• Aucun résultat trouvé

Solutal convection and morphological instability in directional solidification of binary alloys. - II. Effect of the density difference between the two phases

N/A
N/A
Protected

Academic year: 2021

Partager "Solutal convection and morphological instability in directional solidification of binary alloys. - II. Effect of the density difference between the two phases"

Copied!
10
0
0

Texte intégral

(1)

HAL Id: jpa-00210114

https://hal.archives-ouvertes.fr/jpa-00210114

Submitted on 1 Jan 1985

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.

Solutal convection and morphological instability in directional solidification of binary alloys. - II. Effect of

the density difference between the two phases

B. Caroli, C. Caroli, C. Misbah, B. Roulet

To cite this version:

B. Caroli, C. Caroli, C. Misbah, B. Roulet. Solutal convection and morphological instability in direc-

tional solidification of binary alloys. - II. Effect of the density difference between the two phases. Jour-

nal de Physique, 1985, 46 (10), pp.1657-1665. �10.1051/jphys:0198500460100165700�. �jpa-00210114�

(2)

Solutal convection and morphological instability

in directional solidification of binary alloys.

II. Effect of the density difference between the two phases

B. Caroli (+), C. Caroli, C. Misbah and B. Roulet

Groupe de Physique des Solides de l’Ecole Normale Supérieure (*), Université Paris VII, 2 place Jussieu, 75251 Paris Cedex 05, France

(Reçu le 11 mars 1985, accepté le 10 juin 1985)

Résumé. 2014 Nous étendons l’analyse de perturbation du couplage entre déformation de front et convection solu- tale présentée dans un article récent [3] pour tenir compte de l’effet de l’« advection », c’est-à-dire de l’écoulement induit par la variation de densité qui accompagne la solidification. Nous étudions le déplacement de la bifurcation à partir du régime stationnaire à front plan. Nous montrons que, le long de la branche convective, l’effet de l’advec-

tion est du même ordre que celui du couplage convection-déformation, alors que, le long de la branche morpho- logique, il est largement dominant.

Abstract

2014

We extend the perturbative analysis of the coupling between front deformation and solutal convec-

tion performed in a recent article [3] to include the effect of « advection », i.e. of the flow induced by the density change associated with solidification. We study the resulting shifts of the bifurcation from the stationary planar

front regime. We show that, for the convective branch, the effect of advection is comparable to that of the convec-

tion-deformation coupling, while, for the morphological branch, it is by far dominant Classification

Physics Abstracts

61.50C - 47.20

-

64.70D

1. Introduction.

It is well known that when a dilute binary alloy is

submitted to directional solidification (i.e. pulled

at constant velocity V in an external thermal gradient),

the structure of the solid and the morphology of the growth front are primarily controlled, for systems with atomically rough solid-liquid interfaces, by

solute diffusion in the liquid phase. At small pulling velocities, the growth front is planar; beyond a

critical velocity Vc (which depends on the thermal

gradient 6 and on the concentration far from the front in the liquid, C ex)’ the front develops a periodic

cellular structure, associated with the Mullins-Sekerka

instability [1, 2].

However, such a system is never free from hydro- dynamic motion in the liquid phase. This flow is induced by two mechanisms :

(i) the system is submitted to a thermal gradient

(*) Associe au C.N.R.S.

(+) Also : Departement de Physique, UER de Sciences Exactes et Naturelles, Universite de Picardie, 33, rue Saint- Leu, 80000 Amiens, France.

Moreover, solidification of an alloy produces, at the interface, an excess or defect of solute which induces

a concentration gradient in the liquid ahead of the front In the presence of gravity, these gradients give

rise to buoyancy forces and may thus induce a con-

vective flow ;

(ii) the densities of the liquid and solid phases

are different The solidification of a given mass of liquid is therefore associated with a volume change,

which must be compensated for by a hydrodynamic

flow in the liquid In order to make a clear distinc- tion from buoyancy-induced effects, we will conven- tionally call this current an « advective flow ».

