• Aucun résultat trouvé

Capillary instabilities in deep cells during directional solidification

N/A
N/A
Protected

Academic year: 2021

Partager "Capillary instabilities in deep cells during directional solidification"

Copied!
9
0
0

Texte intégral

(1)

HAL Id: jpa-00211119

https://hal.archives-ouvertes.fr/jpa-00211119

Submitted on 1 Jan 1989

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.

Capillary instabilities in deep cells during directional

solidification

K. Brattkus

To cite this version:

(2)

Capillary

instabilities

in

deep

cells

during

directional

solidification

K. Brattkus

Groupe

de

Physique

des Solides de l’E.N.S.

(*),

Université Paris VII, 2

place

Jussieu, 75231 Paris Cedex 05, France

(Reçu

le 2 mars 1989,

accepté

le 8

juin

1989)

Résumé. 2014 Nous

discutons, à la lumière

d’expériences

récentes, la stabilité des sillons cellulaires

profonds

au cours de la solidification directionnelle. Une

analyse

linéaire montre que les sillons

tridimensionnels

axisymétriques

étroits

présentent toujours

une instabilité

d’origine

capillaire,

qui

peut

expliquer

l’émission par les sillons de gouttes

liquides

observée dans les

expériences.

La distance mesurée entre

gouttelettes

successives est de l’ordre de la

longueur

d’onde la

plus

dangereuse.

Abstract. 2014 Motivated

by

recent

experiments,

we discuss a

capillary instability

that arises in

funnel-shaped, deep-cell

roots

during

directional solidification. A linear

stability analysis

of thin,

axisymmetric

roots reveals that due to surface energy, the roots are

always

unstable. The eventual

outcome of this

instability

in

experiments

is a

break-up

of the root into

droplets.

The distance between successive

droplet

centers is in reasonable agreement with the

wavelength

of the fastest

growing

mode. Classification

Physics

Abstracts

61.50C - 81.10F - 81.30F

1. Introduction.

The recent

experiments

of de

Cheveigné et

al.

[1]

have

highlighted

an

interesting

feature of the cellular

growth

regime

in directional solidification.

Here,

thin

samples

of

impure

CBr4

are solidified

by

pulling

them at a constant

velocity

through

a fixed

temperature

gradient

and the

morphology

of the

solid-liquid

interface is

optically

observed. The

interface,

which is

planar

when

pulling speeds

are

small,

develops

a

nearly periodic array of cells when

pulling

speeds

exceed a critical value. If

speeds

are

increased

further,

the cells grow

deep

and

develop

long,

thin roots which

get

so thin that it is not

possible

to

optically

define the exact

point

at

which the root ends. Before

optical

definition is lost

however,

it is observed that some of the thin roots

develop

into a three dimensional « funnel » and

begin

to form small

droplets

of

liquid

which detach from the root

leaving

a trail of

liquid

bubbles behind.

They postulate

that these bubbles are the result of a

capillary pinch-off

driven

by

the surface energy of the

solid-liquid

interface.

(*)

Laboratoire associé au C.N.R.S.

(3)

3000

The transition from

planar

to cellular fronts is

theoretically

well understood

[2, 3]

for directional solidification and the nonlinear

development

of shallow cells has been followed

numerically

in both two and three dimensions

[4-6].

Deep

cells have also been

computed

numerically

in two-dimensions

[7-9]

and the solutions all show smooth

tips

connected to a

deep

root that ends in a

pendant-drop

closure. The closure at the root’s end is small and even

dynamic

numerical codes

[9]

have not been able to observe a

pinch-off

in two dimensions.

A more

analytic approach

in the

description

of

deep

cells is to take

advantage

of the root’s narrowness to

develop

an

asymptotic theory

in the root which may then be matched to a

tip

region.

This

procedure

is feasible

only

if solutions in the

tip region

are tractable

[10-11]

and does not

immediately

account for the

pendant-drop

observed in the numerical studies. There is excellent

agreement

however between the

asymptotic

solutions obtained in the root and those found

numerically.

In this paper we

apply

the

asymptotic approach

to

deep

cells with

axisymmetric

roots. Since the three-dimensional structure of the

tip region

is

unknown,

we do not carry out the

appropriate matching procedure.

A local

stability analysis

of the

asymptotic

solution for

axisymmetric

roots

(funnels)

reveals that

they

are

always

unstable in the presence of

capillarity.

A

relationship

between this

capillary instability

and the observed

break-up

of funnels is

suggested by

the measured values on the

frequency

of

trailing droplets.

The

capillary instability

is three dimensional and is

only

present

for columns of

freezing liquid.

When our

stability analysis

is

repeated

for two-dimensional roots, the

instability

is not

present.

2. An

axisymmetric

model.

