HAL Id: jpa-00208724
https://hal.archives-ouvertes.fr/jpa-00208724
Submitted on 1 Jan 1977
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, estdestinée 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.
Raman scattering and local order in GexSe 1 -x glasses for 1/3 ≤ x ≤ 1/2
P. Tronc, M. Bensoussan, A. Brenac, G. Errandonea, C. Sebenne
To cite this version:
P. Tronc, M. Bensoussan, A. Brenac, G. Errandonea, C. Sebenne. Raman scattering and local or- der in GexSe 1 -x glasses for 1/3
≤x
≤1/2. Journal de Physique, 1977, 38 (12), pp.1493-1498.
�10.1051/jphys:0197700380120149300�. �jpa-00208724�
RAMAN SCATTERING AND LOCAL ORDER IN GexSe 1-x GLASSES FOR 1/3 ~ x ~ 1/2
P.
TRONC,
M.BENSOUSSAN,
A.BRENAC,
G. ERRANDONEACentre National d’Etudes des
Telecommunications, 196,
rue deParis,
92220Bagneux,
Franceand C. SEBENNE
Laboratoire de
Physique
des Solides(*),
UniversitéPierre-et-Marie-Curie,
75230 Paris Cedex05,
France(Reçu
le 2 mai1977, accepté
le 19 août1977)
Résumé. 2014 Après avoir étudié auparavant l’ordre local des verres GexSe1-x entre le sélénium pur et x = 1/3, nous avons
préparé
des échantillons pour 1/3 ~ x ~ 1/2. Des matériaux vitreux ont été obtenus jusqu’à x = 0,44. Pour les valeurs de x supérieures les échantillons sont constitués de cris- tallites de GeSe contenus dans une matrice vitreuseGexSe1-x,
avec x = 0,41 ± 0,01, le pourcentage de verre tendant vers 0quand x
tend vers 1/2. Des mesures de diffusion Raman ont été effectuées surles échantillons à la température ambiante. Les résultats conduisent au modèle de structure locale décrit ci-après. Pour 1/3 ~ x ~ 0,42, les verres présentent un ordre local du type GeSe2 : les coor-
dinances des atomes de Ge et de Se sont égales respectivement à 4 et à 2, les liaisons Se2014Se ainsi que deux liaisons Ge2014Ge par atome de Ge étant statistiquement exclues. Pour x ~ 0,43, il apparait
de fortes indications d’une structure du type GeSe (coordinances des atomes de Ge et de Se toutes
deux
égales
à 3) : un tel type de coordination n’avaitjamais
été signaléjusqu’à
présent pour desverres.
Abstract. 2014 Having
previously
studied the local order of GexSe1-x glasses between pure selenium and x = 1/3, we have nowprepared
samples for 1/3 ~ x ~ 1/2.Glassy
materials have been obtained up to x = 0.44 and, athigher
x values, thesamples
are made of GeSecrystallites
embedded in aglassy matrix of
GexSe1-x,
with x = 0.41 ± 0.01, the percentage of glass going to zero when x goes to 1/2. Raman scattering measurements have beenperformed
on a set of suchsamples,
at room temperature. The results lead to thefollowing
model of local order : for 1/3 ~ x ~ 0.42, theglasses
have a GeSe2-like local order with four-coordinated Ge atoms and two-coordinated Se atoms, Se2014Se bonds and two Ge2014Ge bonds per Ge atom
being statistically
forbidden; for x ~ 0.43 strong indications appear for a GeSe-like local order, with both three-coordinated Ge and Se atoms : such a type of coordination has never beenpreviously
reported for glasses.Classification
Physics Abstracts
61.40 - 78.30
1. Introduction. - The
importance
ofoptical
methods to the
study
of local structure inbinary chalcogenide glasses
has been demonstratedrecently
both for the
GexSel-x [1]
andGexS 1 _ x [2] systems.
In reference
[1],
which will be called 1hereafter,
Ramanscattering
measurements allowed us to drawsome conclusions about the local structure and its
changes
with x for 0 x1/3
in theGexSel _ x glasses.
