• Aucun résultat trouvé

Facet destabilization and macrostep dynamics at the smectic-A smectic-B interface

N/A
N/A
Protected

Academic year: 2021

Partager "Facet destabilization and macrostep dynamics at the smectic-A smectic-B interface"

Copied!
22
0
0

Texte intégral

(1)

HAL Id: jpa-00247523

https://hal.archives-ouvertes.fr/jpa-00247523

Submitted on 1 Jan 1991

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Facet destabilization and macrostep dynamics at the smectic-A smectic-B interface

F. Melo, P. Oswald

To cite this version:

F. Melo, P. Oswald. Facet destabilization and macrostep dynamics at the smectic-A smectic-B in-

terface. Journal de Physique II, EDP Sciences, 1991, 1 (3), pp.353-373. �10.1051/jp2:1991173�. �jpa-

00247523�

(2)

J Phys II1

(1991)

353-373 MARS 1991, PAGE 353

Classification

Physics

Abstracts 82 00

Facet destabilization and macrostep dynamics at the smectic-A smectic-B interface

F. Melo and P Oswald

Ecole Normale

Supbrieure

de

Lyon,

Laboratoire de

physique,

46Allbe d'Itahe, 69364

Lyon

Cedex 07, France

(Received

26

September

1990,

accepted

tn

final form

13 December

1990)

Rksumk. Nous dbcnvons le comportement en croissance directionnelle d'un interface facettd smectique

A-smectique

B lUe matdnau choisi est le 408

(butyloxybenzihddne octylaniline)

Au- dessus du seuil de ddstabilisation qui est bien donna par le critdre dassique du « constitutional

supercooling

», des macromarches se

ddveloppent.

Plusieurs mdcamsmes sont

possibles

suivant la

vltesse de tlrage Ces macromarches sont toujours mstables et se propagent le

long

de l'mterface

macroscopique en

changeant

de forme contlnuellement Nous n'avons jamals observd de solution stationnaire Nous avons

bgalement

mesurb [es effets cindtlques sur la facette que nous attnbuons I un mdcamsme de croissance par dislocation vis Enfin nous ddcnvons le

rdgime

de nuddatlon qui se

ddveloppe quand

on augrnente la vitesse de tirage ainsi que

quelques

effets

vlscodlastiques prdsents

dans la

phase

smectique A.

Abstract. We describe the behavior m directional solidification of the faceted smectic-A smectic-B interface of 408

(butyloxybenzihdene octylaniline)

Above the onset of

instability,

which is given by the classical constitutional-supercooling cntenon, we observe the nucleation of macrosteps Several mechanisms are

possible, depending

on the velocity These macrosteps are

always

unstable and propagate

along

the macroscopic interface, changing shape continuously. We never~observed any stationary solution We also measured the kinetic effects on the facet that we attribute to a screw dislocation

growth

mechanism Finally, we descnbe the nucleation regime, which appears when one increases the

pulling velocity,

as well as some vlscoelastic effects that are present in the smectic A

phase

1. Introduction.

Recently, Bowley

et al

[I]

Studied the Mullins-Sekerka

instab1llty

Of a

binary

mixture for directional Solidification in the case where the basic

planar solid-liquid

interface iS a facet

They

Showed the ex1Stence Of Solutions consisting Of

alternating

cold and hot facets connected

by

curved regions. This crenellated solution appears above the standard Mullins-Sekerka threshold of

instab1llty Later,

Carob et al

[2]

showed that this solution is

always

unstable and

suggested

the existence of another branch of solutions

consisting

of

partly faceted, large- amplitude

cells. For this reason,

they

claim that the cellular bifurcation should exhibit an

anomalously large hysteresis

because one must jump over the unstable branch to reach this

hypothetical

stationary solution.

(3)

In order to test these theoretical

predictions,

we have

performed

an

expenment

on the directional

growth

of a smectic-A smectic-B interface. The matenal studied is 408

(butyloxybenzllidene octylaniline).

Previous

experiments

on the

equihbnum shape

of a smectic-B monodomain

[3]

showed that this interface is faceted

parallel

to the smectic

layers.

The facets match

tangentially

with the contiguous

rough regions

whereas faces

perpendicular

to the

layers

are missing. We showed that such faces are unstable with respect "to the formation of a

hill-and-valley

structure

[4] (Herring's instability [5]).

More

recently,

we

showed some

preliminary

results on the directional

growth

of 408 when the

layers

make an

angle

with the

temperature gradient [6].

We concentrated on what

happens

when the macroscopic front

corresponds

to a missing onentAtion and showed that cells with an

angular discontinuity

at the

tip develop

above the

onsit

of

instability.

In this

article,

we limit ourselves to the case where the basic

planar

interface is a facet

(smectic layers perpendicular

to the

temperature gradient)

We describe how a facet destabilizes in directional solidification and compare our observations to the theoretical

calculations mentioned above

[1, 2].

