The Arabian Journal for Science and Engineering

EFFECT OF BOUNDARY CONDITIONS ON FREEZING IN SPHERICAL POROUS

MEDIA
--Manuscript Draft--

Manuscript Number:
Full Title: EFFECT OF BOUNDARY CONDITIONS ON FREEZING IN SPHERICAL POROUS
MEDIA
Article Type: SCI / ENG - Research Article
Section/Category: ENG-Mechanical Engineering
Keywords: Porous media; Heat Transfer; mass transfer; solidification; cooling; moving
boundary problem; rapid solidification; freezing; phase transformation
Corresponding Author: Rahul Basu, PhD
SAIT
Bangalore, Ka INDIA
Corresponding Author Secondary
Information:
Corresponding Author's Institution: SAIT
Corresponding Author's Secondary
Institution:
First Author: Rahul Basu, PhD
First Author Secondary Information:
Order of Authors: Rahul Basu, PhD
Order of Authors Secondary Information:
Abstract: A model for coupled heat and mass transfer for freezing in a porous matrix with
Dirichlet, and convective boundary conditions and a heat sink is set up and solved.
Variables include porosity, heat transfer coefficients, thermal and mass diffusivity,
density, latent heat and boundary temperatures. A stability criterion for the phase
interface is linked with well-known parameters like under cooling and freezing rate. It is
shown that heat and mass transfer balance at the interface can affect stability. The
effect of time scales is studied with application to rapid quenching
Powered by Editorial Manager and ProduXion Manager from Aries Systems Corporation
Effect of Boundary Conditions on Freezing in Spherical Porous Media
Rahul Basu
Professor
V.T.U., SAIT
M.S.Palya, Jalahalli E
Bangalore 560097

A model for coupled heat and mass transfer for freezing in a porous matrix with Dirichlet, and convective
boundary conditions and a heat sink is set up and solved.. Variables include porosity, heat transfer
coefficients, thermal and mass diffusivity, density, latent heat and boundary temperatures. A stability
criterion for the phase interface is linked with well-known parameters like under cooling and freezing rate.
It is shown that heat and mass transfer balance at the interface can affect stability. The effect of time scales
is studied with application to rapid quenching.
INTRODUCTION:

While earlier technologies used mechanical pre and post treatment at elevated temperatures, later
developments focused on controlled solidification. Analysis of such moving boundary problems has
posed problems due to their non-linearity. The determination of the moving phase boundary has been
historically difficult to achieve, and solutions in closed form are available in very few cases. Some
references that give a historical background and survey earlier attempts are given in Saitoh [1], Duda [2],
and del Giudice [3]. More recent theoretical and experimental work is given in [4-6 ]. Stefan [7] was the
first to give a solution, and others are outlined in Paterson [8] and Carslaw & Jaeger [9]. A review has been
given by Basu & Date [ 10 ] for various models and methods with emphasis in laser melting.

ANALYSIS: Consider diffusion processes (i.e. flow without convection). The energy equation is modeled
as a balance condition at the liquid solid interface. Consider a diffusion model under temperature and
concentration gradients with no pressure assisted diffusion or gravity assisted flow. A porous
microstructure is assumed. We propose to look at different scenarios with types of b.c.s---a spherical
coordinate system with convective, Dirichlet and heat sink B.C., as also a porous slab (half space) under
Dirichlet B.C. (The model used neglects certain physical constraints on the actual freezing. Among these
are the cup shaped bottom profile of the liquid pool, the coupling of z with r, and the finiteness of the outer
radius) The Porosity parameter is a measure of the available volume for diffusion, and includes the
capillary pathways through which diffusion or transport occurs. The effect of the matrix is to provide
pathways for the mass and heat transfer. As a result, the diffusivities and heat transfer parameters are
effective parameters and must be empirically determined for accurate predictions to be made.. It may be
added that the spherical system of coordinates is easily converted to the one-dimensional form by a simple
change of variables.
Assumptions used:
1. porous media of uniform porosity c (Porosity allows mass diffusion)
2. saturated medium (in the sense that available mass does not change by adsorption or reaction)
3. temperature of freezing not constant, can be variable with use of additional equations (included in
cases of slab freezing (Dirichlet and self freezing)).
4. Heat sink at the origin is of delta function type.
The Boundary conditions (B.C.) are (non.dim)

