• Aucun résultat trouvé

A geometrical analysis of limit layers and interfaces of lattices

N/A
N/A
Protected

Academic year: 2021

Partager "A geometrical analysis of limit layers and interfaces of lattices"

Copied!
17
0
0

Texte intégral

(1)

HAL Id: jpa-00246608

https://hal.archives-ouvertes.fr/jpa-00246608

Submitted on 1 Jan 1992

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.

A geometrical analysis of limit layers and interfaces of lattices

Michel Duneau

To cite this version:

Michel Duneau. A geometrical analysis of limit layers and interfaces of lattices. Journal de Physique

I, EDP Sciences, 1992, 2 (6), pp.871-886. �10.1051/jp1:1992185�. �jpa-00246608�

(2)

Classification

Physics

Abstracts

61.50J 61.50K 61.70N 64.70K

A geometrical analysis of limit layers and interfaces of lattices

Michel Duneau

(*)

Centre de

Physique Thdorique,

Ecole

Polytechnique,

91128 Palaiseau Cedex, France

(Received

3J October I99I, revised 3 December I99I,

accepted

13 December

I99I)

R4sum4. Les couches limites d'un rdseau au

voisinage

d'un

plan quelconque

peuvent dtre d£crites par la mdthode de coupe. Les sites

correspondants

sont

proches

d'un r6seau moyen bi- dimensionnel contenu dans le

plan

limite. La modulation assoc16e a la structure d'un cisaillement fractionnaire

simple.

Dans les cas oh deux r£seaux de mdme densit6 sont lids par un double cisaillement nous montrons que pour certaines orientations d'interface les deux couches limites ont le mdme rdseau moyen, favorisant airtsi une coh6rence

optimale.

Abstract. Limit

layers

of a lattice in the

neighborhood

of any

plane

can be described

by

the cut

method. The

corresponding

nodes are close to a 2-dirnensional average lattice contained in the limit

plane.

The associated modulation has the structure of a

simple

fractional shear. ln cases where two lattices of

equal density

are related

by

a double shear we show that for

particular

orientations of the interface both limit

layers

share the same average lattice, thus

favoring

an

optimal

coherence.

I. Introduction.

The cut method was

extensively

used in connection with

quasicrystals

in order to

provide

models of

quasiperiodic

structures with

non-crystallographic symmetries (see

for instance

II).

In such constructions a

higher

dimensional space is introduced

(for

instance R6 in the case of icosahedral

structures)

and the cut method allows one to select nodes of a

non-physical

lattice

(for

instance

26)

and to

project

these nodes in the 3-dimensional

physical

space.

Thus,

quasiperiodic

structures such as the 3-dimensional Penrose

tilings

are

finally

obtained.

In this paper we show that the cut method can be used with benefit in the

ordinary

3-

dimensional space for the

description

of limit

layers,

interfaces and

displacive

transformations

between lattices. The occurrence and the

importance

of

quasiperiodic ordering

in

grain

boundaries was

already

noticed in

[2-4].

The limit

layers

of a half lattice cut

by

a

plane

are obvious instances of the cut method. The tools

developed

in the

quasicrystal

field can be

applied

in this more intuitive situation. As an

example

these limit

layers

tum out to be close to 2-dimensional lattices even when the cut

plane

is irrational. This result is based on the construction of average lattices of

quasiperiodic

structures. We shall see that the bounded transformation from a

(possibly quasiperiodic) layer

(*) CNRS-UPRA 0014.

(3)

of a lattice into its 2-dimensional average lattice is a fractional version of a shear transformation.

The fractional shears are introduced and discussed. We show that

they provide

the

simplest

bounded transformations between lattices and therefore

give

the best

geometrical

candidates for

displacive

transformations. As mentioned above

they

are involved in the construction of

average

lattices,

but

they

also prove to be useful in the

analysis

of more

general displacive

transformations.

Shear transformations can be used to

analyse

the

relationships

between two lattices of

equal density.

Four successive shears are

usually required

to map a lattice

L~

onto a lattice

L~

in 3 dimensions. However for

particular pairs

of

lattices,

or for

particular

relative orientations this number can be lowered to two

(or

even

one).

Such

exceptional

cases are

believed to have a

physical meaning.

For instance it was shown in

[5]

that icosahedral twins of cubic lattices, as observed in

[6]

in the Almnfesi system, are

pair

wise related

by only

two

shears. On the other hand the relative orientations between

grains giving

rise to usual

coincidence lattices can be shown to

correspond

to one or two shears with rational

parameters

[7].

In section 2 we recall the definition and

properties

of linear shear transformations and we introduce fractional shear transformations. Then we show in section 3 that most

simple

bounded transformations between lattices are

given by

such fractional shears. In section 4 we consider limit

layers

of lattices bounded

by

a

plane

of

typically

irrational orientation. We show that such limit

layers

are close to an 2-dimensional average lattice

lying

in this

plane.

