Development of a Heat Transfer

Dimensionless Correlation for a Wide Range

of Prandtl Numbers

Blas Melissari Cassanello

Stavros A. Argyropoulos
Department of Materials Science and Engineering, University of Toronto

Abstract

In this paper a computational approach is employed to derive a dimensionless heat

transfer correlation for forced convection in spheres. This correlation is applicable
to a wide range of ‡uid systems. The lower end of this range includes the Prandtl
number for liquid Sodium (P r » 0:003), whereas the upper end includes the Prandtl
number for water (P r » 10).
1=
N u = 2 + 0:47Re 2 P r 0:36
3 £ 10¡3 · P r · 101 ; 102 · Re · 5 £ 104
The model predictions derived from this research were validated extensively. First,
the model was tested in liquid Aluminum, and subsequently the model was compared
with existing experimental data involving water. Both veri…cation procedures have
shown very good agreement between experimental results and model predictions.

Key words: liquid metals, heat transfer coe¢cient, modeling, experimental, forced
convection, natural convection

Nomenclature
A Area, m2
c Heat capacity, J/(kg.o C)
C1 ; C2 ; ::: Correlation coe¢cients
CV Control Volume
D Diameter, m
g Gravity, m/s2
Gr Grashof number, g:¯:½2 :SP H:D3 =¹2
G Buoyancy parameter, Gr=Re2
I Performance Index
k Thermal conductivity, W/(m.o C)
LF Liquid Fraction, m/m

Preprint submitted to Elsevier Science 6 September 2004

LH Latent Heat, J/kg
MT Melting time, s
n1 ; n2 ; ::. Correlation exponents
Nu Nusselt number, h:D=k
Pr Prandtl number, ¹:c=k
R Resistance, (m2 .oC)/W
Re Reynolds number, ½:u:D=¹
S Source term
SP H Superheat, (T1 ¡ Tm ) in o C
T Temperature, oC
t time, s
¢t time step, s
U Uncertainty
u Velocity, m/s
V Volume, m3
¢x Control volume size, m
Greek symbols
® Thermal di¤usivity, m2 =s
¯ Thermal expansion coe¤., 1=o C
¹ Dynamic viscosity, kg=m:s
½ Density, kg=m3
¾ Standard deviation
Subscripts
0 Initial condition
1 Liquid condition far from sphere
m Melting point (liquidus)
S Solid condition (solidus)
L Wall condition
FC Forced Convection
IN T Interface
i Index
corr Correlation
model Modeling Prediction
exper Experimental Result

1 Introduction

There is a derth of dimensionless convective heat transfer correlations appli-

cable to ‡uids such as liquid metals. Knowledge of heat transfer rates from
particles at high ‡ux levels has become increasingly important to the design
of energy transfer systems and metallurgical processes in general.

Theoretical as well as experimental approaches have been carried out to ana-

lyze the heat transfer in liquid metals. Hsu [1] and Sideman [2] have derived
equations for heat transfer from a sphere to a liquid metal by assuming po-
tential ‡ow around the sphere. Kreith et al [3] performed an experimental
investigation of rotating metallic spheres in liquid mercury and suggested a

2
correlation for forced convection. Witte [4] performed experiments on heat
transfer from a non-melting sphere to liquid sodium and obtained an equation
relating the Nusselt number to Reynolds and Prandtl numbers. Argyropoulos
and Mikrovas [5] [6] immersed spheres in liquid Aluminum and Steel and found
correlations for forced and natural convection based on the measurement of
the melting times of the spheres.

Many investigators studied the heat transfer characteristics of solid spheres

to ‡uids with Prandtl number around unity (P rAir h 0:7; P rW ater h 10).
McAdams [7] compiled numerous experimental results and correlated all of
them into a single empirical correlation valid for Air and Water. Yuge per-
formed pioneering experimental work on the heat transfer from a sphere to
air under mixed convection [8], suggesting procedures for predicting the Nus-
selt number. Vliet and Leppert [9] developed correlations for spheres in water.
Hieber et al [10] studied the spherical system analytically, but their study was
limited to small Reynolds numbers.

Some researchers studied the melting dynamics of ice spheres in water at dif-
ferent convective regimes. Vanier and Tien [11] performed experiments on the
melting of a submerged ice sphere in water, calculating the melting rate by
weighing the sphere. Solomon [12] obtained a solution for the melting of a
sphere in convection as a function of the average diameter and heat ‡ux. Es-
kandari [13] reported on a series of experiments to study the forced convection
heat transfer from a ‡owing stream of water to an ice sphere. Anselmo [14]
[15] undertook an extensive theoretical and numerical analysis of melting of
full and partially submerged ice spheres in a pool of water. Aziz et al [16] and
Hao et al [17] performed measurements of the heat transfer coe¢cient in the
water system by measuring the melting time of ice spheres in forced convec-
tion. Mukherjee et al [18], McLeod et al [19] and Hao et al [20] conducted
visualization studies of ice spheres melting in water under mixed convection
regime.

In terms of mixed convection around a sphere in liquid metals, the work of Kre-
ith and his associates is worth mentioning [21] [3]. By performing experimental
measurements of rotating spheres in media as diverse as Air and Mercury, they
concluded that if the buoyancy parameter (G = Gr=Re2 ) is less than 0:3, then
natural convection is negligible, i.e.: its e¤ect is lower than 5% as far as the
heat transfer is concerned. This value agrees with the theoretical derivation
by Sparrow et al [22]. Numerical model predictions regarding the in‡uence of
natural convection on the total melting time of spheres has been published
by Melissari and Argyropoulos in [23]. They concluded that for values of the
buoyancy parameter lower than the range G = 0:5 » 1:0, the total melting
time is not a¤ected by natural convection e¤ects.