Hydrodynamic flows and front deformations are

coupled In a recent article [3] (hereafter referred to

as (I)), we studied the effect of the coupling between

front deformation and solutal convection on the

position of the bifurcation from the regime with pla-

nar front and quiescent liquid in the case where the

solid is pulled vertically, the thermal gradient is stabilizing, and the solid and liquid densities are

assumed to be equal.

We showed that, close to the convective and mor-

phological bifurcation, the effective coupling between

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0198500460100165700

(3)

1658

convection and front deformation is small at usual values of the thermal gradient, and that a first-order

perturbation treatment is well justified We found,

in particular, that the convection-induced shift of the Mullins-Sekerka (MS) bifurcation is, under usual experimental conditions, extremely small.

In this article, we want to study the ways these results are modified when advection effects are taken into account It is clear that advection couples to

front deformation : indeed, the corresponding flow

must be normal to the solid surface. Thus, a defor-

mation of the front distorts the flow, this modifies the transport of solute from (or towards) the inter- face, and that in turn reacts on the front deforma-

bility.

Of course, the relative density changes 8

=

(ps - PL)/PL are small (typically, of the order 10-2 to, at

most, 10-’) and so are usually considered as negli- gible. However, advection effects are unavoidable

(even in microgravity experiments). Moreover, since

it is found that the bifurcation shifts due to the convec-

tion-deformation coupling are small, the question

arises of comparing them to those induced by advec-

tion.

In § 2 we write down the equations describing the

solidification problem in the presence of both advec- tion and solutal convection and calculate the solution with a planar stationary solid front We set up the linear stability analysis of this solution and write

down the condition of existence of marginal modes.

In § 3 we study the effect of advection on the decou-

pled MS and convective bifurcations.

In § 4, we include the convection-deformation

coupling and calculate the position of the bifurca- tions in a first-order perturbation approximation.

2. Dynamical equations and planar stationary solu-

tion

We assume that the dilute binary mixture is pulled vertically (in the (- 2) direction) at velocity V, and

that the system is quasi-infinite in the (x, y) directions.

Following (I), we also assume that the thermal gra- dient is stabilizing and neglect thermal expansion (i.e.

thermally-induced convection).

We define the following dimensionless variables :

where the tilded variables are the physical ones.

f, ti, dr ,, C are the temperature, liquid velocity,

front velocity, and solute concentration in the liquid

is the relative density change at solidification (for

most materials, s > 0). D is the solute diffusion coef-

ficient in the liquid, TM the melting temperature of the pure solvent, Coo the solute concentration in the

liquid far from the front, and K the equilibrium solute

distribution coefficient

Following (I), we neglect, in the non-dimensionaliz- ed equations of the problem, all terms proportional

to D/Dth and to D/v (where Dth is a heat diffusion coef- ficient and v the kinematic viscosity of the liquid).

We also neglect diffusion in the solid phase, treat the liquid as incompressible, and assume quasi-instan-

taneous local thermodynamic equilibrium on the

front

The system is then described by the following equa- tions :

(i) In the solid phase :

Heat diffusion :

(ii) In the liquid phase:

Heat diffusion :

Solute diffusion, :

(z is the unit vector along Oz)

Mass conservation :

Momentum conservation :

a PL TC- is the solutal expansion coefficient;

PL

the gravity g is positive.

(iii) At the interface (z

=

zs(x, y, t)) : No-slip condition :

where n is the unit vector along the normal to the front pointing into the liquid.

Mass conservation :

Continuity of temperature :

Heat balance :

where n

=

ks/kL is the ratio of the thermal conducti-

vities.

(4)

Concentration balance :

Curvature-induced local interface temperature shift :

M

=

ML CwIKTm, where mL is the slope of the liquid-

us curve of the mixture at (T

=

T M, C

=

0).

r

=

y VIED, where y is the solid-liquid surface tension, C the specific latent heat of fusion, and X

the curvature of the front, defined as positive for a

convex solid

Moreover, we assume that the hydrodynamic flow

is free at infinity in the liquid (z -+ oo), where the

pressure is fixed.

