t h e i m - trostatic template and surfaceenergycalculationsfor
portance of the forms in natural crystalsof halite is all faces should be carried out.
first { 100}, usually the only form present,then { I 1I }, Some oxides with the NaCl structure (e.g. MgO,
MnO, CdO) commonly crystallize in octahedra al-
{l l0}, and {210}.
The unmodified DH rules incorrectly predict that though they have cubic cleavage. Pr€sumably this is
t h e o c t a h e d r o n{ l l l i s h o u l d b e t h e m o s t i m p o r t a n t attributable to adsorption of foreign atoms, or to
prestructuring in the growth medium (e.9. Hartman
form; the forms are listed in DH order in Table 2.
Donnay and Harker (1937) suggestedthat if the Na a n d P e r d o k , 1 9 5 5 b ) .
and Cl atoms are considered to be the same, the For several other.simple crystals with face-centered
lattice becomesprimitive rather than face-centered, lattices, such as fluorite, sphalerite, and diamond,
and the first form is {100}. However, the secondpre- the methods of this paper are not very successful.For
dicted form for a primitive cubic lattic is {ll0} and example, the first-order bonds in sphalerite are the
n o t { l l 1 } . A c c o r d i n gt o t h e P B C m e t h o d , { 1 0 0 }i s t h e same as those in diamond, but the cleavageis {l l0} in
o n l y F f o r m ; { l l 0 } a n d { 2 1 0 }a r e S f o r m s ; a n d { l I I } i s sphalerite and {1ll} in diamond. As pointed out by
Hartman (1959b), the differenceapparently must be
a K form, since the only PBC's in the structure are
p a r a l l e l t o [ 0 0 ] , [ 0 1 0 ] ,a n d [ 0 0 1 ] .T h e a t o m - b y - a t o m attributed to the partial ionic character of bonds in
treatment of Stranski (1928) apparently demon- sphalerite. These crystals will be discussed in detail
when more complete electrostatic computations of
stratedthat {l I I } is lessstablethan {l l0} and presum-
ably would be less important morphologically. The bond energieshave been carried out.
template energymethod ranks the forms in the order
Barite
{ 1 0 0 } ,{ l l l } , { 1 1 0 } ,{ 1 1 3 } ,{ 2 1 0 } .T h i s i s n o t p e r f e c t ,
especiallyas regards the form {l13} which is never This is one of the original examplesgiven by Hart-
observed,but it is in better agreementwith the obser- man and Perdok (1955c) for the PBC method. A
vations than any method previously proposed. The comparison of results for the various methods is
cleavageis correctly predictedto be {100}.Of course, given in Table 3. It was assumed,following Hartman
 ERIC DOWTY