The transformation

mapping

the nodes of the limit

layer

onto the nodes of the average lattice is

proved

to be a fractional shear. In section 5 we consider the most

general

relative orientations between lattices of

equal density.

We show that the transition matrices are the

product

of at most 4 shear matrices and we

give

a

simple algorithm

to compute them. In 2 dimensions the most

general

transition matrix is the

product

of 3 shears. We consider the

example

of two square lattices with a arm relative orientation. In section 6 we focus on the

exceptional

cases of the double shears and we show that

they

can be identified

by

a

simple

criterion. In section 7 the same double shear cases are

proved

to

yield

most « coherent »

interfaces,

the orientations of which are

specified by special

lattice

planes

of an intermediate lattice.

2. Shear transformations and fractional shear transformations.

A shear transformation is a linear

mapping

which

reads, using

Dirac's bracket notation :

+ Sl ~l St ~2 Sl ~3

~

"

~ +

'S) (~

" 52 ~l + 52 ~2 52 ~3

(2.1)

53 ~l 53 ~2 + 53

3~

S(x)

= x +

(« [x)

s

where s is a vector and « is a covector.

Therefore S transforms a

point

x

by

a translation of constant direction s with an

amplitude (« ix) ills ii.

The transformation has an invariant

plane

defined

by (« ix)

= 0 which is

equivalent

to

S(x)

= x.

We shall

only

consider volume

preserving

shear transformations which are characterized

by

the further condition

(« s)

= 0 which is

equivalent

to det

(S)

=

I. In this case the invariant

plane

contains the shear direction s and the inverse of S is

given by

S~ '

=

I

[s) (« [.

Let

L~

=

AZ ~ and

L~

=

BZ~

denote two lattices of

equal density.

The 3 x 3 real structure matrix A is

regular

and its columns

give

a

particular

basis of

L~, la)

=

jai,

a~,

a~)

with

(4)

a,

=Ae;,

where

(e~,e~,e~)

is the standard basis of the 3-dimensional space. Thus A

specifies

the metric and orientation parameters of

L~. Similarly 16)

=

(bi, b~, b~)

with

b,

=

Be;

where B is a structure matrix of the lattice

L~.

Structure matrices map the abstract lattice

23,

in the space of « indices », onto the real lattices in the

physical

space.

However, any lattice has

infinitely

many different basis and

consequently infinitely

many

structure matrices.

Actually,

for any modular matrix U, I.e. with

integral

entries and

determinant ±

I,

we also have

L~

= AUZ 3 in such a way that A'

= AU is an

equally

valid

structure matrix of

L~ corresponding

to the basis

(a')

=

(a(, a(, a()

with

al

=

A'e,

=

£

U~, a~. Therefore a

given

lattice

only specifies

a class of structure matrices

J

defined up to

right multiplication by

a modular matrix.

We shall say that

L~

and

L~

are related

by

a

simple

shear transformation S

= I +

is) («

if :

I) L~

and

L~

have structure matrices A and B such that B

=

AS,

which is

equivalent

to

A B

= S

2)

the vector s is an irreducible vector of the lattice 23 and

(« is)

= 0.

This condition means that the

corresponding

bases

la)

and

16) satisfy

the

following

relation :

b,

=

ASe;

=

A

(e,

+ «,

s)

= a, + «, w

(2.2)

where w

=

As is a lattice vector of

L~.

Notice that since s in invariant

by S,

w

=

As

=

Bs also

belongs

to

L~

so that

L~

and

L~

have a common I-dimensional sublattice.

With more

appropriate

choices of bases

(such

that a~ =

b~

=

As)

this relation can take a

simpler

form where the shear matrix S

simply

reads :

II

o o

S= o I o

«1 «~ l

in such a way that :

b~=a~+«~a~

b~

= a~ + «~ a~

(2.3)

b~

= a~.

The above shear transformation S is defined and acts in the space of indices. In

physical

space, the

corresponding

linear transformation which maps

L~

onto

L~

is

given by

S~

= ASA One

easily

checks that :

S~L~=S~AZ3=ASZ3=BZ3=L~

S~

has the

following expression

:

s~ ~li i i i>j Ii1 (2.4)

with w

= As and

w =

(A

~)~ «.

Thus w

gives

the shear direction and

w defines the invariant

plane

in the

physical

space.

One

easily

checks that

(w [w)

=

0 which is

equivalent

to det

(S~)

=

I. The inverse of

JOURNAL DE PHYSIQUE -T 2, N's, JUNE 1992 34

(5)

S~

is the shear transformation

Si~

=I-

[w)(w[.

It follows from the definition of

S~

that

S~ (a~ )

=

b,

for I

=

1, 2,

3 which means that

S~

maps the basis

(a )

of

L~

onto the basis

(b)

of

L~.

