• Aucun résultat trouvé

Steady poiseuille flow in nematics : theory of the uniform instability

N/A
N/A
Protected

Academic year: 2021

Partager "Steady poiseuille flow in nematics : theory of the uniform instability"

Copied!
11
0
0

Texte intégral

(1)

HAL Id: jpa-00208509

https://hal.archives-ouvertes.fr/jpa-00208509

Submitted on 1 Jan 1976

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.

Steady poiseuille flow in nematics : theory of the uniform instability

P. Manneville, E. Dubois-Violette

To cite this version:

P. Manneville, E. Dubois-Violette. Steady poiseuille flow in nematics : theory of the uniform in- stability. Journal de Physique, 1976, 37 (10), pp.1115-1124. �10.1051/jphys:0197600370100111500�.

�jpa-00208509�

(2)

1115

STEADY POISEUILLE FLOW IN NEMATICS :

THEORY OF THE UNIFORM INSTABILITY

P. MANNEVILLE and E. DUBOIS-VIOLETTE

(*)

Service de

Physique

du Solide et de Résonance

Magnétique,

Centre d’Etudes Nucléaires de

Saclay,

BP

2, 91190 Gif sur-Yvette,

France

Résumé. 2014 Nous étudions la stabilité d’un écoulement de Poiseuille dans un nématique planaire

relativement à une distorsion uniforme du directeur. La théorie hydrodynamique linéarisée que nous

développons permet de caractériser la

symétrie

des modes instables. On s’attend à deux types de solutions suivant la présence ou non d’écoulement transverse

global.

La solution exacte, donnée dans un cas

simplifié,

montre que l’instabilité avec écoulement transverse global a le seuil le plus bas,

en accord avec les résultats expérimentaux.

Abstract. - We study the stability of a Poiseuille flow in a planar nematic, with respect to a uniform distortion of the director. We develop a linearized hydrodynamic theory which allows one to charac- terize the symmetry of the unstable modes. One expects two types of solutions according to the

presence or not of a net transverse flow. An exact solution is given in a simple case which shows that the instability with net transverse flow corresponds to the lowest threshold in agreement with

experi-

mental results.

LE JOURNAL DE PHYSIQUE TOME 37, OCTOBRE 1976,

Classification Physics Abstracts

6.350 - 7.130

1. Introduction. - An

experimental description

of

the Poiseuille flow

instability

in nematics has been

given

in the

preceding

paper

[1] (see

also ref.

[2]).

Experiments

are

performed

on a nematic oriented in a

planar

geometry

(director parallel

to

Ox)

and

subjected

to a pressure

gradient

in the

y-direction (Fig. 1).

Both

steady

and

alternating

Poiseuille flows have been examined. Two

types

of

instability

can

develop

in the

sample.

The first one

corresponds

to a

distortion of the director which is uniform in the

xOy plane.

The second one leads to the formation of convective rolls which have their axis

parallel

to the

flow direction

Oy. Furthermore,

as

long

as

phenomena

above the threshold are not

involved,

the uniform

instability

is found to

develop

first for a d.c. excitation.

FIG. 1. - Geometry : a pressure gradient 0394P/L is applied in the y-direction on a planar nematic (molecules along Ox).

In this paper, we

give

a theoretical

description

restricted to the uniform distortion observed at threshold in the

steady experiments

and we leave to a

forthcoming

paper the case of

alternating

flows.

In this context, we want to

emphasize

that

experiments reported by

Pieranski and

Guyon

in ref.

[1]

and

[2]

correspond

to observations at and above the threshold.

In

particular, they

have seen two types of uniform

distortion,

A and

B,

characterized

by

the absence

or presence of a net transverse flow in the x-direction.

Regime A,

which

corresponds

to a metastable state, is obtained from a

highly

distorted flow

by decreasing

the pressure

gradient

and is outside the scope of this article. On the other

hand, regime

B is obtained from the

unperturbed

Poiseuille flow

by increasing

the

pressure

gradient.

The

instability

which

develops

is

linear and

stationary [3]

and can be

analyzed

within

the framework of a linearized

hydrodynamic theory.

Regime

B is characterized

by

an average twist of the director and

by

a net transverse flow in the x-direction.

