You are on page 1of 5

Materials Science and Engineering A304306 (2001) 235239

A simplied model for gas atomization


G. Vedovato , A. Zambon, E. Ramous
DIMEG, Universit di Padova, 35131 Padova (PD), Italy

Abstract A simplied model describing the cooling behaviour and the solidication of undercooled droplets in gas atomization has been developed. The resulting computer code can predict the cooling behaviour both in the liquid and in the solid state for any powder size. The droplet velocity in the collecting chamber as well as its temperature, solid fraction and cooling rate can be predicted as a function of the characteristics of the processed alloy, the superheating of the melt, the atomizing gas, its velocity at the nozzles and the droplet diameter. The model overcomes the problem of the undetermined value of the undercooling and evaluates the effects of the simplied approach. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Gas atomization; Modelization; Recalescence; Equilibrium; Nucleation temperature

1. Introduction The present model can describe the solidication of droplets in gas atomization. In this process it is important to know the thermal transient of the particles. This is not an easy task because: there is a range of different transients due to different sizes of the particles; solidication occurs with undercoolings, which are variable, unknown and not measurable. Previous numerical models [17] were concerned mainly with the recalescence: undercooling was considered a parameter or it was predicted using heterogeneous nucleation theories. In a strict sense those are valid only for massive nucleation. In a molten batch there are a number of different types of nucleant particles, each one with a different nucleation temperature. In conventional solidication (such as in castings and ingots) undercoolings are generally small and constant because they are originated only by the particles with the highest nucleant potency. In gas atomization those particles can no longer spread solidication to the whole molten bulk, because this is fragmented and the nucleant particles are scattered in a large number of small droplets. In each droplet the nucleation is originated by the particle with the highest nucleant potency that are not necessarily the same: thus undercooling is variable. There are models and data that conrm this hypothesis [811]. Furthermore, there is experimental evidence that there are both single (one

for each droplet) and multiple (several events in a single droplet) nucleation sites. On the other hand, it can be hard to model the variation of the nucleation temperature with just a single parameter. To overcome these problems a model was developed that can preview the thermal history considering two extreme cases: solidication without undercooling; cooling without nucleation (in the liquid phase).

2. Model formulation 2.1. Motion of a spherical droplet To describe the motion of a droplet in the atomizer chamber, the following hypotheses are made: 1. droplets have a spherical shape which is rigid and constant (contractions due to cooling and the phase changes are neglected); 2. droplets form immediately at the nozzle outlet (the spheroidization times were calculated for denite diameters to be very short with respect to the total solidication time); 3. interactions with other droplets are neglected; 4. the velocity eld of gas ow is constant; 5. droplets are supposed to move on a straight path on the ow centreline, therefore: gas uctuations due to turbulence are neglected; gas turbulence, that could alter particles trajectory, is neglected.

Corresponding author.

0921-5093/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 0 9 3 ( 0 0 ) 0 1 4 7 5 - 1

236

G. Vedovato et al. / Materials Science and Engineering A304306 (2001) 235239

Under such hypotheses, the motion of a spherical droplet (in a one-dimensional gas ow) can be described by the following equation: d Vd 1 dud (t) = Vd (d g )g g Ad CD (z)|ud (t) dt 2 ug (z)|(ud (t) ug (z))

(1)

where ud , d , Vd and Ad are the speed, density, volume and cross-sectional area of a droplet, respectively, ug is the axial velocity of the gas, g the density of the gas, g the acceleration due to gravity, t the time. CD is the drag coefcient which can be approximated with an accuracy better than 7% by the following empirical equation [5]: CD = 0.28 + 6 21 + Re Re (2)
Fig. 1. Schematic behaviour of a molten particle: (1) liquid phase cooling, (2) recalescence, (3) solidication, (4) eutectic solidication, (5) solid phase cooling.

where Re g Dd |ud (t) g (z)|/g is the droplet Reynolds number, Dd the droplet diameter, g the dynamic viscosity of the gas. Eq. (2) is applicable over the range of 0.1 < Re < 4000. Analytical relations describing the velocity eld of gas ow are not available. Instead an equation valid for a turbulent circular free jet emerging from an orice into a resting uid is used. It describes the velocity decay prole along the centreline of the incompressible free jet [5,12,13]: ug (z) = 1+ u0 z
20 0.05