More

generally,

if x=

£n;a;

is a lattice node of

L~ (n; integer)

then

S~(x)= £n;b;

is the lattice node of

L~

with the same indices. Notice that since

w is invariant

by S~,w

has the same coordinates with respect to

la)

and

(b).

Now,

if

L~

and

L~

are related

by

a

simple

shear transformation

S,

we can

easily

build a bounded transformation

3,

called hereafter a fractional

shear,

which satisfies the

following properties

:

Ii

3 is a one-to-one

mapping

of

R3

such that

3~

=

A3A maps

L~

onto

L~.

2)

3 is « bounded » in the sense that the

displacement 3(x)

-x remains bounded when

x runs over R3.

3 is

given by

the

following equivalent

definitions

(see Fig, 1)

:

3(x

= x + Frac

j (« [xi j

s

(2.5) 3(x)

=

S(x)

-Rnd

I("lxl)I

S

where

Rnd(t)

is the

(integer valued)

round function

(I)

in such a way that

Frac

(t)

= t Rnd

(t)

lies in the interval

( 1/2, 1/2].

By

definition of

3~

we have :

3~(x)=x+Frac j(«[A~~x)]As=x+Frac j(w[x)]w 3~ (x)

=

S~ (x)

Rnd

j (w [xi

w

~~'~~

Clearly,

3 is bounded since the norm of the

displacement 3(x)

x

= Frac

j(« [x)

s is

bounded

by

i

s11 /2.

Similarly, 3~ (x)

x =

Frac

j(w xi

w is bounded

by

iw11 /2.

Fig.

I. The square lattice L~ is transformed into L~

by

a shear S of direction w

= (2, 1). Both lattices

lay

in the bundle of dashed lines A

parallel

to w. The linear shear involves

diverging

translations between

x in L~ and S(x) in L~ whereas the fractional shear 3 translates x onto the nearest L~ node on the same line.

(1) The round function maps a real number t onto the nearest

integer

(with Rnd (n + 1/2) n for all

integer

n).

(6)

3 is a one-to-one

mapping

for it has an inverse transformation

given by

3~

~(x )

= x Frac

[ (« xi

s.

Similarly, 3~

is also one-to-one with an inverse

given by Ii ~(x)

= x Frac

(w xi

w. The

property

that

3~

maps

L~

onto

L~

then follows from the remark that

3~

differs from

S~ by multiples

of w which also

belong

to

L~ (see (2.6)).

Using coordinates,

the

mapping 3~

is

explicitly given by

:

x =

£x;

a,

-

3~ (x)

=

£

x, b~ Rnd

«;

x;j

w

(2.7)

where

(x~,

x~,

x~)

denote the coordinates of x with respect to the basis

(a).

The coordinates

~y~, y~,

y~)

of y

=

3~ (x)

with respect to the basis

(b)

are therefore

given by

:

y~ = x~ Rnd

£

«; x~ w;

(2.8)

where

(w~,

w~,

w~)

denotes the coordinates of w with respect to both bases

(a)

and

16 )

As a

conclusion,

the fractional shear

3~

has the

following specific property

: it translates a

node of

L~

onto the nearest node of

L~ lying

on the same lattice line

parallel

to

w ; this follows from the

amplitude

Frac

(w ix)

of the translation

being

bounded

by 1/2

and from w

being

irreducible.

3.

Simple

bounded transformations between lattices.

In this section we show that the

simplest

bounded transformations between lattices are

provided by

fractional shear transformations as defined in the above section. The terms

«

simple

» and « bounded »

require

a clear definition which is

given

below. It tums out that such transformations are

possible only

between

particular pairs

of

lattices, namely

those for which the transition matrix is

equivalent

to a shear matrix. More

complex relationships

between lattices will be considered in section 5.

By

a

simple

bounded transformation we mean a

global

one-to-one

mapping 4

of the 3- dimensional space such that

4 (x)

= x + w

(x)

u where q~ is a bounded real function and where u is a fixed vector. Thus

4

transforms any

point

x

by

a translation of constant direction

u and bounded

amplitude

q~

(x )

i u ii.

This definition is devised to

correspond

more or less to

simple displacive

transformations with the difference that in our context no energy consideration is involved.

As will become clear in the

following

the fact that such a transformation maps a

given

lattice

L~

onto some lattice

L~

is not obvious. This

actually corresponds

to

strong

restrictions

on

L~

and

L~.

THEOREM.

Let

L~=AZ~

and

L~ =BZ~

denote

two lattices and assume

4

is a

simple

bounded

transformation,

as defined

above,

which maps

L~

onto

L~.

Then

L~

and

L~

have the same

density,

u is a lattice direction of both lattices and after a convenient

scaling

(q~

- rq~ and

u - r~

u)

which does not

change 4,

u is an irreducible vector of

L~

and

L~.

