Chapter 6

Solidication

Almost all processing of technologically important metals includes a step

where liquid material is cooled to form a solid. It is very common, for exam-
ple, for the components of metallurgical alloys to be mixed in the liquid state
so that, upon solidication, an alloy with the desired composition is formed.
In order to form a solid from an undercooled melt we need formation of crys-
talline nuclei and growth of these nuclei to form a solid. Transformations
which occur by these two processes, nucleation and growth, are examples of
rst order phase transitions. Before beginning our discussion of nucleation
and growth of pure materials, we make some general comments about phase
transitions.
The condition of thermodynamic equilibrium implies minimization of
some free energy. For processes at constant pressure P , we minimize the
Gibbs free energy, G = H T S, while for processes at constant volume V
we minimize the Helmholtz free energy F = U T S, where H and U are the
enthalpy and internal energy respectively and are related by H = U + P V ,
and S is the entropy and T the temperature. For most solid-liquid or solid-
solid transitions it is usually more practical to work at constant pressure,
so G is the appropriate thermodynamic potential. However, at atmospheric
pressure, the changes in the P V term for condensed phase reactions are usu-
ally small compared to changes in the other terms, so either F or G can be
used.
A rst order phase transition involves two distinct phases which are phys-
ically separated by a sharp1 interface, across which there is usually a sym-
1
It is often impossible to distinguish the phase of the atoms right at the interface, so

157
158 CHAPTER 6. SOLIDIFICATION

metry change. Each phase is represented by a free energy which is a function

of temperature, pressure (or volume), and composition. In equilibrium, the
system will congure itself in the phase which has the lower free energy.
Points where the free energies of the two phases cross are transition points.
For example, consider the free energy as a function of temperature for solid
and liquid phases shown in Fig. 6.1. The free energies of the liquid and solid
cross at the melting temperature Tm . In equilibrium, the free energy of the
system is a continuous curve following the free energy of the liquid above Tm
and that of the solid below Tm . There is a slope change in the free energy at
the Tm , so entropy given by:
 
G  F 
S=  = 
T P T V

will have a discontinuity at the transition point:

   


 Gsolid  Gliquid  

SF = S liquid
S solid
 =   
T =Tm T P T P 
T =Tm

A liquid which is cooled innitely slowly will transform to the solid phase
at the melting temperature. However real transitions occur at nite rates,
and thus involve excursions from equilibrium which provide the driving force
for the transition. A liquid cooled below Tm will have a free energy higher
than that of the solid by an amount:

G = Gliquid Gsolid

Since solidication will decrease the free energy of the system, G is said to
be the driving force for the transformation. This is exactly analogous to the
driving force for diusion being provided by gradients in external potentials
or gradients in the non-ideal part of the chemical potential (Sec. 4.3).
We can estimate this driving force with a very simple expression if we are
willing to assume that there is no dierence in the heat capacities of the two
phases. Since G = H T S we nd for the temperature derivatives:
  
G  H  S 
 = S =  T  S
T P T P T P

even sharp interfaces may be spread out over a few atomic distances.
 159

G
G

solid
liquid

TM T
liquid
S
S F
solid

TM T
Figure 6.1: Schematic of free energy versus temperature for solid and liquid
phases.
160 CHAPTER 6. SOLIDIFICATION

So that:   
H  S  2 G 
CP  =T  = T 
T P T P T 2 P
where CP is the heat capacity at constant pressure, which is the change
in enthalpy per change in temperature. We can then nd the dierence in
enthalpy between the liquid and solid states by integrating from the transition
temperature:
 T
H = H liquid
H solid
= HF + CP dT
Tm

where HF is the heat of fusion which is the heat released by solidication

given by:

HF = H liquid H solid 
T =Tm

and CP = CPliquid CPsolid . Similarly we can nd the entropy dierence:

 T
CP
S = S liquid
S solid
= SF + dT
Tm T
Since G = H T S, and at T = Tm G = 0 we see that:

HF = Tm SF (6.1)

If we now assume that CP = 0, that is the curvature of the free energies of

the two phases are zero or equal, then we see that:

G = HF T SF

Inserting Eqn. 6.1 we nd:

HF
G = HF T
Tm


Tm T
= HF (6.2)
Tm
Equation 6.2 is known as the Turnbull extrapolation. It is valid for most
practical solidication situations, except in cases of extremely rapid cooling
where large undercoolings can be present.
In contrast to the behavior of a rst order phase transition, a second
order phase transition has both a continuous free energy and a continuous
 161

entropy (free energy slope). The transition usually occurs over a temperature
range of several degrees rather than at one point. The free energy is usually
represented as a single continuous curve, while the entropy has separate
curves for the two phases. There is a discontinuity in slope of S, which
results in a discontinuity in heat capacity, as shown in Fig. 6.2.
Some structural transitions are second order phase transitions, for ex-
ample the FCC to FCT transition in BaTi03 . Order-disorder transitions,
for example the to  transition in CuZn, can also be second order phase
transitions.

solid
G

liquid

TM T

S liquid

solid

TM T

liquid
cP

solid

TM T
Figure 6.2: Schematic of free energy, entropy and heat capacity as a function
of temperature for a second order phase transition.
162 CHAPTER 6. SOLIDIFICATION

6.1 Nucleation
We now return to solidication of a pure undercooled melt, which, of course,
is an example of a rst order phase transition. The thermodynamic driv-
ing force provided by the undercooling, must accomplish two processes; nu-
cleation and growth. In considering nucleation, we rst establish that the
surface or interface energy will result in an energy barrier which destabilizes
nuclei under a certain critical size. We then examine the population of nuclei
and the rate at which critical nuclei form to nd the nucleation rate.

6.1.1 Nucleation Barrier

We consider the free energy change associated with formation of a solid
particle from an undercooled liquid, as shown schematically in Fig. 6.3. There
are two terms, the rst is the negative, volume term associated with the bulk
condensation energy, and the second is the positive, surface term associated
with formation of the new interfacial area between the solid and the liquid.
Thus, if GT is the total free energy change associated with formation of a
solid particle, we can write:
V
GT = gA + ASL
VA
where V is the volume of the solid particle, A is the solid particle surface area,
VA is the atomic volume, gA is the free energy change per atom associated
with the liquid-solid transformation, and SL is the surface energy. The free
energy change per atom gA is related to the free energy change per volume
GV through:
gA = GV VA
If we take N to be the number of atoms in the solid particle, we can write:

GT = N gA + N 2/3 SL (6.3)

where is a shape factor dened by:

A
=
N 2/3
Since the negative, volume term of Eqn. 6.3 is linear in n while the pos-
itive, surface term varies as N 2/3 , the sum of the two terms will exhibit a
6.1. NUCLEATION 163

Liquid Liquid
Solid

Figure 6.3: Schematic of undercooled liquid and an undercooled liquid with

a solid particle.

peak, as shown in Fig. 6.4. The critical number of atoms, N , corresponding

to this can be found by setting:

GT 
 =0
N N =N

and solving for N , we nd:

 3
2SL
N = (6.4)
3gA
This expression for the critical number of atoms in a solid particle is valid
for arbitrarily-shaped particles.
The free energy, G , of a critical size solid particle can be found by
inserting N from Eqn. 6.4 into the expression for the free energy change as
a function of number of atoms Eqn. 6.3. We nd:

G = N gA + SL (N )2/3
4 3 SL
3
=
27 gA2
SL (N )2/3
= (6.5)
3
If our solid particles are spherical in shape, the number of atoms in a
particle with radius r is:
4
N= r3
3VA
and the shape factor becomes:
A 4r2 2/3
= = 2/3
= (36)1/3 VA
N 2/3 3
[4r /(3VA )]
164 CHAPTER 6. SOLIDIFICATION

Free Energy, GT

N*
surface
volume
total

Figure 6.4: GT versus N showing volume and surface contributions result-

ing in a peak at N = N .

so that the free energy change associated with forming a particle of radius r
is:
4 gA 4
GT = r3 + 4r2 SL = r3 GV + 4r2 SL
3 VA 3