Solid temp u
2
(infinity ,t ) =0
Freezing at temp um, (can be variable and solved for), solid conc. U(,t) =0
Final temp u
2
=0
The Initial conditions (IC) are
Initial molten liquid at temp u
0
(non dim)
(Note: the dimensional values are denoted as T and c, subscript meanings remaining the same)
Here U is the non dimensional concentration, u the non dim temperature, r the radial variable.
For the slab case, the normalization is done wrt the triple point values (fixed). Normalization used is T
i
/T
3

for Dirichlet and self-freezing.
Manuscript
Click here to download Manuscript: revised2.doc
Click here to view linked References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
For the spherical case the following are used
u
i
=(T
i
-T
0
)/(T
0
-T
m
)
U= 1-C/C
0
(note the change of sign for temperature is to allow both curves for temperature and concentration to be
plotted on the same graph)
Henceforth the calculations are in dimensionless variables
COMPUTATIONAL RESULTS Presented below in the appendix are some computational results on
Dirichlet problems (sphere) and the case of a heat sink at the origin for the sphere. Similar results are
obtainable for the slab by a simple change of variable for the spherical equations.

Q leads to a discontinuity at the origin with low temperatures as z- 0. On the other hand, the Dirichlet
and convective solutions are identical in so far as the solid phase is concerned, and also the functional form
of the liquid region solution. We observe briefly that since A2 is an additive constant for the convective
solution the profile will be shifted depending on the sign of A
2
. Comparing B
2
and B
3
it may be seen that
for large a and finite time, A
2
= gB
3
+xB
2
, where
g = SphErf (u ) , x = SphErf ( + ) 23

Since the shapes are similar for the two cases, the temp profile will be shifted up along the family of curves
corresponding to lower L as in the figure attached (Fig 1), implying that the interfacial position has moved
to the right. Hence faster freezing is expected from a mathematical analysis.
There are two time scales, which can be introduced
1. Due to the Fourier number (wavelength effect)
2. That brought about by the latent heat (due to the moving phase front
The two time scales are quite different in most cases, especially if freezing is slow. In effect, the
temperature sensed due to heat conduction of the normal boundary value problem, has superimposed on it a
'wave' of temperature caused by the motion of the phase front. The phase front is regard able as a heat
source, whereas another heat source or sink may be present at the origin, depending on the problem chosen.
Discussion:
The solution is done by the use of a decoupling constant |, which avoids the need to substitute the known
solution for concentration and computer particular solutions. The appendix illustrates the derivation, where
the particular value chosen here is a simpler one. The simultaneous differential equations of heat and mass
transfer can also be solved with IMSL routines (simultaneous solvers), or by algebraic simplification. A
secant solver or a bisection method can then be used to obtain the roots of the resulting transcendental
equation. The secant method is fast and accurate and avoids the need to compute Jacobians as in the
gradient search methods. In this paper Brents technique, outlined in Antia [11] is applied. Zero finding
routines given use bisection with linear and inverse quadratic interpolation with specified tolerance and end
limits [1 ,12 ]. A simultaneous solver used IMSL routines based on the Marquandt Levenberg minimization
method for the slab with coupled sublimation and freezing,[13 ].
It may be mentioned briefly that in using the Fourier number Fo, or q= x/(ot)
.5
( in case of the half space)
as a dependent variable in analyzing the figures, either time or distance can be fixed. In such cases, one
may fix ones attention on the physical process at any point in time varying distance, or at a particular
location as time progresses. The same figure applies to both cases. Hence, if one positions oneself at the
interface, by fixing r =s, the problem converts to a steady state problem in the sense that there is no change
in the solution with time, as the profiles and conditions at the interface remain mathematically fixed ( the
interface may of course become unstable or other changes may occur with time, these cannot be predicted
by the simplified model analyzed here).