The structural model could account for thechanges
in theoptical-absorption edge
as a functionof x.
In the
present
paper, the results of a similarstudy
(*) Associé au Centre National de la Recherche Scientifique.
for
GexSe 1 _ x glasses
with1/3 ,
x1/2
aregiven :
this is of
particular
interest for reasons which aredeveloped
below.In
I,
the current views on thephase diagram
of theGe-Se
system
werereviewed, together
with thecrystalline
structures ofGeSe2
and GeSe. In theparticular
case ofGeSe2,
acomplete knowledge
of thestructure is not
yet
established.However,
some infor- mation has been obtainedrecently [3]
on thecrystalline
form of interest for us, that is the form into which our
glassy samples crystallize
when cooled at a too slowrate.
(The experimental
results which werereported
in 1 about
crystalline GeSe2
both foroptical-absorp-
tion
edge
and Ramanscattering
are relative to this newcrystalline form.)
Inshort, GeSe2 crystals
should beArticle published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0197700380120149300
1494
considered as a set of more or less distorted tetrahedra with a Ge atom at the centre, bonded to four Se atoms
at the corners, each Se atom
being
bonded to twoGe atoms
(Fig. la).
GeSecrystals
are such that each atom, Ge orSe,
has three nearestneighbours
of theother
kind,
one at 2.56A
and two at 2.59A,
the nextthree nearest
neighbours being
at 3.32À
and 3.7A
(Fig. lb).
,FIG. 1. - Local structures of crystalline germanium selenides : 0 Ge atom, 0 Se atom ; a : GeSe2, b : GeSe.
In
Ge,,Se 1 -,, glasses,
as x varies from 0 to1/3,
it hasbeen shown in 1 that the local structure tends towards
a
pseudo-tetrahedral
one, with four coordinated Ge atoms and two coordinated Se atoms, Ge-Ge bondsbeing statistically
forbidden(Fig. 2),
whichimplies
avanishing
number of Se-Se bonds.FIG. 2. - Local structure of GexSe1-x glasses (x 1/3) : 0 Ge
atom ; 1 Se atom.
For x
higher
than1/3
in theGexSe 1 - x system,
sucha situation is no
longer possible
and thequestion
is todetermine its behaviour up to x =
1/2,
where thecrystalline
form shows both three coordinated Ge and Se atoms.Three
possibilities
can be considered :-
First,
if the system does not accept Ge-Gebonds,
the three-coordinated Ge and Se atoms must appear as soon as x >1/3.
2013
Second,
if the system admits any number of Ge-Gebonds,
no three-coordinated Ge and Se atom is necessary.- Third,
if thesystem
admitsonly
a limited number of Ge-Gebonds,
three-coordinated Ge and Se atoms must appear at some intermediate valueof x ;
forexample, if only pairs
of bonded Ge atoms surroundedby
Se atoms areaccepted,
the limit is for theGe2Se3 composition
and three-coordinated Ge and Se atoms must appear for x >2/5.
The local structure of
GexSe 1 - x
thin films athigh x
values has been studied elsewhere
[4-8]
eitherby
electron diffraction or
by
extendedX-ray absorption
fine structure.
Except
for Mikolaichuk andKogut [7]
who claim a GeSe-like local order at x =
0.50,
all the authorssupport
aGeSe2-like
local order with a bondlength
of 2.3 to 2.4A
for x valuesranging
from 0.32 to0.73. Such results
give
another reason tostudy
bulksamples
where the strain can be lower than in thin films.In the present paper, it will be shown that an answer can be obtained to the
question
of local order ofGexSel-x glasses (x
>1/3),
from Ramanscattering
measurements. In section
2,
the main features concern-ing
thepreparation
and the characterization methodsare described. After
giving
theexperimental results,
section 3 is devoted to the discussion of the results which leads to a model of local structure of
GexSe1 _x glasses
for the different values of x in the studied interval.2.