In

particular,

we describe the formation of

macrostep

and their

subsequent propagation along

the interface.

2. Cridcal

pklling velocity.

2,I THEORETICAL PREDICTIONS. It is well known that the

morphological instab1llty

of the front results from the

competition

between the

destabihz~ng

effect of solute diffusion and the stabilization due to the temperature

gradient

and the

capillanty

For an

atomically rough interface,

one can

analyze

front

stability

via linear

stability "theory [7].

This method is

nevertheless not valid when there is a cusp in the Wulff

plot y(9),

say at 9

= 0. This

onentation

corresponds

to a facet in the

equihbnum shape

The

problem

is that the cusp gives

nse to a contnbution

[(dy/d9 lo

~

(dy/d9 lo 3(9

in the surface stiffness y +

y"

and thus in the Gibbs-Thomson

equation

expressing the local

equilibrium

of the front In this

case, the Gibbs-Thomson

equation

must be

integrated

across the facet

yielding

a

global

equihbnum

condition

[8]

Thls

highly

non-local

problem

has not

yet

been

completely

solved.

On the other

hand, Bowley

et al.

[ii

showed there exists a

family

of crenellated solutions when the

pulling velocity

satisfies the condition

(v vc)/vc

»

(do/fc)~'3 (1)

do

is the chemical

capillary length, i~

=

2D/V

the diffusion

length

and V~ the classical constitutional

supercooling

threshold

V~ =

DGk/mC o(k (2)

where k is the

partition coefficient,

m the

slope

of the

hquldus,

G the temperature

gradient

and

Co

the mean concentration of impurity in the

sample.

In our system,

do

m 20

A

and

i~

m 60 ~Lm so that the

inequality (I)

can be rewntten as

(

V

V~)/V~

» 3 x 10~~

(3)

Thus V~ should give to a

good

approximation the cntical

velocity

above which the faceted front destabilizes.

In order to test this theoretical

prediction,

it is first necessary to determine the

phase diagram

and the

partition

coefficient as well as the dilTusion coefficient D

2.2 PHASE

DIAGRAM

DETERMINATION. -Our 408 was

synthetised by

Germam This

molecule exhibits a

plastic

smectlc-B

phase

between 30 °C and 49 9 °C

Thelamples

consist of

(4)

M 3 FACET DESTABILIZATION AT THE Sm A-Sm. B INTERFACE 355

two

parallel glass plates separated by

two

15-~Lm-thick

spacers. To obtain a

planar alignment

of the molecules on the

glass (smectic layers perpendicular

to the

glass)

a

300-A-thick layer

of

poly1mlde

ZLI-2650

(Merck Corp.)

was

deposited

on the inner surfaces Thls

layer

was then

rubbed in a

single

direction in order to onent the smectic

layers perpendicular

to the scratches To measure the

phase diagram

in the presence of a small amount of

impurities,

we

left a

sample

in an oven at 50 °C for several

days.

At this temperature, 40.8

degrades

and it is

possible

to measure at

regular

intervals of time the

liquidus

and the sohdus

temperatures

These temperatures are shown in

figure

I where we have assumed that the

hquldus

is a

straight

line. Under this

assumption,

the solidus is a

straight

line too. The

partition

coefficient k is

approximately

0.56. We also made a

systematic study

of the

degradation

of our

samples.

In

figure 2,

we

reported

the

freezing

range AT

= T~~~

Ts~j

versus time.

Here,

the

sample

is

left in an oven at 70 °C and AT is measured at

regular

intervals of time. Just after

filling

the

sample,

AT

m 0.22 °C. The

degradation

is

rapid dunng

the first

hours,

then it slows down and AT reaches a

l1mlting

value close to 0 6 °C. The ongln of the 1nltial

degradation

is not clear.

one could think that it is due to the surface treatment and to the dilTusion of

poly1mlde

into the

sample,

but

equivalent

effects were observed without surface treatment Another

possibility

is that 408

decomposes

itself in the presence of adsorbed water, an usual

phenomenon

with schiff's bases. As we shall see

below,

the

degradation

of the

samples

is a

severe

expenmental

limitation which must be taken into account in the discussion of the

results

51

50

@

O ~~

@ 48

~

% ~

47

@

~ 46

45

20 40 60 80

Concentration (arbit(afy units)

Fig I Experimental

phase diagram

2.3 MEASUREMENT oF THE DIFFUSION

COEFFICIENi.

The ef§ect of

temperature gradient

on

non-steady

1nltial transient solute redistribution

during

solidification of the smectic-B

phase

can be used to determine the diffusion coefficient of the solute in the

liquid (here,

the

smectic-A

phase).

