HAL Id: jpa-00246438
https://hal.archives-ouvertes.fr/jpa-00246438
Submitted on 1 Jan 1991
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Quasiperiodic models with microcrystalline structures
Michel Duneau
To cite this version:
Michel Duneau. Quasiperiodic models with microcrystalline structures. Journal de Physique I, EDP
Sciences, 1991, 1 (11), pp.1591-1601. �10.1051/jp1:1991227�. �jpa-00246438�
J.
Phys.
I France 1(1991)
1591-1601 NOVEMBRE1991, PAGE 1591Classification
Physics
Abstracts61.50K 61.70N M.70K
Quasiperiodic models with microcrystalline structures
Michel Duneau
Centre de
Physique Th60rique,
EcolePolytechnique,
91128 Palaiseau Cedex, France(Received17
May 1991, revbed 5July
1991,accepted
8 August 1991)Rksumk, Nous montrons comment la mdthode de coupe et
projection
pout dtreadaptbe
I la construction de moddlesgkomktriques prksentant
une structuremicrocrystalline.
La construction est illustrbe parl'exemple
du pavageoctogonal.
Premidrement, une brisurepartielle
de lasymdtrie
pennet de construire de nouveaux pavagesprbsentant
deux bchellesinddpendantes
:l'une est associbe I la taille des mailles et l'autre intervient dans les
parois
entre domaines.L'ensemble des mailles a une structure modulbe
simple.
Ensuite les mailles sont dbcordes selon le motif d'une structurepbriodique approximante,
conduisant ainsi I un moddle microcristallincohbrent. La transformbe de Fourier est calculbe exactement et montre des taches satellites
caractkristiques, dkpendantes
desparamdtres
del'approximant, ajoutdes
au spectrequasipkriodi-
que de rkfbrence.
Abstract. We show how the standard cut and
project
method can beadapted
to buildquasiperiodic geometrical
modelsdisplaying
amicrocrystalline
structure. Theprocedure
is illustrated with theexample
of theoctagonal tilings.
First, apartial
symmetrybreaking
allows to build newfilings
with twoindependent
scales : one is associated to the size of the unit cells while the other occurs in walls between domains. Thecorresponding
array of unit cells has asimple
modulated structure. Then the unit cells are decorated
according
to aperiodic approximant
structure, thusproviding
a coherentmicrocrystalline
model. The Fourier spectrum can becomputed
exactly and shows characteristic satellite spots, added to thequasiperiodic
reference spectrum, whichdepend
on theapproximant
parameters.I. Introducdon.
Since the
discovery
of icosahedralquasicrystals
in temaryalloys
of Almnsi[I]
a number of otherquasiperiodic
structures have beenprepared
ans studiedby
Various methods and othernon
crystallographic symmetries
were indicatedby
diffractionexperiments
andhigh
resolution
microscopy (see
for instance[2~4]).
Morerecently,
thephase diagram
of the Alfecusystem
wasclosely analysed [5]
in theregion
of the icosahedralphase, showing
abiphased
state afterannealing.
Transitions betweenquasicrystalline phases
andmicrocrystal-
line
phases
have beenrecently suggested
or observed[6-1ii.
Recent observations of the AlLicu system showed acomplex
structure with a network of translation walls[12].
All these observations suggest that theQC
states could beclosely related, by
means ofdisplacive transformations,
tomicrocrystalline
states withspecial
kinds of defects and raise thequestion
of the
possible
structuralrelationships
betweenquasicrystals
andcrystalline phases.
1592 JOURNAL DE PHYSIQUE I M 11
From a theoretical
point
ofView, geometrical
models ofdisplacive
transformations betweenQC
andperiodic crystals
have also beenproposed recently [13-17].
In this paper we propose asimple construction,
based on the cut andproject method,
whichprovides
ageometrical
setup for models ofmicrocrystalline
structures with aglobal quasiperiodic,
orperiodic, long
rangeorder. The skeleton of these models is obtained
by
a standard CP method withmerely
a different choice of the orientation of thephysical
space in thehigher
dimensional space, orby
the choice of different lattices and acceptance domains. This
usually implies
apartial
symmetrybreaking,
with respect to thequasiperiodic
referencemodel,
which results in newtilings showing
two different characteristic scales. Thelarge
scale dictates the size of thetypical
unit cells which occur in domainsseparated by
wallsinvolving
the smaller scale. At thispoint
the ratio between the two scales can be chosenfreely.
