2619

Acta Cryst. (1976). B32, 2619

The Geometrical Basis of Crystal Chemistry.
XII. Review of Structures Based on Three-Dimensional 3-Connected Nets

BY A. F. WELLS
Department of Chemistry, University of Connecticut, Storrs, Conn. 06268, U.S.A.

(Received 26 January 1976; accepted 1 April 1976)

For a number of three-dimensional 3-connected nets the symmetry of the most symmetrical configura-
tion of the net is compared with the space group(s) of the actual crystal structure(s) based on the net.
The close correspondence found in most cases indicates that in general a net describes not only the
topology but also the symmetry of a structure.

Introduction this symbol indicates that the smallest polygons meet-

ing at each point are one 6-gon and two 10-gons.
In previous papers of this series a number of three- The uniform nets are derived as topological entities
dimensional 3-, (3,4)-, and 4-connected nets have been (n,p), that is, the symbol gives only the number (n) of
derived. The earlier work, together with much un- links in each (shortest) circuit and the number (p) of
published material, has recently been summarized in links meeting at each point. In contrast to the unique
a monograph (Wells, 1976). In this paper we review two-dimensional 3-connected net (6,3) the uniform
the structures based on three-dimensional 3-connected three-dimensional 3-connected nets are in general not
nets and compare their symmetries with those of the adequately described by the symbol (n,p); there are,
most symmetrical configurations of the nets on which for example, seven known (10,3) nets. It is usually
they are based. The justification for describing certain sufficient to specify also the following purely topologi-
structures in terms of such nets is not simply that this cal properties of a net, that is, properties independent
provides a convenient way of describing the structures. of the particular geometrical configuration of the net:
The only way in which the basic topology of such Zt, the number of points in the smallest (topological)
structures can be correctly and succinctly described is unit cell, "x, the number of n-gon circuits to which
to specify the three-dimensional nets on which they each point belongs, and "y, the number of n-gon
are based. Moreover, the description of a structure in circuits to which each link belongs. These data for the
this way ensures that it can be recognized even if re- seven uniform (10,3) nets are listed in Table 1. It may
ferred to different axes (Strong, Wells & Kaplow, 1971). happen that even the values of Zt, "x, and "y are not
This final stage in the description of a structure is sufficient to define a net uniquely; for example, they
rarely carried out by X-ray crystallographers; a no- do not distinguish (8,3)-a from (8,3)-b (two of the
table exception is the description of the structure of nets listed in Table 2). In such cases it would be
high-SiP207 in terms of its basic net, the (4,6)-con- necessary to determine the values of x and y relating
nected pyrite net (Tillmanns, Gebert & Baur, 1973). to larger circuits.
Polymorphs are conveniently characterized by speci- So far we have considered only the topology of a
fying their nets provided these are topologically dis-
tinct, as is true of the cubic and the two monoclinic
Table 1. Topological data for uniform (10, 3) nets
forms of SiP207. Perhaps more important is the fact
that the correct description of the topology of a struc- z, number of points in smallest cell (topological), ~°x number
of 10-gon circuits to which each point belongs, lOy number of
ture of the type we are considering is a necessary 10-gon circuits to which each link belongs.
preliminary to attempting to understand why the par-
ticular structure is adopted. Net Zt I°X 10y lOymean
Nets are described in terms of their polygonal (10,3)-a 4 15 10 (30/3)
circuits (n-gons) and connectedness (p), and we shall b 4 10 8,6,6 20/3
c 6 5 4,3,3 10/3
be concerned here for the most part with nets d 8 10 8,6,6 20/3
having the special property that the shortest circuit e 12 9,11 8,6,6 20/3
starting from any point and including any two of the f 16 10 8,6,6 20/3
links meeting at the point is an n-gon. Such a net is e 20 12 8 (8)
described as a uniform net. The known uniform three- Note: The net (e) has equal numbers of two kinds of non-equiv-
dimensional 3-connected nets now number thirty: one alent point, and the nets (b)-(f) inclusive have two kinds of
uniform (12,3), seven uniform (10,3), three uniform non-equivalent link, one third of one kind and two thirds of
the other. Note that the number of links per cell is equal to
(9,3), fifteen uniform (8,3), and four uniform (7,3) nets. 3ZJ2 and that the numbers given in the 10y column refer to
They include all the nets discussed later except 6. 102; the three links meeting at each point.
 2620 THE GEOMETRICAL BASIS OF CRYSTAL CHEMISTRY. XlI

net. Before discussing the metrical features (geometry)

of nets it is pertinent to examine the types of crystal
structure based on them. A.
(0) (bl (el
Structures based on three-dimensional 3-connected nets