Table 1 summarizes the correlations for heat transfer around spheres and their

3
Table 1
Nusselt number for Forced convection around spheres
Authors Applicability Nusselt number correlated
1=
Hsu [1] P r < 1; Re · 2 £ 105 0:921(Re:P r) 2

1=
Sideman [2] P r < 1; Re · 2 £ 105 1:13(Re:P r) 2

Kreith [3] P r = 10¡2 ; 7 £ 104 · Re · 106 0:178Re0:375

1=
Witte [4] P r = 10¡3 ; 3 £ 104 · Re · 2 £ 105 2 + 0:386(Re:P r) 2

Argyrop. [5] 10¡2 · P r · 10¡1 ; Re · 3 £ 104 2 + 1:114Re 0:557 P r 0:914

1=
Whitaker [24] P r = 0:7; Re · 8 £ 104 2 + [0:4Re 2 + 0:06Re 0:67]P r0:4
1=
Vliet and L.[9] P r » 10; Re · 105 (2:7 + 0:12Re0:66 )P r 2

Aziz et al [16] P r » 10; 3 £ 103 · Re · 3 £ 104 0:991Re0:527 P r 0:043

Hao et al [17] P r » 10; Re · 3 £ 103 1:015Re0:48 P r 0:23

1= 1=
McAdams [7] P r ¸ 0:7; Re · 2 £ 105 2 + 0:6Re 2 Pr 3

range of applicability. As seen from this summary, all the correlations corre-
spond to either a single or to a narrow range of Prandtl numbers. The present
paper introduces a methodology which allows the development of dimension-
less heat transfer correlations for forced convection around spheres applicable
in a wide range of Prandtl number ‡uids.

2 Mathematical Considerations

2.1 Calculating the Nusselt Number From the Melting Time

By applying a heat balance to the sphere, it is possible to relate the total

melting time to the Nusselt number. It is assumed that the sphere is sub-
ject to uniform melting, meaning that we will calculate the Nusselt number
averaged over the entire surface as opposed to the localized coe¢cient. The
instantaneous heat balance is shown in equation 1.

h:A:SP H:dt = ½:LH:dV (1)

N u:kL
where h =
D

4
In this context, D represents the diameter as a variable and D0 is the initial
diameter of the sphere. The sensible portion of the heat supplied to the sphere
is not included because it is considered that the shell is formed at the expense
of heating the sphere up to its melting point. The volume, surface area and
volume di¤erential of a sphere are as follows:

1
V = ¼D 3 (2)
6
A = ¼D 2
1 1
=) dV = ¼D 2 dD = A:dD
2 2
The Nusselt number can be expressed in terms of the diameter as N u =
1 1
C1 Re =2 = C2D =2 for forced convection. At this stage it would be desirable
to express the exponent in a parametric way since it would be valuable to
…nd the parameter from the equation rather than setting it …xed at a certain
value. Hence these relationships become: N u = C1 Ren F C = C2 D n F C

Substituting Nu and dV in the di¤erential heat balance (1), and by rearrange-

ment we obtain:

1
h:SP H:dt = ½:LH:dD (3)
2
1
=) C2 :kL :SP H:dt = ½:LH:D(1¡n F C ) :dD
2

The LHS of the equation can be integrated in time, whereas the RHS of
the equation has to be integrated between the initial diameter (D0 ) and the
maximum diameter reached by the sphere when the shell is formed (Dmax ),
and then from Dmax to D = 0. This yields the following:

Z ÃZ Z0 !
MT 1 Dmax
(1¡n F C ) (1¡n F C )
C2 :k L :SP H dt = ½:LH D dD + D dD
0 2 D0 Dmax

F C ) ¡ D(2¡nF C ) )
1 (2D (2¡n
max 0
=) C2 :kL :SP H:M T = ½:LH:
2 2 ¡ nF C

Substituting N u = C2:D nF C and rearranging terms, we obtain a relationship

between the melting time and the Nusselt number as a function of the initial
diameter D 0 as seen in equation 4.
à µ !
Dmax ¶(2¡n F C ) ½:LH:D20
Nu = 2 ¡1 (4)
D0 2(2 ¡ nFC ):kL :SP H:M T

5
Assuming that the diameter can be obtained from the mass of the sphere
using a relationship of the form m _ D3 ; we can express the Nusselt number
as equation 5.

½:LH:D02
N u = MTF (5)
2(2 ¡ nF C ):k L :SP H:M T
1 mmax
where M TF = 2mF=3 (2¡nF C ) ¡ 1 ; with mF =
m0

The melting time factor, M TF , a¤ects the total time of the immersed sphere
and is caused by the shell formation upon immersion. It is calculated based
on the mass increase of the sphere. The mass factor, mF , is calculated as the
ratio between the maximum mass of the sphere (mmax ) and the initial mass
(m0 ). In the subsequent sections these factors will be estimated. For a value
of nF C = 21 , the Nusselt number for forced convection is equation 6.

½:LH:D 20
N u = MTF (6)
3k L :SP H:M T
1
where M TF = 2mF=2 ¡ 1 (7)

2.2 Mathematical Modeling of the Melting Sphere

The problem of a melting sphere is modeled as a three-dimensional system in

Cartesian co-ordinates of ‡uid ‡ow and heat transfer coupled by the presence
of natural convection. The model was developed using the SIMPLER algo-
rithm in (x; y; z) coordinates. A detailed explanation of the method can be
found in reference [25].

Continuity equation:
@ux @uy @uz
+ + =0 (8)
@x @y @z

Flow equation in the x¡direction:

à !
@ux @ux @ux @ux @p @ @ux @ @ux @ @ux
½ + ux + uy + uz =¡ + ¹ + ¹ + ¹
@t @x @y @z @x @x @x @y @y @z @z
(9)

Flow equation in the y¡direction:

6
 2.D 6.D

g 2.D g

T∞
u∞
4.D

T∞ u∞

z z
o
y x
o

Fig. 1. Domain used for the numerical simulations.

à !
@uy @uy @uy @uy @p @ @uy @ @uy @ @uy
½ + ux + uy + uz =¡ + ¹ + ¹ + ¹
@t @x @y @z @y @x @x @y @y @z @z
(10)

Flow equation in the z¡direction:

à !
@uz @uz @uz @uz
½ + ux + uy + uz = (11)
@t @x @y @z

@p @ @uz @ @u @ @u
=¡ + ¹ + ¹ z + ¹ z + ½:g:¯:(T ¡ T1 )
@z @x @x @y @y @z @z

Energy equation:

à !
@H @H @H @H @ @T @ @T @ @T
½:c + ux + uy + uz = k + k + k +S (12)
@t @x @y @z @x @x @y @y @z @z

A schematic of the computational domain can be seen in Figure 1.