Physically, as the radius of solidification increases, more area is involved and hence more latent heat
available for removal. The easiest way for the alloy to accomplish the removal is by change of heat flux
through a lowering of boundary temperature or an increase in heat conduction. . In the special case of
Splat cooling, the time scale for the entire transformation is so short and the thickness of the sample also
small that unless parameters are matched, the physics do not match the mathematical solution. Basically, a
moving phase front cannot form in the usual sense. However it may be remarked that for the analytical
solution to hold, the Fourier number should be recoverable (i.e. the ratio of time to distance squared should
be recognizable) and the ratio of latent heat to temperature difference also should be within the physical
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
limits of the usual freezing cases. For extremely small solidification times, a steady state approximation
may be used, where the transient terms and latent heat terms can be neglected. In such a case, the
temperature distribution is easily found. The magnitude and relative sizes of the parameters will determine
if such fast transformations are capable of being modeled by the above physical models.
Effect of b.c. on the rate of freezing
The effect of the boundary conditions on the morphological stability of the interface occurs in an indirect
way through the ability of the moving interface to be compatible with the heat and mass transfer gradients
on either side of the interface. Often, the interface breaks up. A study was done by NASA [14], regarding
optimum gradient parameters for many super alloys. The phase chemistry is also involved, so a purely
physical and thermal/thermodynamic prediction cannot be given in a sweeping manner.
However, it is hoped that the foregoing study points the way to the general directions to be taken and that
some useful results are highlighted and revaluated.

Perturbation Solutions:

Time Scales: Two time scales are possible--one using the Fourier number which influences conduction
through diffusivity, and another through the motion of the molten interface which is inversely related to
time and directly as the strength of the molten source or sink. The solution which is obtained is a
combination or superposition of these two effects, i.e. one from the usual thermal diffusion heat equation,
and the other imposed by the source or sink term.
For small F
0
eqn (2) can be written without the nabla terms, thus becoming one dim in U and u
On the other hand, if c is very small, the reverse holds, and the nabla terms appear. The equations can still
be solved exactly using a similarity transformation. However, some have chosen to solve term by term in a
perturbation expansion (e.g., Pedroso and Domoto [15]) Recently some papers have appeared which try to
analyze the singular terms close to the origin (heat sink), where the gradients are large. In the approach
presented in this paper, we do not focus on such aspects, merely on the phenomenological scaling.
In the case of splat cooling, where an unsteady phase is frozen in, the time scale is not proportionate to the
Fourier number, and in fact it is suggested by the above considerations that such a glassy liquid structure is
possible only if the interface is not allowed time to move forward.
Apart from these two time scales, an exponentially short time scale is treated by Stewartson [ 16 ]
INVERSE PROBLEMS: A large body of literature has appeared over the past decade or two where by
knowing the solution one attempts to work backwards and guess the initial and boundary conditions. Some
recent papers use least square techniques and conjugate gradient methods to recover data about heat
sources or material properties from observed measurements [17].
CONCLUSION In the foregoing, an attempt has been made to present an alternative model to earlier ones
based on convective channeling. Convective and buoyant effects may be appropriate for geometries where
the vertical coordinate is substantial compared to the radial. In the present geometry (foil and low aspect
ratio components in rapid solidification) only a thin region remains liquid - forming a disc or layer over
already solidified regions. Small amounts of gravity assisted flow and micro convection may exist in cells,
but it is doubtful if macro convective currents instigated by gravity and buoyancy would persist in such a
thin layer For dendritic porous media, however, the macroscopic effects of buoyancy and circulation will
be effective in local cells, not in a large region as postulated by some. As such, it is likely that diffusive
and chemical effects are predominant.

REFERENCES:
1 del Giudice S. Comini G, Lewis R W, Int J Num Anal Meth. Geomech. 2, 223-235, 1978
2. Duda J L Malone M F, Notter R H, Vrentes J S, IJHMT, 18, 901-910, 1975
3. Saitoh T L, J Heat Transfer, 100, 294-299,1978
4. Loria E A, Ed., 'Superalloy 718, Metallurgy & Applications',TMS, 1989
5. Loria E A, Ed., 'Superalloy 718,625 & various derivatives',TMS 1991
6. Loria E A , Ed., 'Superalloy 718,625,706 & various derivatives',TMS 1994.
7 Stefan J, Ann Phys u Chem, NF 42,269-86 1891.
8 Paterson S., Glasgow Math Assn. Proc., 1, 42-47, 1952.
9. Carslaw H S & Jaeger J C, 'Conduction of Heat in Solids', 2nd Ed., Clarendon Press, Oxford, 1959.
10. Basu B and Date R W, Sadhana, 13,169-213, 1988
11. Antia H M., 'Numerical Methods for Scientists & Engineers', TMH, Bombay 1991.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
12. Brent R P., 'Algorithms for minimization without derivatives', Prentice Hall, NJ, 1973.
13. Basu R D & Boles M A, IJER, V18, 449, 1994.
14 Lemkey FD, McCarthy G, Quarterly and quinary modifications of eutectic alloys strengthened by delta
Ni3Cb lamellae and gamma prime Ni3Al ppts, NASA CR 134678, Wash DC 1975
15 Pedroso RL and Domoto GA, J Heat Transf, 95,42-46, 1973
16 .Stewartson K and Waechter RT, Proc Roy Soc Lond A,348,415-426
17 Su J and Silva Neto A J, Appl. Math modeling, 25 ,861-872,2001

