HAL Id: jpa-00209519
https://hal.archives-ouvertes.fr/jpa-00209519
Submitted on 1 Jan 1982
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.
The geometrical nature of disorder and its elementary excitations
M. Kléman
To cite this version:
M. Kléman. The geometrical nature of disorder and its elementary excitations. Journal de Physique,
1982, 43 (9), pp.1389-1396. �10.1051/jphys:019820043090138900�. �jpa-00209519�
The geometrical nature of disorder and its elementary excitations
M. Kléman
Laboratoire de Physique des Solides, Université Paris-Sud, Bât. 510, 91405 Orsay, France
(Reçu le 20 avril 1982, accepté le 28 mai 1982)
Résumé.
2014En s’appuyant
sur uneanalyse géométrique d’un ordre à courte distance frustrant,
onpropose
unmodèle de corps amorphe composé de deux réseaux de Frank conjugués, l’un formé de disinclinaisons (dans la ligne des références [2] et [3]), l’autre de domaines cylindriques. Ce modèle peut s’étendre
ausystème de spin des amorphes magnétiques, à la phase bleue désordonnée des cholestériques et, dans
saversion dynamique,
auxliquides.
On discute des conséquences du modèle
surle spectre des vibrations atomiques élémentaires :
onexplique ainsi
le comportement du type Bragg pour k
=03C0/d, les modes de rotons, l’existence de modes localisés de basse énergie
conduisant
auxlois
connuesrelatives à la chaleur spécifique à très basse température.
Abstract.
2014We propose,
onthe basis of
ageometrical analysis of frustating short-range order (in the line of references [2], and [3]),
amodel in which
anamorphous body is made of two conjugated Frank networks,
oneof cylindrical domains, the other
oneof disclinations. A similar model could apply to the spin system of amorphous magnets, to the disordered blue phase of cholesterics and, in
adynamic version, to liquids. Consequences concerning
the phonon elementary excitations
arediscussed, like the Bragg-like behaviour at k
=03C0/d, the rotons modes, and the low energy localized modes invoked in the heat capacity low temperature behaviour.
Classification
Physics Abstracts
1. Introduction.
-It is a well established experi-
mental fact that all amorphous materials (glassy silica,
covalent or metallic glasses, polymers, ...) have com-
mon physical properties; at low temperatures (near
1 K) there is an extra specific heat linear with T, a heat conductibility quadratic with T, saturation of ultra- sonic attenuation at high power inputs, ... ; at high temperatures the transport coefficients obey a Fulcher- Vogel law, ... [for a review see 1]. Spin glasses seem also
to display the same properties, whose general occur-
rence has been found remarkable enough to bring
them the qualification of universal properties. If this is really so, these properties must be rooted in some
general structural model of disorder, in the same way that ordinary specific heat of insulators, heat conduc-
tivity, conductivity,
...have common features which
originate in the three-dimensional translational model
of crystals.
A structural model of amorphous materials has
recently been proposed [2, 3]. It is this model we shall
develop in this paper. This model involves the consi- deration of symmetry groups in spaces of constant
curvature, either positive curvature (3d sphere S3)
or negative curvature (3d Lobatchewskyan
-or hyperbolic
-space H3). Briefly : it is possible to tile
a space of constant curvature with regular polyhedra
whose repetition obey operations of symmetry (rota- tions, mirror planes, ...) which pertain to one of the
subgroups of the continuous group under which either
S3 or H3 are transitive. These subgroups are therefore
similar in nature to the Schoenflies groups of euclidean space R3 ; they could be named the Schoenflies groups of S3 or H3. But they describe a totally different local
order. For example the five fold symmetry which has long ago been advocated by Bernal [4] in his pioneering
work on the structure of liquids, and proposed more recently by Sadoc et al. [5] for non-crystalline solids,
is a symmetry which is perfectly tolerable in S3 or H3 (it pertains indeed to many of their Schoenflies groups)
but is precluded in the conventional 230 Schoënflies groups. It seems therefore very tempting to classify
those amorphous (and liquid) systems in which the local order is anomalous in an euclidean sense by the crystallographic groups of S3 and H3’
Of course, this classification raises a number of
questions, the first one being obviously : a) how to
map a non-euclidean lattice onto ordinary space,
remembering that this operation must lead to a
disordered state. Another question is : b) what are the physical properties which relate to the anomalous
crystallographic properties defined above ?