Samples préparation
and characterization. -GexSel-x glasses
can beprepared
forgermanium
concentrations x
higher
than1/3,
asreported by
Baidakov
[9]
for x0,40,
and laterby
differentauthors
[1,10,11] for x
0.42. In the present case, thesamples
with a concentration up to 0.42 areprepared
and characterized as described in 1 : amor-phous
materials are obtainedby cooling
the quartz cellcontaining
thehomogenized liquid
of propercomposition
at a suitable rate.However,
for x > 0.42 such a method does not lead to anamorphous
material and it is necessary to use adifferent method which is described in detail else- where
[12].
Inshort,
the mixture is heated up to about 30 °C above theliquidus
temperature[1]
for5 minutes in an open silica crucible
encapsulated
with
B203
under an 8 bar argon pressure. Then awater-cooled copper
finger
isquickly plunged
into theliquid.
It ispossible
toget
anamorphous layer
about500 gm thick for x up to and
including
0.44. Theamorphicity
is verifiedby X-ray
diffraction(the
diffractionspectra
areabsolutely flat)
and the average and local chemicalcompositions by colorimetry
andelectron
microprobe.
For x
higher
than0.44, optical microscopy
andelectron and
X-ray
diffraction show that even the surfacelayer
ispartly crystallized :
GeSecrystallites
are embedded in an
amorphous
matrix made ofGexSel _x
with x = 0.41 ± 0.01(as
measuredby
electron
microprobe).
GeSe
single crystals
have beenprepared using
thevapor
phase transport
method asproposed by
Wiedemeier et al.
[13].
3.
Expérimental
results and discussion. -T The Raman spectra of the differentamorphous
and crys- tallinesamples have , been
obtained with theexperi-
mental set up described in I.
Except
forcrystalline GeSe2,
where the laser source was either Kr+ orAr+,
the excitation was obtained
through
a YAG laserat 1.06 gm CW. The
beam
was focused on the sur-face of the
sample,
with an incidenceangle
of about750. The measurements were
performed
at roomtemperature
and the anti-Stokesspectra
were sys-tematically
recorded sincethey give
better resultstaking
account of the variation of thephotomultiplier sensitivity
versusfrequency,
on the onehand,
and ofthe effect of Bose-Einstein statistics upon the line
magnitudes
on the other hand. The results for the different values of x aregiven
onfigures
3 and 4(the
dotted lines represent intensities of the main
peaks
which can be
distinguished
in the Ramanspectra).
When x varied from 0.33 to
0.44,
within the accuracy of anX-ray
diffraction test, thesamples
were amor-phous
and remained so even afterbeing exposed
to thefocused laser
during
the Ramanscattering experi-
FIG. ’3. - Anti-Stokes Raman spectra of GexSe1-x glasses (1/3 x 0.42).
FIG. 4. - Anti-Stokes Raman spectra of GexSe1-x compounds (0.42 x 0.50).
ments. Several
recordings
weretaken,
the location of the laser beam on thesample
or thesamples
itselfbeing changed
eachtime,
for every x value.For the same nominal value
of x,
thereproducibility
is
generally good
if one considersonly
thefrequencies,
the relative intensities and the widths of the différent
peaks.
The absolute intensities canchange easily by
afactor of two
depending
on thequality
of thesample
surface and the details of the
optical alignment.
However,
in the x = 0.42 + 0.01 range, whererapid changes
are observed when xvaries,
agood
repro-ducibility
in the relative intensities of the differentpeaks
is notalways obtained,
which indicates that local.
composition
fluctuations of about 1%
canoccur
(in
the later case, spectra which arepresented
are theaveraged
values of the differentrecordings).
