Acta Materialia 52 (2004) 219231

www.actamat-journals.com

Mechanism of nucleation and growth of hydrogen porosity

in solidifying A356 aluminum alloy: an analytical solution
Kun-Dar Li, Edward Chang *

Department of Materials Science and Engineering, National Cheng Kung University, Tainan 701, Taiwan, ROC
Received 15 July 2003; received in revised form 25 August 2003; accepted 3 September 2003

Abstract

This study derives an analytical solution for the mechanism of nucleation and growth of hydrogen pore in the solidifying A356
aluminum alloy. A model of initial transient hydrogen redistribution in the growing dendritic grain is used to modify the lever rule
for the mechanism of nucleation of pore. The model predicts the fraction of solid at nucleation, the temperature range of nucleation,
the radius of hydrogen diusion cell, and the supersaturation of hydrogen needed for nucleation. The role of solidus velocity in
nucleation is explained. The parameters calculated from the model of nucleation are used for analyzing the mechanism of kinetic
diusion-controlled growth of pore, in which the mathematical transformations of variables are introduced. With the transfor-
mations, it is argued that the diusion problem involving the liquid and solid phases during solidication could be treated as a classic
problem of precipitation in the single-phase medium treated by Ham or Avrami. The analytical solution for the nucleation of pore is
compared with the mechanism of macrosegregation. The predicted volume percent of porosity and radius of pore based on the
mechanism of growth of pore is discussed with respect to the thermodynamic solution, the published experimental data, the nu-
merical solutions, and the role of interdendritic uid ow governed by Darcys law.
2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Aluminum alloys; Solidication; Porosity; Nucleation and growth

1. Introduction late the capillary feeding distance [17,18]. Piwonka and

The advantage of metalcasting for fabricating complex
shapes is often impaired by the porosity defect. The main
causes of defect are attributed to the inadequate feeding of
the volumetric shrinkage of liquid and the insoluble hy-
drogen gas evolution during solidication. Early engi-
neers followed an empirical Chvorinovs rule to design the
steel castings with a sucient size of riser to feed the
shrinkage [1]. Later, it was discovered that the feeding
ability is also dependent on the shape of steel castings [2
5]. Similar results were found for aluminum castings [68].
The phenomena of porosity formation in aluminum
castings were studied but with limited success to explain
the mechanism [915]. Early researchers reasoned the
observations with the thermal parameters of the solidi-
fying castings such as local solidication time (tf ) [16],
thermal gradient (G) [7], or solidus velocity (Vs ) to calcu-
*
Corresponding author. Tel.: +886-6238-7321; fax: +886-6234-6290.
E-mail address: edchanghs@yahoo.com (E. Chang).
Flemings [19] pioneered the mathematics based on Dar-
cys law to address the resistance to the interdendritic uid
that ows from the riser to feed the shrinking liquid at the
solidliquid interface. This mechanism was used to pre-
dict the formation of porosity in steel castings [20], and
aluminum castings [2125].
Another problem of hydrogen redistribution between
the liquid and solid during solidication was treated by
Poirier et al. [26] based on the thermodynamics, the
dendrite structure, and the eect of surface tension.
Sigworth and Wang [27] extended the thermodynamic
problem to include the eect of interdendritic uid ow,
however the latter eect was questioned particularly
under the directional solidication. The authors rea-
soned that the formation of porosity might not be
treated by the thermodynamics alone and that the ki-
netics must be involved. Thus, an understanding of the
phenomenon of porosity formation lies in elucidating
the kinetic mechanism of hydrogen redistribution during
the nucleation and growth of porosity. In the following,

1359-6454/$30.00 2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2003.09.007
220 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231

Table 1
Experimental data of thermal parameters, microstructure, and nucleation and growth events and features of A356.2 alloy solidied directionally [28],
and the predicted Vp (%) by the numerical method using DE model [29]
Exp. C0 Vs G tf d2 rG T~N re rp Vp Vp DE
(ml/100 g) (mm/s) (C/mm) (lm) (s) (lm) (C) (lm) (lm) (%) model (%)
1 0.2 0.10 5.8 62 49 750 577613 880 234 1.8 1.24
2 0.18 0.25 6.3 23 33 510 573615 640 181 2.6 1.19
3 0.18 0.48 5.9 13 29 450 571603 560 137 1.2 0.26
Note: C0 is the initial concentration of hydrogen in the liquid, Vs is the solidus velocity, G is the thermal gradient, tf is the local solidication time,
d2 is the secondary dendtite arm spacing, rG is the radius of grain size, T~N is the temperature range of nucleation, re is the radius of diusion cell, rp is
the radius of pore, Vp (%) is the volume percent of porosity, and Vp (%) DE model is the numerically predicted volume percent of porosity.

the mechanism will be theoretically analyzed. The ana-
lytical solution will be discussed with respect to the
thermodynamic solution [26,27], the published experi-
mental data [28], the numerical solutions [29], and the
role of Darcys law. In the directional solidication ex-
periments of [28], the thermal parameters and micro-
structure of A356.2 aluminum alloy were measured, and
an X-ray temperature gradient stage (XTGS) was used
to in situ observe the nucleation and growth events and
features of hydrogen porosity in the castings as indi-
cated in Table 1. Wherein the solidus velocity, the main
processing variable, was controlled at three values while
the thermal gradient was held essentially constant to
realize the independent eect of solidus velocity on the
porosity formation in the castings. Later, attempting to
model the experimental observation of the same exper-
iments of [28], three numerical solutions were furnished
by the same research group for predicting the hydrogen
porosity formation [29], in which one solution called DE
model is more successful or relevant and is included in
Table 1.
cation [27]. Now adopting C0 0:2 ml/100 g for Exp. 1
of A356 alloy from the literature (Table 1) [28], the
condition of the nucleation of pore is represented by the
dotted curve in Fig. 1 occurring at fsL 0:44 with
T 591 C [32].
Poirier et al. [26] maintained that PH2 should over-
come an additional eect of surface tension, not con-
sidered in Eq. (1), for the nucleation to take place:
PH2
4r=1
fsr d2 C0 =1
fsr 1
kH Kl 2

4r=1
fsr d2 P P x; 2
fsr
where is the fraction of solid at nucleation by lever rule
considering the eect of surface tension, r is the surface
tension, and d2 is the secondary dendrite arm spacing.
Taking the data of kH and r [33] from Table 2, C0 and d2
for Exps. 1, 2 and 3 from Table 1, the condition of the
nucleation of pore can be expressed in Fig. 1 by the solid
curves. In these cases as noted in Eq. (2), nucleation takes
place only when PH2 becomes larger than P x 4r=1

fsr d2 , corresponding to the alloy that solidies at higher
fsr of 0.57, 0.73 and 0.76, respectively.