To question a) we have given
apreliminary answer
in [2, 3], where it is claimed that the introduction of disclinations in the lattice is the natural way of
«
decurving » space, while maintaining a fairly similar
local order (with fairly constant lengths and angles)
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019820043090138900
1390
except on disclinations cores. This mapping « in the
disclination mode » favours probably the H3 model
versus
the S3 model, since in this last case the finiteness
of S3 permits to build only
asmall piece of amorphous
materials (an extension to an infinite body requires
the introduction of walls ; conversely we note here that
the Sg model can be particularly relevant to the small aggregates of Farges et al. [6]). In the first part of this
paper we will proceed further in the investigation of
the geometry of
anamorphous body mapped from an H3 lattice. We show in particular that an amorphous body can be described as
arandom network of small cylindrical domains of finite length, in contact by their
ends (the nodes of the network being at those ends).
The axes of these domains are mappings of reticular
axes
which exist in the H3 lattice, and along which the
lattice repeats itself by an operation of translation.
These reticular axes (R. A. in short) are fundamental features of the hyperbolic lattice, in which there is
nothing like the usual reticular planes and reticular lines of the euclidean lattice, repeated by the elements of
a3d translational group. In an hyperbolic lattice
the translations are strictly limited to the R.A. (one
characteristic translation per R.A.). To the best of our
knowledge these geometrical properties of the Hg
lattice were not described before in such a way,
although the notion of localized translation symmetry has long been known (these translations are the transvections of Cartan [7]).
The network formed by the cylindrical domains in
the amorphous body cannot be accomodated without
disclinations; they run through the interstices left by
the domains and form themselves another network, conjugated to the first. These
arethe disclinations of
our mapping process from H3 to E3, which was at
that stage very superficially described in [2, 3]. There
is
astriking analogy between this model which makes
use of cylindrical domains and the model of the cholesteric blue phases of Saupe [8], recently revived by
Meiboom et al. [9]. We will argue that the cholesteric blue fog [10] is in fact an amorphous state correspon-
ding to our model, and that the spins in amorphous magnetic materials with local anisotropy might also
have a similar topology.
Question b) offers another challenge. In this paper
we
shall stress the implications of our geometrical
model for excitations of -the phonon type (including anharmonicity). In particular we shall gain some insight on the existence of a (Bragg) maximum in the
phonon dispersion curve, the existence of roton-like
modes at twice the (Bragg) wave number, and the
existence of states of small energy with a few number of levels. The two first points relate to experiments and
calculations
onthe phonon dispersion curve which we
shall recall, the last one to the low temperature speci-
fic heat effect (and related effects).
2. The projective representation of hyperbolic space.
-
A major difficulty with hyperbolic space is that
we Wtally lack intuitive insight of its geometrical
Fig. 1.
-Projective representation of H2. The absolute A is
a
conic representing the points at infinity of H2. Its interior
points represent the points of H2. The right lines passing through p (the polar of QQ’)
areall perpendicular to QQ’
(in
anhyperbolic sense). The right lines of the representation
are
right lines of H2. Various hypercycles with axis QQ’
arerepresented.
properties, because of
oureducational background.
Reading Coxeter’s books several times [11, 12] pencil
in hand, and in particular drawing all the
«strange »
figures where two parallels to
agiven line
runfrom
asingle point, proved the best exercise for me to become
acquainted with this geometry. I see no better way of
summing up here this practical knowledge I acquired
than describing
onthe projective representation of H2
and H3 some of the geometrical properties of these
spaces.
In figure 1, the points at infinity in the hyperbolic plane H2 are represented by
aconic A (the absolute),
the points inside the absolute being the image of points in H2. Any straight line
ycis the image of
astraight line in H2. Point p, which is the polar of
with respect to A, is a point common to all the straight
lines perpendicular to
7rin H2.
Two lines which meet on the absolute (i.e. at infinity
in H2) are said to be parallel in a given direction.
There are clearly two parallels to
agiven line passing through a given point. Two lines like pp and pv which meet outside the absolute are said to be ultra-
parallel. Clearly, they have one common perpendi-
cular QQ’ and only one.
There are three families of
«cycles )). I) The circle
is the locus of points at a given distance of a fixed
point. It is also the locus of the mirror images of a given point through all the lines passing through the
centre; 2) The hypercycle is a locus of points equi-
distant to a fixed line n, the axis. A few hypercycles with
7r
as axis are drawn in figure 1. Note that an hypercycle
is also the locus of the mirror images of a given point through all the lines passing by p. p is in this sense the
«
centre » of the hypercycle. The definition of the
hypercycle in term of symmetries will prove to be
important to
us.3) The horocycle is
acycle whose
Fig. 2.