As a model of

deep

roots which

develop during

cellular

growth

in directional

solidification,

we consider an

axisymmetric

tube of

liquid freezing

with

velocity

V

parallel

to an

imposed

temperature

gradient

as shown in

figure

1. The solute C is

rejected

at the

solid-liquid

interface,

R * (z ),

where the ratio of solute in each

phase

is

given by

a

segregation

coefficient,

Fig.

1. - The

axisymmetric

funnel

(located

between the

regions

of the cell

tip

and the root’s

end)

of a

deep

cell

solidifying

with

velocity

V. The local normal to the interface R* is n and its normal

velocity

is

(4)

The

parameterization

of the interface is natural for the root

region

but may not

give

a uniform

description

of the entire cell. We consider a one-sided model that

neglects

solid diffusion and assume the

temperature

field is linear

[12]

so that

DS, L

are the diffusivities of solute in each

phase, To

is a reference

temperature

and G is the

imposed

temperature

gradient.

In the

moving

frame,

the diffusion of solute

through

the

liquid

channel is

governed by

and is

subject

to solute

conservation,

and thermal

equilibrium,

on the interface r = R *

(z).

Here

Vn

is the

velocity

normal to the interface and

equation

(4c)

is the Gibbs-Thomson condition for

dilute,

isotropic

mixtures ;

TM

is the

melting

temperature

of a pure

solid,

m is the

liquidus slope,

y is

the

solid-liquid

surface energy, Lv is the latent heat per unit volume and K is the curvature of the interface.

A dimensionless

description

of the

problem may be derived by scaling

as

follows,

where À is an

appropriate

scale for the root thickness. The scaled version of the diffusion

equation

becomes

(subscripts

denote

differentiation),

with,

on r =

R (z).

The additional dimensionless

parameters

are a scaled channel

thickness,

(5)

3002

For convenience we have chosen as the reference

temperature,

3.

Steady-state

growth

for a thin

axisymmetric

root.

Deep-cell

roots are

characteristically

thin in

comparison

to both the

wavelength

of the cell and the diffusion

length

scale

DL/V.

Therefore,

we

employ a

lubrication

approximation

in

system

(6)

with,

and

expand

both the interface and solute concentration in a power series of A.

The behaviour of the

asymptotic expansion depends

on the relative size of the surface energy, r. The Gibbs-Thomson condition

(6c)

reveals that when T is

large,

the

leading-order

solute concentration for small A is determined

by

a balance with surface energy. If T is

small,

it is determined

by

a balance with

temperature.

We consider the intermediate case which

captures

both

dependencies and relate the surface energy

to A as

follows,

where

F,

is an order one

parameter

as A tends to zero. For the

typical

directional solidification of

deep

cells,

T is between

10- 3

and

10- 4

while

A,

measured a few

cell-wavelengths

back from the

tip,

is

approximately 10- 2.

Under the

relationship

(11)

we can

expand

the

steady

interface and the solute concentration as

and these solutions will remain valid while variations

along

the

length

of the root are slower than radial variations.

When the

expansions

(12)

are inserted into

(6)

we find at

leading

order in A

that,

in 0

r R o (z )

with

on r =

Ro (z ).

The solution for uo is

independent

of r and becomes

(6)

in 0 r

Ro (z )

with,

on r =

Ro (z ).

The existence of a bounded solution to

(15a, b) requires

the

leading

order root

shape, Ro,

to

satisfy

the

following compatibility

condition,

The

boundary

conditions which

complete

this differential

equation

are not

specified

in an infinite root and must come from

matching

to a

region

where the lubrication

theory

no

longer

applies,

i.e.,

a cell

tip

or the root’s end. This

matching procedure

has been

fully developed

in the two-dimensional case

by

Dombre and Hakim

[10]

when k is near one and

by

Weeks and van Saarloos

[11]

when k is zero. In three dimensions the same

procedure

is

impractical ;

we will not be concerned with the exact

shape

of the

axisymmetric

root because it is not critical to

the

stability analysis

of the next section.

When surface energy is

neglected,

equation (16)

has the

following

solution,

This is the three-dimensional

equivalent

of a result

originally

due to Scheil and Hunt

[13]

and since the scaled surface

energies

r s

are

typically

small,

this result should be

applicable

throughout

most of the root. It is not valid in

general

when 1 z

1

is too

large

or too small and

must be modified near the cell

tip

or the root end. The

singularity

(1)

in

(17)

at

z = 0 occurs when the

temperature

is

given by

whereas the

temperature

of a

planar

front

(the

precursor state to

cells)

which

propagates

into a

liquid

with solute concentration

Co is given by

If follows that

Tp

To

when

m(k - 1) V Co/kGDL >

1. This condition is necessary for the linear

instability

of the

planar

front

[2].