This

secondary

flow is different from the one we had

predicted [4]

and which has been observed

[5]

in the

simple

shear flow

experiment.

In that case the trans-

verse velocity 03C5x

was an odd function of z such that :

$$

This

corresponds

to a

long

roll

parallel

to

Oy, occupying

all the width of the cell with no net flow V in the x-direction. In the present case, a net flow V ~ 0 is observed at threshold. This

secondary

flow cannot be sustained all

along

the channel and there must exist

regions

where it

develops

in one

(*) Laboratoire de Physique des Solides, Universite de Paris-Sud,

91405 Orsay, France.

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0197600370100111500

(3)

direction and

regions

where it

develops

in the other

one. This leads to the presence of domains in the

y-direction (in

a frame

moving

with the

primary

flow

(see Fig.

8 of ref.

[1]),

the

secondary

flow stream lines

have to close in the

x0y plane).

Such a domain struc-

ture is

closely

connected with the finite width I of the

experimental

cell but not with the

instability

mecha-

nisms. In the

following,

we shall

neglect

these lateral

boundary

effects

(i.e.

take the limit of infinite

width)

and discuss the

instability by

means of a

simplified

one-dimensional model.

2. Linearized

hydrodynamic equations.

- Let us

consider the cell described in

figure

1 and call

G =

AP/L

the pressure

gradient applied

in the

y-direction.

Molecules are oriented

(through

surface

treatment)

in the x direction. When the flow is very

slow,

due to elastic

effects,

this

alignment prevails

in the bulk. Viscous

torques

exerted on the molecules

are zero and the

viscosity

measured in this geometry is lla

[6, 7].

For a cell of infinite width

I,

the

velocity profile

is :

independent

of x

(the origin

of the coordinates is taken

midway

between the

plates). However,

in an actual cell of finite

width, vy

may be considered as

independent

of x

only

in the central part of the

cell ;

this central

part

gets wider and wider as the width I is

increased,

as can be seen in

figure

2

(calculation

for

an

isotropic liquid).

For narrow channels

(typically lld 10)

the

boundary layers

on each side are wide

and the

velocity profile

has no

longer

the

simple

form

(2 .1 ).

In the

following

we shall

only

consider

wide cells and examine the

stability

of the

unperturbed plane

Poiseuille flow.

FIG. 2. - Transverse unperturbed velocity profile at z = 0 for an isotropic liquid Poiseuille flow in a channel of finite width I. Curves

are given for different values of the aspect ratio r = l/d, r = 2, 5,10, 50.

Eq. (2.1)

defines a shear rate

which is

positive

in the lower

half (z 0)

of the

sample

and

negative

in the upper one. At first

sight,

one could

look for a

description

of the actual flow in terms of

two

superposed

average

simple

shear flows : 7 for

z 0 and - s for z > 0.

Then,

one would use the

analysis developed

earlier

[4].

However,

though

the

instability

mechanisms are

unchanged,

the situation

is somewhat different since the shear rate cannot be considered as

piecewise

constant.

Furthermore,

the

boundary

conditions are different : in the

simple

shear

flow problem,

boundaries were

rigid planes;

here, the simple

shear s would be

applied

between a

rigid

wall at z = -

d/2

and a

free boundary

at z =

0,

the other one - s between the free

boundary

at

z = 0 and the

rigid

one at z = +

dl2

and one would

have a

continuity

condition at z = 0. This

simplified picture

cannot

give

more than a

qualitative descrip-

tion but it stresses the

similarity

of the

instability mechanisms,

and the linearized

hydrodynamic

equa- tions can be

immediately

deduced from those govern-

ing

the

simple

shear flow

problem (1).

Let us denote bn =

(0,

ny,

nz),

ðv =

(vx,

vy,

vz) and p

the

orientation, velocity

and pressure fluctuations.

We do not discuss the roll

instability (ox 1= 0)

and

the domain structure

(Oy 1= 0)

so that we

only

consider

the

z-dependence

of the fluctuations. In this case,

as for the

simple

shear flow

problem,

one can show

that vy = vZ = p = 0

everywhere.

Moreover,

experi-

ments suggest that the

instability

grows at threshold without time oscillations so that we are left

with

a

time

independent problem. Then,

the linearized

equations

reduce to :

Torque equations :

Force

equation :

Boundary

conditions read :

Let us

give

here a

rapid analysis

of eq.