-Projective representation of H3. The absolute A is
a
quadric, whose interior points represent the points of H3.
The two planes tangent to A in QQ’ and containing p define
a
right line QQ’ which is perpendicular (in the hyperbolic sense) to any plane containing p. p and QQ’
arereciprocal polars. The planes of the representation
are(hyperbolic) planes in H3.
centre is at infinity, on the absolute (or whose axis is tangent to the absolute). It is not
astraight line.
All these properties generalize to 3 dimensions (H3). The absolute is now a quadric. Duality by pola- rity couples planes
nto points p (outside A), points
inside to planes M outside, lines A to lines 1. Two
parallel planes intersect in a line which is tangent to A.
Two hyperparallel planes (which intersect 1 outside A)
have
acommon perpendicular A which is the polar of
I. The three types of cycles generalize to three types of spheres (Fig. 2).
The projective representation preserves projective properties, and particularly the cross-ratio of four
points on a line, and of four lines with common point.
The hyperbolic distance d between 2 points p and
vis
defined in function of a cross-ratio (see Fig. 1) :
where R - 1 is
apositive quantity measuring the cur-
vature of space, and the cross-ratio of four points a, p,
y, 6 is
Similary, the angle between two lines (in H2) or the
dihedral angle between two planes (in H3) is defined simply as a function of the cross-ratio of these two lines (resp. planes) with respect to the lines (resp.
planes) tangent to A and having the same intersection.
Note the important fact that the hyperbolic dis-
tances and angles we are discussing are also euclidean distances and angles : it is always possible, by defini-
tion of a Riemannian space (like Hn), to develop it (to
roll it without gliding) on an euclidean space R.
along
aline, with no distorsions of the lengths along the line,
aswell as no relative angular variations of the directions attached to it. (However this isometric
mapping along
aline cannot be extended continuously
outside the line, and it is indeed this difficulty which
led
usto the idea of
amapping in the disclination mode [2]).
3. Reticular axes, hypercylindrical domains and model of the amorphous state.
-Let
usconsider some tiling of the hyperbolic space; to simplify matters, we
start with
a2-dimensional space H2 and take, for the building blocks of the lattice, regular polygons with p
edges, such that q polygons share
acommon vertex.
We note { p, q } the lattice built in this way (Schlafli symbol). The requirement of obtaining an hyperbolic tiling reads [2]
If p is even, it is evident by reason of symmetry that any line passing through the midpoints of two opposite edges is perpendicular to both, and is therefore their
unique common perpendicular QQ’ (Fig. 3a). Opposite
Fig. 3.
-Reticular axis QQ’ in
anhyperbolic lattic. a) p
even
(we have assumed p
=6); b) p odd, q even; c) p odd,
q odd. q, the number of polygons at each vertex, is not defined
in this figure.
1392
edges
aretherefore ultraparallel. This line is also the
common perpendicular to
aninfinite set of ultrapa-
rallel edges, of length do each, which repeat with
aperiodicity d obeying the equation (see the appendix
for formulae in hyperbolic geometry)
Note that R (the radius of curvature of H2) and do
are
related (since there is a natural length in such a space). The relation is [13] :
Such common perpendiculars to infinite sets of edges also exist for p odd, for reasons of symmetry which easily vizualize (Fig. 3b for q even; Fig. 3c for q odd). There too the line QQ’ is an axis of sym- metry of translation, with some periodicity d There
is also no doubt that such lines QQ’ also exist in the
general case, when the building block of the lattice is not a regular polygon.
We call such lines QQ’ reticular axes (R.A.) because
of their analogy with the corresponding euclidean objects. But an essential difference is of course their
uniqueness, for each set of ultraparallel edges.
A number of interesting consequences can be drawn from the existence of R.A. We derive them starting
from the remark that the symmetry group of the
hyperbolic tessellation is generated by reflexions on
the edges of
acharacteristic triangle (Fig. 4) of angles nlp, nlq, n/2.
