Proc. Natl. Sci. Counc. ROC(A) Vol. 23, No. 5, 1999. pp.

622-629

Macro-Micro Modeling of Solidification

L ONG-S UN CHAO
AND

W U -C HANG DU

Department of Engineering Science National Cheng Kung University Tainan, Taiwan, R.O.C. (Received November 3, 1998; Accepted March 31, 1999) ABSTRACT To solve solidification problems, macro-models are generally applied. Macro-micro models, by considering the formation of microstructures in terms of nucleation and growth, can obtain better results than can macro-models. Except for temperature distributions, macro-micro models can offer more information about the solidification process, including undercooling, grain size, grain density etc. These data can be used to predict the mechanical properties of materials directly. In this paper, two macromicro models are built to investigate the equiaxed solidification of eutectic. One macro-micro model considers the nucleation step, and the other does not (assuming that nucleation occurs instantaneously). An experimental example of a cylindrical casting is used to test these models. The finite difference method is utilized to solve the heat transfer problem, and the source term method is employed to handle the released latent heat. From a comparison with the experimental result, the computed cooling curve is found to be very close to the experimental one. From the computational results, it is found that a higher cooling rate yields the greater undercooling, which leads to a smaller grain size or higher grain density. Key Words: macro-micro modeling, solidification, nucleation, cylindrical casting

I. Introduction
For binary alloys, macro-models based on phase diagrams are generally used to solve solidification problems. These models can give only rough predictions of the solidification time, isotherms etc., which do not have any direct relation with the microstructures and physical properties of solidified alloys. Recently, the micro-viewpoint has gradually been incorporated into the solidification models. In these models, the microstructure evolution during the solidification process is considered. The whole process includes three different steps: nucleation (an increase in the number of nuclei), growth (an increase in the volume of the grain) and impingement. These three steps are illustrated in Fig. 1. In the solidification process, as the temperature of the liquid metal falls below the melting point, nucleation begins, as shown in Fig. 2. Crystal clusters (or embryos) are formed. These clusters may melt or grow. When the clusters are big enough, they will not melt any more. At this time, they are called nuclei. At the beginning of nucleation, the number of nuclei increases very slowly. After a critical undercooling value is reached ( T n in Fig. 2), the number increases rapidly. Nucleation proceeds until the decreasing temperature starts to increase, i.e., recalescence occurs. At

Fig. 1. Equiaxed solidification of the eutectic system. [Reprint from Rappaz (1989)]

this point, the number of nuclei reaches its maximum value. After nucleation, there is a long period of growth.

622

Macro-Micro Modeling of Solidification system of grey cast iron by using the concepts of nucleation and growth. Stefanescu and Trufinescu (1974) used this method to investigate the effect of silicon inoculation on microstructure evolution in cast iron. Fredriksson and Svensson (1984) extended this analysis to white and spheroidal graphite cast iron. Su et al . (1984) used diffusion of carbon through the austenite shell to build a growth model. Though they obtained an undercooling value smaller than the experimental one, the general trend of the cooling curves was predicted successfully for spheroidal graphite cast iron. Gradually, the model has been applied to study other eutectic systems, such as Al-Si alloy. Except the eutectic systems, previous research works have also focused on the eutectoid system (Stefanescu and Kanetkar, 1985; Goettsch, 1991). Macro-micro models differ from macro ones in the way they handle latent heat. Before solidification, there are no differences between these models. Once the temperature falls below the melting point, (using the source term method) the methods used to compute f s (the local volume fraction of the solid) are not the same. In this paper, we focus on equiaxed eutectic solidification, and the testing alloy is grey cast iron. A solidification problem of a cylindrical casting is solved by using the following two models: (1) Nucleation is assumed to occur instantaneously. The Johnson-Mehl and Avarami models are applied to the growth and impingement steps. (2) The Gaussian distribution is used in the nucleation step, and the Close-Pack model is applied to the final step.

Fig. 2. A schematic diagram of the temperature, nucleation rate (I), grain density (N ) and growth rate (V ) distributions vs. time.