(2.2)-(2.4) referring

the reader back to ref.

[4]

and

[8]

for a more

complete description

of the mechanisms.

The viscous

destabilizing

contributions to the total

torque

a3 sny and a2 snz

correspond

to the

Guyon-

Pieranski

instability

mechanism

(2).

The corrective

(1) The only difference with equations written in appendix A of

ref. [4] comes from the z-dependence of s(z) in the Ericksen-Leslie tensor term QZx which contains s(z) ny.

(2) This mechanism works only when a3 0 (a2 is always 0).

If a3 > 0, another mechanism leads to a roll instability [4, 8]. In the following we shall assume a3 0.

(4)

1117

term a3

0,-,V-,

in eq.

(2.2)

is due to the presence of a

transverse flow v.,. Two mechanisms induce this transverse flow. One related to the twist distortion

(term s aZny

in eq.

(2.4))

was

already present

in the

simple

shear flow

problem.

The other one due to the

variable shear rate

s z term

n

ds

is characteristic dz)

of the Poiseuille flow

[9].

From a

general point

of

view,

it is

interesting

to

find the symmetry

properties

of eq.

(2.2)-(2.4)

which are

compatible

with

boundary

conditions

(2. 5).

From the form of these relations one may guess that non-trivial solutions will have a definite

parity.

Let us consider for

example

eq.

(2.3)

and look for the solution at two

points symmetric

relative to the

xOy plane.

Since

s( - z) = - s(z)

we

get :

Let us define the even

(e)

and odd

(o) parts

of a fluctuation

w(z) (where w

stands for ny, nZ or

vx) through :

Combining

eq.

which expresses that the even

part

of ny is

coupled

to

the odd one

of nZ

and

conversely.

A similar

separation

can be

performed

on eq.

(2.2)

and

(2.4)

which leads

to the fact that a solution of

given parity (for example

ny, nz, vx)

does not mix with a solution of the

opposite

one

(ny, ny, vx).

This

property

can be extended to the

case of a

general

fluctuation

depending

on x, y, z.

It has been summarized in table I : TABLE I

Two modes are obtained. The first one

(T)

corres-

ponds

to a finite average twist

(ny even)

and a net

transverse flow

(f v,,, dz =A 0 .

The second one

(S) corresponds

to a

finite

average

splay (nZ even)

and no

net transverse flow

(vx odd).

In

general,

these two

modes will have different threshold values.

3. Threshold determinations. -

Performing

the

transformation z

= d

Z 2

we obtain the dimensionless

equations :

where

is the Ericksen number

[6] characterizing

the flow.

E

are two dimensionless numbers related to external fields.

Integrating

eq.

(3.3)

we get :

where K is an

integration

constant.

Replacing D vx

in eq.

(3.1) gives :

and we are

left

with the

problem

of

solving

eqs.

(3.7), (3.2), (3.6) together

with the

boundary

conditions

Ny = NZ

=

vx

= 0 at

Z =

± 1. The

general

case

F # F’ can

only

be solved

using

an

approximation

method. On the contrary, when F = F’ the combi- nations :

further

simplify

the

problem and lead

to an

analytical

solution in terms of

Airy

functions

[10].

Let us turn

to this

simpler

case.

3.1 NO EXTERNAL FIELD APPLIED

(F

= F’ =

0).

-

Combining

eqs.

(3. 7)

and

(3. 2)

we

get :

The

parity property

allows one to separate a solution with average twist

(T)

from a solution with

(5)

average

splay (S).

In the first case,

Ny and Yx

are even

functions of Z so that

ZNy

and

DVx

are odd : from

(3 . 6)

we deduce that K =

0,

i.e. U and V are

decoupled

and that

vx

is

simply given by :

V,(O)

is determined

by

the

boundary

condition

On the contrary

when NZ

is even

(mode S)

it will be

shown that K is different from zero and

actually couples Vx

to

Ny

and

NZ,

3.1.1

Average

twist solution

(Ny

even function

of

Z, K

=

0).