T,t sI-s 3. Observed and predicted morphology of barite importanterrors.Note that someof the formsclassi-
fied as S forms by Hartman and Perdokhave lower
Form DH HP HP'
HP attachmentenergiesthan some of the F forms,
1) 001* 6 0.388F 0.39 0.4r(2) 1.03(2) which seemsto be contraryto the postulates of the
2) 2r0* 7 .396F .39 .42(4) 1 . 0 0( r ) Hartman-Perdok classifi
cation.
3) 101 r .435F .43 .29(r) r.L2(4)
4) 011 3 .945K .53 .43(5) L.29 Although Hartman and Perdok(1955c)originally
5) oro* lL .))/r .)) . s 9( 1 0 ) r . 1 2( 5 ) classifiedthe form {01l} asan .Fform, theycomputed
the high attachmentenergyvalueof 0.945,and sub-
6) zl.r 9 .472F .47 .s0(9) 1.08(3) sequentlyreclassified it as a K form, sinceaccording
7) 100 2 .521F .51 .4r(3) t.27
8) 410 26
to them it is made up of parts of adjacentF faces.
s .50 .81(19) 1.16
9) ro2 8 s .46 .49(7) 1.1s Theyattributedthe anomalously highpositionof this
r0) r11 4 s .52 .46(6) L.26 form in the observedlist to retardationof growth by
adsorptionof foreignions.However,the presentcal-
1r) 2L3 28 S .64 .9r(26) L.2s culationsrevealthat they did not computethe attach-
L2) 230 38 S
13) 2o1 5 .s3 . 4 9( 8 )
ment energyacrossa surfaceof leastbonding.The
S 1.30
L4) 2r4 47 s attachmentenergycomputedby their method across
15) 2L2 13 S .t, . rr-<rrl r-.22 the surfaceof least bonding is 0.53, which is not
particularlyanomalous,and the form is certainlynot
16) 3r0 81 s out of placein the templatefraction list; it is pre-
17) 110 16 s .52 .71(16) 1.15 dicted to be fifth, and its real importanceis fourth.
The location of the two differentplanesis shownin
Form - Lj-sted in order of the persistence values of
Braun (1932). Cleavage denoted by *. the projectionof Figure4.
DH - Ranking accordlng to the unmodified Donnay-
Harker law. Rutile
HP - Attachment energy and face type of llartman
and Perdok (1955c). The DH methodranks {100}first (Table4) with
HP' - Attachment energy computed in this study by
method of Hartman and Perdok,
slightly over half the reticular areaof the next form,
T - Tenplale fraction with ranking in parentheses. {l l0}. However,the cleavage is {l l0}, andthisform is
-
CL Cleavage energy divided by that of the first- probablymoreimportantas a growthform.A recent
ranked form, wit.h rankj-ng in parentheses.
PBC treatment (Woensdregtand Felius, cited by
Hartman,1973)classifies only {ll0} and {l0l} as F
and Perdok, that the sulfate ion can be representedas forms,although{100}and {1ll} are alsoimportant
a point charge, and the formal valences were used. growth forms (Niggli, 1926b).Note alsothat the HP
Both the PBC method and the template energy attachmentenergyin Table 4 for {l0l} ranks this
method seem to improve on the basic DH results. form in sixthplace.The form l22ll is anomalously in
Donnay and Donnay (1962) have also made im- third placein the HP attachmentenergylist.
proved predictions of the morphology of barite, by The first-ordertemplatefraction givesequal rank
considering only the centers of charges, rather than to {l l0} and {100},but by virtueofits largerreticular
individual ions. Two of the three cleavagesare cor- area,\l l0) hasby a considerable marginthe smallest
rectly predicted by the cleavage energy; the third is cleavageenergy.This would certainly be the ther-
ranked fifth in the list. Of course with respectto both modynamicequilibriumprismform, but the equilib-
cleavage and morphology, too perfect an agreement rium terminatingforms are hard to predict.
should not be expected, since sulfate ions were re-
garded as point chargesand only first-order bonds The following examplesare mineralswhich show
were used. Note the lack of large differences in the sectorzoning,as discussed in Part II (Dowty, 1976).
cleavageor surface energies;no forms as far down as As might be expected,in theserelatively complex
thirtieth in the list have relative cleavage energies structuresthe templatefraction method often gives
higher than 1.4. Except for one form, {0ll}, the first- results which are very similar to those of the DH
order attachment energiescomputed by the method method.
of Hartman and Perdok have been reproduced by the
present method (the small differences are due to Diopside-augite
round-offerror). This shows that the use of planesin Structuraldata (for diopside)wereobtainedfrom
the present method probably has not caused any Clark et al. (1969).The DH methodranksthe forms
 CRYSTAL STRUCTURE AND CRYSTAL GROI(TH: I 457

B A R T T E[ r O O ]P R O J E C T T O N fied as F' facesby Hartman (1959c),usingthe PBC

Bo- o S04-o method.The next most important forms, although
usuallysmall,ares {2ttt} and x {5161}(indicesare
for a morphologically right-handed crystal). The
form {21Il} takesfifth placein the DH list and in the
templateenergylist ranks equivalentlywith several
HP other forms after the first three;m, r,and z havejust
SLB half the templatefractionof this next group. How-
(oil ) ever{5l6li is an anomalyby any methodof predic-
tion; evidentlyits prominenceis attributableto retar-
dation of growth becauseof adsorptionof foreign
ions or molecules(Hartman, 1959a).Indeed,syn-
thetic crystalsshow high concentrationsof foreign
Frc. 4. Projectiondown [00] of the structureof barite. The ionsin {5161}sectors (e.g.Cohen,1960).
(011) surfacechosenby Hartman and Perdok(1955c)is denoted
HP; the surfaceof leastbondingis denotedSLB. Andalusite
The DH and templateenergymethodsare not
of diopsideor augitein the order{ll0}, {100},{010}, completelysuccessfulin predicting the morphology
{1li}. The templatefractionmethodranksthem in of andalusite. The most importantforms by far are
theorder{l l0}, {010},{l I I}, {100},andcorrectlyranks pre-
ill0), {001},and {011},with {101}occasionally
{l l0} firstin thecleavage list.The four formslisteddo sent(Niggli, 1926b).The DH methodsranks {ll0},
in fact appear to be the most important forms on
{011},and {101}asthe first threeforms,but {001}is in
augite,althoughNiggli (1926b)alsomentions{130}. eleventhplace.The templatefraction method ranks
Exceptfor this face,which is rankedfourteenth,both the formsin the order{l0l}, {l l0}, {111},with {001}
the DH and template energy methods agree
reasonablywell with the order of the forms as listed
Trsr-B 4. Observed and predicted morphology of rutile
by Niggli down to about fifteenthplace.The first five
forms in the ranking accordingto the HP attachment Form DII HP
energyare {010},l227L{Jl0}, {021}and {10i},in very
poor agreementwith the observedmorphology. 1-rro* 2 0.16(1) r-.oo(1) 4.03(1)
( 0.16 2)4.03(2) 1.40(3)
{l 1
rri
9 9 +} o.42(5) r.70 4.09(4)
Staurolite L101 3 0.33(3) 1.55 4.11(6)