2.3 Parameters of the Numerical Method

2.3.1 Boundary Conditions

Inlet(x = 0), Bottom(z = 0), Top(z = 4D) and Back(y = 2D) :

7
 ux = u; uy = uz = 0; T = T1

Outlet (x = 6:D): @ux =@x = @u y =@x = @uz =@x = @T=@x = 0

Front (symmetry plane y = 0) : uy = 0; @ux=@y = @u z =@y = @T=@y = 0

2.3.2 Initial Condition

Fluid: Liquid at T = T1

Sphere: Solid at T = T0 ; with T0 · Tm · T1 :

2.3.3 Domain Size

The domain size used depends on the diameter of the addition. A minimum
size of 6 diameters in the direction of the velocity (x) and 4 diameters in the
directions perpendicular to the velocity (y; z) was found to be large enough
to obtain a stable and accurate solution. The thermo-physical properties of
Aluminum and AZ91 used in the model are shown in Tables 2 and 3.

2.3.4 Enmeshment

A structured mesh compatible with the SIMPLER algorithm was employed.

Due to the three-dimensional character of the simulation, a cube of side length
1:5D is meshed around the sphere with a control volume (CV) size of D=28
while the rest of the domain is meshed using a CV of size D=7. Figure 2 shows
a schematic of the non-uniform mesh used for this model. This results in 74
CVs in the x-direction, 60 in the y-direction and 30 in the z-direction, totalling
1:3 £ 105 CVs.

2.3.5 Time Step

Due to the high thermal gradients during the initial transient, two di¤erent
time intervals are used: an initial value of ¢t = 10¡4 s is used during the …rst
second and ¢t = 10¡3 s is used for the rest of the simulation.

2.3.6 Other Parameters

As will be detailed in the experimental section of this paper, the spheres are
introduced into the bath by means of a 6 mm diameter stainless steel tube.
For this reason, the holder is introduced in the model. The thermophysical
properties of the stainless steel used can be found in Table 4. An interfacial

8
Table 2
Thermophysical properties of Aluminum in SI units [27]
kS kL cS cL ½ ¹ ¯ Tm LH
220 92 1100 1000 2400 1:2 £ 10¡3 1:3 £ 10¡4 660 3:95 £ 105

Table 3
Thermophysical properties of AZ91 [27]
kS kL cS cL ½ ¹ ¯ TS Tm LH
60 80 1200 1400 1750 1:4 £ 10¡3 1:2 £ 10¡4 437 600 3:7 £ 105

Table 4
Thermophysical properties of Stainless Steel in SI units [28]
kS cS ½
15 500 8000

heat resistance value of 10¡4 m2 K=W is used around the sphere due to the
oxide layer formed [26].

Fig. 2. Meshing around the sphere for the three-dimensional model.

2.3.7 Phase Change Treatment

The Heat Integration Algorithm is used in the Aluminum system due to the
fact that the phase change is modeled as isothermal. When a pure material
is melting/solidifying, the heat supplied/extracted is consumed entirely for
the phase change without a change in its temperature. In the model, when a

9
control volume begins melting/solidifying, its temperature is set at the melting
point until the amount of heat supplied/extracted accounts for the latent heat
of fusion.

Due to the large freezing range of the AZ91 Magnesium alloy, the numerical
integration algorithm is not suitable. The enthalpy method is implemented via
the introduction of a source term, S; in the energy equation. The expression
for the source term can be seen in equation 13.

à !
½:LH @LF
S =¡ + ~u r ¢ LF (13)
c @t

2.4 Results of the Numerical Model

The melting evolution of a D = 5cm sphere in a SP H = 60oC Aluminum

bath and u = 0:2m=s can be seen in Figures 3 through 5. These …eld plots are
represented at the symmetry plane y = 0: It is worth noticing that due to the
di¤erence in thermal conductivity between the holder and the sphere, there is
a marked discontinuity on the isotherms.
002  23 Jul 2 004  sph3D
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 1s
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 0.1s

0.04
0
66

680 6
68 6
50

0
0.12 6
0

6
5
66

0.02 5
0
70 0

5
4
4
0.1
Z

300

4
30 4
Y

0
660 80

0
3
3
6

50

0
0

3
50

6 80

3
2
0.08
-0.02 2
0

2
66

70 1
0 700
1
1
1
-0.04 8
0.06 0.08 0.1 0.12 0.14 5
0.06 0.08 0.1 X 0.12 0.14 0.16
X

Fig. 3. Numerical model result, 5cm Aluminum sphere melting in a 60o C superheat
bath at a velocity of 0.2m/s, t=1s.

10
 002  23 Jul 2 004  sph3D
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 3s
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 0.1s
700
0.04 0
68
6
660 6
0.12 6
6

0
70
5
0.02 5

680
5
4

660
4
0.1
Z

4
680
4
Y

0
3
660

3
3
3
2
0.08
-0.02 2
70

660 2
0

680 1
1
700 1
1
-0.04 8
0.06 0.08 0.1 0.12 0.14 5
0.06 0.08 0.1 X 0.12 0.14 0.16
X

Fig. 4. Numerical model result, 5cm Aluminum sphere melting in a 60o C superheat
bath at a velocity of 0.2m/s, t=3s.

The total melting times of 3cm and 7cm Aluminum spheres as a function of
the bath velocity are shown in Figures 6 and 7 respectively. For u = 0cm=s,
the solution represents the pure natural convection solution.

Figure 8 shows the term (M T:SP HAZ91 ) as a function of the bath superheat
for 3cm and 5cm AZ91 spheres. The importance of this plot is that it can
be used to obtain the melting time for 3cm and 5cm AZ91 spheres in any
condition of bath temperature and velocity.

2.4.1 Melting Time Factor

By running the model for di¤erent sphere initial temperature, T0 ; it is possible

to obtain the melting time factor by …nding the ratio between the melting time
at a given T0 and at T0 = Tm : Figure 9 shows the melting time factor M TF
for Aluminum spheres under di¤erent conditions of diameter, superheat and
velocity, for T0 = 20o C and T0 = 450o C. The values of M TF do not vary
signi…cantly with the diameter and superheat, but change from M TF = 1:5
in natural convection to M TF = 1:3 for u = 0:4m=s, mainly due to the fact
that the melting is non-uniform for the higher values of velocity in forced ‡ow.
If the initial temperature is closer to the melting point of the material, the

11
 002  23 Jul 2 004  sph3D
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 10s
u= 0.2m/s SPH=60.0ºC D= 5.0cm t= 0.1s