Smith et al

[9]

give the

following

formula C

=

(Co/2 k)

I + erf

(1/2) ( V~

t

/D)'/~]

+

(2

k I exp

[-

k

(

I k

V~

t

/D]

x

x erfc

[(1/2) (2

k I

(V~.t/D)~/~]) (4)

(5)

o

- A

~ , A

~06

~

~ A

@ a

Cl

(04

'

~

Cl A

fi

#

A

@ 0 2

~

o

0 5 10 5 2 0

Time t

(hour)

Fig

2

Freezing

range AT versus time. The

degradation

is rapid

dunng

the first hours that follow the

filling

of the

sample

and next slows down

for the solute concentration at the interface in the smectic-A

phase

at time t. This equation

gives, according

to the

phase diagram,

the front temperature

T=m(C-Co)+To (5)

where

To

is the front

temperature

at time t = 0.

Putting

x

=

Vt,

the distance which is covered

by

the

sample

at time t, we

get

T

To

=

(mco/2 k) (1

2 k + erf

[(1/2) ( Vx/D)~'~]

+

(2

k I

)

exp

[- k(

I k

Vx/D]

x

x erfc

[(k 1/2) (Vx/D)~/~]) (6)

In order to measure the front temperature T

accurately,

we have

supenmposed

a

planar sample

of 40.8 and a

homeotropic sample

of the mixture 9CB-10CB

(75

fb and 25 fb in

weight respectively) (Fig. 3).

This mixture of classical

cyanobiphenyl liquid crystals

has a nematic

smectic-A

phase

transition at

TN~~_s~Am5o°C

very close to the smectic-B smectic-A transition

temperature

of 40 8

Furthermore,

this nematic smectic-A

phase

transition is very

weakly

first

order,

so that the

corresponding freezing

range is very small

AT~~~_s~A

m 5 mK.

Since

cyanobiphenyls

are very stable

chemically

too, the nematic smectic-A transition

temperature

can be considered as a reference

temperature

which does not vary with the

pulling velocity (within

a few

mK) Thus,

the front

temperature

is measured

by locating through

the

microscope

the

position

of the two fronts. Because the temperature

gradient

does not

change

at the very small velocities that are used in this expenment, we can calculate from the shift

betwien

the two

fronts,

the real temperature of the SrnA-SmB interface. The main

advantage

of this method is to avoid the errors coming from an accidental shift of the temperature

gradient.

Moreover, we have taken into account the errors due to the vertical

gradient

of

temperature (m

o.5

°C/cm )

that we were able to calibrate

by supenmposing

two

(6)

M 3 FACET DESTABILIZATION AT THE Sm A-Sm B INTERFACE 357

licroscope

hot oven cold oven

Nemotic Smectic A

~ V

Smectic A Smectic B

Fig

3. Double

sample

used for the measurement of the front temperature

identical

samples

of the mixture

9CB-JOCB Finally,

this method allows to measure

temperature

vanations as small as o,ol

°C,

much more

easily

than with a

thermocouple.

In

figure

4 we

reported

T

To

versus x, the distance that is covered

by

the

sample

at time t The

expenment

was

performed

with a

sample

saturated in

impurity

in order to minimize the

parasitic

effects due to the

degradation.

For this

sample

AT

m 0 56

°C,

G

= 56.2

°C/cm

and

Vm02~Lm/s

of the order of

0.2V~.

The fit with equation

(6)

is

good

and

gives

o

-oos

-o i o

o

(

~ _o

-O 25

-O 30

.

200 400 600 800 1000

x

(~m)

Fig

4 Front temperature versus the distance which is covered

by

the sample (transient regime)

Ji

is the initial front temperature The dots are

expenmental

and the solid line is the best fit with

equation

(6)

(7)

D m 5 x 10~?

cm~/s.

Note that D is the

only adjustable parameter

since k and

mco

are known from the

phase diagram.

Note also that D is the diffusion coefficient

along

the director that we called Djj m reference

[6].

Despite

the

good

fit shown in

figure 4,

this method is not very sensitive and

only

gives a

rough

estimation of the diffusion coefficient. We have indeed found a

fairly large dispersion

in

the values of D obtained in this way,

ranging

from 3 x 10~ ?

cm~/s

up to 7 x 10~ ?

cm~/s.

One of the difficulties lies in the control of the initial

conditions,

which differ from one

sample

to

another

Indeed,

the calculation assumes C=

Co

m the smectlc A

phase

at time

t =

0,

a condition that is

certainly

not fulfilled

expenmentally.

Another

difficulty

is,

despite

our

precautions,

the

inhomogeneous degradation

of 408

dunng

the

expenment (which

can be

more important near the hot

side).

This leads to a further drift of the front m the

temperature gradient

On the other

hand,

there is

always

in

practice

a thermal transient which is not taken

into account

by

the

theory

and which will be all the more

important

that the

velocity

is

large.