We restrict here to a tessellation of regular polygons,
for simplicity. Therefore :
-
any vertex, when reflected in any lattice edge perpendicular to QQ’, generates another vertex of the tessellation. All the vertices of a family generated from
a
given vertex lie on an hypercycle of QQ’. Therefore QQ’ is a line of glide-reflexion for the crystal lattice ;
-
any other element of the symmetry group
acting on QQ’ transforms QQ’ to another R.A.
Therefore p R.A. pass through the centre of any
polygon. The whole family of R.A. obtained in this way forms an hyperbolic lattice which can be studied
easily ;
-
since the space is hyperbolic, there is an expo-
nentially growing number of polygons, measured per unit length of QQ’, when one goes away from QQ’
Fig. 4.
-The characteristic triangle OPQ in the hyperbolic
lattice { p, q I.
along one of its perpendiculars. The area spanned
between the hypercycle, at
adistance
xof QQ’, and QQ’, is, for
anunit length along QQ’
and the area of a { p, q }.polygon is
Therefore the density 6 of polygons is growing with x
like
The mapping of the hyperbolic lattice on R2 can
now be better understood. If
wechoose for d some
repeat distance favoured in the amorphous body, the region in the near vicinity of an R.A. constitute clearly
a
domain in which the mapping introduces little dis- torsion, if part of the mapping consists in rolling H2 on R2 along the R.A. itself (on which there is
nodistorsion at all). Consider now
aring made of R.A. segments in H2, and roll H2 on R2 along this ring. This process maps the ring on an open, self-cutting, circuit in R2,
but it is feasible to close this circuit on R2 by removing
the extra R.A. segments, building therefore a ring of
R.A. in R2. The centre of the rings is occupied by a dis-
clination point (a positive disclination : we have removed segments in order to decrease the density of
vertices). Continuing this process,the entire mapping
consists in connected rings of R.A. segments, these segments are axes of domains in which the density
increases away from the axis, enclosing disclination
points whose strength is evidently a multiple of p-1 or q-l (see Fig. 5).
Let us now investigate how there results extend to three dimensions. We restrict too our discussion to a lattice built of regular polyhedra.
If the polyhedron is made of polygons containing an
even number of edges, we can consider the section of
adjacent polyhedra by a plane passing through the midpoints of two ultra parallel edges and perpendi-
cular to them. The intersection obtained in this way is similar to that one of figure 3a, the ultraparallel edges being perpendicular to the figure. Therefore there
exist R.A. passing through the centres of polygons belonging to two adjacent polyhedra and perpendi-
cular to the common polygon of these polyhedra. All
the planes passing through such a R.A. are perpen-
dicular to
aset of ultraparallel faces ; some of them
are
also perpendicular to the edges of these faces,
which are therefore ultraparallel. The consequences drawn before for H2 extend easily; a) all the vertices
lie on hypercycles equidistant to QQ’, which is there- fore so-to-speak an axis of cylindrical symmetry. In fact QQ’ is an axis of glide-reflexion if the constitu-
tive polygon is
asquare, and possibly an axis of
Fig. 5.
-A model for
atwo dimensional amorphous body.
The atoms have translational symmetry along
axesAB, which
arelocal
axesof glide-reflexion. The atomic density
is minimum
onAB and increases towards the centre of each polygon of the network, which is occupied by
adiscli-
nation point L. The polygonal array is random.
helicity if the symmetry of the constitutive polygon is larger; b) the density of polyhedra, measured per unit
length of R.A. increases exponentially with the dis- tance to QQ’ ; c) a whole set of R.A. is obtained by applying all the reflexions generated by the faces of the characteristic tetrahedron; this set forms an hyperbolic lattice; d) the mapping on R3 in the
disclination mode introduces a random network of
cylindrical domains; positive disclinations allowed by
the lattice symmetries pass through the interstices of this network.
If the polyhedron has ultraparallel faces (this is the
case for the dodecahedron 15, 3 }, the icosahedron
{ 3, 5 } and the octahedron { 3, 3 }), the R.A. are lines joining the centres of these faces; they
arelines of
rotation and transvection for the lattice.
The regular honeycombs of H3 fall into one of these
two categories, or in both (for a description of these honeycombs, see [13]). More generally, the search for the R.A. lattices necessitates first to recognize the
families of ultraparallel faces of the hyperbolic crystal.
They should not fail to exist, even if it might be neces-
sary to consider clusters containing more than one building block to find ultraparallel faces.