When

planar

fronts are unstable and the interface is

cellular,

the

singularity

in the Scheil-Hunt solution will occur at a hotter

temperature

than the

temperature

of the

planar

front

(approximately

the

temperature

of the cell

tips).

Therefore,

the

portion

of the Scheil-Hunt solutions that are

physically

relevant for

deep

roots must have

z 0.

Although

the Scheil

approximation

is valid for any

k,

the

experiments

of de

Cheveigné

et al. are for k less

than

unity

and we will

only

consider this case further.

(7)

3004

4. Linear

stability analysis.

Now we consider a local

stability

analysis

of the

steady, axisymmetric

root

by linearizing

about the solution

(12)

at a fixed

position,

zo

0,

with

axisymmetric

disturbances that vary on radial scales.

Specifically,

we let

where

This

scaling

is chosen so that axial variations of the disturbances balance radial variations and the time scale is

appropriate

for radial diffusion.

Locally,

the

steady

root is

cylindrical

and the disturbance

equations

at

leading

order in A

become,

in 0 r

R o (zo )

with

on r =

Ro (zo ).

A normal-mode

decomposition

of

(21)

with,.

results in the

following

characteristic

equation

for the

growth

rate

li,

where

1n

are modified Bessel functions of order n. To show that the

growth

rates are

always

real,

we

multiply equation

(21a)

by

the

complex

conjugate

of

U weighted by

a factor of r and

integrate

over the domain to find that an

eigenvalue

1Ii must

satisfy

Since it follows from the

boundary

conditions

(21b, c)

that

U * (Ro ) Ur (Ro ) is

the

product

of a real

quantity

and

li,

we are

guaranteed

that the

growth

rate f2 is real.

(8)

and consider the case when Z is small

(typically, r s

is

small),

the characteristic

equation

(23)

reduces to the

following

approximate

form,

The

growth

rates w in

figure

2 are

positive

for 0

a 2

1 and

negative

otherwise. The

steady,

axisymmetric

root is therefore

locally

unstable to all

axisymmetric

disturbances with

wavelengths

À

= 2 03C0 /a

greater

than the circumference of the root

Ro (zo ).

Since the

instability

is absent when the surface energy is

identically

zero

[this

can be seen

formally by

taking

Z to be zero in

equation

(25)],

the

instability

is driven

by capillarity.

Fig.

2. - The scaled

growth

rate

ù) IZ

=

f2 IZR’(zo)

versus the wavenumber a =

aRo (zo )

found in

equation

(26).

5. Discussion.

Capillary

instabilities of the

type

found here for funnel

shaped

roots are well-known in other

contexts. A

dynamic description

of

capillary

instabilities was first

given by Rayleigh

[14]

for

inviscid

liquid

jets.

In a

convincing

set of

experiments Rayleigh

showed that disturbances introduced at the nozzle of a

jet

were

magnified by capillarity

and convected downstream until the

jet finally collapsed

into a series of

droplets.

The

spacings

between

droplets

were related to the

frequency

of the disturbances introduced at the nozzle of the

jet.

Surprisingly,

the characteristic

equation (26)

is almost identical to

Rayleigh’s

result for inviscid

liquid jets.

Because the

transport

inside the funnel is diffusive instead of

dynamic

as in the

jet,

the

exponent

of the

growth

rate has

changed ; growth

rates are

squared

for

jets.

Otherwise,

the two results have identical form and the value of the most unstable wavenumber

(scaled

on the local

radius)

is the same as that found for

jets,

ac =

0.69715.

In directional

solidification,

liquid

is convected toward the solid as viewed from the front. When solute disturbances in the

liquid

meet a cellular solidification

front,

they

first affect the cell

tips

and are then convected down the roots at the

pulling speed.

If the root is funnel

shaped,

the

capillary instability

will

operate

and cause it to

eventually pinch

off into

spherical

droplets.

Because the thickness of the root decreases away from the tips, the

wavelength

of the most unstable disturbance decreases and its

growth

rate increases as disturbances are convected downstream.

(9)

3006

find that

they

can excite

droplet

formation when the

frequency

of the modulation is not too

large.

In the

long

roots which appear three

dimensional,

the

break-up

into

droplets

is reminiscent of

liquid jets

[16].

The distance between the centers of successive

droplets

is found

experimentally

to lie within

thirty

percent

of the critical

wavelength

2

03C0 / a c

for a wide range of modulation

frequencies.

The modulation of

growth

over the entire cell

apparently

stimulates a broad band of disturbances which grow due to

capillarity

until the root

finally

collapses

into

droplets.

Although

we have examined

only axisymmetric

roots, the Scheil

approximation

can be extended. The

capillary

instability

will be modified for roots with more

complicated shapes

but it is

expected

to

persist

as

long

as the root remains three dimensional.