Structuraldata wereobtainedfrom Smith (1968).

The formsarerankedin theorder{010},{ll0}, {001}, l2ro s 0.41(4) r.58 s . 0 8( 1 2 )
-l 310 8 0 . 6 7( 9 ) r.80 5.7L(22)
by both the DH and templatefraction methods,and | 320 13 0.58(7) r.37(2) s.08(13)
theseforms are ranked 1,2,4by the HP attachment L410 16 0 . 9 1( r 8 ) 1.90 6 . 8 0( 2 s )
energy.Theseare the first threeforms listedby Niggli
(1926b)as important;staurolitedoesnot commonly 0.67(10) 1.45(4) 4.90(11)
o . s 8( 8 ) I . s 4( 5 ) 4 . 0 8( 3 )
show many forms. The form {l0l}, the only other
form mentionedby Niggli, is anomalouslyfar down 0.67(11) L.62 4.16(8)
the list by all three methods.The indistinct{010}
cleavageis correctlypredictedby the cleavageenergy
calculations. o.Zr<rrt,.Zn ,.;r,to,

Quartz Form - Listed by groups afEer Niggli (L926b).1. main

foms. 2. secondary prism f oms, 3, secondary
The three forms which are almost always.present non-prism foms.
- Ranking according to the unnodified Donnay-
arethe hexagonal prismz {10i0},the positiverhom- DI{
Harker Law.
bohedronr {1011},and thenegativerhombohedron z T - Template fraction, with ranking in parentheses.
{0lil}; the positiverhombohedronis usuallylarger Cl - Cleavage energy, divided by that of the first
ranked fom, with ranking in parentheses.
than the negativeone. Theseare the first forms in HP' - Attachment energy accordinB to the method of
both the DH and templatefraction lists (which rank Hartman and Perdok (in eL/n. Charges: Ti =

the two rhombohedraequally),and they wereclassi- +4, O = -z)