Tg the gas temperature, t the time, fs the solid fraction and cm = cl (cs cs )fs is the specic heat of the mushy alloy (with cl , cs heat capacities of the liquid and solid alloy), h the convective heat transfer coefcient. The gas temperature is assumed to be constant and equal to 20 C. If the solid fraction is a function of the temperature only, Eq. (4) becomes Vd cm d Td (t) dfs (t) Td (t) L t dTd t = hSd (Td (t) Tg ) (5) This can be rewritten as L dfs (t) 1 (Td (t) Tg ) = 1 0 cm d dTd Td (t) t (6)

(3)

where, u0 the gas velocity at nozzle outlet, z the ow axis distance from nozzle, the gas velocity decay constant whose value is = Ag , where is an empirical constant that was estimated as 7.414 [14] for a turbulent free jet and Ag the nozzle throat-area. 2.2. Cooling and solidication The typical cooling behaviour of an atomized droplet (for an eutectic alloy) is shown in Fig. 1. If the enthalpy of a droplet is H, the sum of the internal energy (function of the temperature) and the latent heat 1 (a function of the liquid fraction), its value during ight is given by H (t)/t = q , where q is the heat ux. The radiation contribution can be neglected. During cooling, the temperature gradient inside the particle can be assumed null (because its Biot number (Bi = hd/) is smaller than 0.1). Thus, the thermal behaviour of an atomized particle can be described by a simple Newtonian formulation: Vd fs (Td (t)) Td (t) L cm d t t = hSd (Td (t) Tg ) (4)

where 0 is a time coefcient dened as 1/0 = hSd / cm d Vd . 2.2.1. Heat coefcient The convective heat transfer coefcient h is given by the RanzMarshall relation [15]: h= g 3 (2.0 + 0.6 Re Pr) Dd (7)

where g is the gas thermal conductivity, Pr = cg g /k the gas Prandtl number, cg the gas specic heat (per unit mass). 2.3. Solidication and undercooling After the temperature of a liquid particle reaches the liquidus temperature, there will be a certain degree of undercooling before a nucleation event occurs. Nevertheless, the ideal case of a solidication without undercooling will be considered at rst.

where Vd and Sd are the droplet volume and surface area, L is the latent heat per unit mass, Td the droplet temperature,
1

The pressure contribution is not relevant and is neglected.

G. Vedovato et al. / Materials Science and Engineering A304306 (2001) 235239

237

2.3.1. Solidication with no undercooling The solid fraction behaviour can be calculated making the following assumptions: complete solute mixing in the liquid phase (there is no solute gradient in the liquid phase); kinetic and capillarity effects are neglected. Furthermore, two different extreme assumptions can be considered: (a) Full diffusion in the solid phase. The solid fraction behaviour is given by the lever rule: fs = 1 1 k0 TL T Tm T (8a)
Fig. 2. Different theoretical cooling paths of an atomized particle.

k0 is the equilibrium distribution coefcient, Tm is the melting temperature of the pure solvent, TL is the liquidus temperature. It is interesting to point out that this would be an equilibrium transformation. (b) No diffusion in the solid (Ds = 0). Under this hypothesis, the solid fraction behaviour is described by the Schiels equation [5]: fs = 1 Tm T Tm TL
1/(k0 1)

(8b)