- Let X =

ZE//3. Eqs. (3 . 8), (3. 9)

now read :

the

general

solution may be

expressed

in terms of

Airy

functions

Ai,

Bi

[10]. Recalling

that

and

assuming

that

N,

is an even function of X

(or Z)

we get :

Boundary

conditions

Ny = NZ

= 0 at Z =

+ 1

lead

to

The solution will be the trivial one A = B = 0 unless

Xl

is a root of the characteristic determinant :

and the

instability

threshold will be reached for the lowest root

Xc :A

0 of eq.

(3.14).

This

equation

may be written as :

However,

for

large

X :

so that the I.h.s. of eq.

(3.14) rapidly

decreases to

zero when

Xl

increases.

Neglecting Ai(Xl)lBi(Xl)

in eq.

(3.15)

then leads to the

approximate

solution :

or

a result which is very close to the exact one

The

corresponding

Ericksen number is

in

good

agreement with

experimental

results

given

in

[1].

Once

Xf

is

known,

one can deduce the fluctua- tion

profiles

at threshold from

(3.12)

and

(3.10)

with

B /A

extracted from

(3.13)

To the same

approximation

level as for

(3.16),

we

simply

get :

A

The aspect of the distortion at threshold is

given

within an

arbitrary

constant in

figure

3 which

clearly

reflects the

parity

property. In

particular,

the

velocity profile vx implies

the presence of a net transverse flow

compatible

with the domain structure

experi- mentally

observed.

FIG. 3. - Fluctuation profile for the T-solution with mean twist and net transverse flow (no external fields applied).

3 .1. 2

Average splay

solution

(Ny

odd function

of Z).

- In this case, if we assume K =

0,

we are left with

the

previous problem

after

interchange

of

Ny

and

N.-

The critical value

Xf

is

unchanged but Vx

must be an

odd function

of Z ; then, Vx(Z

=

0)

= 0 and

This solution would no

longer

fulfil the

boundary

condition

on Yx

so that the

assumption

K = 0 is

wrong.

(6)

1119

The

general

solution with

Ny

odd and

NZ

even as

functions of Z is :

Boundary

conditions on

NY

and

Nz

or U’ and V’

at X =

± X1

lead to :

so that A and B are now functions of

Xl.

But

Ny

and

Nz

still

depend

on the

integration

constant K

(or k).

Replacing

in eq.

(3. 6)

we get

and the

boundary

condition

Vx(Xl)

= 0 leads to the

compatibility

condition :

in order to get a non-trivial solution

(k # 0).

In

contrast to the first case, the lowest root of eq.

(3.18) depends explicitly

on a dimensionless parameter

characterizing

the

liquid crystal : l1b/l1a’

The critical Ericksen

number Er

as defined

by

eq.

(3.4)

is

given

as a function of

(1 - l1bll1J

in

figure

4 for 0 llb na’

Figure

5

displays

the aspect of the fluctuations at threshold for MBBA with

It may be noticed that :

a) When na

= qb, the system is unstable

against

a

distortion with average

splay

at the same critical

value as for a distortion with average twist. In fact

vx

vanishes as can be deduced

directly

from eq.

(3.3)

and the two types of distortion become

equivalent.

FIG. 4. - Critical Ericksen number for the S-solution with mean

splay as a function of

(I - 1-b)

for 0 I’/b na For MBBA na

1Jb/1Ja = 0.572 and Er’ = 17.313. The critical Ericksen number as

defined by eq. (3.4) is independent of nb/na in the T-case.

FIG. 5. - Fluctuation profile for the S-solution in zero external field.

b)

The threshold value increases with 1 -

1Jbl1Ja’

In order to account for this variation one has to consi- der the contribution of the transverse flow to the

destabilizing

mechanisms. In ref.

[4],

we discussed

these mechanisms in terms of effective viscous torques. Written in a dimensionless

form,

these viscous torques read :

and the

mechanism

is

locally destabilizing

if :

Consider first a distortion with average twist. Then K = 0 and the shear rate

D Vx

is

proportional

to

ZNy.

The condition

(*)

is fulfilled

everywhere

in the

sample.

On the other

hand,

when K # 0

(average splay distortion)

and the

sign

of the contribution of the shear flow D

Vx

depends

on the

position

Z.

Looking

at the

profile

of

(7)

Ny

in

figure

5 as obtained from the exact

calculation,

one can guess that

which has the

right parity

and satisfies the

boundary conditions,

will lead to a

good approximate descrip-

tion.