The skeleton thus obtained can be shown tocorrespond
to a modulatedlattice,
the nodes of which are attached to the unit cells.It is different from the structures obtained in
[15]
which seems to betopologically
similar to the referencetilings.
As in the usual CP method theglobal
average orientation of thephysical
space can be set in order to have either a
quasiperiodic skeleton,
as assumedabove,
or aperiodic
one withpossibly large parameters corresponding
to someapproximant
structure.For
particular
values of the two characteristicscales,
the skeleton shows a close connection with the referencequasiperiodic
structure. These situationscorrespond
to a characteristic ratio related to the inflation parameter of theQC. Actually
a skeleton of empty unit cells can beexactly superimposed
on afully symmetric quasiperiodic
structure in the sense that thenodes of the skeleton fit with nodes of the reference structure. In this case,
however,
the unit cells cut in theQC
show a finite number ofdifferent
decorations. In order to obtain amicrocrystalline
structure we have toprescribe
aunique
and constant decoration in all unit cells.The Fourier spectrum of these structures can be
computed exactly.
Sincethey
are obtainedby
decoration of aquasiperiodic skeleton,
the Fourier transformrequires
two steps. The skeleton has a nice modulated structure and tums out to be rather close to aperiodic
array of unit cells(the approximant structure)
with a network of defect wallsrestoring quasiperiodici- ty.
On the other hand the unit cellyields
a structure factor which is close to the ideal Fourierspectrum since the unit cell can be obtained
by
a cut in the referenceQC. Finally,
theresulting
spectrum is carriedby
the referencespectrum plus
satellitespots.
Thespectrum
canbe indexed with the same basis but with fractional indices in the directions associated to the
approximant
structure. The characteristic denominator of the fractional indices is linked to theparameters
of theapproximant
and increases with the size of the unit cell.The construction is
developed
in the case of theoctagonal tiling
forsimplicity.
The same ideasapply
for 3 dimensional structures where the referenceQC
is icosahedral and where theapproximant
structures can be either cubic or rhombohedral.2. Notations and
settings.
The
octagonal tiling
is obtainedby
the cut andproject
method(CP method)
fromR~.
In order to haveedges
of unitlength
in thephysical plane
thehypercubic
lattice is definedas
Ao
=/2f~.
The twoorthogonal planes Eo
andE(
are the ranges of the
following orthogonal complementary projectors
:0
II / II /
0
II / II /
~° 2
Ill Ill
0lI / II /
0li° II
QUASIPERIODIC
MODELS WITH MICROCRYSTALLINE STRUCTURES 15930
-ill Ill
,
I o I
-ill -ill
~° 2
-ill -ill
0lll -ill
0The
projections Lo
=
po(Ao)
andL(
=
pi (Ao)
are dense 2f-modulesrespectively generated by (ej,
,
e
~)
and(e(,
,
e
()
where e~=
po(/ e~)
ande)
=
pi (/ e~).
In orthonornlal coordinateset = 1, 0
)
e=
(1,
0~~
/~~~'~~
~~~~ ~' ~~
e~ =
(0,1) e(
=
(0,1)
Thus all vectors e~ and
e)
have unitlength.
The
octagonal tilings correspond
to a cutregion
So=
Eoxwo
where the acceptance domainWo is,
up to atranslation,
theprojection
of the unit cube ofR~
in theperpendicular
space.
Wo
is aregular
octagongenerated by
the four vectors(e(,...,
e() (see Fig, I).
Thetilings
involve six different tiles(two
squares and fourrhombi)
which are theprojections
of the sixpossible
2 dimensional facets of the lattice Ao. The normalization of Aoimplies
that all tiles have alledges equal
to I.The Fourier spectrum of the
octagonal tiling
is cardedby
theprojection
Lo* inEo
of the dual latticeAt
= 2L~ =
Ao.