The simplest types o f structural unit which might be

~l ll
expected to f o r m periodic three-dimensional 3-con-
nected nets (or any other kind of 3-connected system)
are s h o w n in Fig. I. F o r c o o r d i n a t i o n g r o u p s AX3,
A X 4 a n d AX6 in ( b ) - ( d ) the c o m p o s i t i o n s are A2X3, (dl
A2X5 a n d AX3 respectively. C o r r e s p o n d e n c e between Fig. I. Structural units in 3-connected systems.
the s y m m e t r y of a structure a n d that of the most
symmetrical configuration of a net might be expected based on the net (10,3)-b we disregard all the O
to be closest in case (a) a n d progressively less close atoms, some of which are b o n d e d to only one P a t o m
in (b), (c) or (e); no examples of (d) are k n o w n . In while the relnainder link the P a t o m s into a three-
describing the structure of B203 as based on the net dimensional 3-connected net. In replacing
(10,3)-c we suppose that we disregard the (2-con- O O O
nected) O a t o m s which connect the B atoms, leaving / \ ,," \ / \ \ ,/
a link connecting each pair of these a t o ms . Similarly, O=P P=O by P........ P
\ / \ / / \
in describing the structure o f one f o r m o f P2Os as 0 0

Table 2. Structures based on three-dimensional 3-connected nets

Most symmetrical configuration Space
Net Space group Position c'a Compound group Reference a
( l 0, 3)-a (120 °) I4132 8(a) SrSi2 P4132 1
CsBe2Fs P4132 2
H202 P41212 3
(10,3)-a (90 °) 1213 0~-HgaS2Cl2 1213 4
Sn2F3CI a P213 5,6
(10,3)-b (120 °) I4damd 8(e) (0,0,-~-) 21/3 ~-ThSi2 14damd 7
P205 Fdd2 8
(Zn2CIs) (HsO2) Fdd2 9
(CH3)2S(NH)2 a Fdd2 10
[GeS2 Fdd2 11]*
(10,3)-b (90 °) C2/c La,Be2Os C2/c 12
(10,3)-c P3112 6(c) (½,~, ~) 31/3/2 B203 P31 13
(10,3)-d Pnna 8(e) 0c-Resorcinol } 14
B-Resorcinol Pn21a 15
(8,3)-a P6222 6(i) ({, 3, 0) 31/2/5 Caryophyllene
chlorohydrin P31 16
(8,3)-b R3m 18(f) c ({~,0, 0) 1/6/5 C6H4{Si(CHa) 2OH}2 PT 17
6. 102 R3m 1 8 ( ~ ~,c N4(CH2)6.6H20 R3m 18
fl-Quinol R], R3 19
(a) These two examples of the nets were noted after the paper was written and had been submitted for publication, and for that
reason the structures are not discussed in the text. In SnzF3CI each Sn atom forms three bonds to bridging F atoms. Another study
(Donaldson, 1976) shows that the symmetry of SnzFaC1 is only orthorhombic (space group P212121) but markedly pseudocubic,
a = b = c= 7.880 ,~. The dimethyl sulphone diimine structure is similar to that of hydrogen peroxide (though based on a different
net) in that each N atom is hydrogen-bonded to N atoms of two other molecules:
sI
ss
H

N-S(CH3)2-N
H
//

The author hopes to comment on the structures of the tin and mercury compounds listed in Table 2 in a later paper. (b) No
unique configuration is possible with all links of equal length and all interbond angles equal to 120 °. (c) Hexagonal setting.
(d) Key to references: (1) Janson, Schiller & Weiss (1970), Pringle (1972). (2) Le Fur & A16onard (1972). (3) Abrahams, Collin &
Lipscomb (1951). (4) Frueh & Gray (1968). (5) Bergerhoff & Goost (1974). (6) Donaldson (1976). (7) Brauer & Mitius
(1942). (8) de Decker (1941). (9) Foliner (1970). (10) Prince (1975). (11) Zachariasen (1936). (12) Harris & Yakel (1968). (13)
Strong & Kaplow (1968). (14) Robertson (1936). (15) Robertson & Ubbelohde (1938). (16) Rogers & Mazhar-uI-Haque (1963).
(17) Alexander, Northoit & Engmann (1967). (18) Mak (1965). (19) Palin & Powell (1947).
* Note added in proof: A redetermination (Dittmar & Sch/ifer, 1976) shows that low-GeS2 is actually monoclinic (space group
Pc), the structure being essentially the same as that of Zachariasen (1936).
 A. F. WELLS 2621