Then,

and from the

boundary

condition

The effective viscous torque

Fy

reads :

In absence of

secondary

flow

(1Ja

=

nb)

this torque would be :

so that the contribution of

D Vx

not

only

reduces

Ty eff

but also can make its

sign change; leading

to a

stabilizing

mechanism in some

regions

of the

sample.

This is

particularly

the case near the

boundary plates

of the cell

(Z ~ 1)

where a

high

shear leads one to

expect a strong

destabilizing

mechanism. Since

r zl Nz ~ -

Z the

instability

condition

(*) simply

reads :

The

region

where it is fulfilled shrinks

when ’1b decreases,

so that we

qualitatively

understand the increase of

E’

with

1 - 17 b

llain

figure

4.

For all

physical

situations

with ’1b

na the dis- tortion which appears

corresponds

to the lowest

critical Ericksen

number,

that is to say : a distortion with average twist and transverse flow as observed

experimentally.

Our

instability

with average

splay

should not be confused with the metastable

regime

A

described in ref.

[1].

As a matter of

fact, regime

A

displays

an average twist but no transverse flow at all

(Vx(Z)

=

0), keeping

the trace of a feature which appears in the non-linear

domain,

since it is obtained

by starting

from a

large

distortion.

3 . Z MAGNETIC FIELD EFFECTS

(CASE

OF AN IDEAL NEMATIC WITH

K1

=

K2 = K; F

=

F’).

-

Equations governing

U and V now read :

with

where

is a Fredericks field

[6] (3).

For the

T-solution, boundary

conditions

again

lead to eq.

(3.14)

but with the

changes :

Eq. (3.18) corresponding

to an S-solution is also

slightly

modified but we

again

expect a distortion with average twist at threshold and concentrate on this

case.

Eq. (3.15)

now reads :

The

applied magnetic

field

H, parallel

to the unper- turbed

director,

has a

stabilizing

effect and we expect

Erc

to be an

increasing

function of F.

Then, Xl

grows with F and

consequently :

we may

neglect

the I.h.s. of eq.

(3.20)

which

simply

reads :

A;iY - Yci - n

or

Recalling

that

and since

increases

with F,

the

approximation

which leads to eq.

(3.21)

is more and more accurate. No such appro- ximation exists for the other

type

of

instability

and

the

complete

calculation has to be worked out.

(8)

1121

Numerical results are

given

on

figure

6. Once

again

we expect the distortion with average twist which

corresponds

to the lowest Ericksen number for all F.

Fluctuation

profiles corresponding

to this case are

FIG. 6. - Critical

Ericksen

number as a function of the magnetic

field measured the distortion with by F = mean

Xa H 2 (d)2 K 2

twist to for develop the S and T whatever the field is.case. One expects

FIG. 7, 8. - Fluctuation profile for the T-distortion at a given F=AO. Fig. 7.-F= 10 or H ~ 2 Hc. Fig. 8.-F= 150 or

H ~ 8 Hc.

given

on

figures

7 and 8 for H N 2

Hc(F ~ 10)

and

H - 8

Hc(F ~ 150) respectively.

The

aspect

of the distortion at threshold may be understood

qualitati- vely

as follows :

- Opposite

to the case of the

simple

shear flow

[4], here,

the shear rate s =

aZvY

is not constant across

the

sample.

The

corresponding instability

mechanism

is the

strongest

close to the

plates,

then decreases and vanishes

midway

between them.

- On the other

hand,

the

applied

field tends to

align

the director back

along

Ox.

Across the thickness of the

sample

we may then

-

distinguish

the central part, where the

stabilizing

effect of the field

dominates,

from the

neighbourhood

of the

plates,

where the

instability

grows

[6].

This feature is obvious on

figure

8 for H - 8

H,

but the

tendency

was

already

indicated on

figure

7

for H - 2

Hc.

Note however that eq.

(3.21)

which

gives Ec(f)

cannot be derived from a

simplified

model in terms of a constant shear across the unstable

layers.

3. 3 GENRatL CASE

(F # F’).

- Since no obvious

variable or function

change

allows to come back to

a system of the form

(3.19),

we

develop

an

approxi-

mate method in terms of Fourier

expansions [3].