ThereforeLt
=
Lo
isspanned by
the basis/
2 2(et,..,
et)
wheree~* = e~ and all
Bragg peaks
can be indexed with this basis.2
The
octagonal tiling
shows aperfect
inflationsymmetry.
This is a consequence of the existence of a modular transfornlation(')
J inR~
which commutes with the action of the 8- foldsymmetry
group. J has thefollowing expression
with respect to theprojectors
po and
pi
Ii
i o -iJ
= =
(i
+/)po
+
(i /)pi
-i o i
As a consequence the Z-modules
Lo, j
Lo* andLt'
are invariant with respect to thescaling
transfornlations of parameter + 2 or I/.
Inparticular
the Fourierspectrum
iscarried
by
the scale invariant set ofpoints Lt'
Periodic
approximants
of theoctagonal filing
can beeasily
obtainedusing
theunderlying
inflationsymmetry [18].
Theapproximant
of order k is associated to thesymmetric
matrixJ~
which reads:
O~+O~_j O~
0-O~
J~
=
~~
°~ /~
~
/[ /
=
/
i)k pi
+
(
i/)kpi
O~ 0~
~
O~~~~ O~+ (~_j
(')
A modular matrix hasintegral
entries and determinant ±1.1594 JOURNAL DE PHYSIQUE I li° II
e3 3
e4 e2
~~ lleill =
lle'ill
= ~,
2 ~4
Fig.
I.- The setup of the standardoctagonal filing
with the irrationalorthogonal planes
E and E'yielding
8-fold symmetricprojections
of the cubic lattice/ Z~ (top).
Theoctagonal dungs (bottom
left) are obtained with windows
equal,
up to a translation in perpendicular space, to the octagonspanned
by e(, e(, e],e( (bottom right).
where the «Octonacci» sequence
(O~)
is definedby Oo=0, Oj= I, O~=2
and°k
+1 " ~°k
+°k
-1.
3. Par6al symmetry
breaking
of the S-fold sgnmetry.The
following
construction is devised togive
an idealgeometrical
model of aquasiperiodic (or periodic)
array oflarge
unit cells that we call the skeletons. This is obtainedby breaking
the 8- foldsymmetry
in order to have afiling
with two different scales : alarge
scale will dictate the size of the unit cells and the smaller one will be involved in thegrain
boundaries or walls between domains. In the next section we shallgive
aprocedure
to decorate these unit cells with bases such as the bases ofperiodic approximant tilings.
In this case the size of the unit cell will have to match the lattice parameters of theapproximant
lattices.Let A
=
aY~
denote the 4-dimensionalhypercubic
lattice with latticeparameter
a m I and let ~ 0satisfy a~
=
I + ~. The standard basis of A is
~).
Then there exist twoli° II
QUASIPERIODIC
MODELS WITH MICROCRYSTALLINE STRUCTURES 1595orthogonal subspaces
E and E' ofdimension 2 inR~,
withcorresponding orthogonal projections
p and
p' (see Fig. 2)
such that :p(~El)
~
(A, 0)
"
A~l p'(~~l)
"
(', 0)
"
~l
P(~E2)
"
~ (l, 1)
" ~2
P'(~82)
"
~ (~ i,
~"
~~i
P(aE3)
~
(°>
~"
~~3 P'(~E3)
"
(°, 1)
"
e(
P(aE4)
~
~ (~ l, 1)
= e4
P'(~E4)
"
~ (~,
~)
"
~e(
Thus A is the ratio of the two scales involved in the construction. If A is irrational with
respect
to/,
a condition we shall assume in the
following,
theprojections
L=
p(A)
and L'=
p'(A)
are dense Z-modules obtained as allintegral
combinations of these vectors. For A=
I we recover the usual setup of the
octagonal tiling
with full 8-foldsymmetry.
A=aZ4
a2=1+l~~
~~
'3 e3
e~i
e4 e~
l~ei
,
2
J~e4
Fig.