2. Mechanism of nucleation of pore 2.2. Nucleation of pore by the mechanisms of microseg-

2.1. Nucleation of pore predicted by thermodynamics
Consider a solidifying A356 aluminum alloy with the
initial concentration of hydrogen in the liquid of C0 . fl
and fs are the weight fractions of liquid and solid, re-
spectively, while fl fs 1. At fl the concentration of
hydrogen in the liquid is Cl . kH is the equilibrium par-
tition ratio of hydrogen between the solid and liquid [30]
as indicated in Table 2. By lever rule, Cl C0 =1

fs 1
kH , the hydrogen gas pressure (PH2 ) in equilib-
rium with the liquid based on Sieverts law can be
calculated. Nucleation of pore occurs if PH2 is higher or
equal to the local interdendritic uid pressure (P x):
2 and ClL is the concentration of hydrogen in the liquid by
PH2 C0 =1
fsL 1
kH Kl  P P x;
where fsL
is the fraction of solid at nucleation by lever
rule, Kl is Sieverts constant of hydrogen in the liquid
[31], and P x 1:0 atm under the directional solidi-
regation and macrosegregation
Thermodynamics would predict that the nucleation
of pore occurs as late as fsL 0:44, or fsr at 0.570.76 as
indicated in Fig. 1, which is inconsistent with the mea-
sured fraction of solid at nucleation (fsN 0:36) [32,34]
soon after the commencement of coherency at fs 0:23
[35]. Obviously, an enrichment of hydrogen in the liquid
must exist. It hence needs to nd an index of supersat-
uration of liquid for nucleation
I ClN =ClL ; 3
where ClN
is the concentration of hydrogen in the liquid
at nucleation, associated with the fraction of solid at fsN ;
1
lever rule at the same fraction of solid. It has been
postulated that microsegregation of hydrogen during
solidication might be the cause that raises the con-
centration of hydrogen in the liquid at the solidliquid
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231 221

Table 2
Physical constants of A356.2 aluminum alloy
Physical constants Value used in calculation Ref.
Equilibrium partition ratio of hydrogen (kH ) 10680=T
2:008 [30]
Solidication range (DT ) 60 K
 2 [27]
Diusivity of hydrogen in liquid aluminum (Dl ) 3:8  10
6 exp

2315T 
m /s [48]
Diusivity of hydrogen in solid aluminum (Ds ) 1:1  10
5 exp
4922 T
m2 /s [48]
Hydrogen solubility of in liquid aluminum (Cl ) 10
2760=T 2:796 ml/100 g STP [30]
Hydrogen solubility of in solid aluminum (Cs ) 10
2080=T 0:788 ml/100 g STP [30]
Density of solid aluminum (qs 2.685  103 kg/m3 [50]
Solidication contraction (b) 0.057 [33]
Surface tension (r) 0.64 N/m [33]
Tortuosity (s) 1.6 [52]
Constant (C) 11.55 lm/s0:31 [52]
Viscosity (l) 4.7  10
3 Pa s [51]
a 0.77 [35]
fl x 0.02 [23]

Han and Viswanathan [37] reasoned that the enrich-

ment of hydrogen in the liquid occurs at the advancing
eutectic front, which transports the hydrogen from the
chill end to a distant location along the casting (imply-
ing macrosegregation) in the cellular growth as shown in
Fig. 2. This mechanism will be discussed later.

2.3. Nucleation of pore by the mechanism of initial

transient solute redistribution

For normal freezing, Tiller et al. [38] have shown that

the initial transient of solute in the liquid is a function of
plane-front growth velocity (R) and distance from the
chill end (x), under the condition of negligible solute
diusion in the solid and convection in the liquid.
However, the growth of equiaxed grains is generally
found in A356 aluminum alloy castings, while the cel-
lular growth in Fig. 2 is virtually not found [39,40]. The
Fig. 1. Equilibrium of gas pressure, as a function of fraction of solid by equiaxed grains show the features of spherical-like
Sieverts law, with P x for the condition of the nucleation of pore: (a)
the gas pressure of PH2 in Eq. (1) is used per lever rule (round-dotted
curve), (b) the gas pressure of PH2
4r=1
fs d2 in Eq. (2) is used
including the eect of surface tension (solid curves), and (c) an addi-
tional factor of Tiller hydrogen segregation per Eq. (4), in addition to
(b), is considered (broken curves).

interface. See et al. [29] also agreed that the concept of I

with the values 1.72.2 has to be chosen for the condi-
tion of nucleation, without explicitly explaining the
mechanism. Calculation of fsN without and with the ef-
fect of surface tension for the condition of nucleation
per Scheil equation [36] is performed in the Appendix A.
The calculated values of fsN are found 0.43 and 0.56,
respectively, for Exp. 1 in Table 1. Those quantities are
however close to fsL 0:44, or fsr 0:57 by lever rule Fig. 2. Schematic illustration (reproduced from [37]) showing the
formation of a wavelike distribution of pores during directional so-
(Fig. 1), respectively, thus the microsegregation by
lidication. Here, x0 through x5 denote the positions of the eutectic
Scheil equation gives no clue to the needed enrichment isotherm, and C denotes the hydrogen concentration in the liquid
of hydrogen in the liquid aluminum for the nucleation of ahead of the eutectic isotherm. ClN is the concentration of hydrogen in
pore. the liquid at nucleation.
222 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231

growth (e.g. Fig. 392 of [41]). When the grain advances 4.5
with a velocity of Vg , an initial transient of hydrogen
4

Grain growth velocity (V g ), mm/s

boundary layer with a concentration of hydrogen in the
liquid of Cl should occur at the solidliquid interface. 3.5 Exp.1
For the situation, it is proposed that Vg (assumed one- Exp.2
dimensional) and the radius of growing grain (rg ) are 3
Exp.3
substituted for R and x, respectively, in the equation of 2.5
[38]:
     2
1
kH Vg
Cl C0 1
exp
kH rg 1 ; 4
kH Dl 1.5
where Dl is the diusivity of hydrogen in the liquid. 1
Since fs 4prg3 =3=4prG 3
=3, or rg fs1=3 rG , hence
dfs =dt 3rg =rG drg =dt 3fs2=3 =rG Vg , where rG is
2 3 0.5
the radius of grain found from Table 1, and Vg drg =dt. 0
Thus 0 0.2 0.4 0.6 0.8 1
  Fraction of solid (f s )
1 dfs
Vg rG fs
2=3 : 5
3 dt Fig. 4. The deduced grain growth velocity with the fraction of solid for
Exps. 1, 2 and 3.
The weighted average Vg from fs 0 to 1 is denoted as
V
gfs 1 , and the weighted average Vg from fs 0 tofs is
denoted as V
gfs . considered not coherent with those experiments in terms
It is reasonable to assume that V
gfs 1 Vs , the solidus of the dierences in the deduced solidus velocity vs. the
velocity. In Exps. 1, 2 and 3 the calculated V
gfs 1 , based experimented solidus velocity. Hence in Exps. 1 and 2 of
on the respective rG from Table 1 with the particular Fig. 3, the functions of dfs =dt are suggested to be nor-
curve of dfs =dt from [32] in Fig. 3, are 0.784, 0.549 and malized such that when rG , fs and dfs =dt for each case
0.470 mm/s, associated with the measured Vs of 0.1, 0.25 are fed into Eq. (5), the calculated V
gfs 1 shall be equal to
and 0.48 mm/s, respectively (Table 1). It happens that the experimented Vs . Hence for given fs and known rG
the calculated V
gfs 1 ts well with the value of Vs for Exp. from experiments, the values of dfs =dt and Vg have to be
3. Then, with the known rG from Table 1 and the veri- modulated based on Eq. (5) for Exps. 1 and 2 as ex-
ed curve of dfs =dt associated with Exp. 3 in Fig. 3, it hibited in Fig. 3 and Fig. 4, respectively.
follows that for a given value of fs , Vg for Exp. 3 can be 1=3
The relation of rg fs rG , together with the
calculated based on Eq. (5) and plotted as shown in weighted average value of V
gfs as a function of fs re-
Fig. 4. However, a problem of normalization is involved placing the varying Vg in Eq. (5), are then fed into Eq.
for Exps. 1 and 2, since the thermal conditions leading (4). The calculated Cl (denoted as Tiller hydrogen seg-
to the relation between dfs =dt and fs from [32] are regation) can be obtained and plotted as a function of fs
in Fig. 5. The pore nucleates when Cl is larger or equal
to the hydrogen concentration of liquid in equilibrium
with P x 4r=1
fs d2 (denoted as Sievert hydrogen
Exp.1 concentration for nucleation) as shown in Fig. 5. When
2 Exp.2 this occurs, the values of fs at nucleation (fsN ) for Exps.
Exp.3 1, 2 and 3 are calculated as shown in Table 3. In addi-
tion, the range of solid fraction for the nucleation of
1.5
pore (f~sN ) with the values of Tiller hydrogen segregation
df s /dt