The method which will be outlined below is very

powerful

due to the fact that it takes into account

the

parity

property and the

boundary

conditions

simultaneously.

Let us consider the case of a distortion with average twist

(Ny

even function of

Z). Eqs. (3.7)

and

(3. 2)

now read :

N,

is odd and must vanish at Z = ± 1 so that we

assume :

Since eq.

(3.23)

is linear we may consider

Ny

as a

superposition

where

Ny(n)

are the solutions of

which fulfil the

boundary

conditions

... T IMB/r-7 I w -

The

general

solution of the reduced

equation :

hich is an even function of Z reads :

and a

particular

solution of the

complete equation

(3.26)

is :

(9)

so that the solution we are

looking

for reads :

Now

Ny

and

NZ

must

verify

eq.

(3.22).

This

gives :

The I.h.s.

of eq. (3.27)

reads as the Fourier

expansion

of the function defined

by

the series at the r.h.s. In order to

identify

the coefficients we

multiply

both sides

by

sin

(pnZ)

and

integrate

from zero to

1 ;

then we get :

This system may be written as follows :

with À

= Er 2, Then,

the coefficients

are understood as the components of an

eigenvector

A

of the matrix A,

corresponding

to the

eigenvalue À :

and the

highest

real

positive eigenvalue Àc

will corres-

pond

to the critical Ericksen number

given by :

The exact solution is

given by

the infinite series

(3.24)

and

(3.25);

an

approximate

one will be

given by

the truncated series :

- To first order v = 1 we have

- To second

order, Ac(v

=

2)

is the

highest

root

of the characteristic determinant :

- and so on.

Including

more and more harmonics leads to the

exact value. When F = F’ =

0,

the convergence is rather

quick

as can be

appreciated

on table II.

The reason for this

rapid

convergence is

already

apparent from the aspect of

Nz

in

figure

3 which

is very close to a sine wave. For

F,

F’

large,

when the

fluctuation

profiles

look like those of

figure

8 one

expects a much slower convergence but the method remains

quite

efficient.

TABLE II

Average

twist

F and F’ defined

by

eq.

(3.5)

measure the

strength

of the external

stabilizing

fields. In order to compare the ideal case F = F’ to the actual one F -# F’ we

shall take the

geometrical

mean

JFF’

as a

global

measure of the

stabilizing

forces. In absence of elec- tric

field,

this choice is

equivalent

to the

change

from

Ki, K2

to

K

=

J K1 K2,

an

approximation

which

leads to

good

results every time the

anisotropy

of the

elastic constant does not

play a

crucial role. Numerical results

plotted

on

figure

9 are in

good

agreement with the

experimental

ones.

(10)

1123

FIG. 9. - Critical Ericksen number as a function of external fields measured by

for different values of F’/F : 1,1.5,2. All calculated points fall close

to the curve given as S on figure 6 which corresponds to F = F’.

With

while the

experimental

ratio is in the range 16-19.

It may be noticed that the calculated

points

for

Ki

= 1.5

K2

and

Kl = 2 K2

are

practically

on the

same curve as those for

K1 = K2 = k

which were

obtained

much more

simply

from eq.

(3 . 21). However,

this fact does not reduce its

efficiency

to the method

which will be

employed

in cases when no exact solu-

tion can be found

(roll

instabilities described in ref.

[1])

or when the exact solution is too

complicated.

As an

example

let us come back to the case of a distortion

with average

splay.

Then

assuming

we may

deduce Yx

from eq.

(3.6)

and from the

boundary

condition J

and

Nzn)

is the solution of eq.

(3 . 7)

which fulfils the

boundary

condition

N,(Z

= +

1)

= 0.

x

Replacing Ny

and

Nz = Y Nzn)

in eq.

(3.2)

and

n=1

following

the same steps as before one obtains a system similar to

(3.28).

The

only

modifications are :

- the

interchange

of F and F’

(with f

=

,FF),

- the addition of the term :

(which corresponds

to the contribution of the inte-

gration

constant

K)

to the r.h.s. of

(3.28).

In the

limit F = F’ =

0,

this

supplementary

term

simply

reads :

Corresponding results,

when one varies the order of

truncation,

are

given

in table III. Once

again

the

convergence is rather

good,

which could be

expected

from the aspect of

Ny(Z)

in

figure

5 which looks like a

sine wave. However one should observe that the

limiting

value is reached

from below, opposite

to the

case of a distortion with average twist

(table II) (4).