2. A convenient choice of the lattice parameter a and of the orientation of theplanes
E and E'gives perpendicular projections showing
characteristic scales I and A.The new
projectors
p andp'
aregiven by
:A~ All
0-All
_I All
IAll
0~
a~
oAll A~ All
-All
0All
and
~l -All
0All
,_I -All A~ -All
0~
a~
o-All
I-All
All
0-All A~
The
acceptance
domain W in E'is,
up to atranslation,
theoctagon spanned by
the vectorsp'(ae~ )
for I=
I to 4. Notice that W has
only
a four-fold symmetry. Thecorresponding tiling
1596 JOURNAL DE PHYSIQUE I li° II
of E is associated to the cut
region
S= E x W and involves the six files
generated by
the four vectors p(a
e~).
The symmetry
breaking
inducedby
A ~ limplies
that the square tileTj~ generated by
Aej and Ae~ is
larger
than the others. The acceptance domain of this tile dictates itsfrequency
in the
tiling.
Oneeasily
checks that this domain is the squareT(4 generated by Ae(
andAe(.
Therefore thefrequency
ofTj~,
as a function of A, increases with its size. Moreprecisely,
the surface of theoctagon
W is ~ + 2IA
+ I in such a way that the relative
frequency
ofTj~
reads A~/(A~ + 2/
A+ I
).
The total relative surface coveredby
these tiles increases even morerapidly
with A since asimple
calculationgives
A~/(A ~+
2 A ~+
l).
As shown in the
example
offigure 3,
thelarge
tiles arefrequently adjacent
to each other.They
fornl « domains which areseparated by
boundaries made with the smaller tiles. Due to the construction inhyperspace,
theglobal
structure maintains a strictquasiperiodic
order inthe sense that the Fourier transform of the structure is still a sum of Dirac measures which can
be indexed
by
fourintegers.
Therefore the arrangement of the domains can be considered as coherent.Fig. 3. This tiling corresponds to the geometry of figure 2 showing two scales : I and I +
/.
Thelarge
squares ofedge
I +/
represent = 42.68 fb of the tiles and 85.36 fb of the surface.
They
occur in domains of several units.The
example given
infigure
3 wascojuted
with = I +/
for which the relativefrequency
ofTj~
is(3+2/)1(8+4 2)
=0.4268 whereas thisfrequency
isonly 11(2+2/)=0.2071
in theoctagonal tiling.
Thecorresponding
surface represents afraction
(17
+12/)1(20+14 /)
= 0.8536 of the whole
tiling
instead ofI/4
in theoctagonal tiling.
An
important
feature of this construction is that the skeleton thus obtained is a modulated square lattice. This can be seen as follows. The differentlarge
tiles can be referred toby
a localorigin,
for instance the lower leftpoint
of the squaresTj~ spanned by p(aei)
and- p(ae~).
In theperpendicular
space thecorresponding points
fall inside an acceptance domainwhich
is,
up to atranslation,
thecomplementary
squareT(4 spanned by p'(ae~)
andp'(ae4).
It has beenproved
elsewhere[19]
that if anacceptance
domain tiles theperpendicular
space(with
respect to a sublattice of theprojection
L' of thehyperlattice),
thenthe structure associated to this domain has an average lattice and
consequently
can bedescribed as a modulated lattice.
In the present case the acceptance domain of the local
origins, namely T(4
tiles theperpendicular
space withrespect
to the 2D latticegenerated by p'(ae~)
andp'(ae4)
which isbt II
QUASIPERIODIC
MODELS WITH MICROCRYSTALLINE STRUCTURES 1597the
projection
of the 2D lattice A~4spanned by (ae~, ae4).
The skeleton can be indexedby
the
complementary
latticeAj~ spanned by (aej, ae~)
and the average lattice isgiven by
theoblique projection
ofAj~
on Eparallel
to the spaceE~4 containing
A~4. Asimple
calculation shows that this average lattice is a square lattice with basis(a~/Aej, a~/Ae~).
The Fourier
spectrum
of thegrain
skeleton caneasily
be derivedanalytically
since the construction is based on the CP method. The dual lattice of A is A*= a~ 2L~
=
a~~
A. The spectrum is carriedby
theprojection
of A* on E which is the Z-module L*= a~ ~ L
spanned by (a~
~Aei, a~~e~, a~~ Ae~, a~~ e4).
L* is the sum of the dual averagelattice, spanned by
(a~~ Aej,
a~ ~Ae~),
and the square latticespanned by (a~ ~e~, a~~e4)
which isresponsible
for the satellites.