larger than the Sievert hydrogen concentration for nu-

1 cleation in Fig. 5 is recorded, and the corresponding
temperature range of nucleation (T~N ) is obtained (Table
3). The site of nucleation is suggested at the dendrite
0.5 root near the coherency, e.g. the contact of grain 4 and
grain 2 as shown in Fig. 6(a). On further cooling, the
0 nucleation of pores around the growing grain should
0 0.2 0.4 0.6 0.8 1 keep occurring until the pores are surrounded by the
grains, and the pores if more than one may coalesce into
Fig. 3. Variation of dfs =dt with the fraction of solid for A356 alloy. a single pore in the diusion cell as shown in Fig. 6(b).
The curve for Exp. 3 is adopted from [32], while the curves for Exps. 1 Assume that the dendritic grains are treated as spheres
and 2 are the normalized data. with four of them packed into a tetrahedron during
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231 223

1
Exp.1
0.8 Tiller hydrogen segregation
Sievert hydrogen concentration
0.6 f or nucleation

0.4
Hydrogen concentration, ml/100g

0.2

01

0.8 Exp.2

0.6

0.4

0.2

10

0.8 Exp.3

0.6
0.4
0.2
0
0 0.2
Fig. 5. Equilibrium of Tiller hydrogen segregation and Sievert hy-
drogen concentration for the condition of the nucleation of pore.
solidication. The radius of interstice of the tetrahedron
is 0.225 rG , and the radius of diusion cell (re ) is taken as
rG plus 0.225 rG . Hence the diusion cell develops by the
geometrical relation
re 1:225rG :
The radius of diusion cell is calculated in Table 3 using
the data of rG from Table 1. Within the diusion cell, the
surrounding hydrogen in the liquid and solid diuses
into the pore. Also the pore number density per unit
volume (np ) can be determined:
Fig. 6. Schematic representation showing (a) the nucleation of pores
might occur around the dendritic grains at the junction of grains, and
(b) the pores might coalesce into a single pore when encapsulated by
the surrounding grains. The pore and the surrounding liquid and solid
are developing into a hydrogen diusion cell.
0.4 0.6 0.8 1
4p 3
r  np 1: 7
3 e
2.4. Comparison of the present model with the mechanism
of macrosegregation
Han and Viswanathan [37] reasoned that the enrich-
ment of hydrogen in the liquid occurs as a result of
macrosegregation in the cellular growth of directional
6 solidication as shown in Fig. 2. According to the
mechanism as shown in Fig. 7 reproduced from [37], a
higher solidus velocity would generate a higher hydro-
gen concentration at the eutectic front, predicting a
higher porosity content. Comparing Exp. 3 with Exp. 1
from Table 1 shows however that, under virtually the

