International Tables for Crystallography (2006). Vol. C, Section 9.8.1, pp. 907–913.

9.8. Incommensurate and commensurate modulated structures

By T. Janssen, A. Janner, A. Looijenga-Vos, and P. M. de Wolff

9.8.1. Introduction Villain, 1960; Koehler, Cable, Wollan & Wilkinson, 1962). In
the early 1970's, the importance of the irrationality was realized
9.8.1.1. Modulated crystal structures
and the term `incommensurate phase' was introduced for
Lattice periodicity is a fundamental concept in crystal- modulated crystal phases with wavevectors that have irrational
lography. This property is widely considered as essential for components. Such phases were found in insulators as well
the characterization of the concept of a crystal. In recent (Tanisaki, 1961, 1963; Brouns, Visser & de Wolff, 1964). The
decades, however, more and more long-range-ordered con- remarkable morphological properties of calaverite could also be
densed-matter phases have been observed in nature that do not related to the existence of an incommensurate modulation (Dam,
have lattice periodicity, but nevertheless only differ from normal Janner & Donnay, 1985). Structures that are well ordered, show
crystal phases in very subtle and not easy to observe structural sharp incommensurate spots in the diffraction pattern, but cannot
and physical properties. be described as a modulation of a lattice periodic system, were
It is convenient to extend the concept of a crystal in such a way also discovered, indicating that incommensurability is not
that it includes these ordered phases as well. If properly applied, restricted to modulated crystals. Examples are the composite
crystallographic symmetry concepts are then still valid and of (or intergrowth) crystals (Jellinek, 1972) and, more recently,
relevance for structural (and physical) investigation. Instead of quasicrystals (Shechtman, Blech, Gratias & Cahn, 1984). The
limiting the validity of crystallographic notions, these develop- latter not only lack three-dimensional lattice periodicity, but also
ments show how rich the crystallographic ordering can be. show non-crystallographic symmetry elements, such as ®ve-,
Incommensurate crystals are characterized by Bragg re¯ection eight- or twelvefold axes. For a review, see Janssen & Janner
peaks in the diffraction pattern that are well separated but do not (1987).
belong to a lattice and cannot be indexed by three integer indices. We denote a basis for the lattice of main re¯ections by a , b ,

This means that the description in terms of three-dimensional c . The satellites are separated from the main re¯ections by
point and space groups, as given in IT A (1992), breaks down. It vectors that are integral linear combinations of some basic
is possible, however, to generalize the crystallographic concepts modulation vectors denoted by q j j ˆ 1; 2; . . . ; d†. Accord-
on which the usual description is based. The ensuing symmetry ingly, the positions of the Bragg re¯ections are given by
groups and their equivalence classes are different from those of
H ˆ ha ‡ kb ‡ lc ‡ m1 q1 ‡ m2 q2 ‡ . . . ‡ md qd : 9:8:1:1†
three-dimensional crystallography. The main properties of these
new groups and the speci®c problems posed by their application If the modulation is incommensurate, the set of all positions
to incommensurate crystal structures form the subject of the (9.8.1.1) does not form a lattice. Mathematically, it has the
following sections. structure of a Z-module, and its elements are three-dimensional
The tables presented here, and their explanation, have been vectors (of the reciprocal space). Accordingly, we call it a
restricted to cover mainly the simplest case: modulated crystal (reciprocal) vector module.
structures (in particular, those with a single modulation wave). The modulation that gives rise to the satellites can be:
Nevertheless, most of the basic concepts also apply for the other (i) a displacive modulation, consisting of a periodic displace-
classes of incommensurate crystal structures. ment from the atomic positions of the basic structure;
A modulation is here considered to be a periodic deformation (ii) an occupation modulation, in which the atomic positions of
of a `basic structure' having space-group symmetry. If the the basic structure are occupied with a periodic probability
periodicity of the modulation does not belong to the periodicities function.
of the basic structure, the modulated crystal structure is called Mixed forms also occur. Other modulation phenomena found
incommensurate. are, for example, out-of-phase structures and translation inter-
That description is based on the observation in the diffraction face modulated structures. The satellite re¯ections of modulated
pattern of modulated structures of main re¯ections, situated on a structures for the various types of modulation were system-
reciprocal lattice, and additional re¯ections, generally of weaker atically studied by Korekawa (1967).
intensity, called satellites. For an incommensurate structure, at least one component of a
As far as we know, the ®rst indications that complex crystal basic modulation wavevector qj with respect to the lattice of
structures could exist not having normal lattice periodicity came, main re¯ections is irrational. As argued later, even in the case of
at the end of the 19th century, from studies of the morphology of a commensurate modulation, i.e. when all components of these
the mineral calaverite. It was found that the faces of this modulation wavevectors are rational, it is sometimes convenient
compound do not obey the law of rational indices (Smith, 1903; to adopt the description of equation (9.8.1.1) and to apply the
Goldschmidt, Palache & Peacock, 1931). Donnay (1935) showed formalism developed for the incommensurate crystal case.
that faces that could not be indexed correspond to additional The modulated crystal case presented above is only the
re¯ections in the X-ray diffraction pattern. From the experience simplest one giving rise to incommensurate crystal structures.
with manufacturing of optical gratings, it was already known that Another class is represented by the so-called composite or
periodic perturbation gives rise to additional diffraction spots. intergrowth crystal structures. Here the basic structure consists
Dehlinger (1927) used this theory of `Gittergeister' (lattice of two or more subsystems each with its own space-group
ghosts) to explain line broadening in Debye±Scherrer diagrams symmetry (neglecting mutual interaction) such that, in general,
of metals and alloys. Preston (1938) called the `Gittergeister' he the corresponding lattices are incommensurate. This means that
found in diffraction from an aluminium±copper alloy `satellites'. these lattices are not sublattices of a common one. Actually, a
Periodic displacement (modulation) waves were considered by mutual interaction is in general present giving rise to periodic
Daniel & Lipson (1943) as the origin of satellites in a copper± deviations, i.e. to modulations. These crystals are different from
iron±nickel alloy. Complex magnetic ordering giving rise to the previous ones, in the sense that they show more than one set
satellites was found in magnetic crystals (Herpin, Meriel & of main re¯ections. In the case of an intergrowth crystal, the