In the next section we shall consider
particular
values of A for which thelarge
tiles can be identified with theempty
unit cells of anapproximant tiling:
this is achieved ifA =
(1+ /)~
forsome even (2)
integer
k.Figure
4 shows unit cells of the first twoapproximant tilings.
In this case a more convenient construction ispossible. Actually,
if J denotes the inflation matrix thenAej
= po
J~(/
ej and
Ae~
= poJ~(/ e~).
Thissuggest
toconsider the sublattice A
=
M~ Ao
of Ao associated to thefollowing integral
matrixO~+O~_j
0 0 0O~
IO~
0~"
0 0
O~+O~_j 0'
-O~
0O~
k=I k=2
Fig.
4. The unit cells of the first twoapproximant,
k=
I and k
=
2 of the
octagonal tiling
withrespectively
7 and 41nodes.The first and third columns are those of
J~
while columns 2 and 4are those of the
identity
matrix. The natural basis of A has therefore the
projections (Aei,
e~,Ae~, e4)
on E and(A e(, e(,
Ae(, e()
on E'.(~) For odd values of k the construction
gives
rise tooverlappings
of some tiles. This can be handled by aslight
modification of theprocedure (for
instance ej and e~ arechanged
into -ej ande~ in
physical space).
1598 JOURNAL DE
PHYSIQUE
I li° 11The skeleton is obtained
by
the CP method with a squareacceptance
domainspanned by e(
ande(.
The Fourier transfornl is carriedby
theprojection po(A*)
of the dual latticeA*
=
[(M~)~]~ At
ofA,
where the associated matrix isgiven by
:I
-O~
0O~
~~~~~~ ~"O~+~O~_j ~~~~) -~~
0 0 0
O~+O~_i
Consequently
the Fourier spectrum isspanned by
the basiset, et,
°k
+°k
-1
~
et, et
Thuspo(A
* contains the Fourierspectrum
of theoctagonal tiling
which°k
+ k Iis
spanned by
the basis(et, et, et, et).
The newpeaks
are due to the presence of therational fraction and can be understood as satellites in the
et,
andet
directions.O~
+O~
j
4. The
approximant octagonal filings.
Approximants
of theoctagonal tiling
have been studied in[18]
wherethey
were related to the inflationmapping
J. Anexplicit algorithm
wasgiven
to compute the unitcell,
I-e- thesize,
the number of nodes and theirpositions,
as a function of the order of theapproximant.
Moreprecisely,
for anyinteger
k m 0 a rationalplane E~
is defined asJ~ Ej~,
where J is the inflationmatrix and
Ej~
is the 2-dimensionalplane
ofR~ spanned by
ej and e~.E~
and thecomplementary orthogonal plane E( specify
the framework of the CP method where theacceptance
domain Wis,
asusual,
theprojection
of the unit cube onE( (the approximant
of order 0 is thesimple
square lattice of unitedge).
Due to the rational orientation ofE~,
theprojection
of 2L~ onE(
is a 2-dimensional latticeL(. Up
to a modular transfornlation(see [18]) L(
can be identified to a 2-dimensional square lattice. The acceptance domain W isan octagon
spanned by
the four lattice vectors of coordinates(O~
+O~_j, 0), (O~, O~), (0, O~
+O~ j)
and(- O~, O~),
where(O~)
is the Octonacci sequence mentioned insection 2.
The CP method
requires
that theboundary
@W of the acceptance domain W does notcontain any
point
ofL(.
This is achievedby
convenient relative translations t' between W and the lattice. Now on, we shall assume that theoctagon
W is fixed and that theorigin
of theplane
is one of its vertices. The translated lattice is thereforeL(
+ t' and the nodesfatling
inside W
specify
the unit cell of theperiodic tiling
as a function of t'. Now ify denotes the unit square of
L(,
oneeasily
checks that allconfigurations
of the unit cell areobtained when t' runs over y.
Moreover,
theconfigurations
are continuous w-r-t-t'in the sense that modifications of the unit cell
only
occur when theboundary
of theoctagon
meets the translated lattice
L(
+ t'. Thesesingular positions
result inflip-flops
of nodes in the unitcell,
a well knownphenomenon
inquasiperiodic tilings.