Furthermore,

it is well known that the solution given

by

Smith et al. is not exact

(and

even

wrong near t =

0)

because of the

assumption

that the

velocity

of the interface is

equal

to the

pulling velocity Recently, Huang

et al.

[10]

have discussed this effect and have

improved

the

~calculation of Smith et al.

They

show that this

hypothesis

leads one to overestimate the diffusion coefficient and is

only

valid when M

=

mco(I k) V/GD

=

V/V~

« I. In our

expenment, M

m 0.2 so that the error which is made

(overestimation) by

using equation

(6)

is

of the order of 20 fb

(from Fig.

3 of Ref.

[10]).

This calculation also assumes that there is no dilTusion in the solid. This is not

ngorously

true in our

system

since the ratio of the diffusion

coefficients in the two

phases p

=

Ds~~/DS~A

m 0.2. This

phenomenon

leads us again to overestimate the diffusion coefficient. It is

possible

to show

that,

within the

approximation

of Smith et

al.,

the characteristic relaxation time is modified

by

a factor I +

k~ p

m 1,05 so that

the diffusion coefficient it overestimated of about 5 fb.

Taking

into account these two

effects,

we shall take in the

following

D = 4 ± 2 x 10 ?

cm~ Is.

Another way to estimate the diffusion constants

Di

and Djj is to measure the cntical

velocity

and the dnft

velocity

of the cells when the

layers

make an

angle

with the

temperature gradient.

In this case, the interface is

rough

and the classical

constitutional-undercoollng

cntenon

(which

we have

generalized

to the

anisotropic

case, in Ref.

[2])

holds. This method gives

Di

= 6 x

10~? cm~/s

and Djj = 3 5 x

10~? cm~/s.

2.4 CRITICAL VELOCITY. In Order to measure the Onset Of

instability,

We increase Very

slowly

the

pulling Velocity by

increments Of 0. I ~Lm/S. At each

increment,

We Wait for a time Of the order of

D/kV~

which is the

minimum time that is necessary to reach a

quasi-stationary

regime In this way, we obtain a first cntical

Velocity

that we call

VI Then,

we decrease

slowly

the

Velocity

till the front restabilizes. This occurs below a cntical

Velocity Vi.

The

hysteresis VI Vi

is charactenstic of a subcntical bifurcation.

For a temperature

gradient

G

= 50

K/cm

and a

sample

thickness of 15 ~Lm, the threshold

on increasing

speed

is

VI

= 0.55

~Lm/s

whereas it is on

decreasing speed Vj

=

0.4

~Lm/s.

The shift

VI Vi

is

small,

which means that the bifurcation is

only weakly subcntical,

in contrast to what was

expected

from the

theory.

The

large

Value of the

partition

coefficient is

perhaps

responsible

for this small

hysteresis

It is indeed well known that for a

rough

interface the bifurcation becomes

supercritical

as soon as the

partition

coefficient is

large enough, typically greater

than 0.45

[11].

Nevertheless it is not at all clear that this cntenon remains valid for a facet.

In order to test more

quantitatively

the theoretical

predictions concerning

the onset of

instability,

we measured

Vf

for several

samples

of different

composition

and for various

(8)

M 3 FACET DESTABILIZATION AT THE Sm A-Sm. B INTERFACE 359

temperature gradients.

For each

sample,

we measured

carefully

the

freezing

range

AT

=mco(I/k-I).

In

figure 5,

we

plotted VI/G

versus

I/AT

and found a linear

dependence.

The

slope

is

proportional

to a diffusion coefficient and is close to 4, I x 10~ ?

cm~/s.

This value is

equal,

within our

experimental

error

30

fb),

to the diffusion

coefficient Djj given

previously

Thls means that the classical constitutional

undercoohng

threshold gives a

good

estimation of the

stability limit,

even in the faceted case

~_

a

~U~ a

o

~u~

~

~

~

50

~

>

0 2 3 4

1/AT(°K'

Fig

5

VI /G

against

lib

T.

VI

is the critical velocity obtained

by

increasing very slowly the pulling velocity from zero, G

(he

temperature

gradient

and AT the

freezing

range The solid line is the best fit with a linear law Its

slope equals

41 ~Lm~/s.

3.

Description

of some macrostep nucleation mechanisms and measurement of the front

temperature.

Above the cntical threshold

VI,

the front is

composed

from a succession of hot and cold facets

separated by

macrosteps. These macrosteps

always

drift

along

the interface : we have

never observed

penodic

crenellated solutions. This observation is in fact

compatible

with the

stability analysis

of Carob et al.

[2]

who found that crenellated fronts are

always

unstable.

On the other

hand,

the model is unable to

predict

how

macrosteps

occur. Several mechanisms have been

observed,

which we now descnbe.

The first one, which was mentioned in reference

[6], corresponds

to the

heterogeneous

nucleation of a small

bump

on the interface

(Fig 6).