From these
considerations,
some normalizationrule
had to beadopted
forrepresenting
the results.In
figure 3,
for x0.42,
thespectra
aregiven
withthe 200 cm -1
peak intensity proportional
to x. Such achoice will be
justified
later. Infigure 4, for x > 0.42,
the
spectra
aregiven
with thehighest peak
at the sameintensity independent
of thepeak frequency
andthe x value. This choice is
quite arbitrary
and mayeven hide part of the
experimental information ;
itsonly advantage
is togive
a well balanced represen-tation, but,
infact,
as x increases from 0.45 to 0.48 andagain
to GeSecrystal
an increase of the overall intensitiesclearly
appears with amultiplying
factorof 3 to 4. The Raman
spectra
of the GeSecrystal
wereobtained on cleaved
samples
with the samegeometrical
arrangement as for
amorphous samples.
1496
To discuss these results it is necessary to refer to a
model. One can
rely
on the dataprovided by
thecrystallization
behaviour ofGexSel _x glasses
togive
ahint to start
interpreting
the informationsprovided by
Ramanscattering.
The
crystallization
behaviour ofGexSe 1 _ x
com-pounds depends strongly
on the thermal process :- If the
cooling
rate ishigh enough,
as describedabove in part
2, X-ray
diffraction patterns show thatstrictly amorphous samples
are obtained for x 0.44and GeSe
crystallites
appear for x > 0.45.- If the
cooling
rate is a littlelower, partly
crys- tallizedsamples
are obtained : for1/3 x 0.42, only GeSe2 crystallites
areobserved,
while forx >
0.43, only
GeSecrystallites
are detected.-
Upon annealing,
which isequivalent
to a very lowcooling
rate,only GeSe2 crystallites
are observedfor x 0.35 but GeSe
crystallites
start to appear,together with GeSe2,
for x > 0.35.Since the increase of the
cooling
rate seems to deletethe GeSe-like local order and to favor the
GeSe2-like
local order for x
0.42, glasses
in thatcompositional
range can be assumed to have a local structure similar to
GeSe2
with four-coordinated Ge atoms and two- coordinated Se atoms. On the contrary the x ‘> 0.43glasses
show atendency
towards the local order ofcrystalline GeSe,
with three-coordinated Ge and Se atoms. TheGeSe2 crystal
hasonly
Ge-Se bonds.One can assume that no
dangling
bonds exist inchalcogenide glasses [1]. Then,
for x >1/3,
a similarlocal order as
GeSe2 crystal implies
the existence of Ge-Ge bonds. Moreover, Se-Sebonds,
if any, arecertainly
in smallnumber,
asproved by
the Ramanspectra of
figure 3,
where thepeak
at 250cm -1,
whichcharacterizes such bonds and is intense in x
1/3 glasses [1],
is almostnegligible.
From these
considerations,
we shall propose thefollowing
model :- For
1/3 x 0.42,
thesamples
are constituted of trueglassy
material. There is nodangling bond,
Ge atoms are
four-coordinated,
Se atoms are two-coordinated,
Se-Se bonds are forbidden and the local structure is of tetrahedral type.- For 0.43 x
0.44,
thesamples
are still madeof true
glassy materials,
the rules for x 0.43 still hold but local structures with three-coordinated Ge and Se atoms appear which means a dramaticchange
in the first
neighbours configuration.
- For 0.45 x 0.50 the
samples contain
GeSecrystallites
in anamorphous
matrix made of aglassy
material x -
’0.41,
with a similar structure as a trueglassy sample
of similarcomposition.
One has now to check if this model fits with the
experimental
results.a) 1 /3 x
0.42. - In thehigh frequency
range of interest to us, the Raman spectra offigure
3 showthe existence of three
peaks
at175, 200
and 215cm-1
respectively
and of a wider band in the 230-330 cm -1 rangecontaining
a widepeak
at about 290cm-1.
As shown in
I,
thepeak
at 200 cm-1 can be attributed to the tetrahedral structure where a Ge atom is surroundedby
four Se atoms. We shall consider that it still holds if one or more Se atom isreplaced by
a Ge atom at one or more corner of the tetrahedron.