The
singular
translations t' areeasily
identified as those whichbelong
to @W modL( (see Fig. 5).
This subset is the union of theboundary
of y with its twodiagonals.
The squarey
splits
into 4triangular
domains within which the unit cell is a continuous and therefore constant function of t'. As aconclusion,
we see that four different unit cells can result form the construction.Finally,
these four domains can be relatedby symmetry operations
andconsequently,
thecorresponding
unit cells areexchanged by ar/2
successive rotations in such a way that four different lattices relatedby
ar/2
rotations are obtained. In this way theglobal
4-li° 11 QUASIPERIODIC MODELS WITH MICROCRYSTALLINE STRUCTURES 1599
w
I
o
Fig.
5. The acceptance domain for k=
I is an octagon
spanned by
the four vectors(1, 0), (1,
1),(0,
1) and(-I,I).
The square lattice is shiftedby
t' in such a way that nopoint
falls on3W. This condition means that t' does not meet 3W mod
L(
which is the union of theboundary
of y and its twodiagonals.
fold symmetry of the construction is recovered
although
agiven approximant tiling
has noparticular symmetry.
Thereforeglobal phasons,
I.e. a relative translations of thehyperlattice
w-r-t- the cut
region
resultmerely
in rotations of the lattice in thephysical
space.As
expected
from thetheory
ofphasons,
we find that two latticescorresponding
toadjacent triangles only
differby ~periodic)
lines offlip-flops.
5. The
complete nficrocrystalfine
structure.A
complete
structure is obtainedby filling
each empty unit cell of the skeleton with a basis ofan
approximant
structure. In thefollowing
we shall assume that the same basis decorates all unit cells. Thisassumption
is consistent with our purpose which is to getmicrocrystalline
models.
However,
other structures could be obtainedby allowing
different decorations in dilTerent unit cells. Thesepossibilities
could be used for instance to include some randomness in the structures. At theopposite,
the referencequasiperiodic tiling
can be seen as a skeleton of unit cells withspecific
decorations.Finally,
a lot of internlediate structures could beobtained in
principle by
either a random or a deterministicprocedure.
Figure
6 represents a part of acomplete
structure associated to the secondapproximant
of theoctagonal tiling.
The unit cells have a constant decoration of 41nodes and it can be observed that the small tiles can begiven
a constant decoration in such a way that the overallstructure is a
tiling.
In other words the tiles match of course from unit cell to unit cell but alsothrough
thegrain
boundaries.In this
situation,
where the decoration of the unit cells is constant, we canexplicitly compute
the Fourier transform of the whole structure. Thecomputation
is based on thedecomposition
of the structure into a basis and a(uasiperiodic
lattice of translation.Actually,
the basisgives
a structure factorS(q)
which can becomputed explicitly,
once the coordinates of the sites of the basis are known.However, S(q)
can be seen to be close to the Fourier transform of theperfect
reference structure.First,
the basis of anapproximant
can be identified as apart
of thequasiperiodic
reference structure.Consequently
the structure factorcan be recovered as the exact Fourier transform of the infinite structure, smeared out
by
a the Fourier transform of the cut function of the basis volume. This results in a sum of smearedBragg peaks
centered on the nodes of thereciprocal quasilattice.
1600 JOURNAL DE
PHYSIQUE
I li° IIFig.
6. Acomplete microcrystalline
structure is obtainedby filling
thequasiperiodic
skeleton with the unit cell of anapproximant
structure.D q§ ~ q~ a U
D
~ ~
q3 b ~ ~jj u ty P~ ~j'D
~ O
~
p
~
Jt~ ~
fjo ~
n@ ~~j.
~ ~
~ U
D 8
D P
b D q~ u ~ u ~ u
~ ~ O
El
u
e
o ~ ~u
f~
~ a a a q~ n3
n~ ~ ~
~
G U o
@
°~q
o P Qb P u 6~ n q a q
~
j
© t£%~ ~
~ij @
d
~ ~~j°
q3 ~
P j
~ Q
,
D
~ ~li
~ dl b QJ d' & q3fl~ ~
DU O ~
~..
P u
Fig.