907

Copyright © 2006 International Union of Crystallography

908 s:\ITFC\ch-9-8.3d (Tables of Crystallography)
 9. BASIC STRUCTURAL FEATURES
Bragg re¯ections can be labelled by a ®nite set of integral indices Conventionally, a1 , a2 , a3 are chosen to be a , b , and c
as well and the Fourier wavevectors form a vector module as generating the positions of the main re¯ections, whereas a4 ˆ q
de®ned above. The difference from the previous case is that the is the wavevector of the modulation. The corresponding indices
basic structure itself is incommensurate, even if one disregards are usually denoted by h; k; l and m. Also, crystals having two-
modulation. and three-dimensional modulation are known: in those cases,
A third class of incommensurate crystals found in nature is n ˆ 5 and n ˆ 6, respectively.
represented by the quasicrystals (examples of which are the In the case of a composite crystal, one can identify two or
icosahedral phases). In that case also, the basic structure is more subsystems, each with its own space-group symmetry. As
incommensurate, the difference from the previous case being an example, consider the case where two subsystems share a
that the splitting in subsystems of main re¯ections and/or in main and b of their reciprocal-lattice basis, whereas they differ in
re¯ections and satellites is not a natural one, even in the lowest periodicity along the c axis and have, respectively, c1 and c2 as
approximation. third basis vector. Then again, n ˆ 4 and a1 ; . . . ; a4 can be
It is justi®ed to assume that crystal phases allow for an even chosen as a , b , c1 , and c2 . The indices can be denoted by h,
wider range of ordered structures than those mentioned above. k, l1 , l2 . In general, the subsystems interact, giving rise to
Nevertheless, all the cases studied so far allow a common modulations and possibly (but not necessarily) to a larger value
approach based on higher-dimensional crystallography, i.e. on of n. In addition to the main re¯ections from the undistorted
crystallographic properties one obtains after embedding the subsystems, satellite re¯ections then occur.
crystal structure in a higher-dimensional Euclidean space. In the case of a quasicrystal, the Bragg re¯ections require
To present that approach in its full generality would require a more than three indices, but they do not arise from structurally
too abstract language. On the other hand, restriction to the different subsystems having lattice periodicity. So, for the
simplest case of a modulated crystal would give rise to icosahedral phase of AlMn alloy, n ˆ 6, and one can take for a1 ,
misleading views. It what follows, the basic ideas are formulated a2 , a3 a cubic basis a , b , and c and for a4 , a5 , a6 then 
pa ,
 
as simply as possible, but still within a general point of view. b , and c , respectively, with  the golden number ‰1 ‡ 5Š=2.
Their application and the illustrative examples are mainly The Laue point group PL of a crystal is the point symmetry
restricted to the modulated crystal case. The tables presented group of its diffraction pattern. This subgroup of the orthogonal
here, even if applicable to more general cases, cover essentially group O 3† is of ®nite order. The ®nite order of the group
the one-dimensional modulated case [d ˆ 1 in equation follows from the discreteness of the (resolved) Bragg peak
(9.8.1.1)]. positions, implying a ®nite number only of peaks of the same
intensity lying at a given distance from the origin.
9.8.1.2. The basic ideas of higher-dimensional crystallography Under a symmetry rotation R, the indices of each re¯ection
are transformed into those of the re¯ection at the rotated
Incommensurate modulated crystals are systems that do not
position. Therefore, a symmetry rotation is represented by an
obey the classical requirements for crystals. Nevertheless, their
n  n non-singular matrix R† with integral entries. Accord-
long-range order is as perfect as that of ordinary crystals. In the
ingly, a Laue point group admits an n-dimensional faithful
diffraction pattern they also show sharp, well separated spots,
integral representation PL †. Because any ®nite group of
and in the morphology ¯at faces. Dendritic crystallization with
matrices is equivalent with a group of orthogonal matrices of the
typical point-group symmetry is observed in both commensurate
same dimension, it follows that:
and incommensurate materials. Therefore, we shall consider
(1) PL is isomorphic to an n-dimensional crystallographic
both as crystalline phases and generalize for that reason the
point group;
concept of a crystal. The positions of the Bragg diffraction peaks
(2) there exists a lattice basis as1 ; . . . ; asn of a Euclidean
given in (9.8.1.1) are a special case. In general, they are
n-dimensional (reciprocal) space, which projects on the Fourier
elements of a vector module M* and can be written as
wavevectors a1 ; . . . ; an ;
P
n (3) the three-dimensional Fourier components ^ h1 ; . . . ; hn † of
Hˆ hi ai ; integers hi . 9:8:1:2† the crystal density function  r† can be attached to corresponding
iˆ1
points of an n-dimensional reciprocal lattice and considered as
This leads to the following de®nition of crystal. the Fourier components of a density function s rs † having lattice
An ideal crystal is considered to be a matter distribution periodicity in that higher-dimensional space (rs is an n-dimen-
having Fourier wavevectors expressible as an integral linear sional position vector).
combination of a ®nite number (say n) of them and such that its Such a procedure is called the superspace embedding of the
diffraction pattern is characterized by a discrete set of resolved crystal.
Bragg peaks, which can be indexed accordingly by a set of n Note that this procedure only involves a reinterpretation of the
integers h1 ; h2 ; . . . ; hn . structural data (expressed in terms of Fourier coef®cients): the
Implicit in this de®nition is the possibility of neglecting structural information in the n-dimensional space is exactly the
diffraction intensities below a given threshold, allowing one to same as that in the three-dimensional description.
identify and to label individual Bragg peaks even when n is larger The symmetry group of a crystal is then de®ned as the
than the dimension m of the crystal, which is usually three. Euclidean symmetry group of the crystal structure embedded in
Actually, for incommensurate crystals, the unresolved Bragg the superspace.
peaks of arbitrarily small intensities form a dense set, because Accordingly, the symmetry of a crystal whose diffraction
they may come arbitrarily close to each other. pattern is labelled by n integral indices is an n-dimensional space
Here some typical examples are indicated. In the normal group.
crystal case, n ˆ m ˆ 3 and a1 ; a2 ; a3 are conventionally denoted The equivalence relation of these symmetry groups follows
by a ; b ; c and the indices h1 ; h2 ; h3 by h; k; l. from the requirement of invariance of the equivalence class with
In the case of a one-dimensionally modulated crystal, which respect to the various possible choices of bases and embeddings.
can be described as a periodic plane wave deformation of a Because of that, the equivalence relation is not simply the one
normal crystal (de®ning the basic structure), one has n ˆ 4. valid for n-dimensional crystallography.