we are taking no account of the oxygen bond angles structures. We shall see later that it may be convenient
or of the packing of the O atoms, though both factors to recognize more than one standard configuration of
are important in determining the atomic arrangement certain nets.
in the crystal. The representation of the structure of a
crystalline dihydroxy compound such as H202 or
Symmetry of structure and net compared
m-dihydroxybenzene (resorcinol), OH. C6H 4. OH, as a
3-connected net implies regarding the hydrogen- In view of the simplifications made when describing
bonded molecule as a pair of 3-connected points as a structure as based on a particular net it might
in Fig. l(e). In the resulting three-dimensional net one appear that the most symmetrical form of a net is a
third of the links represent the axes of the molecules mathematical abstraction of somewhat dubious inte-
and the remainder O - H . . . O bonds. In these two nets rest, bearing little relation to the actual crystal struc-
the approximate lengths of the two kinds of link are: ture. However, Table 2 shows that in fact there is
usually a close relation between the symmetries of
O-O O - H . • •O
actual crystal structures and those of the most sym-
H202 1-5 A. [ 2"8 A
n o . C6H4. O H 5"5 J metrical configurations of the nets on which they are
based. The most symmetrical arrangement of three
Moreover, the representation of the crystal structure bonds meeting at a point is that with all interbond
of resorcinol in this way implies neglecting the benzene angles 120 °, and if a net can be constructed in three
ring which is responsible for most of the volume of the dimensions with all links of equal length and all bond
molecule. angles 120° this is the configuration whose space group
Evidently, in view of the above facts, it is to be is listed in column 2. Such a configuration is possible
expected that if the structures of two compounds of for three (10,3), two (9,3), and four (8,3) nets, and this
different chemical types are based on the same net the special class includes all the nets in Table 2 except
geometrical configurations of the basic net as found (10,3)-d and 6. 10z.
in the actual structures (for example, by joining to- Both of the nets (10,3)-a and (10,3)-b have four
gether the P atoms in P205 or the Si atoms in ThSi2) points in the smallest (topological) unit cell and eight
may be very different and perhaps difficult to rec- in the unit cells of their most symmetrical configura-
ognize as forms of the same net [in this case (10,3)-b]. tions (equal links and 120 ° bond angles). These con-
It is therefore essential to know the most symmetrical figurations have respectively cubic and tetragonal sym-
configuration of a given net; this is the standard re- metry (Fig. 2). A second special configuration is rec-
ference configuration with which to compare the vari- ognized for these nets, namely, one with equal links
ous distorted configurations found in actual crystal and bond angles of 90 °, because it is found that there
are crystal structures corresponding to both forms of
each net. These special configurations may be derived
.... ~ ---L~:;~ from a portion of the primitive cubic lattice containing
eight points by removing one half of the links. This
operation may be carried out in many ways, to leave
i
I isolated cubes, chains, layers or three-dimensional
nets. Of the numerous possibilities the two shown in
i i
Fig. 3 correspond to (10,3)-a (90 °) and (10,3)-b
L. (90°). In order to assign space groups to these config-
I !
I : I
i i
urations it is necessary to indicate the positions of the
/.~ . . . . . . . ..
links, since the points themselves form a simple cubic
.. f
lattice in all diagrams of this kind. This is done by
• r j "~ "-", i
i
I .........
.; .......................
I
: ...........
i ..........
placing (2-connected) points at the mid-points of the
0..(~:@_ ................................ 1- .... links; these would be the positions of the X atoms in a
compound AzXa based on this net or of the shared X
(o) (b) atoms in a framework of composition AzX5 formed
Fig.'2. The nets (10,3)-a (120°)~and (10,3)-b (120°). from tetrahedral AX4 groups sharing three vertices.
For (10,3)-a (90 °) such points are the position 12(b)
of 1213 for the special value x=½. The structure of
"F"i" ~-Hg3SzClz is based on this net with S atoms in 8(a)
and Hg atoms in 12(b) of this space group.
The unit cell of (10,3)-b (90°), Fig. 3(b), is a cube
but the space group, determined in the above way, is
C2/c, which is that of LazBezOs. The unit cell approx-
imates to a cube: a = 7.54, b = 7.35, c = 7-44 A, fl=
(a) (b) 91°33 ', and the Be atoms occupy the positions 8(f),
Fig. 3. The nets (10,3)-a (90 °) and (10,3)-b (90°). (xyz) etc., with x = 0.47, y = 0.22, z = 0.53. These coor-
2622 THE GEOMETRICAL BASIS OF C R Y S T A L C H E M I S T R Y . XII