In this step, the grain radii increase continuously until the grains come into contact with one another, and this is followed by the third step: impingement. In the final step, though the grain radii can not increase any more, there is still some liquid metal among or inside the grains, which will solidify in this step. Using models which consider the formation of microstructures, we can obtain more information than we can by employing macro-models. Besides a more accurate solidification time and cooling curve, the models can also obtain data about undercooling, the grain number, grain size, nucleation, growth rates, etc., which cannot be obtained using macro-models. There are two main categories of macromodels (Rappaz, 1989): one-domain and two-domain (such as the front tracking method of Crank (1984)) methods. The latter type is numerically difficult and is not good for macro-micro models. Among the one-domain methods, the effective specific heat method (Hsiao, 1985), the enthalpy method (Date, 1991), the source term method (Voller and Swa, 1991) and the temperature recovery method (Tszeng et al ., 1989) are frequently mentioned in the literature. These one-domain methods can also be applied to macro-micro models, but they need some modification for calculating the amount of latent heat released during the solidification process. In this paper, the source term method is applied to macro-micro models since it is easy to see the difference between macro and macro-micro models from the mathematical formulation. The first researcher to propose a macro-micro model was Oldfield (1966), who studied the eutectic

II. Mathematical Model and Numerical Method

In this paper, two macro-micro models are built to simulate the equiaxed solidification of the eutectic system. An experimental example of a cylindrical casting (Kanetkar et al ., 1988) is used as the test case and is shown in Fig. 3. To build these models, the following assumptions are made: (1) There is no melt flow during the solidification process. (2) Since the grain shape of the equiaxed eutectic during nucleation and growth steps is nearly spherical (Kurz and Fisher, 1989), it is assumed to be spherical. (3) Because the testing geometry is that of a cylinder, it is assumed to be axi-symmetric. Before discussing the macro-micro models, the source term method (of the macro model) will be described. (1) Energy equation:

623

L.S. Chao and W.C. Wu

T r

= 0 , and
r =0

(3)

(b) at the air/metal interface ( z =16 cm), q "= h conv( T T ), (4)

where q " is the heat flux, h conv is the convective heat transfer coefficient of air, and T is the environment temperature. (c) At the metal/mold interface: Since the temperature distribution of the sand mold is of less interest and requires a great deal of computing effort, the equivalent heat transfer coefficient is used for the sand mold. The boundary condition at the metal/mold interface can be given by q "= h eff( T T ). (5)

The primary difference between macro and macromicro models is in how the local volume fraction of solid, f s, is computed. In the macro model, Vollers method (Voller and Swa, 1991) is used to calculate f s. In macro-micro models, since equiaxed grains are assumed to be spherical, f s can be expressed as

f s(t ) = 4 R 3(t ) N (t ) , 3

(6)

where N( t ) is grain number per unit volume and R ( t ) is the grain radius. The derivative of fs can be written as

df s = 4 R 2(t ) N (t ) dR + 4 R 3(t ) dN . dt 3 dt dt
Fig. 3. A schematic diagram of the (testing) cylindrical casting.

(7)

T + T ] + L f s , c p T = k [ T +1 r 2 r z 2 t t r
2 2

(1)

Because the radius is very small in the nucleation step, the second term on the right-hand side of Eq. (7) can be ignored, and the equation can be rewritten as

where c p is the specific heat, is the density, k is the thermal conductivity, L is the latent heat, and f s is the local volume fraction of the solid. (2) Initial and boundary conditions: (i) The initial condition is T ( r , z , t =0)=T 0, (2)

df s = 4 R 2(t ) N (t ) dR . dt dt

(8)

Equations (6) and (8) are the basic equations of the micro-macro models. The detailed computing methods used to obtain f s for the two macro-micro models built in this paper are described in the following.

1. Model I
where T 0 is the pouring temperature of the liquid metal. (ii) The boundary conditions are, (a) at the center line ( r =0), Nucleation is assumed to occur instaneously. This means that the nucleation step does not need to be considered in this model, and that N( t ) is a constant.