908

909 s:\ITFC\ch-9-8.3d (Tables of Crystallography)

 9.8. INCOMMENSURATE AND COMMENSURATE MODULATED STRUCTURES
P
In the case of modulated crystals, such an equivalence relation

H mq† ˆ  H mq ha kb lc †:
has been worked out explicitly, giving rise to the concept of the h;k;l
3 ‡ d†-dimensional superspace group.
Consequently, the structure factor SH vanishes unless there are
The concepts of point group, lattice holohedry, Bravais
integers h, k, l, and m such that
classes, systems of non-primitive translations, and so on, then
follow from the general properties of n-dimensional space H ˆ ha ‡ kb ‡ lc ‡ mq: 9:8:1:7†
groups together with the (appropriate) equivalence relations.
A glossary of symbols is given in Appendix A and a list of If q is incommensurate, i.e. if there is no integer N such that Nq
de®nitions in Appendix B. belongs to the reciprocal lattice spanned by a ; b ; c , one needs
more than three integers (in the present case four) for indexing
9.8.1.3. The simple case of a displacively modulated crystal H. This is characteristic for incommensurate crystal phases.
In the diffraction pattern of such a modulated phase, one
9.8.1.3.1. The diffraction pattern distinguishes between main re¯ections (for which m ˆ 0† and
To introduce what follows, the simple case of a displacively satellites (for which m 6ˆ 0†. The intensities of the satellites fall
modulated crystal structure is considered. The point-atom off rapidly for large m so that the observed diffraction spots
approximation is adopted and the modulation is supposed to be remain separated, although vectors of the form (9.8.1.7) may
a sinusoidal plane wave. come arbitrarily close to each other.
This means that the structure can be described in terms of In the commensurate case also, there are main re¯ections and
atomic positions of a basic structure with three-dimensional satellites, but, since here there is an integer N such that Nq
space-group symmetry, periodically displaced according to the belongs to the reciprocal lattice, one may restrict the values of m
modulation wave. Writing for the position of the jth particle in in (9.8.1.7) to the range from 0 to N 1.
the unit cell of the basic structure given by the lattice vector n:
9.8.1.3.2. The symmetry
r0 n; j† ˆ n ‡ rj ; 9:8:1:3†
There is more than one way for expressing the long-range
the position of the same particle in the modulated structure is order present in an incommensurate crystal in terms of
given by symmetry. One natural way is to adopt the point of view that
the measuring process limits the precision in the determination
r n; j† ˆ n ‡ rj ‡ Uj sin‰2q
n ‡ rj † ‡ 'j Š; 9:8:1:4† of a modulation wavevector. Accordingly, one can try an
where q is the wavevector of the modulation and Uj is the approximation of the modulation wavevector q by a commen-
polarization vector for the jth particle's modulation. (This is not surate one: an irrational number can be approximated
the most general sinusoidal modulation, because different arbitrarily well by a rational one.
components Uj may have different phases 'j ; ˆ 1; 2; 3:† In There are two main disadvantages in this approach. Firstly, a
general, the symmetry of the modulated structure is different good approximation implies, in general, a large unit cell for the
from that of the basic structure and the only translations that corresponding superstructure, which involves a large number of
leave the modulated structure invariant are those lattice parameters. Secondly, the space group one ®nds may depend
translations m of the basic structure satisfying the condition essentially on the rational approximation adopted. Consider, for
q
m ˆ integer. If the components , , and of the wavevector example, an orthorhombic basic structure with space group
q with respect to the basis a ; b ; c are all rational numbers, Pcm21 and modulation with wavevector q ˆ c , polarization
there is a full lattice of such translations, the structure then is a along the b direction and positions of the particles given by
superstructure of the original one and has space-group r n; j† ˆ n ‡ rj ‡ U sin‰2q
n ‡ rj †Š; 9:8:1:8†
symmetry. If at least one of the ; , and is irrational, the
structure does not have three-dimensional lattice translation where for convenience the polarization has been taken as
symmetry. Nevertheless, the crystal structure is by no means independent of j. Under the glide re¯ection fmx j m ‡ 12†cg
disordered: it is fully determined by the basic structure and the (integer m) present in the basic structure, the transformed
modulation wave(s). positions are
The crystalline order is re¯ected in the structure factor, which
r0 n; j† ˆ r ‡ U sinf2‰q
r m ‡ 12†Šg;
is given by the expression
P with r ˆ n ‡ rj : 9:8:1:9†
SH ˆ fj exp‰2iH
r n; j†Š
n; j Consider the rational approximation of given by P=Q, with P
P and Q relatively prime integers. Then such a glide transforma-
ˆ fj exp‰2iH
n ‡ rj †Š
n; j tion as given above is (for certain values of m) a symmetry of the
 expf2iH
Uj sin‰2q
n ‡ rj † ‡ 'j Šg; 9:8:1:5† modulated structure only if P is even and Q odd, because only in
that case does the equation P m ‡ 12† ˆ lQ have a solution for
where fj is the atomic scattering factor (which still, in general, integers l and m. Analogously, the mirror my only occurs as a
depends on H). Using the Jacobi±Anger relation, one can rewrite glide plane if P=Q is odd=even, whereas the screw axis along the
(9.8.1.5) as z axis requires the case odd=odd. Hence, if, for example, one
approximates the same irrational number successively by 5=12,
PP P
1
SH ˆ exp‰2i H mq†
n ‡ rj †Š 7=17 or 8=19, one ®nds different symmetry groups. Further-
n j mˆ 1 more, in all the cases, the point group is monoclinic and the
 fj exp im'j † J 2H
Uj †; 9:8:1:6† information contained in the orthorhombic point-group symme-
m
try of the main re¯ections is lost.
where Jm x† is the mth-order Bessel function. The summation These dif®culties are avoided if one embeds the modulated
over n results in a sum of  functions on the positions of the crystal according to the basic ideas expressed in the previous
reciprocal lattice: section. Conceptually, it corresponds (de Wolff, 1974, 1977;