This localized

perturbation quickly facets,

giving nse to a pair of

wide, rough

macrosteps

separated by

a new facet which is hotter than the

preceding

one. This mechanism is common at low

velocity

near the onset of

instability.

It can appear below the onset of

instability but,

in this case, the

bump rapidly

shrinks and

disappears.

There are other mechanisms which are

frequently

encountered.

One of them is associated with the

dynamics

of the macrosteps themselves. As we have

already emphasized,

the macrosteps

propagate along

the interface.

Furthermore,

their

shape changes

m time. As a

rule,

an

initially

wide and smooth macrostep becomes first

step-like

and then

develops

a

sharp edgi

and a cusp

_T(e

drift

velocity

then increases

sharply

and one often

sees on the hot

facet,

behind

thq tip,

a new,

wide,

smooth macrostep

(Fig. 7).

It then evolves

as the former and the

mec%hiiin I@ begin igaii.

This mechanism

seems to be related to the

change

of

velocity

which occurs when a

sharp edge develops.

(9)

~

-~--~

.,,~ ,.,~

~?)j~'~@~@@$$~~i~"'

<~<i»jiji>jt.~~ l~.i,liw~ii~i£wG>#i9<

..~ ~

jjmi~ifl >~ill-'~-

~,

== ~'T~~

<.J,

i,j

r

j

~#jl~(,i~~'~~

~

.~."i~[iii)M (1

#'j'i

i.

~ '

~,. ;., .. t

' ~~~

~4~

il,im~f)I)>..1~~

' i

I,'I(I

i~ j

; tit t'2-'m"=~~- .~'-.~~"

Ul,@l'~.,,ijii !~..P@14fi[)")

ii

~,

~

#$~~(i)4fj!j>~)i~,ii

ii~ii.~~~.?ii.

mini*flilliiliii£.i<,i iri-.z i.lint-1~ ...

Fig. 6 - VI

small

bumps

can onthe interface hey acet, eading to

two

At

large velocity (V

m2

VI ),

another mechanism very

frequently

seen is the sudden nucleation of a small pit on a facet

(Fig. 8)

This pit can either

heal,

giving a small smectic~A inclusion which

rapidly disappears,

or it can grow. In the latter case the two

portions

of the facet evolve

separately

at different velocities. This leads

rapidly

to a

sharp~edged macrostep

that

propagates

towards the colder facet.

We sometimes observed that a

pit

can destabilize

by

giving birth to an

adjacent

hill

(Fig. 9) This'process

is

always asymmetncal,

two hills never appeanng

simultaneously

on the two sides of the pit This hill then

quickly

facets

leading

to an

abrupt

macrostep on the side of the

pit

and to a smooth macrostep on the other Then these two macrosteps

begin

to propagate the first one

(on

the side of the

pit) develops rapidly

a

sharp edge

and a cusp while the other is at first smooth.

The nucleation rate of the

pits

fixes the mean size of facets at

large velocity (Fig. 10).

We found that varies

roughly

with the

pulling velocity

as

V~°~ (Fig 11)

We also noted that the interface often « oscillates

locally,

between stable state » with few macrosteps and one that is unstable

(Fig. 12).

These

cycles

are

irregular

which hinders

their

quantitative analysis. They

are

only

visible near the onset of

instability

and have a

penod roughly equal

to the diffusion time

D/V~.

Finally,

we measured the front temperature versus the

pulling velocity,

using the same method as in subsection 2.3. Our data are

reported

in

figure

13. Measurements were made on

a «clean»

sample (AT

=

0,22 °C

), immediately

after

filling

with 40.8 and as

rapidly

as

possible

in order to minimize the

degradation

effects

(we

measured again AT after the expenment, about half an hour

later,

and found the same value to about ± 0 01

°C).

At very small

velocity,

the front is

planar

and its temperature T is

equal

to the sohdus

temperature

(that

we

previously

measured in a

separate oven).

As the

velocity

increases the front

(10)

N 3 FACET DESTABILIZATION AT THE Sm A-Sm B INTERFACE 361

ioopm

Fig

7.

Shape

evolution of an

initially

rounded macrostep and nucleation of a new macrostep when

the first one becomes

sharp-edged

and

changes velocity

G=31K/cm, AT=12°C and

V

= 012 ~Lm/s The time interval between two

photographs

is I mn.

temperature begins

to decrease

slightly,

then the front becomes unstable : we next measure its

mean

temperature

It does make sense because the

macrostep height always

remains small in

comparison with the retreat of the front it

corresponds,

for

example,

to a

temperature

span

(11)

~

fs~j;

,~~j ~~+

~

< & W~

~l'w fl

~~

~~

-- &~

1"2

'i %

,~#

'~

b&§1

~~ , .~~

~~j.q~i:?ji

W~/! /

J

ioo

pm

Fig.