PROOF.

One

easily

checks that if

L~

and

L~

had not the same

density,

a bounded one to one

mapping

from

L~

to

L~

would be

impossible.

Now for any x and y in

L~

we have :

w(x)- w(y)- w(x-y)

=

iq~(x) q~(y) q~(x-y)i

u.

(7)

First,

we can assume this

identity

does not reduce

always

to 0

=

0 otherwise the

amplitude

q~ would be a linear

(and bounded)

function in this case q~ would vanish and

consequently L~

=

Lb.

If the above

equation

is not trivial

then,

since the left hand side

belongs

to

L~,

u is a lattice direction of

L~.

Since

4

is assumed to be one-to-one from

L~

to

L~,

for any x and y in

L~

one can find

z in

L~

such that 4

(x) 4 (y)

=

4 (z).

This

implies

:

0=w(x)-w(y)-w(z)

=x-y-z+

iq~(x)-q~(y)-q~(z)iu

which in tum

implies

that u is also a lattice direction of

L~.

Consequently

both lattices have a common bundle of lattices lines A

parallel

to u. The

mapping

4 thus transforms the

L~

nodes on such a line into the

L~

nodes

lying

on the same line. Since the

displacements

are assumed to remain

bounded,

the I -dimensional lattices

L~

n A and

L~

n A must have the same

density,

I,e, the same lattice parameter. From this

remark we conclude that

q~ and u can be rescaled

by

some real number

r#0

(q~ -rq~ and u

-r~~u)

in such a way that u is a

generator

of both lattices

L~

n A and

L~

n A. Thus u is an irreducible vector of

L~

and

L~.

At this

point

we can raise a new

question

:

assuming L~

and

L~

are related

by

some

simple

bounded

transformation,

is there a

particular

transformation of this type which minimizes all

displacements?

The answer is

positive

and the solution is

given by

a fractional shear transformation.

Actually,

on any line A the I-dimensional lattices

L~

n A and

L~

n A are

simply

translated from each other

(they

have the same basis vector

u).

So there is

only

one

way to minimize the

displacements

: map any node of

L~

n A onto the nearest node of

L~

n A. But this solution is

precisely

what is obtained

by

the fractional shears as shown at the end of the

previous

section.

Finally,

if two lattices are related

by

a

simple

bounded

transformation, they

are also related

by

a fractional shear which minimizes the

displacements.

This fractional shear is therefore a

simple

and natural candidate for a

possible displacive

transformation between these lattices.

4. Limit

layers

and their average lattices.

In this section we show that limit

layers

of a lattice in the

vicinity

of a

plane

are close to a 2- dimensional lattice.

Furthermore, simple displacements transforming

the limit

layers

into their average lattices are

given by

fractional shears as described in section 2.

Let

L~

=

AZ ~ denote a 3-dimensional lattice with structure matrix A. Assume this lattice is cut

by

a

plane

lI of

equation (v ix)

= c and assume for

simplicity

that c

= 0. lI may be

considered as an ideal free surface of the lattice or as an interface

plane

with respect to an

other structure. In either case different

phenomena

will contribute to

change

the structure in the

neighborhood

of lI so that a definite limit

layer, supposed

to contain all

deformations,

is

hard to

specify.

We shall

only

consider the case where lI is

completely

irrational with

respect

to

L~. Rational,

or

partially

rational orientations of lI

give

rise to a

simpler

situation because the cut

plane

contains at least a I-dimensional sublattice.

For this reason, and for

geometrical

reasons

explained below,

we shall consider different but

particular

limit

layers satisfying

the condition that their width

correspond

to some lattice

vector. In other

words,

we consider limit

layers

bounded

by

two

planes

lI-w/2 and

lI+ w/2 where w is an irreducible vector of

L~ (which

does not

belong

to

10.

Once

w is

chosen,

we can scale the normal vector v of lI in order that

(

v

w)

= I. The limit

layer

thus contains all nodes x such that 1/2 <

(

v

ix)

w 1/2 and each node x has a

decomposition

:

x =

(

v x

)

w + y

(4. 1)

(8)

where y = x

(v ix)

w lies in lI for the above

equation implies ( vi y)

= 0.

Since w is a lattice vector of

L~,

the

mapping

defined

by

x - y = x

(v [x)

w maps all nodes of the limit

layer

on a 2-dimensional lattice contained in lI. This lattice is

simply

the

intersection of lI with the bundle of lattice lines

parallel

to w

(or

the

image

of

L~ by

the

oblique projector parallel

to w onto

10. Moreover,

since the width of the

layer

is

specified by

the irreducible vector w, each lattice line

parallel

to w contains a

unique

node inside the

layer

and

consequently

the

mapping

is one-to-one

(see Fig. 2). Finally,

the 2-

dimensional lattice thus obtained can be considered as an average lattice of the limit

layer

and the

displacements

are bounded

by

iw11 /2 and

parallel

to w.