624

Macro-Micro Modeling of Solidification

Table 1. Constants Used in the Two Micro-Macro Models Constant N (in Model I) T 0 T n max Value 6.0 10 9 m 3 3.0 10 8 m/s C 20 C 4.75 C 1.20 10 11 m 3
2 dN = N max exp[ (T T 0) ]dT , dt dt 2T 2 T

(16)

Accordingly, Eqs. (6) and (8) can be rewritten as

f s(t ) = 4 R 3(t ) N , 3 df s = 4 R 2(t ) N dR . dt dt

(9) (10)

where T 0 is the undercooling at the tip point of the Gaussian distribution, T is the standard deviation of the distribution, and Nmax is the total grain density of the whole distribution (integrating Eq. (16) from zero undercooling to infinity). When T / t >0 (i.e., the recalescence occurs), dN =0. This is the end point of the nucleation step. In the growth step, Eq. (11) is also used to calculate the grain radius. The two-step Close-Pack model (Alexandre et al., 1991) is applied in the impingement step, and the f s is given by

In the growth step, by using the Johnson-Mehl model (Johnson and Mehl, 1939), the growth rate V can be given by V = T 2, (11)

where is a growth constant and T is the undercooling, which is equal to the difference between T e and T . T e is the eutectic temperature. The expression for R can be obtained by integrating Eq. (11):
t

R = r0 +

[T e T (t )]2dt ,

(12)

where r 0 is the critical radius (Kurz and Fisher, 1989). For the impingement step, Eq. (10) is modified by multiplying by a factor F :

df s = 4 R 2(t ) N dR F . dt dt

(13)

Here, F is taken as 1 fs (Avrami, 1940). Equation (13) can be rewritten as

df s = 4 R 2(t ) N dR (1 f s) . dt dt

(14)

By integrating Eq. (14), the expression for f s becomes

f s(t ) = 1 exp[ N 4 R 3(t )] . 3

(15)

2. Model II
In the nucleation step, the Gaussian distribution (Zou and Rappaz, 1991) is used to capture the variation in the trend of the nucleation rate. The equation for dN / dt can be written as

Fig. 4. The computed cooling curves obtained by the macro and macro-micro models. (a) Cooling curves; (b) locally enlarged cooling curves.

625

L.S. Chao and W.C. Wu

df s = 4 R 2(t ) N (t ) dR , dt dt
10.32log10(f s) 1 f s 0.8 1 f s < 0.8 .

(17)

(18)

In this paper, the finite difference method is used to compute the temperature distribution. In formulating the finite difference equations, the central difference is used for the space derivative, and the backward difference is used for the time derivative. The algebraic equations are solved iteratively by using the line S.O.R. method (Anderson et al. , 1984).

III. Results and Discussion

In this paper, two macro-micro models have been built to study solidification processes. The testing material is gray cast iron, whose microstructure is equiaxed eutectic. The constants used in these two models are listed in Table 1. The node number of the uniform grid used in the computation is 61 26. In Model I, the time step can be 1 second. However, in Model II the step can not be larger than 0.1 second, and the relaxation factor is adjusted up to 1.2 since it is more difficult to make it converge than it is in Model I. For convenience, the center point of the casting is used as a reference point (Fig. 3). Figure 4 shows the computed cooling curves of the reference point for the macro and macro-micro models. In Fig. 4(a), there is no big difference between these two curves. However, the undercooling phenomenon can not be obtained from the macro model. In

Fig. 6. The computed cooling curves obtained by the two macromicro models. (a) Cooling curves; (b) locally enlarged cooling curves.

Fig. 5. The cooling curves from the experiment and the macro-micro model.

Fig. 4(b), the undercooling as well as the recalescence phenomenon can be clearly seen from the cooling curve of the macro-micro model. Figure 5 shows the computational and experimental cooling curves of the reference point. From this figure, it can be found that the computational data of the cooling curve and undercooling are quite close to the experimental data (Kanetkar et al ., 1988). Figure 6 illustrates the cooling curves of Model I and Model II. Though these two models are quite different from each other, the computational results are very similar (Fig. 6(a)). Taking a close look (Fig. 6(b)), we can find that Model II has a smaller degree of recalescence (or the maximum undercooling), which is closer to the experimental result. Figure 7 shows f s vs. the grain radius at the ref-