8 At

large velocity,

pits

suddenly

appear on the facets and break them This mechanism leads to new macrosteps when the two parts of the facets evolve later at different velocities G

= 38 K/crn, V

= 2 5~Lm/s The time interval between two successive

photographs

is 1s

of the order of 0.015 °C whereas the retreat of the front is rather 0.06 °C at

i'

= 10

~m/s.

On the other

hand,

it is

possible

to show

theoretically

that the mean position of the unstable front must coincide with that of the

planar front,

if there is no diffusion in the solid

[12].

For this reason, we believe that the

cooling

down of the mean front is the same as the one that would have the

planar

front and

consequently

is due to kinetic effects on the

facets,

which we shall discuss in section 5.

In the

impure samples,

it was not

possible

to measure the kinetics. This is due to the small range of usable

pulling velocities, mainly

because of the

parasitic

nucleation of smectic~B germs ahead of the interface. Since the mean

temperature

of the unstable front is close to that of the

planar stationary

front

(the macrostep height

is much smaller than the diffusion

length too),

it is

possible

to make some theoretical

predictions

on the nucleation rate ahead of the

interface and to test them

expenmentally (see Appendix I).

Let us now describe in more detail the

dynamics

of the

macrosteps

and their interactions,

(12)

N 3 FACET DESTABILIZATION AT THE Sm A~Sm B INTERFACE 363

fi

Fig 9

Asymmetncal

destabilization of a small pit appeanng on a facet G = 32

K/s,

V

= 0 8 ~Lm/s and AT

= 0 2 °C The time interval between two successive

photographs

is 7s

(13)