dinates are close to the values ½, ¼ and ½, and the f.c. cubic and b.c. tetragonal cells of diamond. We may
eight Be atoms lie approximately at the points of a refer cubic diamond to a b.c. tetragonal cell containing
primitive cubic lattice with cell edge a/2 (=b/2=c/2). four atoms in 4(a) of 141/amd and conversely we may
The three sets of (four) shared O atoms occupy posi- refer the net (10,3)-b to a 16-point f.c. cell with space
tions 4(d), 4(e) (0,y,¼) with y=¼, and 4(b), so that group F41/ddm which is the alternative setting of the
they have coordinates respectively ~4,(i14,!a2j,(0, I4, ~),
1 and same space group. In both cases the most symmetrical
(0,½,0), which are the midpoints of one half of the orthorhombic sub-group is Fddd. In fact we find that
'links'. The arrangement of the Be atoms therefore one form of P205 and also (Zn2Cls) ( H 5 0 2 ) both have
corresponds very closely to that of the circles in Fig. the space group Fdd2 with P or Zn occupying the
3(b) and the shared O atoms are at the midpoints of general position 16(b). We have also included GeS2
the links drawn with full lines. This structure based in Table 2. Its crystal structure is a three-dimensional
on the net (10,3)-b (90 °) has room for eight La in the assembly of GeS4 tetrahedra of which each vertex
octants of the pseudo-cubic cell which contains eight (S atom) is common to two tetrahedra. The Ge atoms
Be, and it is therefore suitable for a compound AzB2Xs, are situated at the points of a 4-connected net, this
in contrast to the structure of CsBe2F5 based on net being formed by joining together the midpoints of
(10,3)-a (120 °) which accommodates only one Cs for adjacent links of (10,3)-b so as to form triangles
every two Be in the framework. We now discuss the around the points of the original 3-connected net. In
remaining structures of Table 2 which are less sym- this case also the space group is Fdd2, but whereas
metrical than the most symmetrical forms of the under- P205 and ( Z n z C 1 5 ) ( H 5 0 2 ) have 16 3-connected points
lying nets. (P or Zn) in the unit cell, GeSz has Z = 2 4 since here
In SrSi 2 and CsBezF5 the unit cell contains eight the Ge atoms are situated at the mid-points of the
atoms (Si or Be) corresponding to the eight 3-con- links of the 3-connected net. The relation of these
nected points of (10,3)-a. In SrSi2 the net is formed structures to (10,3)-b (120 °) is also seen from the cell
from Si atoms which are directly bonded together, dimensions of these crystals. For the tetragonal con-
with Sr atoms in the interstices, while in CsBe2F5 the figuration of (10,3)-b, c : a = 2 1 3 = 3 . 4 6 , and fl=90 °.
framework is formed from BeF4 tetrahedra sharing The corresponding figures for the b.c. monoclinic
three vertices (F atoms). Each Be is connected through ( Z = 8) cells of the other compounds are:
F atoms to three others, and the Cs + ions occupy the fl (approx-
interstices. The similarity between the structures is a ( = b) c imate) c: a
evident from Fig. 4. The space group 14132 is not P2Os 4"85/~ 16"3/~, 114 ° 3-36
possible for these crystals because occupation of 8(a) (ZnzC15) (HsO2) 6"43 22"90 92 3-56
by the atoms forming the net would require a fourfold GeS2 6.77 22.34 119 3.30.
position for Sr or Cs respectively and there is no four-
fold position in 14132. (The higher space group would
also imply coplanar Si bonds - see later.) A similar
problem does not arise for ct-ThSi2 because 141/amd
has the required fourfold position.
In crystalline H202 the 4~ helices are formed by the I
hydrogen bonds, which here constitute two thirds of
the links of the basic net. Since their length (2.78 A)
is so much greater than that of the remaining links, o oi 0 ....... o oI Q ..........
which are the intramolecular O-O bonds of length
(a) (b)
1-47 /~, the net is very distorted from the ideal con-
figuration. The 41 axes are retained but the symmetry Fig. 4. Projections on (001) of (a) the Sr and Si atoms of
drops to tetragonal and the space group is P41212. SrSi2, (b) the Cs and Be atoms of CsBe2Fs.
The net is so distorted that it is virtually 5-connected, f.c. cell F41/ddm sheared b.c. cell
for each O atom also has two other O atom neighbours (z=~6) (
at 2.90 A.
The net (10,3)-b, Fig. 2(b), is somewhat more com- """ , - (Z=8) .~', ~'~