The

mapping

x - y defined above is not a shear transformation since

( vi w)

is different from 0.

However,

for x inside the limit

layer,

this

mapping

proves to be

equivalent

to a

fractional shear transformation of the form x

- y = x + Frac

(w [x)

w which is defined as

follows :

Since w

belongs

to

L~

there exists a

reciprocal

vector p in

Lfl (different

from

v),

such that

(p [w)

=

I. Now we define

w = p v and we observe that :

I) (w[w)

=0.

2) (w[x)

=

(p[x) -(v[x)

so that for any lattice node x of

L~

we have Frac

[(w ix)]

=

Frac

[- (v[x) ].

Therefore for any lattice node x inside the limit

layer

we have Frac

(w [x) ]

=

(v [x)

and

consequently

the

mapping

reads :

y=x-

(v[x)

w=x+Frac

[(w[x)]w. (4.2)

This

equation

shows that the

mapping

of the limit

layer

onto its average lattice is

given by

a

fractional shear as defined in section 2. The shear direction is

given by

the lattice vector w and the invariant

plane

is

specified by

w with the condition

(w [w)

=

0 satisfied.

The definition of w

= p v is not

unique

because there are

infinitely

many

reciprocal

vectors such that

(p[w)

=

I. The

corresponding

linear shears

S~

=I +

[w)(w[

are

different but the associated fractional shears

3~

are

equal

because Frac

[(w ix)

] is

independent

of p for x in

L~.

Notice also that all shears

S~

transform the lattice

L~

into a

unique

lattice A

spanned by

w and

by

the 2-dimensional average lattice of the limit

layer

so that lI is a lattice

plane

of A.

fl+w/2

fl-w/2

w

Fig. 2. A layer is specified by the two planes 1I w/2 and 1I + w/2 where w is an irreducible lattice

vector and His an

arbitrary

direction. The nodes (black dots) lying inside the layer have an average

lattice (white dots) in 1I on which

they

map

by

translation bounded

by

ii w [[/2.

(9)

Conversely,

assume A is a lattice such that A

=

SL~

for some shear S

= I +

w) (w

where

w is an irreducible vector of both lattices. If lI is a lattice

plane

of A such that

w and the 2-dimensional lattice A n lI span

A,

then A n lI is the average lattice of the

layer

of

L~

bounded

by

lI w/2 and lI + w/2.

5. Shear factorization of a transition matrix and bounded transformations between lattices.

If

L~

is a

given

lattice and

L~

is a different lattice of

equal density,

the transition matrix T

=

A B cannot in

general

reduce to a

simple

shear matrix even with the free choice of the bases of the lattices.

Actually,

a shear matrix with rational direction

only depends

on 2

independent

real

parameters

whereas the set of lattices of

given density

is an 8-dimensional manifold. For this reason we guess that a

general

transition matrix will be

equivalent

to the

product

of 4 different shears. A

rigorous proof

is

given

below in 3 dimensions

(see [8]

for a

general proof

in n dimensions where n + I shears prove to be

necessary).

The cases where

only

one or two shears can transform

L~

into

L~

are therefore

exceptional

and

correspond

to

particular pairs

of lattices as will be seen in more details in section 6. If

L~

is defined

only

up to a rotation we see that these

exceptional

cases could

possibly point

out

particular

orientational

relationships.

If

L~

is

equal

to

L~,

up to a

rotation,

the same conclusion holds and it can be shown that the relative orientations associated with the existence of coincidence lattices

correspond

to such

exceptional

cases

[7].

In

[5]

it was shown that the relative orientations

giving

rise to icosahedral twins of cubic lattices as observed in

[6], namely

2 ar/5 rotations around

pseudo

5-fold

axis,

also

correspond

to transitions matrices

which are

equivalent

to a double shear.

Let T denote a real 3 x 3 matrix with determinant I. For instance T

=

A B is a transition

matrix between two lattices

L~=

AZ ~ and

L~ =BZ~

of

equal density.

Then T can be

decomposed

as a

product

of 4

simple

shear transformations

So, St,

S~ and

S~.

T

=

so s~ s~ si (5.1)

' So2 So3 o o

So

= o I o S~ = o I o

°

° '

31

532 with

0 0 S12 S13

S~

= s~j s23 S~ = 0 0

(5.2)

o o 0 0

This statement holds for almost all matrices T.

Actually

it

requires

the

following

conditions :

T~i

# 0 and T~~

T~i T~i

T~~ # 0. If these conditions were not met, one

easily

checks that a

permutation

matrix P can be found such that

they

are satisfied for

T'

=

TP

(taking advantage

of the fact that det

(T~

#

0).

The

proof

of the factorization

(5,I)

is

given by

the

following algorithm.

