JMEPEG (2005) 14:604-609

DOI: 10.1361/105994905X66015

ASM International
1059-9495/$19.00

Casting/Mold Thermal Contact Heat Transfer during

Solidification of Al-Cu-Si Alloy (LM 21) Plates in Thick
and Thin Molds
K. Narayan Prabhu, Bheemappa Chowdary, and N. Venkataraman
(Submitted June 16, 2003; in revised form August 4, 2005)
Heat flow at the casting/mold interface was assessed and studied during solidification of Al-Cu-Si (LM 21)
alloy in preheated cast iron molds of two different thicknesses, coated with graphite and alumina based
dressings. The casting and the mold were instrumented with thermocouples connected to a computer
controlled temperature data acquisition system. The thermal history at nodal locations in the mold and
casting obtained during experimentation was used to estimate the heat flux by solving the one-dimensional
inverse heat conduction problem. The cooling rate and solidification time were measured using the computer-aided cooling curve analysis data. The estimated heat flux transients showed a peak due to the
formation of a stable solid shell, which has a higher thermal conductivity compared with the liquid metal
in contact with the mold wall prior to the occurrence of the peak. The high values of heat flux transients
obtained with thin molds were attributed to mold distortion due to thermal stresses. For thin molds,
assumption of Newtonian heating yielded reliable interfacial heat transfer coefficients as compared with
one-dimensional inverse modeling. The time of occurrence of peak heat flux increased with a decrease in
the mold wall thickness and increase in the casting thickness.

Keywords

casting/mold interface, heat conduction, inverse

modeling, Newtonian heating, solidification

1. Introduction
The success of simulation-based process design of castings
to predict accurately the thermal history and to locate hot spots
inside the casting depends to a large extent on a reliable database on the heat transfer boundary conditions specified at the
casting/mold interface (Ref 1-4). Further, when the metal and
the mold have good rates of conductance, the boundary between the two becomes the region of dominant resistance
(Ref 5).
A perfect contact between the metal and the mold surfaces
may not be realized in actual practice as the surface irregularities of the solidifying skin results in irregular contacts to be
established between the die-wall and the skin. The degree of
thermal resistance at the interface is a function of actual contact
area, thermal and physical properties of the materials in contact, and interstitial fluid present in the voids formed by the two
contacting surfaces (Ref 6, 7). The thermal resistance to heat
transfer results in a temperature drop at the interface. The heat
transfer at the interface can be characterized either by interfacial heat flux q or by an interfacial heat transfer coefficient h,
defined as the ratio of the interfacial heat flux to the temperaK. Narayan Prabhu, Department of Metallurgical & Materials Engineering, National Institute of Technology Karnataka, Surathkal, P.O.
Srinivasnagar 575 025, India; Bheemappa Chowdary, Formerly
Graduate Student, NITK, Surathkal, and now with M/s Chowgule
Industries Limited, Panaji, Goa, India; and N. Venkataraman, Professor (Retd.), National Institute of Technology Karnataka, Surathkal,
India. Contact e-mail: prabhukn_2002@yahoo.co.in.

604Volume 14(5) October 2005

ture drop at the interface. Mathematically h is expressed as:

h = q/T.
When conditions are favorable, a skin of solidifying casting
may physically separate from the mold wall resulting in a gap
of finite thickness. The mold configuration and Biot number
have been shown to be important parameters affecting the formation of the gap (Ref 8). Once the air gap forms, the heat
transfer across the interface drops rapidly. Conduction is the
predominant mode of heat transfer through the gap at lower
temperatures. While the radiation heat transfer depends on the
surface temperatures and emissivities, conduction heat transfer
depends on the thermal conductivity of the gas in the gap and
the air gap size as well. The heat transfer coefficients due to
radiation (hr) and conduction (hc) are expressed as:
hr =

2
Tc + Tm
T 2c + T m
1
1
+
1
c m

hc =

kg

Tc and Tm are the casting surface and mold surface temperatures, respectively; c and r are the emissivities of the casting
and mold surfaces; and kg and are the thermal conductivity of
the gas in the gap and the width of the gas gap, respectively.
Complexities of material behavior, metallostatic pressure, and
mold geometry make the calculation of air gap formation very
difficult. Further, the air gap does not start simultaneously at all
points in the casting and the magnitude of the air gap continuously varies with time (Ref 9, 10). Gravity die-casting, pressure
die-casting, investment casting, continuous casting, and twinroll casting are some of the processes where the product quality

Journal of Materials Engineering and Performance

Fig. 1 Schematic sketch of the experimental setup

is affected by interfacial heat transfer conditions (Ref 11-14).