As in
I,
thepeak
at 215 cm-’ can be attributed toa Se atom bonded to two Ge atoms. Both
peaks
at 175
cm -1
and 290 cm -1 have intensities which arevery low for x =
1/3
and increase with x : we shall consider these twopeaks
can be attributed the Ge-Ge bond. To supportthis,
let us remark that the Raman spectrum ofamorphous
Ge[14]
shows a wideband centered at 270 cm -1 with a shoulder around 170 cm-1.
In order to
verify
thepicture
we have drawn for1/3 x 0.42,
let us calculate theintensity changes
which are
expected
as a function of x for the different Ramanpeaks.
If agood
fit with the measured inten- sities isobtained,
it will demonstrate at the same time theconsistency
of the model and thepeak
attribution.Since
only
thehighest frequency
first order Ramanpeaks
areconsidered,
we assume that thefrequency depends only
on the localconfiguration,
that is firstneighbour
interactions. The masses of Ge and Se atoms are very closetogether
and thedensity
of theglasses
variesonly
veryslightly
with x : we cantherefore
safely
assume that a Ge atomoccupies
thesame volume as a Se atom and that this volume does not
depend
on x.Using
all thepreceding
conside-rations,
statistical calculation of the sametype
that is made in 1 shows that theintensity
of the 200 cm-1peak
must vary as x, theintensity
of the 215cm -1 peak
as 1 - x and the intensities of the 175
cm -1
and 290 cm -1peaks
as 3 x - 1.Practically,
a linearvariation of the intensities is
expected,
with apositive
or
negative slope,
the absolute values of the differentslopes,
whichdepend
on Ramanscattering
crosssections and other uncontrolled
parameters, being insignificant.
After a normalization of the intensities where the 200cm-1 peak intensity
isimposed
to beproportional
to x, aspresented
onfigure 3,
theexperi-
mental intensities of the other
peaks
aregiven
onfigure 5, together
with the best linear fitcorresponding
to the
proposed
model. Inspite
of an unavoidabledispersion
of theexperimental
values due tocomposi-
tion fluctuations and local order
distortions,
theagreement
appearsquite satisfying
andjustify
theproposed
model as well as thepeak labelling.
b)
0.43 x 0.50. - As soon as x becomes greater than0.42,
new intensepeaks
which charac- terize a local structure of thecrystalline
GeSetype (Fig. 4),
appear at 150cm -1
and 187cm -1
while thepeaks
at200, 215
and 290cm-’
which characterize alocal structure of the
crystalline GeSe2 type,
almostdisappear
for x = 0.44.Only
thepeak
at 175cm-’
appears all
along
and cannot be consideredby
itselfFIG. 5. - Raman peak intensities for GexSel_x glasses (1/3 x - 0.42) : 1 175 cm - ’ ; ib 215 cm - ’ ; 0 290 cm-1.
as
characterizing
one type of local order.Nevertheless,
let us remark
that,
for x >0.44,
theintensity
of thepeaks
at150, 175
and 187 cm-’ behavesimilarly
andconverge towards the
crystalline
GeSespectrum.
The175
cm-1 peak
therefore agrees with both localconfigurations
and its behaviour does notimpair
ourconclusions. The behaviour of the Raman spectra as a function of x demonstrates the sudden appearance of
a local order of the
crystalline
GeSe type for x = 0.43 in anamorphous sample.
Thequestion
which arises is to knowwhy,
at the sametime,
thepeaks
characteriz-ing
a local order of the tetrahedral type almostdisappear
while it seems that there mustbe,
at theatomic
scale,
a localcomposition corresponding
tox = 0.50
compensated by
a localcomposition
cor-responding
to x =0.41,
as it isproved
forpartly crystallized samples
at x > 0.45(for example,
if thenominal
composition
is x =0.44,
there must be one third of the atoms in a GeSe-likeconfiguration
withx = 0.50 and two third in a tetrahedral like confi-
guration
with x =0.41).