909

910 s:\ITFC\ch-9-8.3d (Tables of Crystallography)

 9. BASIC STRUCTURAL FEATURES
Janner & Janssen, 1977) to enlarging the class of symmetry P
4
transformations admitted, from Euclidean in three dimensions to rs ˆ xi asi ˆ r; rI †;
iˆ1
Euclidean in n dimensions (here n ˆ 4). A four-dimensional
Euclidean operation transforms the three-dimensional structure with
but does not leave distances invariant in general. To illustrate
this phenomenon, consider the effect of a lattice translation of r ˆ x1 a ‡ x2 b ‡ x3 c; and rI ˆ t ˆ x4 q
r: 9:8:1:12†
the basic structure on the modulation: it gives rise to a shift of Some, or sometimes all, transformations of the space group of
the phase. Combining such a translation with an appropriate the basic structure give rise to a symmetry transformation for
(compensating) shift of the phase, one obtains a transformation the modulated structure when combined with an appropriate
that leaves the basic structure and the modulation invariant. phase shift, possibly together with an inversion of the phase. If
Such a transformation corresponds precisely to a lattice fRjvg is an element of the space group of the basic structure,

translation in four dimensions for the embedded crystal is a phase shift and " ˆ 1, then the point r; t† is transformed
structure. In the real space, however, a shift of the phase is to Rr ‡ v; "t ‡
†. The pattern (9.8.1.10b) is left invariant by
not an admitted Euclidean transformation as it does not leave fRjvg; f"j
g† if
invariant the distance between the particles, so that the
combination of lattice translation and (compensating) phase Rn ‡ Rrj ‡ v ‡ Ruj ‰q
n ‡ rj † ‡ "t "
Š; t†
shift also is not admitted as such, even if for an incommensurate 0 0
ˆ n ‡ rj 0 ‡ uj 0 ‰q
n ‡ rj 0 † ‡ tŠ; t†: 9:8:1:13†
modulation it simply corresponds to a relabelling of the atoms
leaving the global structure invariant. The position n0 ‡ rj 0 is the transform of n ‡ rj in the basic
From what has been said, it should be apparent that the four- structure and therefore also belongs to the basic structure.
dimensional embedding of the modulated structure according to One can conclude that the pattern has as symmetry
the superspace approach can also be obtained directly by transformations all elements fRjvg; f"j
g† satisfying
considering as additional coordinate the phase t of the (9.8.1.13). These form a space group in four dimensions.
modulation. The positions in three dimensions are then the The reciprocal to the basis (9.8.1.11) is
intersection of the lines of the atomic positions (parametrized by
t) with the hyperplane t ˆ 0. For the modulated structure of as1 ˆ a ; 0†; as2 ˆ b ; 0†;
9:8:1:14†
(9.8.1.4), these lines are given by as3 ˆ c ; 0†; as4 ˆ q; 1†:
n ‡ rj ‡ Uj sinf2‰q
n ‡ rj † ‡ tŠ ‡ 'j g; t† A general reciprocal-lattice vector is now
any real t; 9:8:1:10a† P
4
Hs ˆ hi asi ˆ h1 a ‡ h2 b ‡ h3 c ‡ h4 q; h4 †:
iˆ1
and for a more general, not necessarily sinusoidal, modulation
by The projection of this reciprocal-lattice vector is of the form
(9.8.1.7), as it should be. Moreover, the projection of the four-
n ‡ rj ‡ uj fq
‰n ‡ rj Š ‡ tg; t†; 9:8:1:10b†
dimensional Fourier transform on the hyperplane t ˆ 0 is exactly
where uj x† is a periodic vector function: uj x ‡ 1† ˆ uj x†. A the Fourier transform of the structure in this hyperplane. If, as a
general point in four-dimensional space is rs ˆ r; t† with consequence of the four-dimensional space-group symmetry, the
r ˆ x; y; z†. four-dimensional diffraction pattern shows systematic extinc-
The notation adopted here for the 3 ‡ 1†-dimensional tions, the same extinctions are present in the diffraction pattern
embedding of the modulated structure is a shorthand notation of the three-dimensional structure.
intended to stress the double role of the variable t as a parameter There are also other ways to describe the symmetry relations
in the three-dimensional description of the structure and as a of incommensurate modulated structures. One is based on