458 ERIC DOWTY

eighth, and incorrectly ranks {l0l} first in the cleav- successfulin general as any other systematic method
age list, insteadof {ll0}. The HP attachmentenergy yet proposed for predicting morphology. Of course,
r a n k s t h e f o r m s i n t h e o r d e r { l 0 l } , { 0 0 1 } ,{ 1 0 0 } ,{ l I l } ,
we can only hope to isolatethe contribution of inter-
{ 0 l l } , { l l 0 l , e t c . w h i c h i s a n i m p r o v e m e n tw i t h r e - nal structure by this approach, and factors such as
spect to {001} but much worse with respectto the composition of the medium, supersaturation,and the
other forms. presence of dislocations have not yet been directly
accounted for. The restriction to first-order bonds,
Bookite
the necessityof using a mathematical plane in the
The three most important forms are {100}, i2l0}, calculations, and the uncertainty of bond strengths
and {lll} (Niggli, 1926b; Arnold, 1929), which are all tend to make the predictionsin any particular case
the first three forms in both the DH and template less than certain, even if the basic hypothesisis ac-
fraction lists. The agreement after these first forms is cepted and external factors are not important. Meth-
not exact, but generallythe common forms are near ods of overcoming these limitations are being in-
the head of the list. The cleavage energy correctly vestigated.
predictsthe indistinct {210}cleavage.The HP attach- What is perhaps of more immediate practical im-
ment energy ranks {100} and i2l0} first and second, portancethan the ability to predict morphology is the
b u t { l l l } i s a t l e a s te l e v e n t h . ability to predict surface structure in complex crys-
Plagioclase tals, and the procedure developed in this study is
probably applicableto this independentlyof any par-
Most igneous plagioclase probably crystallizes
ticular model of the relationshio between structure
with the high-albite structure, the parameters for
and morphology.
which (Smith, 1974)have been used in this example.
According to Niggli (1926b) the most important Acknowledgments
f o r m s a r e { 0 0 1 } ,{ l l 0 } , { l I 0 } , { 0 1 0 } ,a n d { l 0 I } . T h e I thank G. E. Harlow for assistancewith the mathematics of
first four forms are the first four in the ranking ac- projecting crystal structures, and L. S. Hollister and R. J. Kirk-
cording to the DH method, the template fraction and patrick for critical reading of the manuscript.
the HP attachmentenergy.The form {l0I} is anoma-
lously low in all the lists. The cleavageenergy in- References
c o r r e c t l yr a n k s { 1 1 0 }a n d { l I 0 } , w h i c h a r e o n l y p o o r ARNoLD, W. (1929)Beitrage zur Kenntnis desBrookitin mor-
cleavages,aheadof {001}and {010},which are perfect phologischer und optischer Hinsicht. Z. Kristallogr. 11,
and good cleavages,respectively. 344-372.
BrrNrvl, P.eNoG. M. Grluen(1973) Kinetics of crystal growth.
Epidote In, P. Hartman, Ed., Crystal Growth: An Introduction North
Holland, Amsterdam.
Structural data were obtained from Dollase ( l97l ).
Bonu, M. (1923) Atomtheorie desfesten Zustandes.B. G. Tuebner,
Epidote shows a great many forms. The most impor- Leipzig, Germany.
tant zone is [010], on which the principal forms are - AND O. Srpnu (1919) Uber tie Oberfliichenenergie der
{ 0 0 1 } ,{ 1 0 0 } ,{ 1 0 1 } ,{ 1 0 2 } ,a n d { 2 0 1 } ( N i g g l i , 1 9 2 6 b ) . Kristalle und ihren Einfluss auf die Kristallgestalt. Sitzungsber.
T h e D H m e t h o d r a n k s t h e s ef o r m s l , 2 , 3 , 5 , a n d 9 , Preuss.Akad Wiss Berlin.90l-913.
(1932) Morphologische, genetischeund paragenetische
and the template fraction method ranks them 1,2, 3, BneuN, F.
Tracht-studien an Baryt. Neues Jahrb. Mineral., Beil Band, 65A,
4, and 7, correctly assigningfirst rank in cleavage I I 5-lZZ.
energy to {001}, although {100} is essentiallyequal. BucrrrY, H. E. (1951)CrystalGrowrh.John Wiley and Sons,
The HP attachment energy ranks theseforms 2, l, 3, New York.
10, and 17. In almost all casesthe important facesin Buencrn, M. J. (1947)The relativeimportanceof the severalfaces
each zone as described by Niggli (1926b) rank high in of a crystal.Am. Mineral.32, 593-606.
BuRroN,W. K. nxo N C,rsnlnn (1949)Crystalgrowth and
both the DH and template energylists, which do not surfacestructure Disc FaradaySoc.5, 33-39.
differ greatly from each other. Except for the first few Cllnr, J. R., D E. ApplsvnN rNo J. J. Peetrr (1969)Crystal-
forms, the difference in template energy between suc- chemicalcharacterization of clinopyroxenes basedon eightnew
cessiveforms is usually small, which accountsfor the structurerefinements.Mineral Soc Am. Spec.Pap. 2, 3l-50.
large number of forms observed. CosrN, A. J. (1960)Substitutional and interstitial
aluminumim-
purity in quartz,structureand color centerinterrelationships."/.
Phys.Chem.Solids,13,321-325.
Conclusions
Cuntr, P. (1885)Sur la formationdescristauxet sur lesconstantes
Many examplesincluding those given in this paper capillairesde leurs diff6rentesfaces.Bull. Soc. Mineral. France,
show that the template fraction method is at least as E,145-150.
 CRYSTAL STRUCTURE AND CRYSTAL GROII.TH: I 459