plex because the 120 ° configuration can be 'sheared', , ::~ ".....~i..... .....
a process involving rotations about those links which
are parallel to the tetragonal c axis. The fully extended
tetragonal form of the net represents the arrangement
of the Si atoms in a-ThSi2. In sheared forms the 8-
point body-centred tetragonal cell becomes non- (o) (b)
orthogonal but the sheared net may be referred to Fig. 5. (a) Relation between the F and I cells of (10,3)-b
an all-face-centred orthogonal cell containing 16 points. projected along c axes; the dotted circles represent points at
height c/2 above the paper. (b) Relation between the b.c.
These are alternative cells for the space group Fddd cell of 'sheared' version of (10,3)-b (dotted lines) and the
(Fig. 5). This relation is similar to that between the orthorhombic F cell (full lines).
 A. F. WELLS 2623

The reason for the adoption of the less symmetrical (120 °) are retained in the less symmetrical structure of
space group is as follows. In Fddd the 16 tetrahedrally H202, so the essential features of the net are retained
coordinated atoms (P or Zn) would have to 6ccupy in the structures noted above.
two sets of eightfold positions (point symmetry 222) The net (10,3)-c (Fig. 6) is the trigonal (or hexag-
or one of the sixteenfold positions (c)-(g). Of the latter, onal) analog of the (tetragonal) net (10,3)-b (120°).
(c) and (d) are excluded because the point symmetry The characteristic sub-unit in both nets is the planar
is T while in (e), (f) or (g) or in the eightfold posi- zigzag chain. Repetition of such a chain by 21 axes
tions the tetrahedra would lie on twofold axes, which in the plane of the zigzag gives the planar 6-gon net.
is not possible if they are to share three vertices. In In (10,3)-b the chain directions are the a and b axes,
Fdd2 the tetrahedral groups can occupy the general and the chains are repeated around 4, axes which are
sixteenfold position. This simple argument does not parallel to the c axis. In (10,3)-c the chains are
apply to GeSz. In Fddd, with 24 Ge in the cell, it parallel to three equivalent directions perpendicular
would be necessary to utilize an eightfold position, and to c and they are repeated by the operation of 31 axes.
presumably the occupation of one of these special Owing to the special values of the coordinates, namely,
positions (with no variable parameters) would impose x=~-, y = ~ , and z = ~ in the general position 6(e) of
too great a restriction on the packing of the atoms. P3112, there are also 6 4 s c r e w axes. [In the earlier
A feature of several other structures (B203, a- and description of this net (Wells, 1972) the unnecessarily
fl-resorcinol and caryophyllene chlorohydrin) is that complicated coordinates x = ~ , y = ~ and z = ~ were
the symmetry is lower than that of the most symmetrical given for 6(c) in P3z12.] It can be shown that there is
configuration of the net by the absence of one or more no three-dimensional net analogous to (10,3)-b and
sets of twofold axes. At this time, without detailed (10,3)-c generated from planar zigzag chains by the
studies of these structures, we can only suggest that, operation of 61 axes (Wells, 1976).
as in the case of GeS2, the additional symmetry would The net (10,3)-d cannot be constructed with all
not be consistent with satisfactory packing in the crys- links of equal length and all interbond angles 120 °,
tals. Just as the 41 screw axes characteristic of (10,3)-a though it can be built with all links equal in length
and with exact 41 and 43 helices. This and other con-