0.04
6
6
0.12 6

0
6

70
660
5
0.02

68
5

0
5
4

680
4
0.1
Z

4
4
Y

0
3

660
3

66 0
3
3
70 0 680 2
0.08
-0.02 2
2
1
1
1
1
-0.04 8
0.06 0.08 0.1 0.12 0.14 5
0.06 0.08 0.1 X 0.12 0.14 0.16
X

Fig. 5. Numerical model result, 5cm Aluminum sphere melting in a 60o C superheat
bath at a velocity of 0.2m/s, t=10s.
1 Aluminum spheres 1
1.0 Aluminum
Aluminum spheres,
spheres, DD=5cm, SPH=30ºC,
=5cm, SPH =30ºC, NCNC 1.0
1.7 Aluminum
Aluminum spheres.spheres, G <R0.1 =1e-4 m K/W
T 0 =20ºC,
2
1.
INT
0 0.03 0.1 0.04 SPH0.05
0.2 (ºC) 0.30.06 0.07
0.4
30
0.9 0.30 20 40 60 80 10030 0.30
0.9
1.6
1.60.5 MTFC/MT 0.5 1.
RINTP=C 0
0.8 SPH=30ºC D=3cm;MTSPH=30ºC
AID/MT = 3 x 10-5 0.8
RINT PC
SPH=90ºC D=3cm;MTSPH=90ºC -4
250.9 CF>/MT
RINTP=C 1 x 10 25
0.91.
1.5 7Rc = 3 x 10-4
0.7 0.25
0.4 m
INT 0.7
0.25
0.4
1.4 sp
he
20 1.4 MT / MT R=0 re 20 1.
0.6 mmax / m0
0.6
7c
min u (m/s)

MT CF/MTFC m
m / m0

0.5 1.30.3 sp 0.3 1.

MT (s)

0.20
1.2 MT /MT he 0.5
0.20
15 15
u (m/s)

u (m/s)

AID FC
re
0.8 0.8
0.4 1.2 5c 0.41.
m
0.2 sp 0.2
10 he 10
0.3 1.0 re 0.3
1.1 0.15 0.15
1.
3c
m
0.25 sp
0.1 he 50.2
0.1
1 re 1
0.10.7 0.8-6 -5 0.1
-4
0.7-3
100.10 10 1010
0.10
0 0.03 1
0.04R (m
0.05ºC/W) 0.06
29 13 17 5
0.07
0
0 0 0.15 10 0.2G 15
INT 20
0.3 25 0.4 30
0.0 0 D t(m)
(s) 0 0.0
0 50020 1000 40 u (m/s)
1500 60 2000 802500 100
3000
Re
SPH (ºC)

Fig. 6. Melting times of 3cm Aluminum spheres

12
 1 Aluminum spheres 1
1.0 Aluminum
Aluminum spheres,
spheres, DD=5cm, SPH=30ºC,
=5cm, SPH =30ºC, NCNC 1.0
1.7 Aluminum
Aluminum spheres,
spheres. G <R0.1 =1e-4 m K/W
T 0 =20ºC,
2
1.
INT
0 0.03 0.1 0.04 SPH0.05
0.2 (ºC) 0.30.06 0.07
0.4
100
0.9 0.30 20 40 60 80 100100
0.30
0.9
1.6
1.60.5 MT /MT 0.5 1.
FC RINTP=
C 0
0.8 SPH=30ºC MTSPH=30ºC
D=7cm; AID/MT
RINT PC
=3 x 10-5 0.8
SPH=90ºC D=7cm;
MTSPH=90ºC
/MT
R =1 x 10-4 0.9
0.9
80 1.5
CF> INTP C
7Rc = 3 x 10-4 80 1.
0.7 0.25
0.4 m
INT 0.7
0.25
0.4
1.4 sp
he
1.4 MT / MT R=0 re 1.
0.6 mmax / m0
0.6
7c
min u (m/s)

60 60
MT CF/MTFC m
m / m0

1.30.3 sp 0.3 1.
MT (s)

0.5 0.20
1.2 MT AID/MTFC he 0.5
0.20
u (m/s)

u (m/s)
re
0.8 0.8
0.4
40 1.2 5c 0.41.
40
m
0.2 sp 0.2
0.3 he 0.3
1.1
1.0
0.15 re 0.15 1.
3c
20 m 20
0.2 sp 0.2
0.1 he 0.1
1 re 1
0.10.7 0.8-6 -5 0.1
0.7 -4
-3
100.10 10 10
10
0.10
0 0.03 1 R
0.04 (m
0.05ºC/W)
2
0.06 9 13 17 5
0.07
0
0 0 0.15 10 0.2G 15
INT 20
0.3 25 0.4 30
0.0 0 D t(m)
(s) 0 0.0
0 50020 1000 40 u (m/s)
1500 60 2000 802500 100
3000
Re
SPH (ºC)

Fig. 7. Melting times of 7cm Aluminum spheres

0.9
AZ91 Results
18
1400 0.8 1400

16 T∞ = 630ºC
T∞ = 660ºC
1200 0.7 1200
14 3cm model
5cm model
0.6
12
1000 1000
MT.SPH AZ91 (sºC)
MT (s)

LF

10 0.5
800 800
8
0.4
600 600
6
0.3
4
400 400
0.2
2
200 200
0
0.1
0 0.05 0.1 0.15 0.2 0.25 0.3
u (m/s)
0 0.0 0
0 4500.05 475 0.1
500 525 0.15
550 575 0.2 60
u (m/s)
T (ºC)

Fig. 8. Melting times of AZ91 spheres (for all temperatures).

13
shell increase is much lower and subsequently the melting time factor would
be very close to unity. In any case, there is no more than a 20% increase in
the melting time for T0 = 450o C; with little variation in diameter, superheat,
or even velocity.

Fig. 9. Melting time factor, M TF , for di¤erent operating parameters in the Alu-
minum system (numerical model).

The maximum mass increase observed in the numerical simulations was of the
order of 50%; meaning we obtain a mass factor mF = 1:5. Using the expression
1
for the melting time factor, M TF = 2m F=2 ¡ 1, we observe that if we use the
measured values of mass increase, we obtain a value that is in agreement with
1
the results of the numerical simulation: M TF = 2(1:5) = 2 ¡ 1 = 1:45.

Some speci…c measurements were carried out where the spheres were extracted
prior to the complete melting. In this way, the shell formed around the sphere
could be measured and the melting time factor could be estimated. The max-
imum mass increase observed experimentally was of the order of 40% to 60%.