626

Macro-Micro Modeling of Solidification shown in Fig. 8. In the figure, it is shown that the Gaussian distribution can successfully simulate the big change of the nucleation rate in a short time. After the nucleation step, the grain density reaches a constant value, i.e., the final grain density. It can also be found that the nucleation time (about 18 seconds) is very short compared to the local solidification time (about 300 seconds). This proves that the assumption of instantaneous nucleation is reasonable. In general, it is thought that a higher cooling rate ( dT / dt ) yields a higher grain density since the higher cooling rate leads to greater undercooling, which results in a larger number of nuclei (Kurz and Fisher, 1989). This is consistent with the computed results described in the following. In Figs. 9-11, the distributions of the maximum undercooling, grain density and radius (of Model II) are shown. From these figures, it can be found the grain radius and density are strongly related to the undercooling. Closer to the center point (where the cooling rate is lower), the maximum level of undercooling is lower, the grain density is lower and the grain radius is larger. From this study, it can also be seen that the advantages of Model I are that the formulation is simple, and that the undercooling prediction is not bad. However, Model I can not obtain the nucleation rate or the grain density which varies with time, and the grain radius should be infinite, which would make f s equal to one, which is not reasonable. Therefore, Model I is only suitable for rough evaluation of the

Fig. 7. f s vs. the grain radius for different models.

Fig. 8. The nucleation rate and grain density distributions vs. time.

erence point for three different models, Model I, Model II and Model II, without modification of (in Eqs. (14) and (15)). Because of the modification of in Model II (after impingement occurs, fs>0.8), the curve become smoother than the one without modification. In Model I, since fs increases slowly and smoothly with the grain radius, the larger time step can be used, and the convergence rate is faster than that in Model II. However, in Model I, f s will not be one until the radius reaches infinity. This is not reasonable. Accordingly, the end point of solidification is set at the radius when f s = 0.999. At the reference point, the relationship between the nucleation rate (or grain density) and time is that 627

Fig. 9. The distribution of the highest level of undercooling.

L.S. Chao and W.C. Wu

Fig. 10. The grain density distribution after solidification.

Fig. 11. The grain size distribution after solidification.

solidification process. On the other hand, Model II can obtain more information about solidification and a better undercooling prediction than can Model I. However, the computation for Model II is not very stable, so a small time step is needed. The convergence rate of each time step is also low. Consequently, Model II uses much more computation time than does Mode I.

IV. Conclusions
In this paper, two macro-micro models have been built in order to study the equiaxed solidification of eutectic. Model I ignores the nucleation step by assuming that it occurs instantaneously; however, it is considered in Model II. These two models can obtain the undercooling (or recalescence), which cannot be obtained by the macro models. From the computational results, the following conclusions can be drawn: (1) By using Model I or II, cooling curves as well as undercooling can be well predicted, but the computed results obtained by Model II are closer to the experimental results. (2) The formulation of Model I is simple, and the convergence rate is fast. This model is suitable for preliminary evaluation of the solidification process. (3) Model II can obtain the nucleation rate, grain density and radius which varies with time, which

cannot be obtained by Model I. Since the variation of f s vs. the grain radius is not as smooth as in Model I, Model II needs a smaller time step for the computation to be stable. (4) During the solidification process, the cooling rate and undercooling are two important factors influencing the formation of the microstructure and heat transfer. (5) A higher cooling rate yields a higher level of undercooling, which results in greater grain density and a smaller grain radius.

Acknowledgment
This research was supported by the National Science Council, R.O.C., under Contract NSC 85-2212-E006-025.

References
Alexandre, P., M. Castro, and M. Rappaz (1991) Finite element modeling of spheroidal graphite cast iron microstructure formation. In: Modeling of Casting, Welding and Advanced Solidification Processes V . The Minerals, Metals and Materials Society, Warrendale, PA, U.S.A. Anderson, D. A., J. C. Tannehill, and R. H. Pletcher (1984) Computational Fluid Mechanics and Heat Transfer . Hemisphere, Washington D.C., U.S.A. Avrami, M. (1940) Kinetics of phase change, II. Transformationtime relations for random distribution of nuclei. Journal of Chemical Physics, 8 , 212-224. Crank, J. (1984) Free and Moving Boundary Problems . Oxford University Press, New York, NY, U.S.A.