The quantification of heat transfer across the interface depends
on the knowledge of many factors like interface temperatures,
thermal properties of the gas in the gap, surface texture of the
mold surface, coating material and composition, etc. (Ref 1517). Modeling of the casting/mold interfacial heat transfer is,
thus, one of the critical problems in numerical simulation of
casting solidification.
In the present work, thermal contact heat transfer at the
interface between plate type castings of Al-Cu-Si alloy (LM
21), solidifying in thick and thin cast iron molds, was assessed
and investigated.

2. Experimental Method
Cast iron mold halves of thicknesses 10 and 38 mm were
fabricated to cast aluminum alloy plates of dimensions 100
70 12 mm and 100 70 22 mm. Figure 1 shows a schematic diagram of the experimental setup. Two K-type thermocouples (2 and 3) of 0.45 mm diameter were introduced at
different nodal locations inside the mold. For the 10 mm mold,
thermocouples 2 and 3 were located at 2 and 6 mm from the
mold surface in contact with the casting. The corresponding
locations for the 38 mm mold were at 2 and 26 mm, respectively. The geometric center of the casting was also instrumented with a similar thermocouple (1) to monitor the solidification of the casting. The thermocouples were connected by
means of compensating cables to a temperature data logger. All
the experiments were carried out with a coating on the inner
surface of the mold. The mold halves were heated to about
180 C and coating material was sprayed on their surface to a
thickness of 100 m using a spray gun. Alumina and graphite
based die coats were used for this purpose. Table 1 gives the

Journal of Materials Engineering and Performance

Table 1

Composition of mold dressings

Composition, wt.%

Coating
type

Alumina

Graphite

Sodium
silicate

Water

17

16

11
4

72
80

1
2

composition of the coating materials used. The molds were

preheated to 100 C prior to pouring of castings.
LM 21 alloy ingots were melted in an electric resistance
furnace. Table 2 gives the composition of the alloy. The pouring temperature was maintained at around 700 C. The data
logger was activated a few seconds before pouring. After experimentation, the thermocouple signals stored in the temperature data logger were transferred to a computer by an off-line
procedure. Table 3 gives the thermophysical properties of the
alloy and the cast iron mold material.
Becks nonlinear estimation technique (Ref 18) was used to
estimate the mold surface heat flux from knowledge of measured thermal history at nodal locations inside the mold. The
one dimensional heat conduction equation:

T
T

k
= cp
x
x
t

(Eq 1)

was solved inversely subject to the following boundary and

initial conditions.
T 2, t = Yt
T 3, t = Bt
T x, 0 = Tix

Volume 14(5) October 2005605

Table 2
(LM21)

Composition (wt.%) of the Al-Cu-Si alloy

Si

Cu

Ni

Zn

Fe

Mg

Ca

Na

Ti

Al

5.6

3.07

0.31

2.02

0.47

0.47

0.0028

0.0011

0.02

balance

Table 3 Thermophysical properties of the mold material

and casting alloy

Material
Mold:
Cast iron
Casting:
LM21 alloy
Liquid
Solid

Density,
kg/m3

Thermal
conductivity,
W/m K

Specific heat,
J/kg K

7200

72

502

2560
2690

90
194

1254
1164

To find the heat flux at the casting/mold interface, the following function based on a least squares analysis was minimized.

Fig. 2 Casting and mold thermal history during solidification of 12

mm thick casting in a 38 mm thick graphite coated mold

I=mr

Fq =

+i

Y+i2

(Eq 2)

i=1

The variable r is the number of future time temperatures +1,

and m is the /t. and t are the time steps for heat flux
and temperature, respectively. Yn+i and Tn+i are the measured
and calculated temperatures, respectively, at the temperature
node (TC1) located in the mold near to the casting/mold interface. The future temperatures are the calculated temperatures at
time steps greater than the present time steps estimated using
the measured temperatures at location TC3 as the known
boundary condition. The unknown heat flux q(0,t) was approximated by some arbitrary value in the initial time step.
Minimizing Eq 2 with respect to q, by setting the partial
derivative to zero, the correction term for heat flux was obtained as:
I

n+i

1
qM+1

l1
Tn+i
l1
i

i=1

(Eq 3)

2
l1
i

i=1

where qM+1l qM+1l qM+1l1.