. Find a shear matrix

So,

of the form

given

in

(5.2),

such that

(So)~~

T

= U satisfies

Uii

= I and

Uii

U~~

Ui~ U~i

=

I. This is

easily

achieved since :

1) lfll

"

l~ll

S021~21 S031~31

~

l

~) ~fll ~22 ~12 ~21

"

l~ll

1~22 1~121~21 503(1~221~31 1~21 1~32) ~ l.

(10)

The second

equation gives

so~ and the rust one afterwards

gives

so~,

provided

the condition mentioned above is satisfied :

l~ll1~22

+ 1~121~21

~~~

1~221~31 1~21 1~32 1-

Tii

+

so~T~i

SW "

~m 21

. Find a shear matrix

Si

such that

U(Si )~

=

V satisfies

Vi

j =

V~~

= I and

Vi~

=

Vi~

= 0.

These

equations

read :

1) V22

"

~22

S12

~21

"

~) l~12

"

~12

S12

" °

3) ~'13

~

~13

S13

" °

They

are satisfied

by Uking

si~

=

Ui~

and si~

=

Ui~.

Now V reads :

0

1

~'"

~'21 ~'23

'31

~'32 ~'33

. Find a shear matrix S~ such that

V(S~)~

=

W satisfies

W~i

= W~~ = 0 and W~~ = I.

These conditions read :

1)

f~'21

"

~'21

S21 ~ °

~)

f~'23

"

~'23

523 " °

3)

f~'33 "

"33

523~'32 "

They

are satisfied

by taking

s~i

=

V~i

and s~~ =

V~~.

Now W reads :

Ii

o o

w= o i o.

W31

W32

. The last

step yields

a shear matrix S~ such that

W(S~)~~

=

I,

I,e. S~ =

W,

so that

S31 ~ f~'31 ~tld 532 ~

~'32.

Finally

we have obtained

(So)~ T(Si)~ (S~)~ (S~)~

=

I which reads :

T

=

SOS~S~SI.

The above factorization of the transition matrix T=A

~~B

shows that

any two lattices

L~

and

L~

with the same

density

are

usually

related

by

the

product

of four shear

transformations :

Lb

~ $A,O~A,3 ~A,2~A, I

~a

with

S~,

I = AS

; A

By replacing

each linear shear transformation

S~,, by

the

corresponding

fractional shear

3~,

; = AZ ;A ~, as defined in section

2,

we see that a bounded transform- ation from

L~

to

L~

is

given by

the

composition

of four fractional shears :

3~,

o

3~,

~

3~,

~

3~,

1.

As an

example

we shall consider here the 2-dimensional case of two square lattices with a arm relative orientation.

(11)

First the factorization theorem in 2 dimensions reads as follows :

If T is a 2x2 matrix with determinant

I,

there exist 3 shear transformations

So, Si

and S~ such that

T=

SOS~SI (5.3)

where

I

oy

i

oj

I

sjj

o 1

~~

s2

~~ o

~~'~~

The

product

of these shear matrices reads

So S~ Si

=

~

~ ~°~~ ~° ~ ~~ ~ ~° ~~~

(5.5)

s~ I + s~ si

The three

parameters

s~ si and s~ are

easily

derived from the entries of T

(notice

however that if

T~i

=

0,

T must be

replaced

with TP where P is a convenient

permutation matrix).

If

L~

and

L~

denote the two square lattices related

by

a arm

rotation,

the transition matrix T is

simply

the arm rotation and the above factorization reads :

/jl -lj [I I-/j[

0j[1 1-/j ~~6)

~ ~~

l 0 II

fi

0

According

to the definitions of section

2,

the

corresponding

fractional shears are

given by

30(x)

= x + Frac

[(I /) x~]

ei

~2(Y

" Y + Frac ~yl~

II

e2

(5.7)

Ii (z

= z + Frac

[(I /) z~]

ei

% / j~ I /~m ~ ~

j

' ~ *-w ~ *+ ~ ~a

/+m

~ ~

~ d ~~m

f / ~ j

w~ , ~ ~ ~

GO

W~G*~

. OW C~ ~B

+~ ~ +~ ~ +~

~ ~ ~ /

~*Q~

~R

'~

~ ~

*9e~

~

wv

j~ , j--~, ~

~, ~/, ~i /,

Fig.

3. The one-to-one

mapping

between the two square lattices at arm is

given by

the

composition

of

3 successive fractional shears.

(12)

The

composition

of these three

mappings

therefore maps

L~

onto

L~ (see Fig. 3)

and writes :

30 3~ 30(x)

= x + Frac

[(I /) x~]

ei + Frac

[(xi

+ Frac

[(

I

/) x~])/ ~fi]

e~

+ Frac

[(

I

/)(x~

+ Frac

[(xi

+ Frac

[(

I

/) x~] )/ II )]

ei

(5.8)

This formula shows that the

displacements

are

always

bounded

by

iet +

e~/2[[

=

l12.

This transformation can also be

computed explicitly using

coordinates associated to the successive

lattices,

like in section 2. As a

result,

x

= xi at +x~a~ in

L~

is

mapped

onto

y = yj

bi

+ y~

b~

in

L~

with :

yj = xi Rnd

[(I /)x~]

Rnd

i(

i

/) (x~

Rnd

(/(x~

Rnd

( (i /) x~))/2))1 (5.9)

y~ = x~ Rnd

[/ (xi

Rnd

( (I /) x~) )/2]

Such a bounded one-to-one

mapping

is not obvious. At first

glance

a natural candidate for