Table 3
The parameters predicted in this study for the mechanism of nucleation of pore in the solidifying A356 aluminum alloy
Exp. fsN f~sN fsNp ClN (ml/100 g) ClL (ml/100 g) I T~N (C) TNp (C) re (lm)
1 0.30 0.300.49 0.40 0.45 0.278 1.62 584606 595 919
2 0.30 0.300.51 0.41 0.49 0.248 1.97 581606 594 625
3 0.29 0.290.51 0.40 0.51 0.248 2.06 581607 594 551
Note: fsN is fraction of solid at nucleation, f~sN is the range of solid fraction for the nucleation, fsNp is the fraction of solid at peak nucleation rate,
ClN is the concentration of hydrogen in the liquid at nucleation at fsN , ClL is the concentration of hydrogen in the liquid by level rule at fsN , I is the
index of supersaturation of liquid for nucleation, T~N is the temperature range of nucleation, TNp is the temperature of peak nucleation rate, and re is
the radius of diusion cell.
224 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231
Fig. 7. Calculated total hydrogen concentration (reproduced from [37])
at the eutectic front for C0 0:1 ml/100 g. G 10 C/mm, and solidus
velocities at 0.01, 0.1 and 1 mm/s. Pore nucleation is neglected.
same C0 and G by the controlled experiments, Exp. 3
with higher Vs fails to record a correspondingly higher
porosity content. In addition, the phenomenon of hy-
drogen enrichment at the eutectic front leading to a
wavelike distribution of pores in Fig. 2 predicted by the
mechanism of macrosegregation was not observed in
aluminum castings [915,42]. Several causes could act
adversely to the mechanism of macrosegregation: (1)
The growth of equiaxed dendritic grains in aluminum
castings [3741] would impede the macrosegregation
and the solidliquid interface boundary layer of hydro-
gen in Fig. 2. (2) The solidication contraction will draw
the remote uid from the riser with much lower hy-
drogen concentration to the solidliquid interface. (3)
The nucleation and growth of pore will consume hy-
drogen locally to interrupt the macrosegregation. Based
on the experimental ndings and the rationale, the
mechanism was not adopted in the present study.
In our present mechanism of initial transient solute
redistribution, a directional cellular growth is not nee-
ded. The enriched hydrogen layer in the liquid at the
solidliquid interface associated with a growing grain
should not be disturbed by other independent grains.
When a pore nucleates, the surrounding hydrogen will
be consumed locally, however other smaller grains
continue to grow and build-up the Tiller hydrogen seg-
regation. It is interesting to nd that the dendritic grain
grows in a spherical-like equiaxial manner (e.g. Fig. 3
92 of [41]), and that the initial grain growth velocity at
lower fs is much more rapid than the remaining solidi-
cation periods (Fig. 4). This phenomenon induces the
initial transient high concentration of hydrogen solute
needed for pore nucleation as shown in Fig. 5. The
calculated temperature range of nucleation (T~N ) for
Exps. 13 is 581607 C (Table 3), which predicts well
with the experiments of 571615 C (Table 1). The
temperature with the maximum deviation of Tiller hy-
drogen segregation from Sievert hydrogen concentration
in Fig. 5 is dened as the temperature of peak nucleation
rate (TNp ). The calculated TNp of 594595 C (Table 3) is
associated with the fraction of solid at peak nucleation
rate (fsNp ) of 0.410.42, which is close to fsL of 0.44
calculated by lever rule in Fig. 1.
In Fig. 2, ClN is undetermined. This value can be
determined from Fig. 5, and the result is shown in Table
3. The hydrogen gas pressure in equilibrium with the
Sievert hydrogen concentration in Fig. 5 can be re-
plotted vs. the fraction of solid as shown in Fig. 1 by the
dashed curves, indicating that the pore is predicted to
nucleate at fsN of 0.30, 0.30 and 0.29 for Exps. 1, 2 and
3, respectively (Table 3). The predictions are against the
experiment at 0.36 [32,34], shortly after the coherency
of dendritic grains at fs 0:23 [35]. The earlier occur-
rence of nucleation at fsN 0:30 in this study compared
to the experiments might be due to the present simplied
model that assumes the spherical-like equiaxial grain
growth as a one-dimensional event in the initial tran-
sient segregation by Tiller et al.
The index of supersaturation required for nucleation
(I) is calculated per Eq. (3), and the result is shown in
Table 3. The predicted values of 1.62, 1.97 and 2.06,
corresponding to Vs of 0.1, 0.25 and 0.48 mm/s, respec-
tively, t well with the chosen quantities of 1.72.2 by
See et al. Our present mechanism indicates that a high
solidus velocity could eect a higher index of supersat-
uration that causes a large shift of fs for the nucleation
of pore as shown in Fig. 1. In addition, as Vs increases, I
increases, more nucleation events should occur around a
growing grain, resulting in a smaller re (Fig. 8); however
as Vs increases, rG should also decrease due to the faster
cooling. Hence re is correlated with rG as shown in Fig. 8,
and the relation is reasoned by Eq. (6). Fang and
Granger observed that np is only slightly aected by C0
[43]. This is reasoned that although more nuclei may
form around the growing grains due to the higher C0 ,
similar to the situation of higher I value due to the
900 900
800 800
700 700
600 600
rG , um
re , um
500 500
400 400
300 300
cell radius (re )
200 200
radius of grain size (rG)
100 100
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 8. Variations of re and rG with Vs (data from Table 2 [28]).
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231
Table 4
The parameters predicted in this study for the mechanism of growth of pore in the solidifying A356 aluminum alloy
Exp. Pg (atm) ts (s)
1 1.07 2.10
2 1.09 0.97
3 1.10 0.71
Note: Pg is the pressure inside the pore, ts is the relaxation time, Vp (%) is the volume percent of porosity, Vpmax is the maximum allowed volume
percent of porosity, and rp is the radius of pore.
higher Vs , the survival nuclei of pore are limited by the
diusion cell and the surrounding grains as shown in
Fig. 6, and Eqs. (6) and (7). It was also agreed in another
study that the high solidus velocity should increase the
hydrogen supersaturation but this is postulated that
there is less time for hydrogen to diuse into the pores
[44]. Furthermore, the present mechanism would not
predict that a higher solidus velocity should result in a
higher porosity content (Table 4 as detailed here), which
is compatible with the experiments in Table 1.
3. Mechanism of growth of pore
3.1. Growth of pore predicted by thermodynamics
Thermodynamics was also used for treating the hy-
drogen redistribution during the growth of porosity
[26,27]. Sigworth and Wang [27] formulated the fol-
lowing equation to predict the volume percent of po-
rosity formed:
1 1 T
Vp % C0
Csi q ;
w l PH 273
where Csi is the hydrogen concentration of initially so-
lidied solid, ql is the density of liquid aluminum, PH is
the pressure inside the gas pore when it rst forms, and
T is the absolute temperature. In Eq (8), 1
kH 1 is
assumed. However, the amount of porosity predicted
was too high as noted in [27], and the ignored diusion
of gas into the pore was considered by an undetermined
kinetic correction factor w.
3.2. Kinetic growth mechanisms of pore reported in the
literature
Han and Viswanathan [37] have not suggested a
mechanism for the diusion of hydrogen into the pore.
The kinetic growth of porosity as a diusion-controlled
process of redistributing hydrogen into the pore was
solved numerically by Atwood et al. [28,29]. Alterna-
tively, a semi-empirical solution was formulated [45],
asserting that the growth of porosity in the solidifying
A356 aluminum alloy could be described by Avramis
equation [46]. In the following the latter assertion is
analyzed, and the solution will be discussed with respect
225
Vp (%) Vpmax (%) rp (lm)
1.38 1.40 220
1.16 1.23 142
1.10 1.22 122
to the thermodynamic solution [26,27], and the numer-
ically solved kinetic mechanism.
3.3. Diusion-controlled growth of pores in the solidifying
aluminum alloy with two phases
3.3.1. Mathematical formulation for diusion of hydrogen
in the pseudo single-phase medium comprising liquid and
solid phases
The classic theory of diusion-controlled precipitation
in the single-phase medium was solved by Ham [47]. The
particles with a high concentration of solute are in equi-
librium with the matrix. The diusion problem is simpli-
ed as such no ux is transported across the cell boundary
(Fig. 9(a)). The diusion of hydrogen from the solidifying
matrix into the pore resembles in several aspects to the
problem described in Fig. 9(a). However, as shown in
Fig. 9(b) the present problem involves two phases, liquid
and solid, in the matrix. By the numerical method, the
growth of a pore was surrounded by the liquid with the
enriched concentration of hydrogen due to the rejection
from the growing solid. The present analytical model,
8 however, assumes that both the liquid and the solid are in
contact with the gas pore (Fig. 9(b)). The following sim-
plications are made in the analysis: (1) The temperature
of peak nucleation rate (TNp ), and the fraction of solid at
p
peak nucleation rate (fsNp ) at the time of tN , are taken as
the reference conditions for the start of the diusion-
controlled growth process. (2) The nucleation of pore
reduces the local hydrogen segregation that brings the
equilibrium to the system [37]. During the growth stage,
the liquid is assumed in thermodynamic equilibrium with
the solid at the solidliquid interface. Also, at the alloy
gas pore interface, both the liquid and the solid are in
equilibrium with the gas pore. (3) The concentration of
hydrogen in the liquid and solid, in excess of the hydrogen
solubilities in the liquid (Cl ) and the solid (Cs ), respec-
tively, will be diused into the pore.
Consider a local small mass consisting of the liquid
and solid with the hydrogen concentrations of Cl and Cs ,
respectively. The weighted average of hydrogen con-
centration in the mass (C
) can be expressed as
C
fl Cl fs Cs : 9
Referring to Fig. 9(b), dAl and dAs are the small areas of
liquid and solid in contact with the pore, respectively,
226 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231

 
~ oC

C 
J rp
D : 11
or rrp t

Before letting the diusion ux of hydrogen through

the physical liquid and solid by Eq. (10) be equal to that
through the pseudo single-phase medium by Eq. (11),
the term C  in Eq. (11) should be assessed. Comparing
Eqs. (9)(11) shows the following. At t 0; fs
0; C
Cl ; dAs 0; dAl =dA 1; D
~ Dl , hence C  C  .
l