628

Macro-Micro Modeling of Solidification

Date, A. W. (1991) A strong enthalpy formulation for the stefan problem. International Journal of Heat and Mass Transfer , 34 , 2231-2235. Fredriksson, H. and I. L. Svensson (1984) Computer simulation of the structure formed during solidification of cast iron. In: The Physical Metallurgy of Cast Iron , pp. 273-283. H. Fredriksson and M. Hillert Eds. North-Holland, New York, NY, U.S.A. Goettsch, D. D. (1991) Modeling the Microstructure Development in Gray Iron Castings . Ph.D. Dissertation. Department of Mechanical and Industrial Engineering, University of Illinois, Champaign, IL, U.S.A. Hsiao, J. S. (1985) An efficient algorithm for finite difference analysis of heat transfer with melting and solidification. Numerical Heat Transfer , 8 , 653-666. Johnson, W. A. and R. F. Mehl (1939) Reaction Kinetics in Process of Nucleation and Growth . AIME Technical Publication No. 1089, AIME, Warrendale, PA, U.S.A. Kanetkar, C. S., D. M. Stefanescu, and N. El-Kaddah (1988) A latent heat method for macro-micro modeling of eutectic solidification. Transactions ISIJ , 28 , 860-867. Kurz, W. and D. J. Fisher (1989) Fundamentals of Solidification , 3rd Ed. Trans Tech Publications, Aedermannsdorf, Switzerland. Oldfield, W. (1966) A quantitative approach to casting solidification: freezing of cast iron. Trans. Am. Soc. Met ., 59 , 945-960. Rappaz, M. (1989) Modelling of microstructure formation in solidification processes. International Materials Reviews , 34 (3), 93123. Stefanescu, D. M. and S. Trufinescu (1974) Crystallization kinetics of gray iron. Zeitshuft fur Metallkunde , 65 , 610-615. Stefanescu, D. M. and C. Kanetkar (1985) Computer modeling of the solidification of eutectic alloys, the case of cast iron. In: Computer Simulation of Microstructural Evolution , p. 171. D. J. Srolovitz Ed. The Metallurgical Society, Warrendale, PA, U. S.A. Su, K. C., I. Ohnaka, I. Yaunauchi, and T. Fukusako (1984) Computer simulation of solidification of nodular cast iron. In: The Physical Metallurgy of Cast Iron . H. Fredriksson and M. Hillert Eds. North-Holland, New York, NY, U.S.A. Tszeng, T. C., Y. T. Im, and S. Kobayashi (1989) Thermal analysis of solidification by the temperature recovery method. Int. J. Mach. Tools Manufact ., 29 (1), 107-120. Voller, V. R. and C. R. Swa (1991) General source based method for solidification phase change. Numerical Heat Transfer , Part B, 19,175-189. Zou, J. and M. Rappaz (1991) Experiment and modeling of gray cast iron solidification part ii: unidirectionally solidified castings. In: Material Processing in the Computer Age, p. 349. V.R. Voller, M.S. Stachowicz, and B.G. Thomas Eds. The Minerals, Metals and Materials Society, Warrendale, PA, U.S.A.

!"#$%&'
!"#$
!"#$%&#'

!"#$%&'()*+,-./012345*6789:;<=>?@&'!"ABC;DEF@G !"#$%&'()*+ ,()-./0123045678 9:()-;6<=>?@A@BC@D !"#$%&'()*+,-./ 012345678$9:;<.=>?@AB$CDEFGHIJ$K !"#$%&'()*+,-./0123456789:;<'+,=>%?@ABCDEFG?@H9I !"#$%&'()*+,-./0123456/7*8390:-.;!<=>?@A@2BC'DE$ !"#$%&' ()*&+,-&.!"/012345678)9:;<=>?@ABC?@DEFGH !"#$%&'()*+,-./$01234,$ 56789:67; <$=>?@!ABC$D !"#$%&'()%*"+ ,-."/012%34-."567%89:

629