It is convenient to minimize the error involved in the estimation of these

factors. A possible way is to perform experiments with heated spheres thus re-
ducing the mass increase, as observed numerically in Figure 9. Ideally, a solid
sphere at its melting point will not form a shell. However, it is highly imprac-
tical to immerse such spheres mainly due to their poor mechanical integrity.
Nevertheless, some experiments were carried out with Aluminum spheres pre-
heated at T0 = 450o C simply to determine its experimental feasibility. These
were performed using 3cm and 5cm spheres in 60oC superheat bath in the
velocity range 0m=s » 0:3m=s.

14
3 Validation of the Numerical Model

In order to validate the numerical model developed, a series of tests are carried
out to run under the characteristics of a set of established solutions found in
the literature.

Paterson Point Heat Source Analytical Solution. First, a one dimen-

sional heat di¤usion/melting system is developed in spherical co-ordinates to
solve for Paterson’s point heat source analytical solution [29]. The conditions
chosen emulate the very high heat transfer rates found in liquid metals. This
tests the Heat Integration Algorithm as well as the ability of the code to handle
a large heat transfer rate under a hypothetical heat source. A more detailed
explanation about this validation can be found in [30].

Experimental Results in Melting Gallium. Second, a two-dimensional

convective/melting system in rectangular coordinates is developed in order
to compare it with the experimental study of the melting of Gallium in an
enclosure by Gau and Viskanta [31]. This tests the Heat Integration Algorithm
as well as the natural convection induced ‡ow in a liquid metal (complete
validation results can be found in [30]).

Experimental Results in Melting Ice. Third, the three-dimensional model

developed for liquid metals is set to run using the thermophysical properties of
water to simulate the melting of ice spheres in liquid water. The ‡ow patterns
around the spheres obtained by the model are compared with images from
visualization studies in water performed by Hao et al in [20]. The melting
times in forced convection obtained experimentally in the ice/water system by
Aziz et al [16] and Hao et al [17] are compared with results of our numerical
model.

Finally, the model will be compared with experimental results in Aluminum

and AZ91 carried out in the present work.

3.1 Melting of Ice Spheres in Water

Aziz et al [16] and Hao et al [17] performed numerous investigations in the

water/ice system. A series of experiments were conducted to measure forced
convection heat transfer rate from submerged ice spheres in water.

The three-dimensional model was used to obtain the melting times of ice
spheres under pure forced convection. The thermophysical properties of water
used are shown in Table 5. The meshing and time step scheme used is similar

15
Table 5
Thermophysical properties of water in SI units [28]
kS kL cS cL ½ ¹ ¯ Tm LH
2:2 0:6 2100 4200 1000 1:2 £ 10¡3 4:0 £ 10 ¡4 0 3:4 £ 105

to the one used for the Aluminum system.

There was no mass increase observed in any of the spheres, due to the fact
that their initial temperature was only 10o C below zero (melting point). The
expression to calculate the Nusselt number is equation 14 (a particular case
of equation 6).

½:LH:D02
=) N u = (14)
3kL :SP H:M T

A series of runs were performed using a 3:6cm ice sphere initially at ¡10oC
immersed in a water bath with superheats ranging between 10o C and 30oC
and velocities between 0:01m=s and 0:1m=s. The calculated dimensionless heat
transfer coe¢cient was obtained from equation 14 and was plotted along with
the experimental results by Aziz et al (Figure 10) and Hao et al (Figure 11).
The thin solid line in Figure 11 represents the results obtained by Aziz et al in
1995 and were included in the original paper by Hao et al for comparison [17].
The deviation in the Nusselt number reported on both investigations is of the
order of 20%: Good agreement is observed between the numerical model and
the experimental results from both experimental investigations.

The ice melting system is implemented using the three-dimensional model to

study the mixed convection regime ‡ow around the sphere. Flow visualization
images of the work by Hao et al in 2001 [20] are compared with ‡ow plots
of our numerical model at di¤erent convective conditions. Figures 12 and 13
show the ‡ow around a 3:6cm in diameter ice sphere immersed in water at
30oC and an imposed velocity of 0:01m=s as a result of the visualization study
and our numerical model, respectively. It is worth noticing the strong e¤ect
of the natural convection induced ‡ow (G = 50). The ‡ow pattern in mixed
convection predicted by the model is almost identical to the one observed
experimentally.

16
Fig. 10. Comparison between the numerical model and the experimental results by
Aziz et al [16].

4 Experimental Measurements In Liquid Metals

4.1 Experimental Work involving the Revolving Liquid Metal Tank

The apparatus used to immerse spheres consists of a cylindrical Stainless Steel

tank that can rotate inside a heavily insulated electrical resistance furnace.
The Revolving Liquid Metal Tank (RLMT) is connected to a DC motor ca-
pable of controlling the rotating speed to a resolution of 1 RP M . The interior
diameter of the RLMT is 380 mm, the height is 200 mm and it has a usable
capacity of 20 lit (50 kg of Aluminum, approximately). Figure 14 shows a
schematic of the RLMT inside the electrical resistances furnace. Figures 15
and 16 show the RLMT in the furnace and the complete setup, respectively.

A 41 HP Bodine Electric Company DC motor with velocity control is assem-

bled and connected to the RLMT shaft via a belt and a wheel. To adapt to
di¤erent velocity ranges, the rotation ratio could be modi…ed by changing the
belt/wheel assembly. The centre of the perforation on the lid is located at a
radius of approximately 12cm and one side of the lid is secured to the struc-

17
Fig. 11. Comparison between the numerical model and the experimental results by
Hao et al [17].

Fig. 12. Ice melting in water, D = 3:6cm; u = 0:01m=s [20]

(Re = 2 £ 102 ; Gr = 2 £ 106 ; G = 50)

18
 02  25 Nov 2003  sph3D
Ice at -20ºC in water 0.05m/s
u = 0.01m/s SPH=30ºC D= 3.6cm t= 3.0s
Ice at∞-20ºC in water
u= 0.01 m/s SPH=3ºC D= 3.6cm t= 1 .0s
0.03
0.09
0
25 2

15
-1 0
25 2

15
0.02 5 2
-1 -1 2
5
0.08
2
1

-100
1
0
0.01
1
0.07 1
Z

1
-10
25
15

8
Y

0 6

15
25
-1 5 4
5 -1
0.06 2
-1 0 0 0
0
-0.01 -2
15
25 -4
-6

25
0.05 -8
-1
-0.02 -1
-1
-1
0.05 0.06 0.07 0.08 0.09 0.1 -1
0.04 0.05 0.06 0.07
X 0.08 0.09 0.1
X

Fig. 13. Ice melting in water. Numerical model result.