~
While at t tf ; fs 1; C Cs ; dAl 0; dAs =dA 1; D
Ds , hence C  Cs . Thus C  in Eq. (11) is a variable
dependent on time or fs , however this should be avoided
for the mathematical purpose which requires a xed
boundary condition as shown later.
For the purpose of having a xed boundary as de-
scribed above, let the rst term on the right-hand side of
Eq. (10) be
oCl
Cl oCl
Cs
Dl  dl ; 12
or or
where dl is the diusivity of hydrogen associated with a
pseudo liquid aluminum having the pseudo hydrogen
solubility (Cs ) on the right-hand side of Eq. (12). If Dl
and dl are not functions of r, Eq. (12) is rewritten as
Cl
Cl
dl Dl : 13
Cl
Cs
In the equation, dl is a function of fs via the relation
Fig. 9. The geometries of: (a) solute diusion cell for the classic Cl C0 =1
fs 1
kH . The diusion of hydrogen in
problem of precipitation involving the diusion of solute in the single-
the liquid, described on the left-hand side of Eq. (12), is
phase medium, and (b) hydrogen diusion cell involving the diusion
of hydrogen in the solidifying aluminum alloy with the liquid and solid associated with a xed diusivity of Dl and solubility of
phases. Cl at the liquidpore interface. Mathematically, this is
equivalent to the diusion of hydrogen in the pseudo
and dA is the total area of contact. When t P tN p
, the ux liquid aluminum, with a xed hydrogen solubility of Cs
of hydrogen J rp at r rp t diusing into the pore at at the liquidpore interface, and a varying diusivity of
the interface should be dl . With dl and Cs replacing Dl and Cl , respectively, in
  Eq. (10) via Eq. (12), C  in Eq. (11) should also be
dAl oCl
Cl dAs oCs
Cs
J rp
Dl Ds ; substituted by Cs since both the solid and the pseudo
dA or dA or rrp t liquid possess the same solubility of Cs at the medium
10 pore interface. Eq. (10) can now be rewritten, assuming
dAl =dA fl and dAs =dA fs :
where Ds is the diusivity of hydrogen in the solid. The
 
dimension of the precipitate, the pore in the present oCl oCs
case, may be permitted to vary with time [47]. J rp
fl dl fs Ds
or or rrp t
A concept is proposed in which the weighted average 

of hydrogen concentration in the mass (C
) is diusing
D ~ oC : 14
into the pore through an imaginary pseudo single-phase or rrp t
medium comprising liquid and solid as a whole mathe-
matically but not physically. The pseudo single-phase The unknown D ~ is solved by dierentiating C
from
medium has a pseudo diusivity of D ~ . The hydrogen Eq. (9) with respect to r:
concentration in the pseudo single-phase medium at the
oC
ofl oCl
mediumpore interface (C  ) is in equilibrium with the C l 1
k H kH 1
kH fl  : 15
or or or
hydrogen concentration of gas pore (C
g ) as discussed
~ and C  are unknown. It follows that the
later. Both D Eq. (15) is further simplied by assuming that ofl =or 0
ux of hydrogen diusing through the single-phase for a random distribution of solid phase in the solidi-
medium is fying aluminum. Hence
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231 227

oCl 1 oC
3.3.2. Mass equations, boundary conditions and Hams

; 16 solution
or kH 1
kH fl or
Eq. (14) implies the following. The physical problem
oCs kH oC
of hydrogen diusion from the solidifying liquid and
: 17 solid phases into the gas pore can be mathematically
or kH 1
kH fl or
transformed into a problem in which hydrogen diuses
~ can be
By inserting Eqs. (16) and (17) into Eq. (14), D through an equivalent pseudo single-phase medium with
solved as the pseudo diusivity of D ~ . With the mathematics, the
1 quantity of hydrogen leaving the medium per unit time
~
D fl dl kH fs Ds : 18
kH 1
kH fl can be expressed both as the volume of the cell times the
rate of change of C
in the cell, and as the ux into the
The relation holds for fs P fsL . When fs < fsL ; D ~ 0,
pore times the surface area of the pore. For the radius of
meaning that diusion has not commenced. Further- pore (rp t), the relation is
more, Dl and Ds are measured as [48]
  4pre3 dC


6 2315 J rp 4prp2 ; 22
Dl 3:8  10 exp
m2 =s; 19 3 dt
T
  where Jrp is the ux at r rp of the mediumpore
4922
Ds 1:1  10
5 exp
m2 =s: 20 interface in the pseudo single-phase medium. The mass
T diusion around the pore is governed by Ficks second
A plot of D ~ as a function of fs at the eutectic tem- law in a spherical coordinate system:
perature of 565 C [32] (838K) is shown in Fig. 10. When 
oC
~ 1 o

~ is low since the excess
fs is only slightly larger than fsL , D 2 oC
D 2 r : 23
concentration of hydrogen in the liquid is still low; On or r or or
further solidication, the value of D ~ would increase to a
maximum, and then decreases to Ds at the end of so- Referring to Fig. 9(b), the boundary conditions are
lidication to reect the increasing fraction of hydrogen C
Cs at r rp t cf: Eq: 2:2 in 47; 24
being diusing through the solid with a low diusivity. It
is desired to have a weighted average D ~ (denoted as D ~ ave ) oC

to replace the varying D ~ as a function of fs in Fig. 10. A 0 at r re cf: Eq: 3:1 in 47: 25
or
solution of the weighted average diusivity for the
varying diusivity as a function of time has been solved The above mass equations and the boundary condi-
[49]. Since fs t=tf can be assumed [32], D ~ ave can be tions infer that Hams solution [47] can be invoked for
derived: solving the present diusion problem directly. Thus, the
Z 1 concentration of hydrogen diused into the gas pore
D~ ave 1
f L ~ dfs :
D 21 (Cp , ml/100 g), compared to the initial concentration of
s
fsL hydrogen minus the hydrogen solubility in the matrix, is
the transformed fraction of hydrogen (f t) diused into
3.E-07 the pore at the end of solidication:
  p n 
Cp tf
tN
Pseudo diffusivity (D ), m /s

f t 1
exp
; 26
2

C0
Cs ts
2.E-07 where tf is the local solidication time, and ts is the re-
laxation time as will be substantiated below. n will de-
pend on the precipitate shape [47]. The relation of
p
tN fsNp tf holds when fsNp P fsL ; if fsNp < fsL the relation
1.E-07 p
of tN fsL tf should be used instead based on Fig. 10.
However fsNp fsL in this study as stated, the relation of
fs L p
tN fsL tf will be approximated in the following.
Ds

0.E+00 3.3.3. Relaxation time and pseudo precipitate of gas pore

0 0.2 0.4 0.6 0.8 1 The relaxation time ts in Eq. (26) can also be obtained
Fraction of solid (f s ) from Hams solution [47]