coordinate in the 3 ‡ 1†-dimensional embedding. In the latter representation theory. This method has in particular been used
case, it implicitly assumes the choice of a fourth basis vector d when the modulation occurs as a consequence of a soft mode.
perpendicular to the physical space of the crystal, so that the Then the irreducible representation of the space group to which
following four-dimensional vector notations are considered to be the soft mode belongs gives information on the modulation
equivalent: function as well. Heine & McConnell have treated the symmetry
with a method related to the Landau theory of phase transitions
rs ˆ r; t† ˆ r; td† ˆ r; t†: (Heine & McConnell, 1981; McConnell & Heine, 1984).
The pattern of lines (9.8.1.10a; b) has lattice periodicity. Perez-Mato et al. have given a formulation in terms of three-
Indeed, it is invariant under the shift t ! t ‡ 1 and, for every dimensional structures parametrized by what are here the
lattice translation n from the basic structure, there is a additional coordinates (Perez-Mato, Madariaga & Tello, 1984,
compensating phase shift: the pattern is left invariant under the 1986; Perez-Mato, Madariaga, ZunÄiga & Garcia Arribas, 1987).
combination of the translation n and the phase shift Other treatments of the problem can be found in Koptsik (1978),
t ! t q
n. Therefore, (9.8.1.10a; b) is invariant under Lifshitz (1996), and de Wolff (1984).
translations from a four-dimensional lattice with basis
9.8.1.4. Basic symmetry considerations
as1 ˆ a; q
a†; as2 ˆ b; q
b†;
9:8:1:11† 9.8.1.4.1. Bravais classes of vector modules
as3 ˆ c; q
c†; as4 ˆ 0; 1†:
For a modulated crystal structure with a one-dimensional
In accordance with the equivalent descriptions given for rs , modulation, the positions of the diffraction spots are given by
equivalent descriptions for as4 are as4 ˆ 0; 1† ˆ 0; d† with dual vectors
basis vector as4 ˆ q; 1† ˆ q; d †, where d is reciprocal to d in P
3
the fourth direction. With respect to the axes (9.8.1.11), a Hˆ hi ai ‡ mq: 9:8:1:15†
general point in four dimensions can be written as iˆ1

910

911 s:\ITFC\ch-9-8.3d (Tables of Crystallography)

 9.8. INCOMMENSURATE AND COMMENSURATE MODULATED STRUCTURES
This set of vectors is a vector module M  . The vectors a1 ; a2 ; a3 where [cf. (9.8.1.14)] m has now been replaced by h4 and
form a basis of the reciprocal lattice  of the basic structure and
asi ˆ ai ; 0†; i ˆ 1; 2; 3; as4 ˆ q; 1†: 9:8:1:19†
q is the modulation wavevector. The choice of the basis of 
has the usual freedom, the wavevector q is only determined up to As will be explained in Section 9.8.4, these vectors span the
a sign and up to a reciprocal-lattice vector of the basic structure. four-dimensional reciprocal lattice for a periodic structure
A vector module M  has point-group symmetry K, which is having as three-dimensional intersection (say de®ned by the
the subgroup of all elements R of O 3† leaving it invariant. hyperplane t ˆ 0† the modulated crystal structure (a speci®c
In the case of an incommensurate one-dimensional modula- example has been given in Subsection 9.8.1.3). In direct space,
tion, M  is generated by the lattice  of main re¯ections and the the point group Ks in four dimensions with elements Rs of O 4†
modulation wavevector q. It then follows that K is characterized then acts on the corresponding dual basis vectors (9.8.1.11) of
by the following properties: the four-dimensional direct lattice as
(1) It leaves  invariant. (Only in this way are main
re¯ections transformed into main re¯ections and satellites into P
4
Rs asi ˆ R†ji asj i ˆ 1; 2; 3; 4†; 9:8:1:20a†
satellites.) jˆ1
(2) Any element R of K then transforms q into q (modulo
reciprocal-lattice vectors of  ). where R† is the transpose of the matrix  R 1 † appearing in
An element R of K then transforms the basic vectors a1 , a2 , a3 , (9.8.1.17) and therefore for incommensurate one-dimensionally
q into ones of the form (9.8.1.15). If one denotes, as in modulated structures it has the form
 