If thesuperposition
.of thetwo
corresponding
Raman spectra is not observedexperimentally
this is due to the difference in Ramanscattering
cross sections : as we remarked in the pre- sentation ofexperimental results,
theintensity
of themain
peak
increasesqualitatively by
a factorof
3 to 4as x goes from 0.42 to 0.50. It proves that the
peaks correspond
to thecrystalline
GeSe-likeconfiguration
dominate the spectrum as soon as a notable propor-
tion of that
configuration
is present in thesample
andhide the
peaks corresponding
to the tetrahedral-likeconfiguration.
One can wonder whether the GeSecrystal peaks appearing
in Ramanspectra
of x = 0.43 and x = 0.44compounds
are not accountable to GeSemicrocrystals.
Scherrer formula[15]
showsthat, provided
the concentration ishigh enough (a
fewat.
0/00
atleast),
ratherlarge microcrystals (let
us saymeasuring
at least one hundredA,
to fix thepoint)
would
produce sharp peaks
inX-ray
diffraction spectra, which is not the case. Therefore the amount in thecompounds
of GeSemicrocrystals
with süch a sizeis very low
(less
than 0.5 at.% approximately)
andcannot justify
intensepeaks corresponding
to the GeSeconfiguration
in Ramanspectra
of x = 0.43 andx = 0.44
compounds :
the intensities of thesepeaks
should indeed be at least one hundred times less intense than for pure
crystalline
GeSe. If the com-pound
would contain smaller GeSemicrocrystals,
theX-ray
diffractionpeaks
would be wider(up
to tendegrees approximately
for tenA microcrystals).
Allthe theoretical studies
[16-18]
undertaken up to nowshow that the intensities of the
peaks
would be great.This allows to conclude that the amount of very little
microcrystals
is not either sufficient toexplain
theRaman
peaks.
Let us remark at last thatedge
effectsprobably
disturbangles
andbondlengths
on thecrystal
boundaries : for this reason, below 10À,
theconcept of
crystal
isprobably meaningless.
Another
question
arises which concerns the number of Ge-Ge bonds in theglassy
material with x 0.42.If
only
one Ge-Ge bond wasaccepted by
each Geatom, the maximum value of x
compatible
with thetetrahedral-like
configuration
is0.40,
and for x =0.40,
we would have a local structure as
given
infigure
6a.If two Ge-Ge bonds were
accepted by
each Ge atom,a tetrahedral-like
configuration
would bepossible
up to x =0.50,
with a local structure of the typegiven
in
figure
6b.However,
one can prepareglasses
with apractically
pure tetrahedral-likeconfiguration only
up to x = 0.42 while at
higher x
the GeSe-like confi-guration
appears. It provesthat, statistically, only one
Ge-Ge bond is
accepted by
a Ge atom. Such a resultshows that the Ge-Se
system
is similar to the Ge-S system where Feltz et al.[19]
demonstrated from ESCA measurements that all the Ge atoms wereequivalent
in
Geo.40SO.r,,O,
which meansonly
one Ge-Ge bondper Ge atom is
accepted.
FIG. 6. - Hypothetic local structures of GexSe 1 _ x glasses :
0 Ge atom, 0 Se atom ; a : Ge2Se3, b : GeSe.
1498
4. Conclusion. - A model of local structure,
supported by
Ramanscattering
measurements, has been established forGexSel _x glasses.
It is confirmedby X-ray
measurements on thecrystallized samples.
In
I,
we have shown that for 0 x1/3
the coordi- nation-numbers ofgermanium
and selenium atomsare
respectively equal
to four and two, the Ge-Ge bondsbeing statistically
forbidden. Moreover thegermanium
atoms tend topart
from one another as faras
germanium
content of the mixture allows.Now,
wesee
that,
for1/3 x 0.42,
theglasses
are constituted of four-coordinated Ge atoms and two-coordinated Se atoms, Se-Se bonds and two Ge--Ge bonds per Ge atombeing statistically
forbidden. For x > 0.43a GeSe-like local order appears, with three-coordinat- ed Ge and Se atoms the material
being
stillcompletely amorphous.