DoLLAsE, W. A. (1971) Refinement of the crystal structures of KLEBER, W. (1957)Zur Adsorptionstheorie der Exomorphose. Z.
epidote, allanite, and hancockite. Am. Mineral. 56,44'7-464. Kristallogr 109, I 15-128.
DoNNly, J. D. H. eNo D. Henxsn (1937)A new law of crystal KossEL,W. (1927)The theoryofcrystalgrowth.Nachr.Ges.Il'iss.
morphology extending the law of Bravais. Am. Mineral. 22, GbttingenJ ahresbe r. math-physik.Klasse,135-143.
446-467. KuBAscHEwsKI,O. AND E. L. EveNs (1967) Metallurgical
- AND G. DoNNny (1961) "Assemblage liaisons" et structule Thermochemistry. 3rd Ed. PergamonPress,Oxford, England.
cristalline. C. R Acad. Sci. Paris,252.908-909. MooRE, C. E. (1970)Ionizationpotentialsand ionizationlimits
- AND - (1962) Center of charges inferred from barite derivedfrom the analysesof optical spectra.NSRDS-NBSJ4'
morphology Sou. Phys. Crystallogr. 6, 6'19-684. U.S. GovernmentPrinting Office.
Dowrv, E. (1976) Crystal structure and crystal growth: II Sector NIGGLT,P. (1920) BeziehungzwischenWachstumformenund
z o n i n g i n m i n e r a l s .A m . M i n e r a l . 6 l , 4 6 0 - 4 6 9 . Structurder Kristalle.Z. Anorg.Chem.ll0' 55-81.
Estnup, P. J. (1975) The geometry of surface layers. Physics To- - (1926a)Baugesetze kristallinerMaterie.Z. Kristallogr.63'
day,28, 33-41. 49-121.
Gtses, J. W. (1961) The Scientific Papers Dover, New York. - (1926b)Lehrbuchder Mineralogie,2nded. Borntraeger,
Gorrsctnrrot, V. (1923) Atlas der Kristallfurmen.9 Vols. Winter, Berlin.
Heidelberg, Germany. Penrrn, R. L. (1923)Zur Kristallmorphologie von Anatasund
HlnrneN, P. (1956). An approximate calculation of attachment Rutil. II. Teil. Die Anatasstructur. Z. Kristallogr.59' l-55.
energies for ionic crystals. Acta Cristallogr 9,569-572. - (1930)Die Kristallmorphologie im Lichteneuereranalytis-
- (1958) The equilibrium forms of crystals. Acta Cristallbgr. cher U ntersuchtngen. Fort schr. M ineral. 14, 75-93.
rt, 459-464. ScHNrnn,C. J. (1970)Morphologicalbasis for the reticularhy-
- (l 959a) The effect of adsorption on the face development of pothesis.Am. Mineral.55, 1466-1488.
the equilibrium form. Acta Cristallogr. 12,429-439. Srratrs,J. V (1968)The crystalstructureof staurolite.Am. Min-
- (1959b) Sur la structure atomique de qeulques faces des eral.53, I 139-1155.
cristaux du type blende et wurtzite. Bull. Soc.fr. Miner Cristal- - (1974) Feldspar Minerals. Yol l. Crystal Slructure and
logr.82, 158-163. PhysicalProperties.Springer,New York, Berlin.
- (1959c) La morphologie structural du quartz. Bull. Soc fr. StnnNsrr, I. N. (1928)Zur Theorie des Kristallwachstums. Z.
M iner. C ristallogr. t2, 335-340. Phys.Chemie,138, 259-278.
- ( 1 9 7 3 ) S t r u c t u r e a n d M o r p h o l o g y . I n , P . H a r t m a n , E d . , Wnrrs, A F. (1946)Crystal habit and internal structure.Pftil.
Crystal Growth: An Introductior. North Holland, Amsterdam. Mag. 37, 184-199,217-236, 605-630.
- A N D W G . P E R D o K( 1 9 5 5 a , b , c )O n t h e r e l a t i o n s b e t w e e n Wor-Rr,G. A. eNo J. G. Guelrtent (1962)PBC vector,critical
structure and morphology of crystals. Acta Cristallogr 8, (I) bond energyratio and crystalequilibriumform. Am. Mineral.
49-52; (il) 52r-524; (III) 525-529. 47, 562-584.
JAcKSoN, K. A. (1958) Mechanism of growth. ln, Liquid Metals Wuun, G. (1901)Zur Frageder Geschwindigheit desWachstums
and Solidification. American Society for Metals, Cleviland. und der Auflcisung der Kristallfliichen. Z. Kristallogr. 34,
K n n N , R . ( 1 9 5 5 ) I n f l u e n c e d u m i l i e u d e c r o i s s a n c es u r l a c o r r e - 449-530.
spondance entre morphologie et structure cristalline. Bull. Soc. Wvcrorr, R. G. (1968) Crystal Structures,2nd ed. Vols 1-4.
fr. Miner C ristallogr 78, 461-474, 49'7-506. Interscience, New York.
- (1968) Crystal growth and adsorption . Growth of Crystals, Manuscripl receiued,July 28, 1975;acceptea
8.3-23. for publication,February24, 1976.