The procedure was then repeated for a new heat flux value.
The iteration was continued until:
l
qM+1
l1
qM+1

0.005

(Eq 4)

the final iterated value of q was used as the initial heat flux for
estimating the heat flux for the next time step. The calculation
of the heat flux was continued for the desired time period.
The partial derivative in Eq 3 is called the sensitivity coef-

606Volume 14(5) October 2005

ficient and is a measure of the change in the estimated temperatures with a small change in the boundary condition.
It was calculated by using Eq 4:
l1
i =

l1
l1
Tn+iqM+1

1 + Tn+iqM+1
l1
qM+1

(Eq 5)

where the numerator is the difference in temperatures calculated using an explicit finite difference scheme at the monitored
node at the same time step for temperature (T), using the
boundary conditions q and q + . is a small number and was
taken as 0.001 in the present investigation.

3. Results and Discussion

Figures 2 and 3 show the typical thermal histories in the
casting and the mold during solidification of the alloy in the
mold coated with graphite based dressing. The liquid metal
cools rapidly from the pouring temperature to the liquidus temperature. At the liquidus and solidus temperatures, the cooling
curve shows a rapid change in the slope due to the evolution of
the latent heat. It is also observed that during solidification of
the alloy the locations in the mold near the mold surface in
contact with the solidifying casting heated rapidly to the maximum temperature.
Analysis of casting and the mold thermal history indicated
the following:

The peak temperature attained by the mold during solidification increased with a decrease in the mold wall thickness. This is attributed to the lower volumetric heat capacity (CpV) of the thin walled molds.
The peak mold temperature increased with an increase in
the casting thickness due to its higher heat content.
The change in the coating material had a significant effect

Journal of Materials Engineering and Performance

Fig. 3 Casting and mold thermal history during solidification of 12

mm thick casting in a 10 mm thick graphite coated mold

Fig. 4 Estimated heat flux transients for 12 mm thick casting solidifying in graphite coated molds

on mold thermal history. The molds coated with the graphite-based mold wash exhibited higher temperatures compared with those coated with the alumina based mold coat.

possible contraction away from the mold wall. This could result in a nonconforming contact in the casting/mold interfacial
region causing a sharp decrease in the heat flux transients. If
the conditions are favorable, a gas gap may form at the interface. The formation of the gas gap during solidification of an
aluminum alloy against a metallic chill has been experimentally proved by Prabhu and Campbell using real-time x-ray
imaging (Ref 19).
The graphite coated molds showed higher heat flux transients compared with that obtained with alumina-coated molds.
This is due to the higher thermal conductivity of the graphitebased mold coats. It was also observed that the 10 mm thick
mold resulted in high heat flux transients and the heat flow was
found to be higher compared with 38 mm thick molds. On the
contrary, the solidification time and the cooling rates increased
with increase in the mold wall thickness. A similar result was
observed by Prabhu and Campbell (Ref 19) during solidification of an aluminum alloy against metallic chills. Copper and
cast iron chills of 10 mm thickness appeared to show abnormally high values of heat transfer coefficients. It was inferred
that the thin 10 mm chills, being relatively flexible, expand as
their front face heats up on contact with the solidifying casting.
The improved contact will cause the front face to heat up
further, enhancing the distortion toward the casting. On the
other hand, the thick chills, having greater rigidity, keep their
shape and location while the casting contracted, leading to a
nonconforming contact/gas gap. It is also likely that the heat
transfer might not be truly unidirectional and one-dimensional
inverse analysis may not be applicable in the case of thin chills/
molds. In light of contradicting results obtained with thin
molds, it is necessary to reassess the heat transfer at the casting/
thin mold wall interface.
In the present investigation, for thin molds, the thermal
histories at all locations inside the mold wall were approximated by a single mold heating curve determined by taking the
mean of the temperatures at two different locations (2 and 3)