'g,-q

)m)<~

'"j~

,

~~~~~~~~~'~~'~~

'

~

,i;vi~i=iii:.ill.

~

~" ~ ~@Wlfliiilflt

~.

~

»

.

ioopm

Fig 10 Front aspect at increasing

velocity (indicated

in ~Lm/s on each

photograph)

G

=

38 K/cm and AT

= 0 2 °C

30

ji

~i

~ 20

~j

w

i~

~

o

2 4 6 8

Velocity (~m/s)

Fig 11 Facet size versus the

pulling velocity

(14)

N 3 FACET DESTABILIZATION AT THE Sm A~Sm. B INTERFACE 365

sopm

Fig.

12 Sequence of pictures

showing

the cyclic behavior of the interface near the critical velocity.

G

= 31 K/cm, AT

= 0.2 °C and V

=

0 8 ~Lm/s 0) t = 0, 1) t = 85 s, 2) t = 230 s

on

80

11

©~ 60

~o

~i 4

E40

20

o

° ~

~veloci(y

V

(~n~/s)

~~

Fig

13 Front

undercoohng

8T~~ = Ts~j T versus

pulling velocity

V The solid line is the best fit

wtth a

@

law

4.

Macrostep dynamics.

We measured the

velocity

of the macrosteps at a fixed temperature

gradient

and found that it

strongly depends

upon their

shape,

the

sharp-edged

macrosteps

propagating

faster than the rounded ones. We found that the

velocity

of a rounded macrostep

is,

within our

expenmental

error, close to 0 3 v

irrespective

of their width On the contrary, the

velocity

of a

sharp-edged

(15)

macrostep lies between V and 2 V

depending

on the tip

sharpness.

The

sharper

the

tip,

the

larger

is its

velocity

In

figure14

we show the dnft

velocity

Vd~i~ of the

sharp~edged macrosteps

as a function of the

pulling velocity

V We see that Vd~i~ is

proportional

to V in spite of a

large dispersion.

The same result holds for rounded

macrosteps.

Another

interesting question

is how the macrosteps interact. We have thus observed the

dynamics

of an isolated pair of

macrosteps

of

opposite

signs,

approaching

each other.

« Isolated means that the other macrosteps are far away, at a distance which is greater than

twice the diffusion

length

Such situation occurs

only

near the onset of

instability.

Expenmentally,

either the two

macrosteps

are rounded and

they

remain

unchanged

till

they

join

together,

or

they

are

pointed

with a cusp where the

impurity

molecules collect.

5

(Sharp~edge macrosteps(

4

o

3

>

~ l

~

* ~

~ ~ . ~ #

i

~ *

$ ~

$

* *

o

~

o o

o

f O

I

o o

o

0 2 4 6 8 10

V(~m/S)

Fig

14 Dnft

velocity

of

sharp~edged

macrosteps versus the puffing

velocity.

The second case is

by

far the more

frequent. Figure

15 shows two

sharp~edged

macrosteps

approaching

each other.

Here,

the two

macrosteps

are

slightly asymmetrical.

For macrosteps

having exactly

the same

height,

the distance d separating them

(Fig. 16a)

decreases

linearly

versus time wuich means that the two

macrosteps

move at constant

velocity. Figure

16b shows

the position h+

(resp h~)

of the hot

(resp. cold)

facet versus time h+ is constant, which

means that the hot facet is quasi immobue in the

temperature gradient,

while h~ decreases so

that the other facet recedes and

consequently

cools down This is the consequence of a

global

increase of the

impurity

concentration at the cold facet, As a

general rule,

the colder a

facet,

the faster it recedes in the temperature

gradient

and cools

The

situation

is s1mllar when two rounded macrosteps meet before

deve16ping

a

sharp edge, except

for one point, in tills case, the cold facet retreats faster than

before, mainly

because there is no cusps into wuich

impurities

can escape

Finally,

we

emphasize that,

in both cases, the dnft

velocity

starts to decrease when the two

macrosteps

are very close each

other, namely

a few ~m, i e at a distance much smaller than

(16)

N 3 FACET DESTABILIZATION AT THE Sm A-Sm. B INTERFACE 367

loopm

Fig

15. Two

sharp~edged

macrosteps

approaching

each other

They

leave a rounded

macrjstep

after

partial

annihilation

the diffusion

length.

In

general,

the two macrosteps anneal in the

end,

either

completely

if

they

are of the same

height

or,

partially, by leaving

a new rounded macrostep, if

they

are not

exactly

of the same

height (see Fig. 15)

(17)

0 loo 200 300

t(s)

~~

° h+

. h~

~ a o o ~ o o o a o o Da a ~ ~ ~ o a

E

/5

~

~~

jf

o o.

°

o ° o o

o ~

~~

o o.

o o

o o

5

0 loo 200 300

t(s)

Fig 16 a) Distance between two

sharp~edged

macrosteps propagating m opposite directions versus time b) Position h+ (resp. h~ of the hot (resp

cold)

facet m the temperature

gradient

versus time (the ongln has been chosen

arbitranly)

5. Discussion.

One of the main results of tuis

study

is that there is no

stationary

regime, even near the onset of

instability.

The facet advances

by

generating macrosteps which propagate

along

the

macroscopic interface. The

leading edge

of these macrosteps is

rough just

after formation. It

is

always unstable,

the bottom reentrant corner

advancing

more

slowly

than the

top

corner

and results in an

overhang

with a

sharp tip

and an

impurity~nch

cusp.

Let us mention that similar

behavior, namely

loss of

stability

of

macrosteps

and formation of a flat inclusion under the

overhang,

is sometimes observed

during growth

from solution of

faceted materials

(sucrose

for

example [13]).

(18)

N 3 FACET DESTABILIZATION AT THE Sm A-Sm B INTERFACE 369

So

far,

there is no unified

theory describing

the behavior of the

macrosteps

in directional solidification To understand

qualitatively

their

dynamics,

we focus on opposing

physical

mechanisms

[14]

Let us first assume that an isolated macrostep has been formed on the

growing

interface and that G

= 0. We

neglect capillarity

and attachment kinetics. The

equiconcentration

lines of the solute may be

represented

as in

figure17a. Here,

the surface is an

equlconcentration

line.

Over the top comer of the macrostep, the

equlconcentration

lines

shrink, enhancing

the

growth

rate

there,

while in the vicinity of the bottom comer, the lines move aside so that the

growth

rate decreases This is the consequence of the conservation law of the solute.

V~

AC

=

D aC

Ian (7)

Thls diffusion

mechanism,

which is

responsible

for the dnft of the macrostep, its

amplitude

increase, and its

shape instability

is just the well-known Mullins~sekerka

instability adapted

to the geometry of a

macrostep.

On the other

hand,

the

temperature gradient

and the surface free energy both oppose the

growth

of the

macrostep.

In this case, the

equiconcentration

lines cut the solid as shown