4. Elementary excitations ; the Bragg case.
-Is it possible to recognize the nature of the H3 crystallo- graphic arrangement by some sort of diffraction method in the resulting amorphous body ? It is this
kind of question which led
meto consider the Bragg problem in an amorphous material, i.e. to find under
which conditions (on the wave length A and the dif- fraction angle) an horospherical wave (issued from
Fig. 6.
-The geometry for the calculation of the Bragg
diffraction conditions; the figure displays
aplanar section of H,, containing the incident and diffracted directions (Q
1and O2), and a diffracting R.A.
some point Q 1 at infinity) is diffracted by an R.A.
towards some other point Q2 at infinity.
The geometry is illustrated figure 6, which shows
some planar section of H3 by a plane containing 921, Q2’ and the R.A. here QQ’. Points a and b
areat
atransvection distanced. hi and h2 are the feet of the
perpendiculars to QQ’ erected from 92, and Q2’ Call h
the length hl h2. It is easy to prove, by using the right- angled asymptotic (Ql and Q2 at infinity !) triangles Q I hl a, Q2 h2 a,..., that :
a
the + or - sign depending on the relative values of a 1
versusa2, fl,
versusP2’
Formula (8) are obtained by using the fundamental
equation for
aright-angled asymptotic triangle, which
here read (i
=1, 2)
As soon
asthe lattice node considered (a, b) is a few
intersites distances d from hi, it appears clearly on equation (9) that the angle cxi is very small (remember
that d and R are comparable, from equations (3) and (4)). Therefore
wecan safely replace sin x,, tg cxig etc...
by cxi for most of the nodes, the exception being a very
few nodes in the vicinity of hl and h2 (which are fixed points).
Now consider the asymptotic triangles Qi ab, 02 ab. The differences in length ð1 = 01 a - Q2 b, 62 = Q2 a - Q2 b, can be calculated using hyper-
bolic trigonometry. We find
1394
(we have assumed that, like on figure 6, (XI
1PI’
(X2 P2)’
Hence the total path difference for the two rays issued from Q 1 reads :
Taking now equation (9) into account with the same approximation, we find
We are now able to state the Bragg condition in an hyperbolic lattice.
From equation (12) it appears that we have to
satisfy
a condition which is similar to the well-known Bragg
condition in crystals (2 d sin 0
=nÀ.) for 0
=7r/2.
This is in fact what happens here, approximately,
since ai and fli are so small 0 - 7E - oci, p) so that,
when considering the corresponding amorphous body,
the corresponding rays are practically parallel to the
R.A. The hyperbolic
«limit » of the Bragg law is then obtained for 0
=7r/2.
However equation (13) states also that the Bragg
condition is very isotropic in an hyperbolic lattice :
this is not very surprising, since there is a natural
length in such a lattice (the radius of curvature), and we
expect nA to be of the same order of magnitude.
Note that our demonstration does not require that ol Q2 and QQ’ be in the same plane. For 0, and the
R.A. given, 02 is at infinity on a cone which is well defined by equations (8) and (13) and whose axis is
along the R.A.
A natural extension of these considerations con-
cerning the building of coherent waves (but not only
the horospherical wave have discussed here) would
lead to an investigation of the question of the
«reci- procal lattice » of a lattice on H3 We leave this ques- tion for a further study and limit our discussion to the consequences on the Bragg law we have found (Eq. 13) on the phonon dispersion curve.
In our model of an amorphous body consisting of cylindrical domains centred round R.A., we expect that
planar acoustic waves with wave number k
=n/4
whatever their direction may be, are Bragg diffracted, because there are many cylindrical domains at Bragg position (0~ 7r/2). If there is some correlation in the orientations of the cylindrical domains, some direc-
tions of diffraction might be favoured.
Now, comparing to experimental results, it appears
that it is a very general feature of the phonon disper-
sion curve of liquid materials that it bends down from
the linear behaviour
a) =ck (expected in a model of
disorder at all scales), and displays a maximum at
some value of k which is always comparable to
areciprocal lattice parameter. This has been observed and computed in liquid Rb [14, 15] and Pb [16], for longitudinal modes. Experiments in amorphous mate-
rials are rather scarce; the maximum itself has never
been observed, to the best of my knowledge, but it has
been computed in a-Mg70Zn30 [17] for longitudinal
waves; also various computed results on transverse
waves have been obtained for k n/d [17] and
k > n/d [18], which clearly let us expect a maximum for transverse waves, inasmuch as the descending part (k > 7r/d) of the dispersion curve has been
observed [19]. There are all reasons to believe that this
behaviour is related to our model of cylindrical domains, which favors Bragg diffraction along R.A.