2 6
(Re = 2 £ 10 ; Gr = 2 £ 10 ; G = 50)

ture to avoid any shifts. Before beginning a set of experiments, the radius of
rotation is measured (this radius is denoted as rrot ). The rotation speed of
the tank is measured (in revolutions per minute, RP M ) and the velocity u is
calculated using equation 15.

2¼
u= £ rrot £ RP M (15)
60

For the experiments carried out in AZ91, a protective atmosphere is used,

consisting of a mixture of CO2 and 0.5% SF6 . A 1 =8 " Stainless Steel tube
was connected through the lid at approximately the centre of the RLMT
and connected to the protective gas cylinder. The regular ‡ow rate for the
experiments was 2 lit/min, although the system was calibrated to a maximum
of 10 lit/min of gas to be used in an emergency situation. The gas inlet can
be seen in Figure 14.

The melting time of the immersed spheres is measured by means of the change
in electrical resistance between the tip of a wire inside the sphere and the bath.
Figure 17 shows a schematic of the sphere/holder assembly. A 5 =32 " hole is
drilled to the centre and threaded up to 41 " from the centre of the sphere.
The insulated wire runs inside the holder, leaving the tip in the centre. A 5V
DC is supplied to the internal wire and the bath is grounded by the holder

19
 400
220
190
120

lid gas inlet

70 insulation

25

metal bath

90

370 200
sphere
250

180

50 insulation CL
roll bearing
Notes:
- Not to scale rotating shaft
- All distances in mm

Fig. 14. Schematic of the apparatus used to immerse spheres

when the sphere is immersed. The voltage potential between the wire and the
bath is measured; when the sphere melts, the voltage drops to the short circuit
voltage (0V ). The immersion of the sphere is also recorded in a similar way, by
measuring the potential di¤erence between the sphere holder and the grounded
bath. Figure 18 shows a schematic of the electrical circuit used to measure the
melting time of the immersed sphere, as well as the voltage evolution during
an immersion.

The Aluminum spheres were manufactured by Brampton Foundries using sand

cast commercially pure Aluminum; the AZ91 spheres were machined. There
was no visibly observed porosity in the spheres cut. The 41 " Stainless Steel
holder is painted with a layer of Boron Nitride to prevent dissolution, as well
as to prevent the Aluminum from adhering to the holder, particularly to the
threaded portion. The spheres and holder assembly can be seen in Figures 19
and 20. The RLMT is coated with a layer of refractory paint (901 Alumina

20
 Fig. 15. Photograph of RLMT inside the furnace.

Fig. 16. Photograph of the complete apparatus.

ceramic from Centronics) to avoid exposure of the Stainless Steel to the liquid
metal.

The output voltages and the bath temperature are recorded using a National

21
 internal
insulated
wire

holder s.s.
1/4" diam.

Sphere

Fig. 17. Schematic of the sphere/holder assembly.

Power supply 5V

R R

Vh Vs
to internal
Vh Vs wire

0
immersion
(bath grounded) time
sphere sphere
immersed melted

Fig. 18. Schematic of the circuit used to measure the melting time of the spheres.

Instruments 6034E data acquisition PCI board via an 8 channel isolated input
module with internal temperature compensation. The bath thermocouple is
type K and the sampling rate is set to 5 measurements per second. A Labview
code is implemented to acquire the data from the PCI board. Figure 21 shows
a typical result for a 5cm Aluminum sphere in a bath with a superheat of 60o C
and a velocity of u1 = 0:33m=s. The holder voltage Vh drops instantaneously
as the sphere is immersed (A). Due to humidity and oil residues inside the
cavity and the holder assembly, there is a drop in the sphere’s voltage Vs before
the end of the melting (point B). However, when the liquid front reaches the
centre of the sphere, a sudden drop occurs (point C, at t = 15s).

22
 Fig. 19. Photograph of 5cm and 7cm Aluminum spheres.

Fig. 20. Photograph of 3cm Aluminum sphere and holder detail.

4.2 Experimental Results in Liquid Metals

Using the relationship between the Nusselt number and the melting time of a
sphere, all the experimental results are grouped in a single graph. The results
for the preheated Aluminum spheres will also be included in the graph (the
melting time factor used is M TF = 1:1; as obtained in the numerical model).
Also, the results for AZ91 will be plotted in the same graph; the melting time
factor used is M TF = 1:5 because AZ91 and Aluminum show very similar
shell formations due to the fact that their thermophysical properties are very

23
 5
Aluminum
Al 5cm, 0.33m/s
725
5
4 B
700 700
A
720
4 3 A
Voltage (V)

B
C
C
3 6802 680
715
Voltage
T (ºC)

T∞
Vh D
V0
1
V+
2 710
V-
660 T∞ 660
Vh
1 0 Vs
0 5 10 15 20 25 30 35 40
705
45
t (s)

0 640 640
0 2 4 6 700
0 5 t (hr)10 15
t (s)

Fig. 21. Typical data acquired for the one-probe sphere.

similar. Moreover, the Prandtl numbers of both systems are also of the same
order of magnitude (P rAl = 0:015 and P rAZ 91 = 0:024), hence the Nusselt
number expected should also be similar for the same convective condition
(Nu = f (Re; Gr; P r)).

All experiments carried out under forced convective conditions are grouped
together in Figure 22. This Figure includes experimental results for 3cm, 5cm
and 7cm Aluminum spheres. In these experiments, the spheres were at room
temperature prior to immersion. Experimental results in which the Aluminum
spheres where preheated at T0 = 450o C are also shown in Figure 22. Finally,
experimental results for the AZ91 Magnesium alloy are also depicted in the
same Figure.