When comparing equations (1-12) with the corres- ponding equations of (I), it is seen that taking advec-

tion into account results in the following :

-

a redefinition of the non-dimensional variables that is, a rescaling of the diffusion length and time of the system which become, respectively, D/(1 + 8) V and D/(1 + 8)2 V2 ;

-

a factor (1 + s)- 3 in the coefficient of the second term of the hydrodynamic equation (7). As shown in

(I), this coefficient is directly related to the Rayleigh

number for our problem

where

and the (1 + E)-3 factor naturally accounts for the rescaling of the diffusion length;

-

the presence of a term proportional to 8 in equation (9), which describes how the volume defect

resulting from solidification must be fed by advection.

We first look for a stationary solution of the above

equations corresponding to a planar front Choosing

the origin of the z-coordinate at the front position,

one finds that this solution is described by

GL is the positive dimensionless temperature gradient

in the liquid :

with 6 the physical thermal gradient

We are interested in the thresholds of instability

of this solution, so we want to study its linear stability.

That is, we assume that the planar front undergoes

a small deformation of wavevector a (chosen along the arbitrary x-direction)

This induces variations of wavevector a and growth

rate a of the various fields. One then standardly expands equations (3.12) to first order in these varia- tions. The corresponding equations, which are written

and solved in the Appendix, provide the « dispersion

relation » a

=

f (a). We assume that the bifurcations

we study are not oscillating. As shown in (I) and

confirmed by numerical results [4], this is indeed true in the absence of advection, except at unrealistically

small thermal gradients or for very small values of the distribution coefficient K. Since the parameter e, which measures the strength of advection effects, is

small (s 10-1), one may reasonably expect that the introduction of these effects does not alter the character of the bifurcations. This can be proved explicitely [5], to first order in 8, with the help of a perturbation expansion.

The equation determining the neutral modes of the system then reads (see Appendix) :

where

with

The functions A and B which appear in equation (19)

are the same as those appearing in (I) ; the presence of advection changes their argument from R,, into ,k.

The new quantity C(.,k, a) is defined in the Appendix.

3. The uncoupled bifurcations.

Following the analysis performed in (I), we first

define the « pure » or uncoupled bifurcations from the planar solution, namely :

- the pure morphological one, i.e. the MS bifur- cation at zero gravity, and

- the pure convective one, Le. the onset of convec-

tion in the liquid ahead of a planar non-deformable solid front.

In this section, we study how these uncoupled

bifurcations are shifted by advection.

(5)

1660

3.1 ADVECTION-INDUCED SHIFT OF THE MS BIFUR- CATION.

-

The equation for the corresponding

neutral modes is obtained by taking the l§’ -+ 0

limit of equation (19). In this limit (see Appendix)

where

The neutral mode equation (19) then becomes

The determination of the bifurcation proceeds

in the standard way : one looks, at fixed p, for the position CbMS, aMS) of the minimum of the neutral

curve 1Jc(a).

The critical wavevector is the solution of

Equations (24) and (25) provide a parametrization

of the bifurcation curve -Gms(P).

Since E 1, they can be, to a good approximation,

solved to first order in £. One obtains :

where the critical wavevector is, to the first order, unshifted It is given by

With the help of equations (20), (21), and (27),

we obtain from equation (26) the relative shift of the bifurcation at constant G, V :

As discussed in (I), for not-too-small thermal gra- dients (typically 6 > 1 K/cm) and not-too-large pulling velocities (V 1 cm/s)

and

That is, on the small velocity side of the bifurcation

curve CII(V), which corresponds to usual directional solidification conditions,

-

if 8 > 0 (ps > PL) the effect of advection is

destabilizing : the MS bifurcation for given Coo occurs

at smaller velocities,

-

if 8 0, it is stabilizing.

At very large velocities (ams 1 ; V in the cm/s

range or more (’)), it can be checked from equations (26), (27), and (28) that

i.e., advection becomes stabilizing (resp. destabilizing)

for 6 > 0 (resp. E 0).

This is illustrated in figure 1, where we plot the

pure MS bifurcation curve in the (Coo, V) plane for a PbSn alloy (s

=

0.04) pulled in a thermal gradient i7

=

200 K/cm.

3.2 ADVECTION-INDUCED SHIFT OF THE UNCOUPLED CONVECTIVE BIFURCATION.

-

This corresponds to the

limit of a non-deformable front or, equivalently,

of infinite interface tension r, i.e. to oo. Equa-

tion (19) then reduces to

The relative density change s only appears in equa- tion (32) via the definition of the Rayleigh number Rs. That is, the corresponding marginal curve is

(1) Note that, at such large velocities, the present model should be improved Thermal transport, latent heat pro- duction and interface kinetics in particular, should be taken

into account

(6)

Fig. 1.