(9.8.1.2), q by a4 , this implies E R† 0
R† ˆ : 9:8:1:20b†
P
4
 M R† " R†
Rai ˆ R†ji aj ; i ˆ 1; . . . ; 4; 9:8:1:16†
jˆ1

with  R† a 4  4 matrix with integral entries. In the case of an 9.8.1.4.3. Four-dimensional crystallography
incommensurate modulated crystal structure, only two vectors Let us summarize the results obtained in the previous
with the same length as q are q and q. As  is left invariant, it paragraph. The matrices R† form a faithful integral repre-
follows that for a one-dimensionally modulated structure  R† sentation of the three-dimensional point group K with a four-
has the form dimensional carrier space Vs . It is a reducible representation
  
 having as invariant subspaces the physical three-dimensional
 E R† M R†
R† ˆ ; where " R† ˆ 1: 9:8:1:17† space, denoted by V (or sometimes also by VE ), and the
0 " R†
additional one-dimensional space, denoted by VI . In V, the four-
This matrix represents the orthogonal transformation R when dimensional point-group transformation acts as R (sometimes
referred to the basis vectors ai i ˆ 1; 2; 3; 4† of the vector also denoted by RE ), in VI it acts as one of the two one-
module M  . As in the case of lattices, two vector modules of dimensional point-group transformations: the identity or the
modulated crystals are equivalent if they have bases (i.e. a basis inversion. Therefore, the space Vs can be made Euclidean with
for the reciprocal lattice  of the basic structure together with a R† de®ning a four-dimensional point-group transformation Rs ,
modulation wavevector q) such that the set of matrices  K† which is an element of a crystallographic subgroup Ks of O 4†.
representing their symmetry is the same for both vector modules. The four-dimensional point-group transformations are of the
Equivalent vector modules form a Bravais class. form R; "†, with " ˆ 1 and they act on the four-dimensional
Again, as in the case of three-dimensional lattices, it is lattice basis as
sometimes convenient to consider a vector module that includes
P
4
as subset the one spanned by all diffraction spots as in (9.8.1.15). R; "†asi ˆ R†ji asj ; i ˆ 1; . . . ; 4; 9:8:1:21†
Within such a larger vector module, the actual diffraction peaks jˆ1
then obey centring conditions. For a vector module associated
where " stands for " R† as in (9.8.1.20b). So the point-group
with a modulated structure, centring may involve main
symmetry operations are crystallographic and given by pairs of a
re¯ections (the basic structure then has a centred lattice), or
three-dimensional crystallographic point-group transformation
satellites, or both. For example, if in a structure with primitive
and a one-dimensional " ˆ 1, respectively. The case " ˆ 1
orthorhombic basic structure the modulation wavevector is given
corresponds to an inversion of the phase of the modulation
by a1 ‡ 12 a2 , one may describe the diffraction spots by means of
function.
the non-primitive lattice basis a1 ; 12 a2 ; a3 and by the modulation
As in the three-dimensional case, one can de®ne equivalence
wavevector a1 .
classes among those four-dimensional point groups.
Crystallographic point groups are denoted generally by the
Two point groups Ks and Ks0 belong to the same geometric
same letter K.
crystal class if their three-dimensional (external) parts (forming
the point group KE and KE0 , respectively) are in the same three-
9.8.1.4.2. Description in four dimensions
dimensional crystal class [i.e. are conjugated subgroups of O 3†]
The matrices  R† form a faithful integral representation of and their one-dimensional internal parts (forming the point
the three-dimensional point group K. It is also possible to groups KI and KI0 , respectively) are equal. The latter condition
consider them as four-dimensional orthogonal transformations implies that corresponding point-group elements have the same
leaving a lattice with basis vectors (9.8.1.14) invariant. Indeed, value of ".
one can consider the vectors (9.8.1.15) as projections of Such a geometric crystal class can then be denoted by the
four-dimensional lattice vectors Hs ˆ H; HI †, which can be symbol of the three-dimensional crystal class together with the
written as values of " that correspond to the generators.
P
4 Also, the notion of arithmetic equivalence can be generalized
Hs ˆ hi asi ; 9:8:1:18† to these four-dimensional point groups, as they admit the same
iˆ1 faithful integral representation K† given above. This means

911

912 s:\ITFC\ch-9-8.3d (Tables of Crystallography)