For x > 0.45completely amorphous
material cannot be
prepared
and GeSecrystallites
appear embedded in a
glassy
matrix ofGexSel _x
withx - 0.41. To our
knowledge,
it is the first time that the existence of a three-three coordination is observed in achalcogenide glass :
it is obtained in aquite
narrowinterval and needs a very fast
quenching
rate. This isdifferent from the thin film case where a tetrahedric local order has
generally
beenreported
for any x value : such a difference suggests that the structure of the lattice isprobably
closer to that of thecrystal
ofneighboring composition
inglasses
that in thin films.The results we have
reported
about the three-three coordination arecompletely
new. About thisquestion
we believe that our present work is to be considered
only
as astarting point.
It will have to beimproved
and
developed by
otherexperimental techniques
before the existence of a new type of coordination may be considered as
definitely
established inchalcogenides glasses.
Acknowledgments.
- We are verygrateful
to theCCM
Department
of Centre National d’Etudes desTélécommunications,
where thesamples
were pre-pared
andcontrolled,
andparticularly
to G. Le Rouxfor the
X-ray
measurements, A. M.Pougnet
for thechemical
analysis
and C.Daguet
for the electronmicroprobe
measurements. We are indebted to R.Beserman,
J. F.Morhange
and M. Massot wholet us use their Raman set up.
References
[1] TRONC, P., BENSOUSSAN, M., BRENAC, A. and SEBENNE, C., Phys. Rev. B 8 (1973) 5947.
[2] LUCOVSKY, G., GALEENER, F. L., KEEZER, R. C., GEILS, R. H.
and Six, H. A., Phys. Rev. B 10 (1974) 5134.
[3] BURGEAT, J., LE Roux, G. and BRENAC, A., J. Appl. Crystallogr.
8 (1975) 325.
[4] FAWCETT, R. W., WAGNER, C. N. J. and CARGILL, G. S., III, J. Non-Cryst. Solids 8-10 (1972) 369.
[5] MOLNAR, B. J. and DOVE, D. B., J. Non-Cryst. Solids 16 (1974) 149.
[6] UEMURA, O., SAGANA, Y. and SATOW, T., Phys. Status Solidi (a)
26 (1974) 99.
[7] MIKOLAICHUK, A. G. and KOGUT, A. N., Sov. Phys. Crystallogr.
15 (1970) 294.
[8] SAYERS, D. E., LYTLE, F. W. and STERN, E. W., J. Non-Cryst.
Solids 8-10 (1972) 401.
[9] BAIDAKOV, L. A., in Solid State Chemistry (Z. U. Borisova Editor, Consultant Bureau, New York) 1966, 63.
[10] BORISOVA, Z. U., SHOKOL’NIKOV, E. R. and KOZHIMA, I. I., in ref. [9], p. 164.
[11] FELTZ, A., BÜTTNER, H. J., LIPPMANN, F. J. and MAUL, W., J. Non-Cryst. Solids 8-10 (1972) 64.
[12] BRENAC, A. and THOMAS, J., To be published in J. Mater. Sci.
[13] WIEDEMEIER, H., IRENE, E. A. and CHAUDHURI, A. K., J.
Cryst. Growth 13-14 (1972) 393.
[14] WHIL, M., CARDONA, M. and TAUC, J., J. Non-Cryst. Solids
8-10 (1972) 172.
[15] GUINIER, A., Théorie et Technique de la Radiocristallographie (Dunod, Paris) 1964, 482.
[16] DIXMIER, J. and GUINIER, A., Fizika (Zagreb), Suppl. 2 (1970)
9.1.
[17] Moss, S. C. and GRACZYK, J. F., Phys. Rev. Lett. 23 (1969)
1167.
[18] MALAURENT, J. C. and DIXMIER, J., To be published in Proc.
Symposium on the Structure of Non-Crystalline Materials, Cambridge (U.K.) (20-23 sept. 1976).
[19] FELTZ, A., VOIGT, B. and SCHLENZIG, E., Proc. 5th Internat.
Conf. on Amorphous and Liquid Semiconductors (J. Stuke
and W. Brenig, Editors, Taylor and Francis, London), 1974, p. 261.