By

replacing

the

curvature in

(6c)

by

its two-dimensional

equivalent,

we find that the

instability

is not

present

in two-dimensional roots and is not

expected

to arise in the two-dimensional numerical

simulations.

Finally,

our

stability analysis

of section 4 can

easily

be extended to the case of

non-axisymmetric

disturbances. As in

cylindrical

liquid

jets, non-axisymmetric

modes are

always

stable. The observed

droplets

which result from the

break-up

of funnels are

nearly

spherical.

In

closing

we note that the one-sided model which

neglects

solute diffusion in the solid is

not

physically

relevant at the root end where

back-melting

is

present

in the

pendant-drop.

However,

the

capillary instability predicts

that a

large portion

of the three-dimensional root,

including

the

region

where the one-sided Scheil

approximation

is

valid,

is unstable. The

instability

which leads to

droplet

formation in the roots of

deep

cells does not

require

the existence of a

pendant-drop.

Acknowledgments.

The author would like to

acknowledge

several

interesting

conversations on

deep

cell

growth

with B.

Caroli,

C. Caroli and B. Roulet. Discussions with S. de

Cheveigné,

C.

Guthmann,

G. Faivre and P. Kurowski on the

experimental

results were very helpful. The work was

supported through

an NSF/NATO Postdoctoral

Fellowship

in Science.

References

[1]

DE CHEVEIGNÉ S., GUTHMANN C., FAIVRE G. and KUROWSKI P., J.

Phys.

France 50

(1989)

3007.

[2]

MULLINS W. W. and SEKERKA R. F., J.

Appl. Phys.

35

(1964)

444.

[3]

WOLLKIND D. J. and SEGEL L. A., Philos. Trans. R. Soc. London 268

(1970)

351.

[4]

UNGAR L. H. and BROWN R. A.,

Phys.

Rev. B 29

(1984)

1367.

[5]

MCFADDEN G. B. and CORIELL S. R.,

Physica

D 12

(1984)

253.

[6]

MCFADDEN G. B., CORIELL S. R. and BOISVERT R. F., J.

Cryst.

Growth 84

(1987)

371.

[7]

KESSLER D. A. and LEVINE H.,

Phys.

Rev. A 39

(1989)

3041.

[8]

RAMPRASAD N., BENNETT M. J. and BROWN R. A.,

Phys.

Rev. B 38

(1988)

583.

[9]

SAITO Y., MISBAH

C.,

MÜLLER-KRUMBHAAR H., KUROWSKI P., DE CHEVEIGNÉ S. and GUTHMANN C.,

preprint (1989).

[10]

DOMBRE T. and HAKIM V.,

Phys.

Rev. A 36

(1987)

2811.

[11]

WEEKS J. D. and VAN SAARLOOS W.,

preprint

(1988).

[12]

LANGER J. S., Rev. Mod.

Phys.

52

(1980)

1.

[13]

HUNT J. D., Solidification and

Casting

of Metals, Metals. Soc. London

(1978).

[14]

LORD RAYLEIGH, Scientific papers, i

(Cambridge, England)

1899, pp. 361-71.

[15]

CHANDRASEKHAR S.,

Hydrodynamic

and

Hydromagnetic Stability

(Dover)

1981.

Références

Documents relatifs

We show that the corresponding shape of the phase diagram induces a strong reduction of the scale of the bifur- cation curve which makes its complete exploration

Si on suppose que le domaine géométrique Ω du système (2.1) est connu avec une certaine précision, alors de cette précision prés, le choix d’un seul actionneur ( p = 1) peut

The, the implémentation of a law of non- linear control, high performance adaptive applied to a synchronous generator connected to an infinite network, to improve

ﺺﺨﻠﻤﻟا Ostreopsis Ovata يﺮﮭﺠﻣ ﺐﻠﺤﻃ ، ﺔﯿﻠﺨﻟا ﺪﯿﺣو ﻢﺳ ﺞﺘﻨﯾ palytoxin ﺔﯿﺋاﻮﺘﺳﻻا رﺎﺤﺒﻟا هﺎﯿﻣ ﻲﻓ ةدﺎﻋ ﻢﯿﻘﯾ ،. ﮫﻧا ﻦﻣ ﺔﻋﻮﻤﺠﻣ ﻰﻟا ﻲﻤﺘﻨﯾ طﻮﺴﻟا

It can be shown that a linear stability analysis of the standard equations of directional solidification modified for high pulling velocities exhibit a Hopf bifurcation, I-e-

agrees better with the prediction of Karma and Pelc4. For this value of k, our numerics agree better with the solid curve for small values of pe. This is expected since their

In another type of secondary instability, known as a tilt wave, a zone of inclined cells propagates through normal ones, as has been clearly observed in eutectic solidification

Figure 6 shows two typical recoil curves measured at the same pulling velocity in two different samples of undoped CBr4.. These curves have been corrected for the instrumental