 9. BASIC STRUCTURAL FEATURES
that two such groups are arithmetically equivalent if there is a Tables for Crystallography, a symbol that determines ", and one
basis transformation for the reciprocal-vector module, which for the corresponding intrinsic internal translation .
transforms main re¯ections into main re¯ections and satellites
into satellites and which transforms one of the matrix groups 9.8.1.4.4. Generalized nomenclature
into the other. The arithmetic classes are determined by the
In Section 9.8.4, the theory is extended to structures
arithmetic equivalence class of the three-dimensional group KE
containing d modulations, with d  1. In this case, each point-
[i.e. by E K†] and by the components of the modulation
group transformation in internal space is given by RI and the
wavevector with respect to the corresponding reciprocal-lattice
associated internal translation by the (d-dimensional) vector vI .
basis. This is because the elements " are ®xed by the relation
Thus,
Rq  "q (modulo reciprocal-lattice vectors
9:8:1:22† gs ˆ f R; RI †j v; vI †g:
of the basic structure).
The transformations R and RI are represented by the matrices
Note that these 3 ‡ 1†-dimensional equivalence classes are not
E R† and I R†, respectively. In the following discussion, this
simply those one obtains in four-dimensional crystallography, as nomenclature (but with vI rather than vI ) is sometimes also
the relation between the higher-dimensional space Vs and the applied for the 3 ‡ 1†-dimensional case. The usual formulae are
three-dimensional physical space V plays a fundamental role. obtained by replacing RI by " and vI by
.
The embedded structures in four dimensions have lattice
periodicity. So the symmetry groups are four-dimensional space 9.8.1.4.5. Four-dimensional space groups
groups, called superspace groups. The new name has been
introduced because of the privileged role played by the three- Four-dimensional space groups were obtained in the
dimensional subspace V. A superspace-group element gs consists 3 ‡ 1†-reducible case by Fast & Janssen (1969) and in the
of a point-group transformation R; "† and a translation v;
†. general case by Brown, BuÈlow, NeubuÈser, Wondratschek &
The action of such an element on the four-dimensional space is Zassenhaus (1978). The groups were determined on the basis of
then given by algorithms developed by Zassenhaus (1948), Janssen, Janner &
Ascher (1969), Brown (1969), and Fast & Janssen (1971). In the
gs rs ˆ f R; "†j v;
†g r; t† ˆ Rr ‡ v; "t ‡
†: 9:8:1:23† book by Brown, BuÈlow, NeubuÈser, Wondratschek & Zassen-
It is important to realize that a superspace-group symmetry of an haus, quoted above, a mathematical characterization of the basic
embedded crystal induces three-dimensional transformations crystallographic concepts is given together with corresponding
leaving the original modulated structure invariant. Correspond- tables for the dimensions one, two, three, and four. One ®nds
ing to (9.8.1.23), one obtains the following relations [cf. there, in particular, a full list of four±dimensional space groups.
(9.8.1.13)]: The list by Fast & Janssen is restricted to space groups with
3 ‡ 1†-reducible point groups. The four-dimensional groups in
uj 0 ‰q
n0 ‡ rj 0 †Š ˆ Ruj ‰q
n ‡ rj † "
Š 9:8:1:24† the work of Brown et al. are labelled by numbers. For these
with same groups, alternative symbols have been developed by
Weigel, Phan, Veysseyre and Grebille generalizing the princi-
n0 ‡ rj 0 ˆ R n ‡ rj † ‡ v: ples of the notation adopted by International Tables for
Crystallography, Volume A, for the three-dimensional space
These are purely three-dimensional symmetry relations, but of groups (Weigel, Phan & Veysseyre, 1987; Veysseyre & Weigel,
course not Euclidean ones. 1989; Grebille, Weigel, Veysseyre & Phan, 1990).
In three-dimensional Euclidean space, the types of space- The difference in the listing of four-dimensional crystal-
group transformation are translations, rotations, rotoinversions, lographic groups one ®nds in Brown et al. and in Weigel et al.
re¯ections, central inversion, screw rotations, and glide planes. with respect to that in the present tables is not simply a matter of
Only the latter two transformations have intrinsic non-primitive notation. In the ®rst place, here only those groups appear that
translations. For superspace groups, the types of transformation can occur as symmetry groups of one-dimensional incommensu-
are determined by the point-group transformations. By an rate modulated phases (there are 371 such space groups).
appropriate choice of the basis in Vs , each of the latter can be Furthermore, as already mentioned, a ®ner equivalence relation
brought into the form has been considered that re¯ects the freedom one has in
0 1
cos ' sin ' 0 0 embedding a three-dimensional modulated structure in a four-
B sin ' cos ' 0 0 C dimensional Euclidean space. Instead of 371, one then obtains
B C; ";  ˆ 1: 9:8:1:25†
@ 0 0  0A 775 inequivalent groups for which the name superspace group
0 0 0 " has been introduced. A 3 ‡ 1†-dimensional superspace group is
thus a four-dimensional space group having some additional
By a choice of origin, each translational part can be reduced to its properties. In Section 9.8.4, the precise de®nitions are given.
intrinsic part, which in combination with the point-group In the commensurate one-dimensionally modulated case, 3833
element R; "† gives one of the transformations in V indicated four-dimensional space groups may occur, out of which 320
above together with the inversion, or the identity, or a shift in VI . already belong to the previous 371. The corresponding additional
So, for phase inversion (when " ˆ 1), the intrinsic shift in VI 3 ‡ 1†-dimensional superspace groups are also present in the
vanishes. When " ˆ ‡1, the intrinsic shift in VI is given by . It listing by Fast & Janssen (1969) and have been considered again
will be shown in Subsection 9.8.3.3 that the value of  is one of (and applied to structure determination) by van Smaalen (1987).
1 1 1 1 The Bravais classes for the commensurate 3 ‡ 1†-dimensional
0; ; ; ; : 9:8:1:26† case are given in Table 9.8.3.2 b†.
2 3 4 6 The relation between modulated crystals and the superspace
Therefore, a superspace-group element can be denoted by a groups is treated in a textbook by Opechowski (1986). That
symbol that consists of a symbol for the three-dimensional part between the superspace-group symbols of the present tables and
following the conventions given in Volume A of International those of Weigel et al. is discussed in Grebille et al. (1990).