-

Bifurcation diagram for the PbSn system with

G

=

200 K/cm (the values of the parameters are taken from reference [4]). Full line : Bifurcation curves in the presence of advection and of the convection-deformation

coupling; the dots correspond to the perturbative result, the

crosses to numerically computed results. Dashed line : Pure Mullins-Sekerka and pure convective bifurcation curves

without advection. Dotted line : Pure convective bifurcation

curve with advection. (Note that the corresponding dotted

line for the Mullins-Sekerka branch is not distinguishable

from the full line on the scale of the figure.)

-given by

where the function & is the same as for the advec- tion-free system. Let (Rs*, a*) be the coordinates of the minimum of R.,,(a); they depend only on the dis-

tribution coefficient K (see (I)). The position of the

convective bifurcation is then given, with the help

of equations (14) and (15), by

So, to first order in s, the relative shift of the convec-

tive bifurcation

is constant along the bifurcation curve.

4. Effect of the convection-deformation coupling.

It has been shown in (I) that for advection-free systems under usual experimental conditions, the effective

coupling between the two above « pure » bifurcations is small enough for a first-order perturbation approxi-

mation to be justihed

This results primarily from the large mismatch

between the critical wavevectors a* and a’s which characterize convection and deformation. In typical

situations a* N 0.3, while, close to the point of

parameter space where the uncoupled bifurcation

curves cross (C.,O, Yo), aMS > 1.

We have seen in the preceding section that, in the

presence of advection a* is independent of 8 and

the shift of aMS is at most of order s’. Thus, the effec-

tive strength of the convection-deformation coupling obviously remains small, and its effects can be calculat- ed within the perturbation approximation developed

in (I). Since, moreover, due to the smallness of s, advection effects are small and can be treated to first

order, we choose to rewrite equation (19) under the equivalent form

where A - A(,k, a), etc.

Following (I), we then solve equation (35) for the

shift of each bifurcation by first-order iteration around the neutral modes associated with each of the l.h.s. factors.

4.1 THE MORPHOLOGICAL BIFURCATION.

-

It has been shown in (I) that, along the MS bifurcation curve, A, B and C can be calculated to first order in

Rg to a good approximation. This is obviously true

-far from the crossing point (C 00,0’ Yo) of the two

bifurcations, since, then, R; I along the MS

curve. Close to the crossing point, R; -- ]?;’ =- 10.

But in this region, aMS > 1 and the relevant Rayleigh

number for a perturbation of wavevector aMS

-

which is the true expansion parameter

-

is not Rs’, but R s 11(aM3 1.

We therefore calculate the shift

(where -6’ corresponds to the pure MS bifurcation of the advection-free system) to first order in s and in R;, neglecting terms of order E2, Rs’2 and ERs’.

We thus find (with the help of- equations (A. 21

and A. 22) of (I))

The shift of the bare MS bifurcation is given thus, naturally

-

since we keep to first-order expansions

-

by simply adding the convection-induced shift studied in (I), corrected for advection by the rescaling

and the pure advective correction (Eq. 26).

(7)

1662

Equation (37) can be rewritten as :

In order to estimate the relative orders of magnitude

of the two contributions to the shift, we rewrite equation (38) in the vicinity of the crossing point (Coo ,°, Vo), where aMS » 1 (for not-too-small 6’s).

Using equation (30), and equation (57) of (I), one

gets :

For PbSn alloys, using the material parameters given in reference [4], this gives

at the crossing point and, for all reasonable values of 6(6 > 1 K/cm), the convective (first-term) cor-

rection is usually negligible compared with the

advective (second term) one.

As one moves away from the crossing point along

the MS curve, the convective shift decreases while the advective one remains quasi-constant (for usual pulling velocities), and thus dominant

4.2 THE CONVECTIVE BIFURCATION.

-

Its shift is obtained by iterating equation (35) to first order around the solution of

i.e., around the convective bifurcation (R,,*, a*) of

the non-deformable system in the presence of advec- tion.