Of course, this is no longer an equilibrium transformation. The temperature will be (in each instant) lower than in case (a) because a higher quantity of enthalpy would increase the energy of the resulting metastable structure (structural enthalpy). Eqs. (8a) and (8b) can be introduced in Eq. (6). 2.3.2. Solidication with undercooling In a real transformation, a droplet will cool in the liquid phase until a nucleation event is obtained. Then the temperature Td of the droplet quickly rises due to the latent heat release (recalescence phase). It is not easy to quantify the temperature rising, because recalescence is not an equilibrium transformation. We will evaluate this value with consideration to the total enthalpy. We made the assumption that the droplet geometry (and therefore the motion) does not change: the heat transmission coefcient h (Eq. (7)) is always the same, independent of the cooling path. In both cases (with or without undercooling), the heat loss by convection during ight is derived from Newtons law: q = hSd (Td (t) Tg ); the terms h, S and Tg are the same, the temperature Td is similar, so the heat loss is very similar. As a consequence, the undercooled droplet will start solidication with an enthalpy just slightly higher than it would have had if it begun to solidify without undercooling. The difference is rather small as will be seen in Section 4. The recalescence will stop in the temperature interval between the two lines of full diffusion and no diffusion (see Fig. 2). Also the temperature difference between the two lines is small. We point out that, due to the high solidication rate it is more realistic to consider that a complete

segregation occurs. Thus the recalescence will stop at the temperature given by Schiels equation. To conclude, a solidication with a generic undercooling can be studied by calculating just two extreme cases: solidication without undercooling (and full segregation) and a solidication always in the liquid phase. The only indeterminate factor is the nucleation temperature that can be determined in a second time. The possible nucleation of different phases must be considered (just like in FeNi alloys). We believe that in this case, the model could be used with more important connection to experimental data. Anyway, in this case the nal temperature will be lower than in the case of simple segregation because a higher quantity of enthalpy will go to increase the resulting metastable structure. The model will work with pure metals and alloys without nucleation of metastable phases (such as NiCu). An important limit is given by the hypercooling (or critical nucleation temperature): when the undercooling exceeds this limit, all the liquid fraction solidies during recalescence. The hypercooling is given by Thyp = H /Cl . We point out that according to the model just developed: The recalescence arrest temperature is a function of the solid fraction corresponding to a certain enthalpy. It does not depend on parameters such as the heat coefcient, time, nucleation rate and density, etc. The criterion just exposed is valid even if the recalescence step is not adiabatic: in any case, it will stop when the Schiels line is reached (see Fig. 2).

3. Model assessment The model previously described was solved by means of a computer code, where Eqs. (1) and (6) were solved simultaneously by means of nite differences schemes. It is possible to change the following simulation parameters: processed alloy (all the characteristics), atomizing gas (all

238

G. Vedovato et al. / Materials Science and Engineering A304306 (2001) 235239 Table 2 Thermophysical properties of Al4.5 wt.% Cu alloy 1039 J kg/K 2.6 102 W/m K 1.78 105 N s/m 1.16 kg/m3 Specic heat of liquid alloy Specic heat of solid alloy Density Latent heat per unit mass Thermal conductivity Equilibrium solute distribution coefcient Slope of liquidus Melting temperature of pure solvent Liquidus temperature Solidus (or eutectic) temperature Hypercooling 982 J kg/K 900 J kg/K 2800 kg/m3 348 000 J/kg 164 W/m K 0.14 3.4 934 K 919 K 821 K 354.37 K

Table 1 Thermophysical properties of nitrogen Specic heat (or heat capacity) Thermal conductivity Viscosity Density

the characteristics), gas velocity at nozzle outlet, droplet diameter, superheating of the molten alloy. The main parameters that the program can predict (as a function of the time or the axial position) are: droplet temperature, solid fraction, gas velocity, droplet velocity, Reynolds number, heat coefcient, cooling rate. The characteristics of the atomizing gas are shown in Table 1. To test the model, the Al4.5 wt.% Cu alloy was used (Table 2), because its properties are well known, and it has already been studied in other papers.

4. Results and discussion In Fig. 3, the results obtained by the computing of the two extreme cases of thermal behaviour (no undercooling and no nucleation) are traced. A generic solidication behaviour of an undercooled particle can be traced by means of these two cases. Initially a particle cools according to the lower line (no nucleation): when a generic recalescence occurs the temperature rises until the limit given by the Schiels law (see Section 2.3.2). This applies to any possible recalescence behaviour, instantaneous or not. It is possible to see the maximum error involved, assuming that the heat ux is invariant with undercooling. Thus, when

solidication following path A is completed, the difference between values of curve A and B should be exactly equal to Thyp . A certain discrepancy has instead to be expected due to the lower heat ux experienced by a particle solidifying according to path B. In Fig. 3 this discrepancy is about 30 K, in the present example this is the maximum possible error. The hyperundercooling (Thyp = H /Cl ) limit is traced. When the undercooling exceeds this limit, all the liquid fraction solidies during recalescence.