912

913 s:\ITFC\ch-9-8.3d (Tables of Crystallography)

 9.8. INCOMMENSURATE AND COMMENSURATE MODULATED STRUCTURES
Note that no new names have been introduced for the that transforms every diffraction peak into a peak of the
underlying crystallographic concepts like Bravais classes, same intensity.y
geometric and arithmetic crystal classes, even if in those cases
also the equivalence relation is not simply that of four- As PL leaves invariant the subset of main re¯ections, this Laue
dimensional Euclidean crystallography, an explicit distinction group belongs to one of the 11 Laue symmetry classes.
always being possible by specifying the dimension as 3 ‡ 1† Accordingly, the Laue group determines a three-dimensional
instead of four. holohedral point group which determines a crystallographic
system.
9.8.1.5. Occupation modulation
(2) The second step consists of choosing a basis according to the
Another type of modulation, the occupation modulation, can
conventions of IT A for the main re¯ections and choosing a
be treated in a way similar to the displacive modulation. As an
modulation wavevector.
example consider an alloy where the positions of the basic
structure have space-group symmetry, but are statistically
From the centring extinctions, one can deduce to which Bravais
occupied by either of two types of atoms. Suppose that the
class the main re¯ections belong. This is one of the 14 three-
position r is occupied by an atom of type A with probability p r†
dimensional Bravais classes. Notice that the cubic Bravais
and by one of type B with probability 1 p r† and that p is
classes do not occur because a one-dimensional (incommensu-
periodic. The probability of ®nding an A atom at site n ‡ rj is
rate) modulation is incompatible with cubic symmetry. For this
PA n ‡ rj † ˆ pj ‰q
n ‡ rj †Š; 9:8:1:27† same reason, only the nine non-cubic Laue-symmetry classes
occur in the one-dimensional incommensurate case.
with pj x† ˆ pj x ‡ 1†. In this case, the structure factor becomes The main re¯ections are indexed by hkl0 and the satellite
PP

SH ˆ fA pj ‰q
n ‡ rj †Š ‡ fB f1 pj ‰q
n ‡ rj †Šg re¯ections by hklm. The Fourier wavevector of a general
n j re¯ection hklm is given by

 exp‰2iH
n ‡ rj †Š ; 9:8:1:28† H ˆ ha ‡ kb ‡ lc ‡ mq: 9:8:2:1†
where fA and fB are the atomic scattering factors. Because of the Note that this step involves a choice because the system of
periodicity, one has satellite re¯ections is only de®ned modulo the main re¯ections.
P
pj x† ˆ wjm exp 2imx†: 9:8:1:29† When a satellite is in the vicinity of a main re¯ection, it is
m reasonable to assign it to that re¯ection. But one has, especially
Hence, when deciding whether or not situations are equivalent, to be
 aware of the fact that each satellite may be assigned to an
P P arbitrary main re¯ection. It is even possible to assign a satellite
SH ˆ fB
H† exp 2iH
rj † ‡ fA fB †
H ‡ mq†wjm
j m to an extinct main re¯ection. One takes by preference the q
 vector along a symmetry axis or in a mirror plane. According to
 exp‰2i H ‡ mq†
rj Š ; 9:8:1:30† equation (9.8.2.1), the fourth basis vector a4 is equal to the
chosen q, the modulation wavevector.
where
H† is the sum of  functions over the reciprocal lattice
of the basic structure: (3) In the third step, one determines the space group of the
  average structure (from the main re¯ections).
P P
3

H† ˆ  H hi ai :
h 1 h2 h 3 iˆ1 The average structure is unique but possibly involves split atoms.
The space group of the average structure is often the symmetry
Consequently, the diffraction peaks occur at positions H given group of the undistorted phase. That helps to make a good choice
by (9.8.1.7). For a simple sinusoidal modulation ‰m ˆ 1 in for the basic structure and also gives an insight as to how the
(9.8.1.29)], there are only main re¯ections and ®rst-order satellite re¯ections split from the main re¯ections at the phase
satellites m ˆ 1†. One may introduce an additional coordinate transition.
t and generalize (9.8.1.27) to
PA n ‡ rj ; t† ˆ pj ‰q
n ‡ rj † ‡ tŠ; 9:8:1:31† (4) Step four is the identi®cation of the 3 ‡ 1†-dimensional
Bravais lattice type. In superspace also, centring gives rise
which has 3 ‡ 1†-dimensional space-group symmetry. General- to centring extinctions, and that corresponds to making the
ization to more complex modulation cases is then straight- choice of a conventional unit cell in 3 ‡ 1† dimensions.
forward.
The previous three steps establish a , b , c , the three-
9.8.2. Outline for a superspace-group determination dimensional Bravais class and q ˆ a ‡ b ‡ c , where the
components , , and are given with respect to the three-
In the case of a modulated structure, the diffraction pattern dimensional conventional basis.
consists of main re¯ections and satellites. The main re¯ections
span a reciprocal lattice generated by a1 ; a2 ; a3 . Considerations ˆ q
a; ˆ q
b; ˆ q
c: 9:8:2:2†
are here restricted for simplicity to the one-dimensional The 3 ‡ 1†-dimensional Bravais class is ®xed by that three-
modulated case, i.e. to the n ˆ 4 case. Extension to the more dimensional Bravais class and the components ; ; of q.
general n ˆ 3 ‡ d case is conceptually not dif®cult and does not Just as for three-dimensional lattices, a conventional cell can
modify the general procedure outlined here. be chosen for 3 ‡ 1†-dimensional lattices. To this end, the
(1) The ®rst step is the determination of the Laue group PL of the y Except for deviations from Friedel's law caused by dispersion; see IT B (1993,
diffraction pattern: it is the point group in three dimensions p. 241, Subsection 2.3.4.1).

913

references
914 s:\ITFC\ch-9-8.3d (Tables of Crystallography)