Given below are calculations specific to the spherical geometry. For the slab and self freezing cases,
the operator is replaced with the appropriate form. Dirichlet b.c.s similar to the foregoing were used
with similarity variable q = x/2(ot)
0.5


In the liquid region ( 0<r<s)

cu
1
/ct = o
1
/r
2
c /cr (r
2
cu
1
/cr ) 1
In the solid region ( s<r< )

cu
2
/ct = o
2
/r
2
c/cr (r
2
cu
2
/cr ) + cL/C
p
cU/ct 2
oi=thermal diffusivity in region i, c=porosity , =density
Cp =sp.heat, U =nondim conc., t = time
ot/r
2
= Fo=Fourier no.
o
12
= o
1
/o
2
The thermal properties are mainly those of the material pervading the pores, and due to this,
the discontinuity occurs at the phase interface. The main contribution of the matrix is in allowing porosity
and mass transfer. The second term in eqn 2 describes the remelt effect due to
solute rejection at the interface, also termed constitutional super cooling.

The equation for concentration is

cU/ct = o
m
/r
2
c/cr (r
2
cU/cr) 3
Where o
m
= mass diffusivity


The above equations are a set of coupled diffusion equations with Dirichlet boundary conditions

( Solid temp u
2
(infinity ,t ) =0
Freezing at temp um, (can be variable and solved for), solid conc. U(,t) =0
Final temp u
2
=0
The Initial conditions (IC) are
Initial molten liquid at temp u
0
(non dim))

The solution is effected by means of a decoupling variable
| (=cL/C
p
/(o
m
1)),
Obtaining a new equation in terms of Z,
where Z = u
2
+ |U (U=1-c/c
o
) 4
Knowing the solution for c, the liquid phase temperature is decoded. The solutions are obtained in terms of
the spherical error function used by Paterson (13) and also in Carslaw & Jaeger (14) A Bessel series
solution is also possible.

Thus cZ/ct = 1/r
2
c/cr (r
2
cZ/cr) 5
The solutions are obtained in terms of constants of integration Ai and Bi as follows:

u
2
=Z - |U = A
2
+B
2
[ exp(-r
2
/4t)/r -(t/2)
0.5
erfc(r/(2o
12
t)
0.5
]
|B
3
[(o
m2
t)
0.5
exp(-r
2
/4o
m2
t)/r- (t/2)
0.5
erfc(r/2(o
m2
t)
0.5
] 6
( 2 denoting Solid Region here for outward solidification)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Applying u
2
( , t) = 0 yields A
2
=0
Liquid region temp. u
1
= u
0
+ B
1
[ ( o
12
t)
0.5
Exp( - r
2
/4o
12
t )(t/2)
0.5
erfc( r/2(o
12
t )
0.5
) 7

Concentration solution U = A
3
+ B
3
[ (o
m2
t)
0.5
exp (-r
2
/4 o
m2
t)/r (t/2)
0.5
erfc( r/2(o
m2
t)
0.5
] 8

Using the conditions: A
2
= A
3
=0
B
1
=(u
m
-u
0
)/SphErf(-
2
), B
2
= (u + | B
3
SphErf(-o
1m

2
)/ (-o
12

2
) 9
B
3
is as shown below.
Where the terminology Sph Erf(x) is used to denote [Exp (x)-erfc(x)]
In addition we have one more condition at the moving phase interface which renders the problem non linear
due to the discontinuity in properties:

Energy balance: K
1
cu
1
/cr - K
2
cu
2
/cr = L(1-U)cs/ct , s = 2 \o
12
t 10
Mass balance: o
m2
cU/cr = (1-U)cs/ct 11

Convective Boundary conditions:

The same diffusion coupled system is posed, with the change that at r=a, the boundary, the radiation or
convection boundary condition is satisfied. Hence
generally
Kcu/cr ,at r=a = h ( u
2
- u
1
) 12

Solutions are obtained as before, the only difference being that A2 need not be zero. All other parameters
are similar to the Dirichlet solution in form and value, excepting A
2
and B
2
.
Thus:
A
1
= u
0

A
2
= kB
3
/h[ 2 |/a + h/k | Sph Erf ( u ) ] - B
2
k/h[2/a -h/k Sph Erf( + )] 13

B
2
= ( u
m
+ | B
3
SphErf( - o
1m

2
) -A
2
)/SphErf (-o
12

2
) 14


B
3
= o
12

2
/[o
12

2
SphErf(-o
1m

2
)- o
m2
exp (-o
1m

2
)] 15

s = 2 \o
12
t

+ and u are given as -a
2
/(4 o
m2
t) and - a
2
/(4 t) respectively

Note: For asymptotic time and large r, the Dirichlet and convective solutions are the same. Only for finite
time or radii, when A2 is not zero, is it possible to have control of the interface by variation of A2. Change
in A2 will affect the mass and heat balance at the phase interface. Carslaw and Jaeger (9 ) give another
solution of a line sink of strength Q, and claim this to be the only known solution for the cylindrical
freezing case. In addition, a constant freezing velocity solution is given.
HEAT SINK AT ORIGIN SOLUTION:
Here, a heat point source at the origin having a delta function form occurs. In effect there is a sphere of
solidifying, or melting, liquid or dendritic porous matrix and liquid.
The same equations are set up as before, the only change in the boundary conditions being due to the heat
sink Q at the origin.
lim [t r
2
cu / c r] = Q \( o
12
t) , 16
r 0
the solution for temperature is checked using the above relation.
The boundary conditions being:

u
2
(, t ) =0
U(,t )= 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Initial conditions:
u(r,0) =0, U(r,0) =0
where we have used the normalization
u
2
= (T
2
-T
0
)/(T
0
-T
m
)
U = 1-C/C
o
Solutions are:

Solid u
1
= u
m
+ Q/4t [(o
12
t)
0.5
Exp(-r
2
/4o
12
t] -(t/2)
0.5
Erfc (-
2
)] 17


liquid u
2
= Z -|U 18

Z = A Sph Erf (-R
2
/4t ) 19

where A = ( u
m
+ |D Sph Erf (-o
1m

2
))/SphErf(-o
12

2
) 20

D = o
12

2
/[o
12

2
SphErf (-o
1m

2
)- o
m2
exp (-o
1m

2
)] 21
s = 2 \o
12
t
Explicitly, u
2
= A SphErf (-r
2
/4t)- |D SphErf(- o
2m
r
2
/4t ) (liquid region) 22
These equations are similar to the Dirichlet solutions with the exception of the region 1 solution (solid)
because of the heat sink singularity.
SYMBOLS
A, B, D constants of integration
c concentration
cp specific heat
G thermal gradient
Fo Fourier number
h heat transfer coeff.
a boundary length
K thermal conductivity
L latent heat
m mass
R growth rate
r radial variable
s interfacial position s = 2 \o12t
T temperature
t time
U non dim concentration
Z decoupled solution variable
z axial coordinate
o, k diffusivity
| decoupling constant ( depends on the problem )
c porosity
q similarity variable
velocity of interface in similarity form (lambda) (lambda and eta are identical)
t nondimensional time
u nondimensional temperature
density
SphErf Spherical Error function
IMSL commercial software package for engineering and scientific programmes. Many of the routines are
listed in similar form by Antia, Numerical methods for Scientists and Engineers (ref 13)
NORMALISATION: U = 1 -c/co
u = (T-To)/( To- Tm) where To, Co are boundary temperatures and concentrations
Subscript 3 refers to triple point values-considered fixed for any substance, whereas melting point values
change with equation of state
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
SUBSCRIPTS:
0 for reference (initial or boundary)
3 steady state reference
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65