5 Predicting the Nusselt Number for Other Material Systems

The model has been validated with experimental results in Aluminum and
AZ91 (P r » 10¡2 ) as well as with experiments in the water/ice system (P r »
101 ) obtained from the literature. The Nusselt number can then be predicted
for various ‡uids having di¤erent Prandtl numbers. This procedure was carried
out by running the numerical model for several conditions corresponding to
di¤erent Prandtl number ‡uids. With the Nusselt number obtained for values
of the Reynolds and Prandtl numbers, a correlation of the form N u = 2 +
C:Ren Re P rnP r is sought. The range of values of the dimensionless parameters

24
 Mixed Forced convection;
Convection. Forced
Numerical
Experimental Convection.
Forced
Model
results Experimental
convection; results
Numerical Model
Forced Convection;
Natural Numerical
Convection, Model, Al
Experimentalspheres; T 0=20ºC; MT F =1.5
Results.
Natural
30 Convection;
10 -2 Numerical
10
3 Model,
10 -1 Al spheres; 10T00=20ºC;
10
4 MT F=1.5
101
40
35 40 -3 4
15 25 D=3cm Pr=3x10 2 +2σ
1030 D=5cm Re=3x10
D=3cm -23
Pr=1x10 +2σ
D=3cm -1
Re=3x10
D=7cm Pr=1x10
D=5cm 4
10
25
10
2
202 D=5cm 1
Re=3x10
2
10102
20 AZ91Pr=1x10
D=7cm 2
D=7cm
(Nuexper-Nucorr FC)/Nuexper ,%

corr NC
correlation
correlation
10 20
15
0 0
Nu-2 Nu

15
Nu-2
Nu

-2σ
1 1
10 1 -20 10 1-
10
10 10
10
Al, D=3cm
5 Al, D=5cm
-2σ Al, D=7cm
-40 Al, T0 = 450ºC -
AZ91
5 Numerical Prediction

1050 -60 -2 -1 0 1010 -

3 10 10 4 10 10105 0
10010 10 10
Re G
3 4 5

Fig. 22. Numerical model predictions and experimental results in Aluminum and
AZ91 in forced convection (P r » 10¡2 ).

studied are 3 £ 10¡3 · P r · 101 and 3 £ 102 · Re · 3 £ 104 .

The sphere is set to be solid at the same thermal properties of the liquid
and at an initial temperature equal to its melting point (T0 = Tm ) in order
to avoid the shell formation and the subsequent error in estimating the mass
increase and the melting time factor, M TF . For each condition, the melting
time is obtained and the Nusselt number is calculated using equation 6 with
M TF = 1; as done for the water/ice system.

Figure 23 shows the Nusselt number as a function of the Prandtl number for
the values of the Reynolds number studied. The slopes of the curves are almost
identical among each other in the log-log plot, meaning that the exponent of
the Prandtl number (nP r ) is independent of the Reynolds number. Figure 24
shows the relationship between the Nusselt and the Reynolds number for four
of the total of eight values of the Prandtl number studied. The exponent of
the Reynolds number (nRe ) appears to be independent of the Prandtl number.

By performing a regression analysis of the data from the results of the model,
an expression for the dimensionless heat transfer coe¢cient is obtained in the
range 3 £ 10¡3 · P r · 101 and 3 £ 102 · Re · 5 £ 104 for the 24 points

25
 Mixed Convection. Forced
Experimental
Forced convection; Convection.
Forced
results
Numerical Model Experimental
convection; results
Numerical Model
Forced Convection;
Natural Numerical
Convection, Model, Al
Experimentalspheres; T =20ºC; MT F=1.5
Results.
Natural Convection; Numerical Model, Al spheres; 0 0
10T 0=20ºC; MTF=1.5
-2 -1 1
10 103 10 10 4 10
40
25
35 40 -3
15 D=3cm Pr=3x10 2 +2σ
1030 D=5cm Re=3x10
D=3cm -23
Pr=1x10 +2σ
20 D=3cm -1
Re=3x10
D=7cm Pr=1x10
D=5cm 4
10
25
2
102 D=5cm 1
Re=3x10
2
10102
20 AZ91 Pr=1x10
D=7cm
(Nuexper-Nucorr F C)/Nuexper ,% D=7cm
corr NC
correlation
correlation
10 20
15

0
Nu-2 Nu

15
Nu-2

-2 σ
Nu
Nu
Nu

1 1
10 101 -20 10 1
10 10
10
Al, D=3cm
5 Al, D=5cm
-2σ Al, D=7cm
-40 Al, T0 = 450ºC
AZ91
5 corr FC n = ½
Model
1050 -60 100
3 10-2 10-1 104 100 101
100 10 100
Re G
3 -2 -1 4 0 5
10 10
6
10 10
3
7
10 10 10 8
10 10 4 8 10 1
107
10
Pr
GrRe 10
Gr Re

Fig. 23. Nusselt number as a function of the Prandtl number for forced convection
on spheres.

considered. The problem is as follows: …nd the constants fC; nRe ; nP r g ¸ 0

such that I is minimized. The performance index I is given by equation 16.
The resultant correlation is equation 17, with a standard deviation ¾ corr = 4%:

N
X
I= (Numodel;i ¡ (2 + C:Rein Re P rin P r ))2 (16)
i=1

N u = 2 + 0:47Re0:5 P r 0:36 (17)

Equation 17 is used to predict the dependence of the Nusselt number on the

Reynolds number for the experimental results shown in Figure 22. In this
Figure, the straight solid line shows predictions based on equation 17. As
seen, there is a good agreement between the various experimental results and
predictions based on equation 17. The observed deviation on the lower range
of Reynolds numbers is partly attributed to buoyancy e¤ects.

Uncertainty analysis on the Nusselt number estimation from equation 17 has

shown that the error involved is of the order of 20%.

26
 Mixed Forced convection;
Convection. Forced
Numerical
Experimental Convection.
Forced
Model
results Experimental
convection; results
Numerical Model
Forced Convection;
Natural Numerical
Convection, Model, Al
Experimentalspheres; T =20ºC; MT F=1.5
Results.
Natural Convection; Numerical Model, Al spheres; 0 0
10T 0 =20ºC; MTF=1.5
-2 -1 1
10 103 10 10 4 10
40
3525 40 -3 40
15 D=3cm Pr=3x10 2 +2σ
1030 D=5cm Re=3x10
Pr=1x10
D=3cm 3
-2 +2σ
20 D=3cm -1
Re=3x10
D=7cm D=5cm 4
Pr=1x10
102 2 D=5cm 2 2
25
10 20 AZ91 Re=3x10
D=7cm 1
Pr=1x10 1010
20
D=7cm
(Nuexper-Nucorr FC )/Nuexper ,%
corr NC
correlation
correlation
10 2015

0 0
Nu-2 Nu

15
Nu-2
Nu

-2σ
Nu

1 1
10 101 -20 10 1 -2
10 10
10
Al, D=3cm
5 Al, D=5cm
-2σ Al, D=7cm
-40 Al, T0 = 450ºC -4
AZ91
5 corr FC n = ½
Model
0 0
105 -60 -2 -1 0 101 -6
3 10 10 104 10 10
100 10 100
Re G
3 4 5
106 10 10 -2 10 7
3 7 10 -110 100
10 8 10 4 8 10 1
10
10 10
GrRe Pr
Gr Re

Fig. 24. Nusselt number as a function of the Reynolds number for forced convection
on spheres.
6 Conclusions

A mathematical model was developed to describe the various transport phe-

nomena involved when a melting sphere is immersed in a moving ‡uid. This
model was validated with various experimental results involving liquid metals
and water.