such a map would consist in

mapping

a

point

x of

L~

onto the

origin

y of the square of

L~

which contains x.

However,

squares of

L~

may contain

0,

or 2

points

of

L~

so that the one- to-one

property

could not be satisfied.

6. The double shear case.

The cases where the transition matrix

T=A~~B

between two lattices

L~ =AZ~

and

L~

=

BZ~

is the

product

of

only

two shear transformations seem to be of

physical

interest.

This

algebraic property

of the transition matrices has a more intuitive

meaning

as we shall see below.

Moreover,

a

geometrical

characterization is welcome since the transition matrices

between two lattices are defined up to

right

and left

multiplication by

modular matrices

(owing

to

changes

of

bases).

Let us first consider the case where T

= A B = S is a

single

shear transformation for an

appropriate

choice of the structure matrices A and B. Then, as

explained

in section

2,

both lattices have a common lattice vector w

=

As

=

Bs and

they lay

on a common bundle of lattice lines.

In Fourier space, the dual lattices

Lfl

and

Lf

have structure matrices A*

=

(A~~)~

and B*

=

(B~~)~.

If the bases

la)

and

16)

are such that a~ = b~ = w, like in

(2.3),

the

corresponding

dual bases

la *)

=

jai, at, at

and

16 *)

=

16(,

b~*,

b?)

are related

by

the

following equations

:

at

=

b? (6.1)

~~"~~+~l~l+~2~~.

All lattice

planes corresponding

to

reciprocal

vectors nj

al

+ n~

at belong

to both lattices.

These

planes

are those which are built

by

means of any

pair

of lattice lines

parallel

to

w. The dual lattices

lay

on a stack of

planes parallel

to the

plane spanned by

at

and

at

and their intersection

generically

reduces to the 2-dimensional lattice

spanned by

the same basis.

The double shear case is characterized

by L~

and

L~ having

structure matrices

A and B such that T

= A B is the

product

of two shear matrices :

~

~2 ~l (~

+

'52) (~2' )(~

+

'Sl) (~l )

(~ ~)

Tx

= x +

(«i ix)

si +

(«~[x)

s~ +

(«~[sj) («~ ix)

s~

(13)

where si and s~ are two different irreducible

integral

vectors

defining

the directions of the shears in the space of indices. The invariant

planes

are defined

by

«~ and «~ and

satisfy

(«~[s~)

=

(«~[s~)

=

0.

Since A B

=

S~ St,

we may define the structure matrix r

= AS

~ = BS

p

of a new lattice A

=

rZ~.

A is

a

simple

shear deformation of both

L~

and

L~

and can be considered as an

intermediate lattice between them: if

S~=AS2A~'

and

S~=BSi~B~'

we have

r

=

S~A

=

S~

B so that A

=

S~ L~

=

S~ L~.

These shears are

explicitly given by

:

s~

= I +

jw~j jw~j

~ ~

s~

= i +

iw~i iw~i

with w~ = As

z, w~ =

(A

~)~ «~, w~

=

Bsi

and w~ =

(B~

~)~ «~.

Now, using

the above relations we see that :

rsi

= AS

~ si =

As1

+

(«~[si) As~

=

BSj

si

=

Bsi

= w~

(6.4) rs~

= AS

~s~ =

As~

= BS

j

s~ =

Bs~ (« ii s~) Bsi

= w~.

Consequently

the three

pairs

of vectors

(rsj, rs~)

in

A, (As1, As2)

in

L~

and

(Bsj, Bs2)

in

L~

span the same lattice

plane

W which therefore

belongs

to the three lattices. W is invariant

with

respect

to both shears S~ and

S~

since the directions of the shears w~ and w~ are in W. Therefore

L~, L~

and A

lay

on a common stack of lattice

planes parallel

to W.

Finally,

the double shear

Sp

'S~ which maps

L~

onto

L~

transforms the

points by

translations

parallel

to W.

In Fourier space, the existence of a common sequence of lattice

planes implies

that the

reciprocal

lattices of

L~

and

L~

have a I-dimensional intersection

spanned by

a

reciprocal

vector w

orthogonal

to W. The Miller indices of

w are

proportional

to sj x s~

(and equal

if this vector is

irreducible). Actually,

the dual lattices

Lfl

and

Lt

have

respective

structure matrices A*

=

(A

~)~ and B*

=

(B~

~)~ which

satisfy B*~~A*

=

T~=

(I

+

i«i) isii)(I +1«2)iS21),

therefore si x s~ is invariant

by

T~ and A* si x s~

=

B* si x s~

belongs

to both

reciprocal

lattices.

7. Interfaces and smooth transitions in the double shear case.