7. The Fourier transforrns of theoctagonal
tiling(circles)
and ofthe1nicrocrystalline
structure(squares).
Satellites with rational shifts in the ej and e~ directions can be seen.On the other
hand,
the skeleton is obtainedby
a standard CP method where theacceptance
domain is the squareT(4
and where thehyperlattice
is the sublattice of/
2L~ associated to thematrix
M~,
where k is the order of theapproximant
unit cell. Asexplained
in section 3 the Fourierspectrum
of this structure is carriedby
theZ-module generated by
the basis' ~*
~i
' ~*~/
°k
+°k-1
~ ~°k
+°k-1
~bt 11
QUASIPERIODIC
MODELS WITH MICROCRYSTALLINE STRUCTURES 1601Finally
the Fourier transform of thequasiperiodic
array of decorated unit cells isgiven by
theproduct
of the structure factor of the unit cellby
the Fourier transform of the skeleton.Figure
7 shows asuperimposition
of the referencespectrum
and themicrocrystalline
spectrumin the case of the second
approximant
k=
2. Since
O~
+O~
= 3 we see that the Dirac
peaks
can be indexed w.r.t, the basis
(ef, et, et, e?)
withintegral
indices for theet
andet
vectors and rational indices with a denominatorequal
to 3 for theef
andet
vectors.Acknowledgments.
The author is indebted to
P.Donnadieu, C.Oguey
and G.Coddens for many fruitfuldiscussions.
References
[ii
SHECHTMAN D., BLECH I., GRATIAS D. and CAHN J.,Phys.
Rev. Lent. s3(1984)
1951-1954.[2] The
Physics
ofQuasicrystals,
P.J. Steinhardt and S. Oslund Eds.ovorld
Scientific,Singapore, 1987).
[3] Introduction to Quasicrystals, M. Jaric Ed.
(Academic
Press, SanDiego, 1988).
[4]
Quasicrystalline
Materials, C. Janot and J. M. Dubois Eds.ovorld
Scientific,Singapore, 1988).
[5] CALVAYRAC Y., QUIVY A., BESSIERE M., LEFEBVRE S.,
CORNIER-QUINQUANDON
M. andGRATIAS D., J.
Phys.
France sl(1990)
417-431.[6] BANCEL P.,
Phys.
Rev. Lent. 63(1989)
2741-2744.[7] AUDIER M. and GUYOT P., in Third Intemational
Meeting
onQuasicrystals
«Quasicrystals
and incommensurate structures in condensed matter »~ivorld
Scientific, Vista Herrnosa(Mexico),
1990), p. 288-299.[8] AUDIER M. and GUYOT P., in
Proceedings
of the Anniversary Adriatico Research Conference onQuasicrystals ovorld
Scientific, Trieste,Italy, 1990),
p. 74-91.[9] JANOT C., AUDIER M., DE BoissiEU M. and DUBois J. M.,
Europhys.
Lent. 14(1991)
355-360.[10]
DtNOYER F., HEGER G., LAMBERT M., AUDiER M. and GUYOT P., J.Phys.
France sl(1990)
651.[I
Ii
LAUNOiS P., AUDiER M., DtNOYER F., DONG C., DUBOiS J. M. and LAMBERT M.,Europhys.
Lent. 13
(1990)
629.[12] DONNADiEU P.,
Organization
of defects in the first cubicapproximant
of thequasicrystal Al~L13Cu,
to bepublished
in Phil. Mag. B(1991).
[13] TORRES M., PASTOR G. and JIMtNEz I., Phil.
Mag.
Lent. 61(1990)
319-325.[14] ARAG6N J. L., REYES-GASCA J. and Jost-YACAMAN M., Phil. Mag. Lent. 62
(1990)
337-347.[15]
TORRES M., PASTOR G. and JiMtNEz I., Phil. Mag. Lent. 62(1990)
349-355.[16] CODDENS G., J.
Phys.
I France 1(1991)
523.[17] CODDENS G. and LAUNOIS P., J. Phys. I France 1
(1991)
993.[18] DUNEAU M., MOSSERI R. and OGUEY C., J.
Phys.
A 22(1989)
4549-4564.[19] DUNEAU M. and OGUEY C., J.