Using equations (58) and (59) of (I), one obtains

where j = (K/RS*) (dRS*/dK) has been tabulated in

(I) as a function of K. Typically, for most materials, 2 x 10-1.

The starred quantities A *, C* and m* are to be

calculated for Rs = and a

=

a*.

Developing equation (42) to the first order in E, one obtains for the relative velocity-shift of the convective bifurcation

where

As shown in (I), along the convective curve A is maximum at the crossing point, where, for aoms >> 1,

it is of order K(m + K - 1 ) -1. We have calculated the quantity C*IKA* R.* numerically. Its variations

with the distribution coefficient K are tabulated in table I. One may then check that, since s 10- 1 and (/3 10-1, the cross-correction proportional

to BÇ appearing in equation (43) is negligible, except for mixtures with very small values of K (2).

Table I.

(2 ) This is, for example, the case of succinonitrile-etha- nol [6]. Note that, for such mixtures, the effective convec-

tion-deformation coupling is not small, and our pertur-

bative approximation is not valid

(8)

For 8 > 0 (PS > PL), which is the case for most

materials, both the advection effect and the coupling

to deformation are stabilizing.

5. Conclusion.

Thus, it appears that the shifts due to convection and advection are, to a good approximation additive.

This results from the fact that each of them is small and may thus be calculated in a first-order pertur- bation approximation.

The smallness of the advection-induced shifts

simply results from the size of the relative density change during solidification.

The physical origin of the smallness of the con-

vection-induced shifts has been discussed at length

in (I). Let us briefly recall that it is due to the fact that, except for very small thermal gradients or segre-

gation coefficients, the wavevectors of the uncoupled

convective and morphological instabilities are widely separated The mode that is becoming marginal

e.g., to fix ideas, the surface deformation mode, only couples to the convective mode with the same wave-

vector, which is always rapidly relaxing, and thus

follows almost adiabatically the slow mode. Therefore,

the marginal deformation (resp. convection) mode only « drags a small amount of convection » (resp.

deformation). Moreover, one finds that, due to the larger curvature of the convective spectrum, the convection-deformation coupling affects the convec-

tive instability more strongly than the MS one, the shift of which is, under usual conditions, negligible

in practice.

We have shown here that the shifts of the two instabilities induced by advection are of comparable

orders of magnitude.

For the convective branch, in the vicinity of the crossing point, this shift is comparable to that due to

the coupling to surface deformation. For example,

for PbSn at a thermal gradient of 2000/cm (see (I))

the deformation-induced shift is of the order of 9 %,

while the advection-induced one is 4 %. When one

moves away from the crossing point, the deformation- induced shift decreases as ( v/ vo)213, while the advec-

tive one is constant and thus becomes dominant at small velocities.

The reason why advection stabilizes the system against convection, for 8 > 0, is apparent from equation (14). For a non-deformable interface, the only effect of advection is to rescale the thickness of the « convective box » ahead of the front and the concentration gradient by factors (1 + E)-1 and (1 + ~); this rescales the Rayleigh number by (1 + S)3,

thus reducing the critical velocity at given C..

For the Mullins-Sekerka branch, as explained above, the shift due to convection is, in general, negligible compared to that due to advection. For

positive 8, in a wide range of values of V above the

crossing point (corresponding to usual experimental

conditions in directional solidification), advection

is destabilizing. This may be analysed qualitatively

in the following way.

a) On the one hand, as already mentioned, advec-

tion increases the value of the concentration gradient

at the interface. This increases the amplitude of the destabilizing effect of diffusion.

b) On the other hand, when the interface deforms,

the streamlines, being constrained to remain per-

pendicular to it, must curve. When E > 0, this amounts

to the creation of a current along the average direction of the surface which carries concentration towards the regions where the solid bumps into the liquid

This is equivalent to increasing the local production

of concentration excess (or defect, depending on

the sign of K - 1) at the interface on the bumps.