5. Summary and conclusions In order to predict the solidication behaviour of particles in gas atomization, a mathematical model was developed. A new theoretical approach to determine the arrest recalescence condition is proposed: when a nucleation event occurs in an undercooled alloy, the temperature rises until a value given by the Schiels equation.

Fig. 3. Cooling behaviours of Al4.5 wt.% Cu droplet. Just the two extreme cases (no undercooling and no nucleation) were computed. Some possible examples of recalescence are drawn (dotted lines). Diameter 50 m, superheating: 40 K, gas initial velocity: 260 m/s. (1) Instantaneous recalescence (and adiabatic), (2) non-instantaneous recalescence, (3) nucleation event under hypercooling limit: all the liquid fraction solidies during recalescence. The maximum error made considering the enthalpy independent of undercooling is shown.

G. Vedovato et al. / Materials Science and Engineering A304306 (2001) 235239

239

Thus, the solidication of an undercooled particle can be described by computing two extreme cooling paths: 1. solidication without undercooling; 2. cooling occurring entirely in the liquid phase (with no nucleation event). A particle will initially cool through the second path. When a nucleation event occurs the temperature will increase until the rst path is reached. The solidication will be completed through the rst path (see Fig. 3). We point out that this approach can be used even if the recalescence phase is neither adiabatic nor instantaneous. In this way it is no longer necessary to determine the nucleation temperature exactly. Therefore the model can be useful in industrial applications, in the designing of a gas atomizer or to evaluate whether an existing atomizer can be used to process a specic alloy. In this case an approximate undercooling range can be xed. In an experimental eld it can be used to study the solidication evolution. With the aid of experimental data it allows evaluation of an existing undercooling extent and prediction of the inuence of the process parameters on the microstructure.

References
[1] C.G. Levi, R. Mehrabian, Metall. Trans. B 11 (1980) 2127. [2] C.G. Levi, R. Mehrabian, Metall. Trans. A 13 (1982) 221223. [3] P. Mathur, D. Apelian, A. Lawley, Acta Metall. 36 (2) (1989) 429443. [4] E. Gutierrez-Miravete, E.J. Lavernia, G.M. Trapaga, J. Szekly, N.J. Grant, Metall. Trans. A 20 (1989) 7184. [5] E.S. Lee, S. Ahn, Acta Metall. Mater. 42 (1994) 32313243. [6] H. Liu, H. Rangel, E.J. Lavernia, Acta Metall. Mater. 42 (10) (1994) 32773289. [7] Y.H. Su, C.Y.A. Tsao, Metall. Mater. Trans. B 28 (1997) 12491255. [8] M. Libera, G.B. Olson, J.B. Vander Sande, Mater. Sci. Eng. A 132 (1991) 107118. [9] P.G. Hockel, H. Sieber, J.H. Perepezko, in: Proceedings of the 1998 TMS Annual Meeting, TMS, 1998. [10] J.H. Perepezko, in: R. Mehrabian, B.H. Kear, M. Cohen (Eds.), Rapid Solidication Processing: Principles and Technologies, Vol. 11, Claitors, Baton Rouge, LA, 1980. [11] J.H. Perepezko, B.A. Mueller, K. Ohsaka, in: E.W. Collings, C.C. Koch (Eds.), Undercooling Alloy Phases, The Metallurgical Society of AIME, Warrendale, PA, 1987. [12] B.P. Bewlay, B. Cantor, Mater. Sci. Eng. A 118 (1989) 207222. [13] B.P. Bewlay, B. Cantor, Metall. Trans. B 21 (1990) 899912. [14] H. Schlitchting, Boundary-Layer Theory, Mc-Graw Hill, New York, 1979, pp. 747749. [15] W.E. Ranz, W.R. Marshall, Chem. Eng. Proc. 58 (1952) 141154.

You might also like