5. Elementary excitations : the limit of long wave- lengths ; the rotons.
-By extrapolating the short wavelength behaviour of these observed transverse waves (in a-Mg7oZn3O), it appears that a soft mode near k
=2 nld, or a roton of very small energy, should exist. We want to argue that our model predicts such
a roton, on a physical basis very similar to that which
explains the
«solidon » recently proposed for 4He [20].
The small cylindrical domain is indeed a small ordered region which can trap phonons whose wave- length compares with d or some multiple of d. These
are the Bragg phonons we discussed above. Such
phonons can be transferred to the
«crystallon » by Umklapp processes involving waves k1 and k2 near Bragg diffraction, which certainly can interact strongly
in the cylindrical domain. They would decrease their energy by creating an excitation of small energy and small k (we discuss afterwards of their nature), and
a localized phonon of momentum
with Kc
=2 nld. The corresponding energy of the
crystallon would be
where V is the volume of the cylindrical domain. If E c = nw c wither - 1012 S-1 one get V - 1O - 2 2 cm3 (this value of We has been measured in liquid argon and seems typical of a small roton energy).
To the difference with H6ritier et al. solidon, the
«
crystallon » described here is not a dynamical object,
but this does not really make a difference. The excita- tion of small energy and small k appearing in the pro-
cess would simply replace the hydrodynamic part which is essential in the solidon process.
These excitations of small energy are presumably
related to the modes of vibration of the cylindrical
domain. Vibrations of a cylinder of small section compared to length, with boundaries free of traction,
have been studied in details (for ex., see Love [21]
and Rayleigh [22]). The present situation is different : the cylindrical domain has
adensity which grows
exponentially from axis to exterior : this condition,
which is difficult to treat exactly, is probably correctly
fitted by taking
acylinder with fixed boundaries. The calculation is easy for torsional and longitudinal vibrations; the frequency spectrum contains in both
cases an acoustic branch
co= ck, where
cis some
sound velocity, and
«optical » branches, starting at
w =
cla, where a is the radius of the cylinder and
csome (other) sound velocity. There are vibrations of
possibly smaller frequencies : those related to flexural
(transverse) deformations of the cylinder, propagating along the cylinder or helically. They might be favoured because they correspond truly to the crystallography
of the cylindrical domain. But I have not made a
calculation of these modes. In all these cases, one
expects that the smallest excited k would be related to the finite length of the cylinder, corresponding to a
localized phonon of very small energy trapped in the cylindrical domain. This could be a model for the soft
zones so often invoked to explain the low temperature behaviour of the specific heat [1].
Another possibility for this low temperature beha- viour, not exclusive of that one just described, is with
the regions where cylindrical domains meet; the atoms
sitting there have practically equal energetic choices
between the various positions related
«crystallogra- phically » to the various domains. These atoms can
therefore be easily excited on a finite number of levels,
either by the phonons generated in the cylindrical domains, which cannot propagate very far without
interchanging energy with the lattice, or by the Bragg
diffracted beams.
6. Discussion.
-In this paper, I have proposed,
on the basis of a geometrical analysis of the conse-
quences of frustrating short-range order, a model
in which the amorphous body is divided in cylindrical
domains forming themselves a sort of Frank network of domains, through which disclinations (all of the
same sign) are finding way. I have discussed some
effects of this model on elementary excitations. Other consequences should of course be discussed, among them the simplest would concern plastic behaviour;
since this model predicts some underlying network,
this should facilitates the understanding of processes which have, by many aspects, much resemblance with those observed in ordinary crystals, like slip (shear)
bands and slip lines, brittleness, etc...
Let us also notice that the present model reconcile,
in some way, the supporters of a model of an amorphous body made of microcrystallites, and others : the
microcrystallites would be the cylindrical domains,
but they are not separated one from the other by grain boundaries.
Also, a number of experiments are required to
check the model :
moremeasurements of the phonon dispersion curve, especially near n/ d and 2 nld, firstly; also careful microscopic examination of frac-
ture areas (by scanning electron microscopy) and
attention to density contrast in transmission electron
microscopy would be rewarding. Finally, other mate-
rials than conventional amorphous bodies should
display similar structural properties,
asalready stated;
the blue fog [10] is an obvious candidate, on which light scattering experiments on one hand, electron microscopy observation of freeze-etched specimens
on