Fig. 10. Variation of pseudo diusivity (D ~ ) based on Eq. (18) with the 2=3
C
g1=3 re2 C0
fraction of solid (fs ) at the eutectic temperature of 565 C for A356 ts ; 27
alloy. ~ ave C0
C 
3D s
228 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231

where C
g (ml/100 g) is the concentration of hydrogen in bilities in the liquid and solid are given by the equa-
the pseudo precipitate of gas pore which should be ex- tions [30]
plained. In the classic solution of the diusion-limited 2760
growth of precipitate, the second phase containing the log Cl
2:796 ml=100g STP; 32
T
solute and the solvent is precipitated from the matrix that
also contains the solute and solvent of dierent compo- 2080
log Cs
0:788 ml=100g STP; 33
sitions. By Eq. (14), it has been shown that the present T
diusion problem can be simplied as one of diusing where T is the temperature in K. If K~ave deduced from Eq.
hydrogen through a pseudo single-phase medium. Like- (30) and Pg from Section 3.3.5 are inserted in Eq. (28), the
wise for the mathematical consistency inferred by Ham concentration of hydrogen in the pseudo precipitate of
[47], the solved C
g in Eq. (27) obviously should be in unit gas pore (C
g ) can be calculated. The calculated C
g can then
of ml/100 g identical to C0 and Cs in the equation. be fed into Eq. (27) to solve the relaxation time (ts ), which
The value of C
g in the pseudo precipitate of gas pore is then used to calculate Cp in Eq. (26). Cp will be employed
to be fed into Eq. (27) must be elucidated. Since the in Section 3.3.4 for the quantitative calculation of the
pseudo precipitate of gas pore is a single-phase medium, volume percent of porosity.
C
g must be in equilibrium with and dependent on the gas
pressure inside the pore (Pg , atm) by Sieverts law. Hence
p 3.3.4. Prediction of volume percent of porosity and pore
C
g K~g Pg ; 28 size
The volume of porosity (Vp ) is assumed to be deter-
where K~g is a proportional constant (ml/100 g/atm
1=2 ) mined by the ideal gas law
dened as Sieverts constant of hydrogen in the pseudo
precipitate of gas pore, and Pg is the gas pressure inside PSTP T V^cell  ql
the pore. Similar to Eq. (9), the pseudo precipitate Vp Cp ; 34
Pg TSTP 100
should also comprise of the liquid and solid phases
mathematically, hence C
g in the pseudo precipitate can where V^cell is the volume of the hydrogen diusion cell, ql
be expressed as is the density of liquid, PSTP is the atmospheric pressure
under the standard condition, T is the formation tem-
C
g fl C g fs C g ; 29
l s perature of gas pore, TSTP 273 K. The volume percent
where Clg and Csg are the imaginary concentration of of porosity can therefore be represented as
hydrogen in the liquid and solid of the pseudo precipi-
100  VP
tate of gas pore in ml/100 g, respectively. At the alloy VP % ; 35
gas pore interface, the hydrogen solubilities of Cl and Cs VP V^l V^s
in the liquid and solid are assumed in thermodynamic
where V^l and V^s are the volume of liquid and solid, re-
equilibrium with the concentration of hydrogen in the
spectively. Furthermore, V^l fl V^cell , V^s 1
fl V^cell ,
physical gas pore, or with the equivalent pseudo pre-
and V^cell =Vp is much larger than 1, Eq. (35) can therefore
cipitate of gas pore having the hydrogen concentrations p be rewritten as
of Clg and C sg , respectively. Clg and Csg are equal to Kl Pg
p
and Ks Pg , respectively. Kl and Ks are Sieverts con-  
1
fl V^cell 1
fl V^cell Vp
stants of hydrogen in the liquid and solid aluminum in VP % 100 1  100 :
VP ^
Vcell
ml/100 g/atm
1=2 , respectively. Equating Eqs. (28) and
(29), and since Ks kH Kl [31], we can obtain the pseudo 36
Sieverts constant of hydrogen in the pseudo precipitate The volume percent of porosity at the end of solidi-
of the gas pore cation can be obtained by inserting Eqs. (26) and (34)
into Eq. (36) as
K~g fl Kl 1
kH kH Kl : 30    n 
1
fsL tf
Following the same technique in Eq. (21), the weighted- VP % C0
Cs 1
exp

ts
average of K~g (denoted as K~ave , ml/100 g/atm
1=2 ) from
Eq. (30) could be deduced for the solidication process. 1 T
 ql ; 37
Sieverts constant of hydrogen in the liquid or solid Pg 273
aluminum can be expressed as [31] where T could be rationally adopted as the eutectic
Kl or Ks Cl or Cs  10 0:017CMg
0:0269CCu
0:0119CSi
; temperature [26,27,35]. Since 4p=3rp3  np Vp %, this
relation can be combined with Eq. (7) to derive the ra-
31
dius of pore as
where CMg , CCu and CSi are the concentrations of Mg,
1=3
Cu and Si in wt%, respectively. The hydrogen solu- rp re Vp % : 38
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231
3.3.5. The gas pressure inside the pore
It is believed that the interdendritic uid will en-
counter a resistance while owing from the riser to feed
the solidication contraction at the solidliquid inter-
face. Applying Darcys law, the local interdendritic uid
pressure becomes [23]
DT 0:38 Vs1:62 a is found signicantly inuenced by the radius of pore
P x P0 ql gh
24ps3 lb0 ln ; 39
C 2 G0:38 fl x
where P0 is the atmospheric pressure, ql gh is the riser
hydrostatic pressure, ql is the density of liquid as cal-
culated from the density of solid [50], h is the riser
height, l is the viscosity of interdendritic liquid [51], b0 is
equal to b=1
b, b is the solidication contraction
[33], a is the fraction of liquid at the coherency [35], and
fl x is the fraction of liquid [23]. DT is the solidication
range [27], s is the tortuosity [52], C is the alloy constant
[52]. At the end of solidication, the nucleated pore is
assumed to reside in the intergranular space [53], hence
the gas pressure inside the pore (Pg ) is calculated by the
relation
Pg P x 2r=rp : 40
Comparing Eqs. (37), (38) and (40) indicates that none
of Vp %, rp or Pg can be determined a prior without
knowing the others, hence the quantities are solved by
the iteration method.
3.4. Physical mode, reaction kinetics, thermal parameters
and microstructure
In this analysis as presented in Fig. 9(b), the pore is
assumed in contact with the liquid and solid during
solidication. The metallography apparently supports
the phenomenon (e.g. Fig. 3 in [11]). In spite of the low
value of Ds , the diusion of hydrogen from the solid into
the pore should play a role of decreasing D ~ at the end of
solidication as shown in Fig. 10 per Eq. (18). This
equation has incorporated the factor of the commence-
ment of hydrogen diusion and the diusion of solid
and liquid in the pseudo single-phase medium. The
justication of the assumption of ofl =or 0 in Eq. (15)
should be examined. The metallography (e.g. Fig. 3 of
[11]) suggests that this assumption could apply for a
randomly distributed solid phase in the solidifying alu-
minum alloy.
The reaction kinetics is determined by the relaxation
time relative to the local solidication time per Eq. (26).
The relaxation time based on Eq. (27) and n 0:56 [45]
can be calculated as shown in Table 4. A longer local
solidication time would allow more hydrogen to diuse
into the pore to achieve a thermodynamic equilibrium,
whereas larger values of ts would mean a slow kinetics of
hydrogen diusion. The relaxation time in Eq. (27) is
controlled by several parameters. A smaller radius of
hydrogen diusion cell (re ) would imply a short diusion
229
path, which decreases the value of ts and accelerates the
accomplishment of hydrogen diusion. In Eq. (27) if Cs