An ideal

planar

interface between two lattices

L~

and

L~

is defined

by

5

parameters

: two of them

specify

the orientation of the interface

plane

and the others

give

the relative translation

between the lattices. The double shear case has

interesting

consequences with

regards

to the

description

of

possible

interfaces

planes

between

L~

and

L~.

More

precisely

we show now that in this situation

pa~icular

interface orientations are

pointed

out for

simple geometric

reasons.

Assume His a lattice

plane

of the intermediate lattice A defined in the above section which satisfies the

following

conditions : w~ and A n lI span

A,

and

similarly

w~ and A n lI span A

(see below).

We consider the

layer

of

L~

bounded

by

lI

wJ2

and lI +

wJ2

and the

layer

of

L~

bounded

by

lI-

w/2

and lI+

x/2.

NOW these two

layers

have the same average

lattice, namely

the 2-dimensional lattice

A n lI. This

actually

follows from the construction of average lattices of

layers explained

at the end of section 4.

(14)

More

precisely,

the fractional shear

3~

which maps the

L~

nodes

lying

between lI-

wJ2

and lI+

wJ2

on the

plane

lI is

given by

:

3~(x)

= x + Frac

[(w~[x)

w~

(7,1)

3~

maps the nodes of limit

layer

of

L~

onto the 2-dimensional average lattice A n lI.

Similarly,

the fractional shear associated to

S~

is

given by

:

3~(x)

= x + Frac

[(w~[x)

w~

(7.2)

3~

maps the nodes of limit

layer

of

L~

onto the same 2-dimensional lattice A n lI

(see Fig. 4).

The condition

bearing

on the

planes lI,

mentioned at the

beginning

of this

section, requires

that the 2-dimensional lattice An lI spans A with both irreducible vectors w~ or

w~. Therefore if

v is the

reciprocal

vector of A* associated to lI we must have

~~' ~a)

~ ~ ~'~~

~~' ~b)

" ~ ~.

~ o o

~ o

~ o °

~ o

~ o

~ o

~ o

~ o

~ o o

~ o °

~ o °

a

~ o o

~ o

~ o °

~ o ~ a a

~ o o

o o °

~ o °

a D a

~ o o

o o °

~ o °

a a a a

~ o

~ o °

~ o D a a a D D

~ o

~ o

o o D D a D a a a a

o o

o o °

a a a a a a a a a

o o

'I

a D a D D D a a a a

a D D a D D a D a a a a

a D D D a D D a D D D D a

Fig.

4. The limit

layers

of L~ and L~ arc

parallel

to a lattice

plane

of the intermediate lattice A.

They

have the same average lattice on which

they

map

by simple displacements

of constant direction

and bounded

amplitude.

Finally

the double shear case is characterized

by

the

following prope~ies

:

I) L~

and

L~

have a common lattice

plane

W

spanned by respective

bases

(As

i,

As~)

and

(Bsi, Bs~)

such that

Bsj

=

As i +

r~AS~

and

As~

=

Bs~

+ ri

Asi

with si and s~ in

Z~,

and

ri, r~ real numbers

(see Eq. (6.4)).

Références

Documents relatifs

This simple example illustrates the fact that the use of a weighted or a non weighted graph for the join count statistics can have a strong influence on the result, depending on

It is shown that if the monolayer- to-bilayer transition occurs with only small changes in the nearest neighbour spacing, the chemical potential there depends on the

In the macroscopic geometric limit (that is the macroscopic, rigid and quasistatic limits), with either volume or normal stress controlled simulations, static and dynamic

Schr¨ odinger equation, Classical limit, WKB expansion, Caustic, Fourier integral operators, Lagrangian manifold, Maslov

We now study the scaling limits of discrete looptrees associated with large conditioned Galton–Watson trees, which is a model similar to the one of Boltzmann dissections: With

However, volumes of limit and market orders are equal in the two latter cases, and we have shown in section 3 that these different shapes can actually be obtained with different

A vector bundle with metric connection (E, h, ∇ E ) → (M, g) over a Riemannian manifold with boundary is of bounded geometry if (M, g) is of ∂- bounded geometry and the curvature

We prove that a uniform, rooted unordered binary tree with n vertices has the Brownian continuum random tree as its scaling limit for the Gromov-Hausdorff topol- ogy.. The limit