:Y figurations are derivable from the general (eightfold)

position in Pnna with appropriate values of x, y, and z.
This net may be derived in two ways. It is the second
member of a family of nets which project as the planar
4:8 net. The first member is (10,3)-a in which all
the fourfold helices are of the same chirality, while in
(10,3)-d the helices which are directly connected are
of opposite chirality (Fig. 7). Alternatively we may
use the operation
1 ,: . 3 1<--0
~ , ~ ' ~ instead of ,~
2/ \ 0 2--+3
described in an earlier paper as a '412 axis' (Wells,
1956), in which case (10,3)-d is the second member of
the family which starts with (10,3)-b (Fig. 8). These
two ways of deriving this net correspond to the use
of the net in the two polymorphs of resorcinol. In the
Fig. 6. The net (10,3)-e. [The unit cell is not that of Interna- form the hydrogen bonds form the fourfold helices,
tional Tables but has its origin at (~,~-,~-).]
the broken lines in Fig. 7, while in the fl form they
I I
correspond to the '412 axes'. These configurations of
\-/
',4,:
:S"',--"
: ,~ ~,:
(10,3)-d afford a unique example of the use of the
same net in two quite different ways by the same com-
pound with the same space group. The space group
,,./\../ Pn21a has no diad axes, whereas there are two sets
of diads in Pnna (full symbol P2/n21/n2/a). In the
latter space group molecules of m-C6H4(OH)2 could
not be placed in fourfold positions 4(a) or 4(b) with
point symmetry ] and would have to be situated in
4(c) or 4(d) on diad axes. This is presumably not
(IO,3)-o (JO,3)-d consistent with efficient packing of the benzene rings
Fig. 7. Projections of nets (]0,3)-a and (10,3)-d along the and simultaneously the formation of chains of hydrogen
fourfold screw axes. bonds.
2624 THE G E O M E T R I C A L BASIS OF C R Y S T A L C H E M I S T R Y . XII

We are aware of only one structure based on each meeting at each point (for details see Wells, 1976).
of the nets (8,3)-a and (8,3)-b, the relation between The structures based on this net have lower symmetry
which can be seen from the projections of Fig. 9. than R3m. In the hexahydrate of N4(CH2)6 the hexa-
The unsymmetrical molecule of caryophyllene chloro- methylenetetramine molecules are situated in the
hydrin, C14HzlCl(OH)z, contains two OH groups, and interstices of the net formed by the hydrogen-bonded
the basic topology of the crystal structure is deter- water molecules, and the most symmetrical space
mined by its behavior as a dihydroxy compound. The group compatible with the point symmetry (3m) of the
hydrogen bonds form threefold helices all of the same N4(CH2)6 molecule is R3m. In 'fl-quinol' there are two
chirality, and the whole structure has trigonal sym-
metry P31 (Fig. 10). In crystalline
HO(CH3)zSiC6HaSi(CH3),OH the trigonal symmetry
R3m of the most symmetrical configuration of the net
~,..'...,,~ ").d
(8,3)-b is not retained, and the structure has only
triclinic symmetry (with Z = 3 , to give six OH in the
unit cell). However, the helical arrangement of the
hydrogen bonds is evident in the projection of Fig.
11, and in the description of the structure it is noted
that 'the geometry about the hydrogen-bond axes
shows a distinct tendency in the direction of 31 helical
symmetry'. The molecules are situated at three of the (10,5)- b (10,5)-6
four symmetry centres in the unit cell so that alternate Fig. 8. Projections of the nets (10,3)-b and (10,3)-d along
helices are of opposite chirality as required in the net the 4/2 axes.
(8,3)-b.
The most symmetrical configurations of the net
6 . 1 0 z, one of which is illustrated in Fig. 12, have the
symmetry of the space group R3m with Z = 6 for the
rhombohedral cell or 18 for the hexagonal cell. There
is no configuration with all links equal in length and
all interbond angles equal to 120°, but the hexagons
are planar (and regular) and all links of equal length
for appropriate ranges of values of c'a and of x in (8,5)-o (8,3)-b
position 18(f) (hexagonal cell) or puckered in a range Fig. 9. Projections of the 3-connected nets (8,3)-a and
of configurations having three equal coplanar links (8,3)-b.

Fig. I0. Projection of the structure of caryophyllene chlorohydrin.