1=3
is much smaller than C0 , ts is then proportional to C0 .
As stated increasing the initial hydrogen concentration
only slightly changes the np [43], however C0 can aect ts
and the growth kinetics. The volume percent of porosity
[43]. This is evidently suggested by Eq. (38), and the
predicted rp is exhibited in Table 4.
The data in Tables 1 and 2 are used in Eq. (39) to
calculate the eect of the thermal gradient and solidus
velocity on the local interdendritic uid pressure. Under
the present directional solidication, with high values of
essentially constant thermal gradient and low values of
solidus velocity (Table 1), those parameters should not
exert a signicant inuence on the interdendritic uid
pressure, the calculated Pg in Table 4, or the porosity
formed. The mild variation of Pg in Table 4 is due to the
variation of rp per Eq. (40). It is agreed that the eect of
interdendritic uid pressure is insignicant under the
directional solidication [27]. However, the typical A356
aluminum castings are solidied under the thermal
gradient less than 0.4 C/mm, and the solidus velocity
larger than 2.0 mm/s [54]. Recently, the factor of inter-
dendritic uid pressure was dierentiated from the eect
of kinetic hydrogen redistribution, hence it could still
play a role in the alloy solidied non-directionally [45].
3.5. Comparison of the present mechanism with the
thermodynamic mechanism
By comparing Eq. (37) with Eq. (8), the discrepancy
of the predicted volume percent of porosity based on the
present kinetic mechanism and the thermodynamics
becomes clear. Since PH and Csi in Eq. (8) are close in
values to Pg and Cs , respectively, in Eq. (37), the major
dierence lies in the quantity of Cp , the hydrogen con-
centration diused into the gas pore in Eq. (26). In
thermodynamics, the equilibrium redistribution of hy-
drogen between the solid and liquid has reached, while
the kinetic hydrogen diusion to reach the equilibrium is
corrected by a factor in Eq. (8) [27]. In our present ki-
netic model, the correction factor w could be explicitly
expressed as
  n 
1 1
fsL tf
1
exp
: 41
w ts
As the local solidication time approaches innite
(w 1:0), the maximum allowed volume percent of
porosity predicted by the kinetic model is
1 T
Vpmax % C0
Cs ql : 42
Pg 273
These values for Exps. 1, 2 and 3 in Table 1 are calcu-
lated as 1.40%, 1.23% and 1.22%, respectively (Table 4).
Comparison of those values with the measured volume
230
percent of porosity in Table 1 infers that the hydrogen
redistributions in the experiments are approaching the
thermodynamic equilibrium.
3.6. Comparison of the present mechanism with the
numerical solution
For predicting the porosity formation in the experi-
ments in Table 1, See et al. [29] used three computa-
tional schemes employing the nite dierence technique;
however, only the DE model was more successful or
relevant, and the result is also shown in Table 1. In DE
model, the nucleation of pore is not predicted, hence the
growth of pore has to start from the stipulated condi-
tions. In our present analytical solution based on Eq.
(37), both the nucleation and growth process are mod-
eled. The initial concentration of hydrogen in the liquid,
the thermal variables, the microstructures, and the
physical properties of A356, from Table 1 and Table 2,
together with the calculated parameters for the nucle-
ation of pore from Table 3 were used for all the pre-
dicted results as shown in Table 4. The predicted volume
percent of porosity at the eutectic temperature of 565 C
[32] (838 K) could be evaluated by incorporating the
data as stated above and the regression value of n 0:56
for A356 aluminum alloy [45] into Eq. (37). Table 4
shows that the value of predicted volume percent of
porosity is slightly inuenced by Pg that is in turn in-
uenced by rp per Eq. (40), where rp as shown in the
same table is calculated per Eq. (38). The predicted
volume percent of porosity compares reasonably well
with the experiment as shown in Fig. 11. In general, it
also agrees with the mechanism of diusion-controlled
Volume percent of porosity (V p (%))
3.5
3 DE model
2.5
2 phases during solidication could be treated as a classic
1.5
1 general Avramis equation. The derived analytical so-
0.5
0 drogen in the liquid, the thermal parameters associated
0 0.2
Solidus velocity (V s ), mm/s
Fig. 11. The calculated volume percent of porosity by the present
model and the DE model [29], for predicting the experimental data in
Table 2 [28].
K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231
growth of porosity solved by the numerical method with
DE model as shown in the same gure.
4. Concluding remarks
In this study, an analytical solution is developed for
modeling the mechanism of nucleation during the solidi-
cation of A356 aluminum alloy. The initial transient
hydrogen solute segregation in front of the fast growing
dendritic grain of the alloy is reasoned as the source of
supersaturation needed for the early nucleation of pore
that occurs soon after the coherency of grains. The phe-
nomena are otherwise unexplainable based on lever rule
or the microsegregation of hydrogen solute by Scheil
equation. From the present mechanism, the fraction of
solid at nucleation, the temperature range of nucleation,
and the radius of hydrogen diusion cell, etc. can be
predicted. The analytical solution can also calculate the
hydrogen concentration of liquid at nucleation and the
index of supersaturation of liquid needed for the nucle-
ation of pore in A356 aluminum alloy. Increasing the
solidus velocity is reasoned to aect the transient super-
saturation for nucleation and to shift the fraction of solid
for the nucleation of pore. The derived mechanism ex-
plains the experimental ndings [28], however the mech-
anism does not involve the macrosegregation of hydrogen
solute transport as discussed in the text.
This study also aims to study the growth of porosity
during the solidication of A356 aluminum alloy. It is
intended to prove a previous assertion of a semi-em-
pirical formulation, without a rigorous scientic basis,
that the growth of pore of the alloy is the one of a more
general phase transformation in materials as described
by Avramis equation [46]. In the derivation, the sim-
plication of assumptions has to be made to allow the
experimental mathematical analysis possible. The calculated param-
eters based on the nucleation mechanism are directly
present model
employed as the reference conditions for the beginning
of the diusion-controlled growth of hydrogen into the
pore. With the present analytical derivation, it is argued
that the diusion problem involving the liquid and solid
problem of precipitation in the single-phase matrix
treated by Ham [47]. It has been proved by Christian
[55] that Hams theory is compatible with the more
lution is compared with the thermodynamic solution to
explain the kinetic factor between the two solutions.
Employing the data of the initial concentration of hy-
0.4 with the cooling condition, and the microstructure of the
casings, our present analytical solution can predict the
features of the porosity including the volume percent of
porosity formed and the radius of pore. The predicted
porosity content compares reasonably well with the
 K.-D. Li, E. Chang / Acta Materialia 52 (2004) 219231 231