Figures 4 and 5 show the computed heat flux transients for

typical casting/mold combinations. As the liquid metal filled
the mold, superheat is dissipated from the liquid metal to the
casting. This is shown in the cooling curve as a rapid decrease
in casting temperature. Below the liquidus temperature a solid
skin is formed. The casting skin may be pushed against the
mold wall by the metallostatic pressure of the liquid metal
resulting in a conforming contact. The good contact at the
casting/mold interface increases heat transfer from the solidifying casting to the mold, which causes a drop in the casting
skin temperature. The peak in the heat flux transient can be
associated with the formation of a stable shell at the casting/
mold interface.
The initial heat flux transients were low for all of the experiments. It is expected that the true heat flux begins at a high
value and then declines with time. One of the methods adopted
to approximate the true value of the peak heat flux is to extrapolate the heat flux transient curve at the point of peak heat
flux to meet the ordinate at zero time. However, due to the
asymptotic nature of the curve it results in very high values of
the peak heat flux to be justified by the type of experiment.
Further, the peak in the heat flux transient curve is always
identified with the solidification of the casting surface in contact with the chill. For the aluminum alloy LM21, the thermal
conductivity of the solid (ksolid 194 W/m K) is more than
twice the thermal conductivity of liquid (kliquid 90 W/m K).
Hence, the heat flux at the point of casting skin solidification is
greater than the heat flux at the liquid metal/chill interface. An
increase in the heat flux is, therefore, expected during the formation of a stable solidified shell.
As the thickness of the solidified shell increases, its strength
increases, which can resist the metallostatic pressure, allowing

Journal of Materials Engineering and Performance

Volume 14(5) October 2005607

Fig. 5 Estimated heat flux transients for 12 mm thick casting solidifying in alumina coated molds

Fig. 6 Heat transfer coefficients estimated assuming Newtonian

heating for 10 mm thick molds

thick mold. Further, the Biot number corresponding to the peak

heat transfer coefficient is calculated as:
inside the mold. The casting/mold interfacial heat transfer coefficients were then estimated by adopting Newtonian heating
analysis. The following equation is used to estimate the heat
transfer coefficient for 10 mm molds.
T
t
h=
ATc Tm
VCp

(Eq 6)

, Cp, and V/A are the density, specific heat, and volume-tosurface area ratio of the mold, respectively. Tm is the mold
temperature obtained as the mean of the temperatures at mold
locations 2 and 3. T/t is the heating rate of the mold calculated by taking the derivative of the heating curve Tm(t). Tc is
the casting temperature measured at location TC1.
Figure 6 shows the interfacial heat transfer coefficients estimated using Newtonian heating for the alloy solidifying
against 10 mm thick molds coated with graphite- and aluminabased dressings. The peak heat transfer coefficient was nearly
700 W/m2K for the graphite-coated mold. The corresponding
mold temperature was 162 C. Assuming the peak to be associated with the formation of solid shell, the peak flux at this
point was calculated as:
Q A = q = h Tsolidus Tm = 700 525 162 = 254,100 W m2
= 254.1 kW m2
This value is significantly less compared with the peak heat
flux (380 kW/m2) obtained by inverse analysis for the mold of
same thickness and also the peak heat flux estimated for the

608Volume 14(5) October 2005

Bi =

hl 7000.01
=
= 0.0972
k
72

The corresponding peak heat flux and the Biot number for the
alumina-coated mold are found to be 154 kW/m2 and 0.055,
respectively. In both cases, the Biot number is less than 0.1
indicating that in 10 mm thick molds, the temperature at the
surface of the mold in contact with the solidifying casting did
not differ greatly from the temperature at the center by more
than 5%. Hence, the assumption of a uniform temperature for
the 10 mm mold is valid. The heat transfer boundary conditions
calculated by using Newtonian heating analysis seem to be
more reliable as compared with the heat flux transients estimated by inverse analysis for thin molds. However, the Newtonian heating analysis cannot be applied for thick molds where
the Biot number is greater than 0.1 due the higher thermal
resistance (l/k) offered by the mold.
It was also observed that the time for the occurrence of peak
heat flux increased with the decrease in mold wall thickness
and increase in casting thickness. For example, the increase in
the casting thickness from 12 to 22 mm resulted in an increase
in the time of occurrence of the peak value from 16 to 21 s.
With a decrease in mold wall thickness, the ability to extract
heat from the casting decreases. The solidifying metal at the
interface remains a liquid for a longer time. With an increase in
mold wall thickness, a solid shell of the metal forms early at the
interface, corresponding to the peak in the heat flux transients,
and as the solidified shell gains strength it contracts away from
the mold wall. Similarly, with an increase in casting thickness
the total heat content of the casting to be extracted by the mold
of similar thickness increases, leading to a delay in the formation of the solid shell at the casting/mold interface.