For a given concentration gradient, this must be compensated for by a decrease in the local growth velocity. So, for E > 0, this effect is stabilizing, It

increases with the amplitude of the advective current, which is proportional to V, but it decreases when the

wavelength of the instability increases, which occurs

when V increases.

Thus, it is seen that the overall stabilizing or destabilizing character of the advection effect results from the rather subtle interplay between these effects.

At not-too-large V’s, the MS wavelength is of order

V-’I’, and one finds that the overall effect is destabi-

lizing. A numerical inspection of equation (38)

shows that it changes sign at large velocities (typically

V > 1-10 cm/s).

Finally, in order to check the accuracy of our

perturbative results, we have calculated numerically. ’

the bifurcation curves from the exact equation (19)

for the case of PbSn alloys in a thermal gradient

G

=

200 K/cm and compared them with the approxi-

mate curves obtained from our perturbation approxi-

mation. The results are plotted in figure 1. It is seen that, at such a value of G, the accuracy of the pertur- bative prediction is excellent. Of course, the agreement would become poorer for very small distribution coefficients and, as discussed in (I), for very small thermal gradients.

Appendix.

We assume that the planar front undergoes a small

deformation :

To first order in (, this deformation induces responses of the concentration, temperature and velocity fields

of the form

(with f

=

(C, T L, T g, u)).

Equations (3-12) can then be written, to first

order in C.

(9)

1664

(i) In the solid phase :

(ii) In the liquid phase :

(iii) At the interface :

with the boundary conditions

Equations (A. 3), (A. 4), (A. 8) and (A. 9) give

Elimination of Cl(z) between equations (A. 5) and (A. 7), and of C between (A. 10) and (A. 11) finally

reduces the problem to solving

with the boundary conditions

and

where

and band p are defined in equation (20).

The general solution of equation (A. 17) can be

written

where the three functions u(i)(s) have been defined in the Appendix of (I). Plugging this solution into the interface conditions (A. 19, 20, and 21) and writing the compatibility condition of the resulting equations, we obtain, for a

=

0, the equation defining

the neutral modes :

The functions A, Bare those defined in equations (25)

of (I). C can be written as

the coefficients of which are defined in equation (26)

of (I). Then, using the small-R" expansions given in equations (A. 8-20) of (I), we easily obtain, for the

limit R.’ -+ 0

(10)

References

[1] MULLINS, W. W. and SEKERKA, R. F., J. Appl. Phys.

35 (1964) 444.

[2] WOLLKIND, D. J. and SEGEL, L. A., Philos. Trans. R. Soc.

268 (1970) 351.

[3] CAROLI, B., CAROLI, C., MISBAH, C., ROULET, B., J.

Physique 46 (1985) 401.

[4] CORIELL, S. R., CORDES, M. R., BOETTINGER, W. J., SEKERKA, R. F., J. Cryst. Growth 49 (1980) 13.

CORIELL, S. R., private communication.

[5] MISBAH, C., Thèse de Doctorat de 3e Cycle, Université

Paris VII (1985).

[6] SCHAEFER, R. J. and CORIELL, S. R., Materials Pro-

cessing in the Reduced Gravity Environment of

Space, Guy E. Rindone ed. (Elsevier Science

Publishing Co.) 1982, p. 479.

Références

Documents relatifs

The calculated temperature at a depth of 100km and the continental/mantle velocities (after the initial time step) are shown in Fig. Similar to the free convection,

In particular, the time and wavelength corresponding to the onset of convection are determined for given non-dimensional numbers. In section 5, comparisons of these results

The shift of the convective bifurcation, though much larger, can also be calculated with very good accuracy in the same range of values of the thermal gradient with the

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

Abstract 2014 We study, for directional solidification or fusion of binary mixtures, the stability of planar fronts against one-dimensional deformations.. We extend the

The solar convection does not appear to be in strict thermal wind balance, Reynolds stresses must play a dominant role in setting not only the equatorial acceleration but also

[10] have carried out an asymptotic study of Rayleigh–B enard convection under time peri- odic heating in Hele–Shaw cell and have shown that modulation has an effect on the threshold

(1997) studied numerically the fluid flow and heat transfer induced by natural convection in a finite horizontal channel containing an indefinite number of uniformly spaced