experiments by Atwood et al. [28] and the numerical [9] Kim JM, Kwon HW, Loper Jr CR. AFS Trans 1996;104:743.
solution with DE model by See et al. [29]. In DE model [10] Fuoco R, Correa ER, Goldenstein H. AFS Trans 1996;104:1151.
[11] Fuoco R, Goldenstein H, Gruzleski JE. AFS Trans 1994;102:297.
for the numerical solution, some dierent assumptions [12] Fuoco R, Correa ER, Correa AVO. AFS Trans 1995;103:379.
were made. Both the present and the numerical ap- [13] Fuoco R, Goldenstein H. AFS Trans 1997;105:883.
proaches appear to solve the kinetic diusion-controlled [14] Kim JM, Kwon HW, Kim DG, Loper Jr CR. AFS Trans
growth of hydrogen pores in the aluminum alloy. The 1997;105:825.
numerical solution may have the leeway in mathematics. [15] Fuoco R, Correa ER, Bastos MA. AFS Trans 1998;106:401.
[16] Entwistle RA, Gruzleski JE, Thomas PM. Met Soc 1979:345.
Our present analytical solution however has the attri- [17] Davies VL. AFS Cast Met Res J 1975;11:33.
bute of understanding the functional relationships in- [18] Davies VL. Met Soc 1979:357.
volved in the mechanism of porosity formation of the [19] Piwanka TS, Flemings MC. Trans Met Soc AIME 1966;236:1157.
solidifying A356 aluminum. The role of interdendritic [20] Niyama E, Uchida T, Morikawa M, Saito S. 49th International
uid ow governed by Darcys law is found not im- Foundry Congress, Chicago, April 1982, p. 1.
[21] Argo D, Gruzleski JE. AFS Trans 1988;96:65.
portant under the directional solidication, however it [22] Combeau H, Carpentier D, Lacaze J, Lesoult G. Mater Sci Eng
could still play an important role when the thermal 1993;A173:155.
parameters become unfavorable. [23] Kao ST, Chang E. Cast Met 1995;7(4):219.
[24] Anson JP, Stucky M, Gruzleski JE. AFS Trans 2000;108:419.
[25] Chen Q, Ravindran C. AFS Trans 2000;108:297.
[26] Poirier DR, Yeum K, Maples AL. Metall Trans A 1987;18A:1979.
Appendix A [27] Sigworth GK, Wang C. Metall Trans B 1993;24B:349.
[28] Atwood RC, Sridhar S, Zhang W, Lee PD. Acta Mater
Instead of using the equilibrium hydrogen concen- 2000;48:405.
tration in the liquid for calculation of the condition of [29] See D, Atwood RC, Lee PD. J Mater Sci 2001;36:3423.
nucleation, the hydrogen concentration as a result of [30] Ransley CE, Neufeld H. J Inst Met 1948;74:599.
[31] Alscan model F Users Manual Ver. 1.2, Bomem Inc., 1994.
solute segregation by Scheil equation is employed. In the [32] Backerud L, Chai G, Tamminen. Solidication characteristics of
equation, no hydrogen diusion in the solid and a aluminum alloys. In: Foundry alloys, vol. 2. The American
complete hydrogen diusion in the liquid are assumed. Foundrymens Society; 1990. p. 139.
For case 1: the segregated hydrogen gas pressure is in [33] Emadi D, Gruzleski JE, Toguri JM. Metall Trans B
equilibrium with P x 1:0 atm without the eect of 1993;24B:1055.
[34] Atwood RC, Sridhar S, Lee PD. Scr Mater 1999;41(12):1255.
surface tension. The condition of the nucleation of pore [35] Arnberg L, Backerud L, Chai G. Solidication characteristics of
should be aluminum alloys. In: Dendrite coherency, vol. 3. American
k
1 2 Foundrymens Society; 1996. p. 95.
C0 1
fsN H =Kl  1:0: A:1 [36] Flemings MC. Solidication processing: international edition.
New York: McGraw-Hill; 1974.
In Eq. (A.1), fsN
is to be solved, however kH (Table 2)
[37] Han Q, Viswanathan S. Metall Mater Trans A 2002;33A:2067.
and Kl (Eq. (31)) are dependent on the unknown tem- [38] Tiller WA, Jackson KA, Rutter JW, Chalmers B. Acta Metall
perature. By the iteration method for C0 0:20 ml/100 g 1953;1:428.
from Exp. 1 of Table 1, the followings are obtained: [39] Chai G, Backerud L, Rolland T, Arnberg L. Metall Mater Trans
T 592 C, kH 0:060, Kl 0:34 and fsN 0:43. A 1995;26A:965.
[40] Chai G, Backerud L, Arnberg L. Z Metallk 1995;86:54.
For case 2: the segregated hydrogen gas pressure is in
[41] Arnberg L, Backerud L, Chai G. Solidication characteristics of
equilibrium with P x 1:0 atm with the eect of sur- aluminum alloys. In: Dendrite coherency, vol. 3. American
face tension. The condition of the nucleation of pore Foundrymens Society; 1996. p. 101.
should be [42] Huang H, Berry JT. AFS Trans 1993;101:669.
[43] Fang QT, Granger DA. AFS Trans 1989;97:989.
C0 1
fsN kH
1 =Kl 2
4r=1
fsN d2 1:0: A:2 [44] Lee PD, Hunt JD. Scr Mater 1997;36(4):399.
[45] Li KD, Chang E. Mater Trans JIM 2002;43(7):1711.
Similarly, for C0 0:20 ml/100 g from Exp. 1 of Table 1, [46] Avrami M. J Chem Phys 1930;7:1103.
the followings are obtained: T 573 C, kH 0:063, [47] Ham FS. J Phys Chem Solids 1958;6:335.
Kl 0:29 and fsN 0:56. [48] Eichenauer W, Markopoulos JZ. Z Metallk 1974;65(10):649.
[49] Shewmon P. Diusion in solids. second ed. The Minerals, Metals
& Materials Society; 1989. p. 37.
[50] Hatch JE, Aluminum properties and physical metallurgy. Amer-
References ican Society for Metals; 1984.
[51] Procast Build-in Database, The Professional Casting Simulation
[1] Chvorinov N. Giesserei 1940;27:177. System UES Inc., 1988.
[2] Caine JB. AFS Trans 1949;57:66. [52] Lee YW. Ph.D. Thesis. National Cheng Kung University, ROC, 1992.
[3] Adams CM, Taylor HF. AFS Trans 1953;61:686. [53] Kubo K, Pehlke RD. Metall Trans B 1985;16B:359.
[4] Bishop HF, Myskowski ET, Pellini WS. AFS Trans 1955;63:271. [54] Lee YW, Chang E, Chieu CF. Metall Trans B 1990;21B:715.
[5] Pellini WS. AFS Trans 1953;61:61. [55] Christian JW. The theory of transformations in metals and alloys.
[6] Taylor HF, Fleming MC, Piwonka TS. Foundry 1960:217. In: An advanced textbook in physical metallurgy part 1: equilib-
[7] Johnson WH, Kura JG. AFS Trans 1959;67:532. rium and general kinetic theory. second ed. Oxford: Pergamon
[8] Kuo YS, Chang E, Jeng RR, Ho HH. AFS Trans 1990;98:801. Press; 1975 [Chapters 1 and 12].