 A. F. W E L L S 2625

identical interpenetrating nets. Here the vertical sym- structure is based on a 3-connected net, (b) why this
metry planes are absent because of the relative orienta- should be a three-dimensional net, (c) why one partic-
tions of the benzene rings, which are situated along ular net is chosen in preference to others, and (d)
the links of the framework. As in the former structure why in some cases the symmetry of the structure is
there are molecules of a second compound X (or in lower than that of the most symmetrical configuration
the argon compound, atoms of argon) in the inter- of the net. These questions are inter-related, as we shall
stices of the framework, and the composition is ap- show for certain silicides. We may distinguish (a) as
proximately C6Ha(OH)z.~X. In each structure the in- essentially a chemical problem, a matter of the elec-
terstitial molecule (atom) is situated at the origin of tronic structures of atoms and of their interactions,
the rhombohedral cell, where the point symmetry cor- and this is not the place to discuss the reasons why the
responds to the space group symbol. In units of Fig. 1 form 3-connected nets. We shall com-
C6H4(OH)2.½SO2 the SOz molecule becomes effectively ment briefly here on (b) and (c); (d) has already been
centrosymmetric by rotation and the space group discussed.
is R3, but in C6H4(OH)E.1CHaOH the methanol mol- The second question, (b), is perhaps the most dif-
ecule apparently cannot rotate about a horizontal axis ficult one to answer. For structural units which are
and the space group is R3. to be linked to three others the resulting 3-connected
system could be finite or infinite in one, two, or three
Discussion dimensions. As regards freedom to adapt to the bon-
ding requirements of the atoms a three-dimensional
One of the reasons for studying crystal structures is framework would appear to be the least favorable type
to understand why particular structures are adopted of structure. A chain or layer structure would seem
by the various elements and compounds. For the com- to be more adaptable and a structure built of finite
pounds under discussion we might enquire (a) why a units still more so. Phosphorus pentoxide is remarkable
for crystallizing with structures of three of the four
main types, namely, finite (P4010 molecules), layer,

"L_2 and three-dimensional framework. However, almost

all compounds AmBzX5 have layer structures: M2Si205
( M = L i , Na, K, Rb, Ag, TI), BaSizOs, MBe2F5
(M = Rb, Cs, T1 and probably K and NH4). Only
La2Be205 and one polymorph of CsBe2F5 are at pres-
ent known to form structures based on three-dimen-
sional nets. The 6-gon layers in the silicates buckle to
various degrees to accommodate a range of cations
(for a review see Liebau, 1968). Tetrahedral Si44- ions
can pack with cations ranging in size from Na ÷ to
Cs ÷ (in NaSi etc.) or with one half the number of
Fig. 11. Projection o f the structure o f C6H4[Si(CH3)2OH]z. Ba z+ ions in BaSi2. In fact the only sizeable group of
F o r all molecules except one only the H O - O H axis is shown. compounds with structures based on a three-dimen-
The letters c a n d a indicate clockwise and anticiockwise
helices. The small black circles indicate s y m m e t r y centres sional net are the 4 f a n d 5 f disilicides with the ThSi2
and the larger open circles O H groups. structure, and three of these compounds also crys-
tallize with the A1B2 structure:
M in MSi2
ThSi2 structure: Th, U, Pu; La - Ho
j AIB2 structure: Th, U, Pu ; Er, Tm, Yb, Lu.
Elementary silicon does not have a graphitic form
but three coplanar bonds are formed by Si in the
closely related A1B2 and ThSi2 structures. The 12-
coordination of the metal atom is extremely similar in
these two structures, which must have very similar
energies; there are no diborides with the ThSi2 struc-
ture.
The next question, (c), concerns the choice of a
particular net. Of the structures listed in Table 2 most
are based on the simplest and most symmetrical three-
dimensional nets. The types of compound with struc-
F
Fig. 12. The 3-connected net 6 . 10 z.
tures based on particular nets and the distortions (if
any) from the most symmetrical configurations sug-
gest that the atomic (electronic) interactions lead to