Journal of Materials Engineering and Performance

4. Conclusions
Based on the preceding results and discussion the following
conclusions were drawn:

The mold wall, casting thicknesses, and the coating material of the mold significantly affect the thermal behavior of
the mold, and hence, the solidification of the casting.
The interfacial heat flux transients (q) estimated by inverse
analysis were found to be higher for thin molds, although
the solidification times were lower. The contradictory results obtained with thin molds were attributed to mold
distortion due to thermal stresses.
For thin metallic molds, where the Biot number is small,
assumption of Newtonian heating yielded realistic and reliable assessment of the casting/mold interfacial heat transfer compared with one-dimensional inverse modeling.
The time of occurrence of peak heat flux associated with
the formation of a stable solid shell at the interface increased with decrease in the mold wall thickness and increase in the casting thickness.
The use of graphite coating on the inner surface of the
mold increased the peak heat flux by about 20%.

References
1. H. Huang, O. Gurdogan, H.U. Akay, and W.W. Fincher, Thermal
Transport Phenomena in Metal Casting Simulations, AFS Trans, Vol
103, 1995, p 243-252
2. M. Jolly, Casting Simulation, How Well do Reality and Virtual Casting
Match? State of the Art Review, Int. J. Cast Met. Res., Vol 14, 2002,
p 303-313
3. K. Ho and R.D. Pehlke, Metal-Mold Interfacial Heat Transfer, Metall.
Trans. B., Vol 16, 1985, p 585-594
4. T.S.P. Kumar and K.N. Prabhu, Heat Flux Transients at the Casting/
Chill Interface during Solidification of Aluminum Base Alloys, Metall. Trans. B., Vol 22, 1991, p 717-727

Journal of Materials Engineering and Performance

5. J. Campbell, Castings, 1st ed., Butterworth-Heinemann, Oxford, UK,

1991, p 125
6. H. Fenech and W.M. Rohsenow, Prediction of Thermal Conductance
of Metallic Surfaces in Contact, J. Heat Transfer, Vol 85, 1963, p
15-22
7. Y.P. Shlkov and Y.A. Ganin, Thermal Resistance of Metallic Contacts,
Int. J. Heat Transfer, Vol 7, 1964, p 921-929
8. A.I. Veinik, Thermodynamics for the Foundryman, Maclaren, London,
UK, 1969, p 1-18
9. J. Isaac, G.P. Reddy, and G.K. Sharma, Variations of Heat Transfer
Coefficients during Solidification of Castings in Metallic Molds, The
British Foundryman, 78, 1985, p 465-468
10. Y. Nishida, W. Droste, and S. Engler, The Air-Gap Formation Process
at the Casting-Mold Interface and the Heat Transfer Mechanism
through the Air Gap, Metall. Trans. B., 16, 1986, p 833-844
11. D.J. Browne and D. OMahoney, Interface Heat Transfer in Investment Casting of Aluminum Alloys, Metall. Mater. Trans. B., Vol 32,
2001, p 3055-3063
12. D.R. Gunasegaram and T.T. Nguyen, Comparison of Heat Transfer
Parameters in Two Permanent Molds, AFS Trans., Vol 105, 1997, p
551-556
13. J.S. Kim, M. Isac, R.I.L. Guthrie, and J. Byun, Studies of Interfacial
Heat Transfer Resistances and Characterization of Strip Microstructures for Al-Mg Alloys Cast on a Single Belt Casting Simulator, Canadian Met. Quarterly, Vol 41 (No. 1), 2002, p 87-96
14. L. Strezov and J. Herbertson, Experimental Studies of Interfacial Heat
Transfer and Initial Solidification Pertinent to Strip Casting, ISIJ Int.,
Vol 38, 1998, p 959-966
15. C.A. Muojekwu, I.V. Samarasekara, and J.K. Brimacombe, Heat
Transfer and Microstructure during Early Stages of Metal Solidification, Metall. Trans. B., Vol 26, 1995, p 361-381
16. W.D. Griffiths, The Heat Transfer Coefficient during Unidirectional
Solidification of an Aluminum Alloy, Metall. Mater. Trans B., Vol 30,
1991, p 473-482
17. C.P. Hallam, W.D. Griffiths, and N.D. Butler, Interfacial Heat Transfer between a Solidifying Aluminum Alloy and a Coated Die Steel,
Mater. Sci. Forum, Vol 329-330, 2000, p 467-472
18. J.V. Beck, Nonlinear Estimation Applied to the Nonlinear Heat Conduction Problem, J. Heat Transfer, Vol 13, 1970, p 703-716
19. K.N. Prabhu and J. Campbell, Investigation of Casting/Chill Interfacial
Heat Transfer during Solidification of Al-11% Si Alloy by Inverse
Modeling and Real-Time X-ray Imaging, Int. J. Cast Metals Res., Vol
12, 1999, p 137-143

Volume 14(5) October 2005609