Based on this model, a dimensionless correlation for convective heat transfer

to spheres was developed.

1=
N u = 2 + 0:47Re 2
P r0:36 (18)

3 £ 10¡3 · P r · 101 ; 102 · Re · 5 £ 104

This correlation has applicability in ‡uids with wide range of Prandtl numbers,
and it was compared with experimental results derived in liquid Aluminum
and Water. The comparisons have shown good agreement between predictions
from the derived correlation and experimental results.

27
References

[1] C. Hsu, Heat transfer to liquid metals ‡owing past spheres and elliptical rod
bundles, International Journal of Heat and Mass Transfer, Vol 8 (1965) 303–315.

[2] S. Sideman, The equivalence of the penetration theory and potential ‡ow
theories, Industrial and Engineering Chemistry Research, Vol 58 (2) (1966)
54–58.

[3] F. Kreith, L. Roberts, J. Sullivan, S. Sinha, Convection heat transfer and

‡ow phenomena of rotating spheres, International Journal of Heat and Mass
Transfer, Vol 6 (1963) 881–895.

[4] L. Witte, An experimental investigation of forced convection heat transfer from

a sphere to liquid sodium, ASME Journal of Heat Transfer, Vol 90 (1968) 9–12.

[5] S. Argyropoulos, A. Mikrovas, An experimental investigation on natural and

forced convection in liquid metals, International Journal of Heat and Mass
Transfer, Vol 39 (1995) 547–561.

[6] S. Argyropoulos, A. Mikrovas, D.A.Doutre, Dimensionless correlations for

forced convection in liquid metals: Part i. single-phase ‡ow, Metallurgical and
Materials Transaction, Vol 32B (2001) 239–246.

[7] McAdams, Heat Transmission, 3rd edition, McGraw-Hill, 1954.

[8] T. Yuge, Experiments on heat transfer from spheres including combined natural
and forced convection, ASME Journal of Heat Transfer, Vol 82 (1960) 214–220.

[9] G. Vliet, G. Leppert, Forced convection heat transfer from an isothermal sphere
to water, ASME Journal of Heat Transfer, Vol 83 (2) (1961) 163–175.

[10] C. Hieber, B. Gebhart, Mixed convection from a sphere at small reynolds and
grashof numbers, Journal of Fluid Mechanics, Vol 38 (1969) 137–159.

[11] C. Vanier, C. Tien, Free convection melting of ice spheres, AIChE Journal, Vol
16 (1970) 76–82.

[12] A. Solomon, On the melting time of a simple body with a convection boundary
condition, Letters in Heat and Mass Transfer, Vol 7 (1980) 183–188.

[13] V. Eskandari, G. Jakubowski, T. Keith, Heat transfer from spherical ice in

‡owing water, ASME 82-HT-58.

[14] A. Anselmo, V. Prasad, J. Koziol, Melting of a sphere when dropped in a pool

of melt with applications to partially-immersed silicon pellets, Heat Transfer in
Metals and Containerless Processing and Manufacturing, ASME HTD, Vol 162
(1991) 75–82.

[15] A. Anselmo, V. Prasad, J. Koziol, K. Gupta, Numerical and experimental study

of a solid-pellet feed continuous czochralski growth process for silicon single
crystals, Journal of Crystal Growth, Vol 131 (1993) 247–264.

28
[16] S. Aziz, W. Janna, G. Jakubowski, A comparison of correlations for forced
convection heat transfer from a submerged melting ice sphere to ‡owing water,
ASME Heat Transfer Division, Vol 334-3 (1995) 329–334.

[17] Y. Hao, Y. Tao, Heat transfer characteristics in convective melting of a solid

particle in a ‡uid, ASME Heat Transfer Division, Vol 364-2 (1999) 207–212.

[18] M. Mukherjee, J. Shih, V. Prasad, A visualizations study of melting of an ice

sphere in a pool of water, ASME 94-WA/HT-14.

[19] P. McLeod, D. Riley, R. Sparks, Melting of a sphere in hot ‡uid, Journal of

Fluid Mechanics, Vol 327 (1996) 393–409.

[20] Y. Hao, Y. Tao, Melting of a solid sphere under forced and mixed convection:
Flow characteristics, Journal of Heat Transfer, Vol 123, No 5 (2001) 937–950.

[21] R. Nordlie, F. Kreith, Convection heat transfer from a rotating sphere,

International Developments in Heat Transfer, ASME (1961) 461–467.

[22] E. Sparrow, R. Eichhorn, J. Gregg, Combined forced and free convection in a

boundary layer, Physics of Fluids, Vol 2, No 3 (1959) 319–328.

[23] B. Melissari, S. Argyropoulos, The identi…cation of transition convective regimes

in liquid metals using a computational approach, Progress in Computational
Fluid Dynamics, an International Journal, Vol 4, No 2, Inderscience Publishers
(2004) 69–77.

[24] S. Whitaker, Forced convection correlations, AIChE Journal, Vol 18 (2) (1972)
361–371.

[25] S. Patankar, Numerical Heat transfer and Fluid Flow, McGraw-Hill, 1980.

[26] S. Argyropoulos, N. Goudie, M. Trovant, The estimation of thermal resistance

at various interfaces, Minerals, Metals and Materials Society/AIME, Fluid Flow
Phenomena in Metals Processing (USA) (1999) 535–542.

[27] E. Turkdogan, Physical Chemistry of High Temperature Technology, Academic

Press, 1980.

[28] P. R. H. C. Touloukian, Y.S., P. Klemens, Thermophysical Properties of Matter,

The TPRC Data Series, Vol 1, IFI/Plenum Press., 1970.

[29] S. Paterson, Propagation of a boundary of fusion, Proceedings Glasgow

Mathematical Association 1 (1952) 42–47.

[30] B. Melissari, The Sphere Melting Technique: Mathematical Modelling,

Experimental Measurements and Applications, Ph.D. Thesis, University of
Toronto, 2004.

[31] C. Gau, R. Viskanta, Melting and solidi…cation of a pure metal on a vertical

wall, ASME Journal of Heat Transfer, Vol 108 (1986) 174–181.

29