TABLE III

Average splay (*)

(*) MBBA : nb/na, = 0.572.

4. Conclusion. - In this paper, we have

given

a

theoretical

description

of a Poiseuille flow

instability corresponding

to a uniform distortion of the director.

As for the

simple

shear flow

instability,

the Poiseuille

case is well described

by

means of a linearized

theory.

Locally,

the mechanisms are similar in both cases, which is revealed in the

expression

of torques and forces

(Eqs. (2.2)-(2.4)). However,

the presence of a

non-constant shear modifies the symmetry

properties

of the unstable modes. In the

simple

shear case two

(4) This seems to remove part of the interest for a variational principle governing this kind of problem since such a principle

would not have a definite extremum property.

(11)

types of solutions

(ny, nz, vx)

and

(n y, nz, vx)

could

exist. The

instability

threshold was obtained for the first one which was associated with a lower elastic energy. In the present case the

parities

of ny

and

are

opposite

in a

given

mode. Solutions are of the form

(ny, n’, vD

and

(ny, nz, v’)

and the elastic energy no

longer plays

a

prominent

part.

Actually,

it is the

contribution of the transverse flow to the viscous torque which separates the two solutions as discussed

in §

3.1.2. When no transverse flow is present

(1Ja

=

nb)

the

instability

thresholds are

equal

for solutions with average twist

(nye)

or

splay (nz).

For the more realistic

case nb na’ due to the presence of transverse flow

one expects an average twist distortion associated with

+d/2 a net transverse flow

(f+d/2 dz Vx =1= 0 .

Exact solu-

- d/2

tions may be derived when

Ki

=

K2

and no electric

field is

applied.

For the

general

case we have

develop-

ed a

powerful approximate

method which is

being

extended to the case of the roll

instability

in alternat-

ing

flows. Threshold values and the

predicted

symme- try of the unstable modes are in

good

agreement with

experimental

results

presently

available.

Acknowledgments.

- The authors would like to thank E.

Guyon

and P. Pieranski at

Orsay,

N. Boc-

cara, and J. P. Carton at

Saclay

for very

stimulating

discussions.

References

[1] Preceding paper.

[2] GUYON, E., PIERANSKI, P., J. Physique colloq. 36 (1975) C1-203.

[3] CHANDRASEKHAR, S., Hydrodynamic and Hydromagnetic Sta- bility (Clarendon Press, Oxford) 1961.

[4] MANNEVILLE, P., DUBOIS-VIOLETTE, E., J. Physique 37 (1976)

285.

[5] PIERANSKI, P., Thesis Orsay (1976).

[6] For a review see : DE GENNES, P. G., The Physics of Liquid Crystals (Clarendon Press, Oxford) 1974.

[7] GÄHWILLER, Ch., Mol. Cryst. Liq. Cryst. 10 (1973) 301.

[8] PIERANSKI, P., GUYON, E., Phys. Rev. A 9 (1974) 404.

[9] PIERANSKI, P., GUYON, E., Phys. Lett. 49A (1974) 237.

[10] Handbook of Mathematical Functions, Abramowitz M., Stegun I. A., ed. (National Bureau of Standards) 1972.

Références

Documents relatifs

Calculations have been made for the case in which the transition is reduced to a break of symmetry characterized by the appearance of a roll set in the direction

hydrodynamic flame model based on the equations for purely irrotational flow of the

2014 The nature of the flow alignment of nematic liquid crystals is discussed within the hydrodynamic theory of Leslie, Ericksen and Parodi.. It is shown that there is a

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

Abstract 2014 We present a microscopic theory for the flow alignment in biaxial nematics and nematic discotics and we propose three simple geometrical configurations to

- The Poiseuille cell was formed by two parallel lucite plates polished to provide the molecular alignment (Fig. Parallel metallic wires of uniform diameter d = 200

The cross section of a pair of adjacent rolls will be distorted into the form of squinting eyes (Fig. Hence the term. The periodicity of the domain pattern, with wave-

In the stationary case the magnetic field does not affect the rate of dissipation of kinetic and elastic energy but has a strong influence on the heat focusing