A C 32B - 6
2626 T H E G E O M E T R I C A L BASIS OF C R Y S T A L C H E M I S T R Y . XII

some preferred bond arrangement which in turn leads that the net selected is one of the two (topologically)
to a particular type of sub-unit of structure. For simplest three-dimensional 3-connected nets.
example, the 'graphitic' behavior of Si is satisfied in a The object of this paper and in particular of these
planar 6-gon net or in a three-dimensional structure concluding remarks is to indicate some of the pro-
built from planar zigzag chains, that is, (10,3)-b or blems raised by the structures of comparatively simple
(10,3)-c rather than (10,3)-a or (10,3)-d etc. We should compounds. If we agree that the structures of solids
then associate different nets with various types of sub- form an integral part of structural chemistry then this
unit: type of crystallochemical enquiry is the logical sequel
to studies of the structures of finite molecules in terms
Sub-unit: Planar zigzag Helical arrays
(21 axis) ^ of the electronic structures of the component atoms.
(planar 6-gon net) 3"1 41 6~
Probable f (10,3)-b (8,3)-a (10,3)-a (12,3) References
three-
dimen- ABRAHAMS,S. C., COLLIN, R. L. & LIPSCOMB,W. N. (1951).
sional nets (10,3)-c (8,3)-b (10,3)-d. Acta Cryst. 4, 15.
ALEXANDER, L. E., NORTHOLT, M. G. ~ ENGMANN, R.
For example, if threefold helices are the preferred sub-
(1967). J. Phys. Chem. 71, 4298.
units the (8,3) nets with Z = 6 are to be expected rather BERGERHOFF,G. ~ GOOST,L. (1974). Acta Cryst. B30, 1362.
than the (10,3) nets with Z = 4 , 8 , etc. It then follows BRAUER, G. ~ MITIUS, A. (1942). Z. anorg. Chem. 249, 325.
that for a particular net certain sub-groups may be DECKER, H. C. J. DE (1941). Rec. Tray. Chim. Pays-Bas, 60,
expected to be less likely than others. In the scheme 413.
for less symmetrical versions of the cubic (10,3)-a DITTMAR, G. ~ SCH~,FER, H. (1976). Acta Cryst. B32,
net (where we use Sch6nflies symbols for brevity) we 1188-1192.
should not expect to follow the D3-C3 route if we DONALDSON, J. D. (1976). Private communication.
associate the net with fourfold helices since this would FOLLNER, H. (1970). Aeta Cryst. B26, 1544.
FRUEH, A. J. & GRAY, N. (1968). Acta Cryst. B24, 156.
imply the breakdown of four equivalent points into
HARRIS, L. A. & YAKEL, H. L. (1968). Acta Cryst. B24, 672.
(3+ 1) points. Accordingly the space group of H202 JANSON, K. H., SClaXEER,H. &WElSS, A. (1970). Z. anorg.
is D~: Chem. 372, 87.
LE FUR, Y. & AL/~ONARD,S. (1972). Acta Cryst. B28, 2115.
_D3 ~ C3 LIEBAU, F. (1968). Acta Cryst. B24, 690.
MAK, T. C. W. (1965). J. Chem. Phys. 43, 2799.
PALIN, D. E. & POWELL,H. M. (1947). J. Chem. Soc. p. 208.
o PRINCE, E. (1975). Acta Cryst. B31, 2536.
PRINGLE, G. E. (1972). Acta Cryst. B28, 2326.
D4 C4 Cz ROBERTSON, J. M. (1936). Proc. Roy. Soc. A157, 79.
ROBERTSON, J. M. & UBBELOHDE,A. R. (1938). Proc. Roy.
Soc. A 167, 122.
In contrast to the formation of three coplanar bonds ROGERS, D. & MAZHAR-UL-HAQUE(1963). Proc. Chem. Soc.
in the 4 f and 5 f disilicides, silicon forms three pyra- p. 371.
midal bonds in combination with the most electro- STRONG, S. L. & KAPLOW,R. (1968). Acta Cryst. B24, 1032.
positive elements, the structural chemistry of Si- being STRONG, S. L., WELLS, A. F. & KAPLOW, R. (1971). Acta
very similar to that of the isoelectronic neutral P: Cryst. B 27, 1662.
TILLMANNS, E., GEBERT, W. & BAUR, W. H. (1973). J.
Discrete tetrahedral groups: white P: (Si4)4- in BaSi2 Solid State Chem. 7, 69.
Planar 6-gon net: black P: CaSiz WELLS, A. F. (1956). Acta Cryst. 9, 23.
Three-dimensional net: - : SrSi2. WELLS, A. F. (1972). Acta Cryst. B28, 711.
WELLS, A. F. (1976). Monograph Three-dimensional Nets
It is not obvious why SrSiz adopts a structure based and Polyhedra. New York: Wiley-Interscience. In the
on a three.dimensional net rather than, say, a two- press.
dimensional net as in CaSi2, though it is satisfactory ZACHARIASEN, W. H. (1936). J. Chem. Phys. 4, 618.