schematically

in

figure

17b. The main effect is to reduce the

growth

rate of the top corner and

conversely

to increase that of the bottom corner

Nevertheless,

the

expenment

shows that these two effects cannot suppress the

destabilizing

effect of the diffusion field

Concernlng

the kinetic

effects,

we first

emphasize

that

they

are much smaller than in usual

plastic crystals

such as salol

[15].

On the other

hand,

the

macrostep

behavior is

qualitatively

the same in all the

samples

near the onset of

instability

and does not

change significantly

with

the

velocity.

As the kinetics are

naturally strongly dependent

on the

velocity,

we conclude that

they

do not

play

a dominant role in the drift and the destabilization of the macrosteps

By

d)

b)

Fig 17

-

a) Isoconcentratlon lines

near a rounded macrostep

dashed lines) hen G = 0 and

capillanty is egligible, the interface

itself is

(19)

contrast,

they

are

perhaps

more

important

in the

high~velocity

regime which is observed in the

purest samples, favonng

hot over cold facets as was shown in reference

II

Finally,

we discuss the microscopic

growth

mechanism that is

responsible

for the observed kinetics. Two mechanisms are a priori

possible.

the first one is set

by

the

nucleaiion

and the

growth

of two~dimensional nuclei on the facet. Th1s is a slow process which controls the facet

growth only

if the

crystal

is

nearly perfect.

This is not,

by far,

the case in our

expenment

;

the second one is a screw dislocation

growth

mechanism It is much more

likely

for

many reasons

first,

we know that screw dislocations are numerous in our

samples

because

there

always

exists a small

angular

mismatch 0 between the scratches of the surface treatment on the two

glass plates

Thls mismatch induces a twist deformation of the

layers

which can be

relaxed

by introducing elementary

screw dislocations with a

density A~'

=

o/b

where b is the

layer

thickness and A the mean distance between two dislocations

Expenmentally,

0

=10~~

rd and b

=

3

x10~~

cm so that A~ '

= 3

x10~~ cm~'.

In other

words,

there is a

screw dislocation every 0.3 ~m wuich means that even the smallest facet is

pierced by

several

dislocations.

Second,

there

ale

theoretical

predictions according

to which the

resulting

kinetic

undercooling 8T~,~

is

proportional

to

vi [16].

This behavior

is observed

expenmentally,

the best fit with the data of

figure

13 giving 8

T~,~(°C)

= 0.023

@ (~m ). Finally,

it

is

possible

to evaluate the minimum

undercooling 8Tz~

that is necessary in order that a

step joining

two

screw dislocations of

opposite

signs works as a Frank Read source. A

simple

calculation

gives

8Tz~=2flT~/Lbd

where d is the distance between the two

opposite dislocations,

fl

the free energy of the step

(fl =3x10~~erg/cm[3]),

L the latent heat

(L

= 5 x

10~ erg/cm~)

and T~ the transition temperature

(T~

= 323

K). Expenmentally,

the shortest distance d is of the order of 0.3 ~m from which we calculate

8Tz~

= 0 05 °C. Of course, this is an overestimation of the

undercoolmg

that is necessary to make grow a facet because there is a

large

distribution of distances d and in

particular

some values of d much

larger

than 0.3 ~m. In

conclusion,

a very small

undercoolmg

is sufficient to make

grow a facet via a screw dislocation

mechanism,

m agreement with our measurements.

6.

Concluding

remarks.

So

far,

we have treated the smectic~A

liquid crystal

as an

ordinary isotropic liquid

In

fact,

this

is incorrect for at least two reasons :

first,

diffusion m the sniectic~A is amsotropic We showed in reference

[2]

that this effect is

responsible

for the dnft of the cellular

pattern

when the

layers

make an

angle

with the

temperature gradient

We tuink that this

amsotropy (D~ /Djj

=

7

)

must accelerate the dnft of

the"macrosteps

because

impurity

diffuses faster inside the

layers

than

perpendicular

to

them

second,

the smectic~A is an

organized phase

Its lamellar structure inhibits the solute~

dnven convection

present

in many expenments Such convection

leads

to

an increase of the

critical

velocity

with the thickness of the

sample (see,

for

instance,

the case of the nematic-

isotropic

interface

II 7])

We checked that in our system the cntical

velocity

is

independent

of the thickness

(which-we

varied from 10 to 50

~m).

On the other

hand,

there are elastic effects

which are due both to the

change

of

density

and to the difference of

layer

spacing between the two smectic

phases [18].

In

particular,

the smectic A

phase

can sustain a stress normal to the

layers. However,

one can

easily

show

(see Appendix II)

that in the situations encounted in this

article,

elastic effects may be

neglected

In

conclusion,

we have shown that the faceted Sm.A~Sm.B interface is unstable above a critical

velocity,

which is given

by

the classical constitutional

undercoohng

criterion We have

Références

Documents relatifs

Ne croyez pas que le simple fait d'afficher un livre sur Google Recherche de Livres signifie que celui-ci peut etre utilise de quelque facon que ce soit dans le monde entier..

The schemes are con- structed by combing a discontinuous Galerkin approximation to the Vlasov equation together with a mixed finite element method for the Poisson problem.. We

Our general strategy will be to reduce by a specialization argument (See Subsection 5.2) to the case where the base field k is finite in order to exploit the fact that, over

A Hardware Platform is a family of architectures that satisfy A Hardware Platform is a family of architectures that satisfy a set of architectural constraints imposed to allow the re-

After some preliminary material on closed G 2 structures in §2 and deriving the essential evolution equations along the flow in §3, we prove our first main result in §4:

We know that it begins when the felted beech coccus or beech scale, Cryptococcus fagisuga Lindinger (C. fagi Bderensprung) infests the living bark of Fagus

La plénitude dans le ciel, ou cercle de Gwenved ; et sans ces trois nécessités, nul ne peut être excepté Dieu. îS Trois sortes de nécessités dans

Maintaining himself loyal to Hegel, he accepts the rel- evance of civil society, but always subordinate to the State, the true “we”, as the interests of the civil society and the