We can nd the critical radius r by dierentiating this directly or by sub-
stituting the above expressions for and n into Eqn. 6.4:
2SL VA 2SL
r = =
gA GV
The energy change associated with formation of a particle of this radius is:
16VA2 SL
3 3
16SL
G = = (6.6)
3gA2 3G2V
If we consider the process of adding an one additional atom to a solid
particle of a given size, we see that for solid particles with N < N the addi-
tion of one more atom will increase the free energy, while for solid particles
with N > N , addition of an atom will decrease the free energy. Hence, solid
particles with N < N tend to shrink and disappear. These solid particles
are known as embryos. Solid particles with N > N will grow, since this
6.1. NUCLEATION 165

decreases the free energy. The process of adding one additional atom to a
solid particle with N = N is the nucleation process, since this results in a
nuclei, the growth of which will decrease the free energy.

6.1.2 Distribution of Solid Particle Sizes

The equilibrium distribution of solid particles is described by xN which is
the number fraction of solid particles with a given size. The concentration of
a particle of a given size can be found in a manner exactly analogous to that
used in nding the equilibrium concentration of vacancies. The formation of
clusters in the collection of atoms increases the number of ways the system
can arrange itself, and hence increases the entropy. Thus, even though the
enthalpy will increase if a sub-critical cluster is formed, there will be some
concentration of clusters of a given size which will minimize the free energy.
Thus xN is given by:

GT
xN = exp
kB T
where GT is given by Eqn. 6.3. The actual number fraction of solid particles
of a given size will be dierent from the equilibrium distribution since solid
particles with N > N will grow and will be removed from the distribution.
The Volmer-Webber theory of nucleation assumes that the actual number
fraction of solid particles with N atoms, ZN , is given by:

xN if N < N
ZN = (6.7)
0 if N > N

In this theory, the number of critically sized nuclei is given by:


G
Z = exp = xN
kB T
with G given by Eqn. 6.5 or 6.6.
The Becker-Doring theory of nucleation goes beyond the unrealistic as-
sumptions of Eqn. 6.7 and considers the eect of the nucleation process on
the population of particles. By considering the evolution of the population
distribution via the reactions:

N + 1 N+1
N+1 1 N
166 CHAPTER 6. SOLIDIFICATION

they nd (after some mathematical shenanigans which are beyond the scope
of this course) the number of critical sized nuclei:



1 G 1/2 G
Z = exp
N 3kB T kB T

1/2
1 G
= xN
N 3kB T

The factor:

1/2
1 G
z
N 3kB T

is known as the Zelldovich factor and was rst derived by Farkas. A com-
parison of the population of particles predicted by the Volmer-Webber and
Becker-Doring theories is shown in Fig. 6.5.

Volmer-Webber
xN
ZN

Becker-Doring

0 1
N/N*

Figure 6.5: Schematic of number of nuclei as a function of number of atoms

for Volmer-Webber and Becker-Doring nucleation theories. Dotted line is the
equilibrium number assuming no depletion due to growth.
6.1. NUCLEATION 167

6.1.3 Nucleation Rate

A nuclei will be formed each time an atom sticks to a critical sized particle.
Hence the nucleation rate, I, per volume is given by:
Z
I = RA
VA
where R is the sticking rate per site, is the sites per area on the critical
nucleus, and A is the area of a critical nucleus, and VA is the atomic volume.
In the case of a liquid-solid transformation, the attachment rate is exactly
analogous to the exchange frequency in the diusion process, and can be
expressed as:  
GM
R = 0 exp
kB T
where 0 is some vibration frequency, and GM is the activation energy for
the jump process. Thus we have for the nucleation rate:

1/2   
A 0 G G + GM
I= exp
VA N 3kB T kB T
The temperature dependence is dominated by G which is given by Eqn. 6.5
or for a sphere, Eqn. 6.6. We can use the Turnbull extrapolation to nd gA
for the liquid-solid case:
T
gA = hF
Tm
where hF = hL hS is the dierence in enthalpy per atom between the
liquid and solid states. For spherical nuclei we nd:
3
16SL
G =
3G2V
16VA2 SL
3
Tm2
=
3h2F (T )2
Relative to this 1/(T )2 dependence in the exponent the other terms are
nearly constant, and Turnbull has suggested that homogeneous nucleation of
the solid from an undercooled liquid can be approximated by:



1 G
I 10 exp
39
(6.8)
m3 s kB T
168 CHAPTER 6. SOLIDIFICATION

6.1.4 Heterogeneous Nucleation

In practice, nucleation of the solid phase rarely takes place homogeneously
in the bulk of the liquid. In most situations, nucleation initiates on the
walls of the containment vessel or on particles of high melting point oxides
inadvertently dispersed in the melt. These inhomogeneous nucleation sites
reduce the nucleation barrier by lowering the surface energy cost of forming
a nucleus. In fact, only by dispersing the liquid into droplets small enough
so that there is an appreciable chance of not having any nucleation sites in
a droplet, can homogeneous nucleation be achieved.
To examine how heterogeneities eect the energy of forming a solid phase
particle in an undercooled liquid, we consider formation of a spherical solid
cap of radius r and contact angle on a mold wall, as shown in Fig. 6.6. Here
we have three dierent interfacial energies to consider: SL is the surface
energy between the solid and liquid, SM is that between the solid and the
mold, and LM is that between the liquid and the mold. The contact angle
is determined by the surface energy balance:

LM = SM + SL cos

This expression, derived in Appendix C can be viewed as a force balance

considering each of the surface energies as a surface stress, although strictly
speaking is the surface energy and not the surface stress. It is straightfor-
ward geometry to show that the volume VS of the spherical cap is:
r3

VS = 2 3 cos + cos3
3
and the area ASL of the solid liquid contact is:

ASL = 2r2 (1 cos )

and the area ASM of solid mold contact is:

ASM = (r sin )2

The formation of this spherical cap involves: changing the liquid to the
solid, creating new interface area ASL and changing the interface energy from
LM to SM in the contact area ASM . Hence the free energy of formation is:

GT = VS GV + ASL SL + ASM (SM LM )

6.1. NUCLEATION 169

SL
liquid

solid

LM SM
r mold

Figure 6.6: Schematic of a solid spherical cap with radius r and contact angle
forming on a mold wall.

which can be shown to be given by:

 !
4
GT = r3 GV + 4r2 SL S()
3
where S() is a contact angle dependent function:
S() = (2 + cos )(1 cos )2 /4
which is shown plotted in Fig. 6.7.
Apart from the factor of S(), this is the same expression we obtained for
homogeneous nucleation. The energy of formation will again have a peak at
a the critical radius r given by:
2SL
r =
GV
but the energy barrier will be reduced by a factor of S():
3
16SL
G = S()
3G2V
This is shown schematically in Fig. 6.8.
Since Ghet is much less than Ghomo , nucleation can occur at much
smaller undercoolings. This is a result of the increased ratio of volume to
surface energy for the spherical cap compared to a sphere.
170 CHAPTER 6. SOLIDIFICATION

1.0
S()

0.5

0.0
0.0 0.5 1.0
/

Figure 6.7: Plot of the reduction factor S() for nucleation barrier for het-
erogeneous nucleation.

Homogeneous
GT

Heterogeneous

0 1 2
r/r*

Figure 6.8: Formation free energy as a function of particle radius both homo-
geneous and heterogeneous nucleation of a solid particle in an undercooled
liquid.
6.1. NUCLEATION 171

Recall that the probability of a critical nucleus occurring on a site is:


G
exp
kT
so that even though the number of possible sites for nucleation is much
smaller for heterogeneous nucleation, the probability of it happening on a
given site is so much higher that heterogeneous nucleation usually domi-
nates.

6.1.5 Numerical Estimation of Nucleation Rates

So far we have discussed nucleation with only passing reference to any real
numbers. We can make some good headway toward estimating nucleation
rates by using a couple of simple assumptions which are in the spirit of
those often made this course. The result of this simple analysis is a good
starting point for estimations of nucleation rates and helps to solidify our
understanding of the underlying materials science.
We rst consider the critical radius r , given by
2
r =
GV
where is the interface energy, in this case between liquid and solid, and
GV is the volume driving force for the transition, given by
T hF T
GV HF =
Tm VA Tm
where VA is the atomic volume, HF and hF are the enthalpy of fusion per
volume and per atom respectively, and T and Tm are the undercooling and
melting temperature respectively. The interface energy can be estimated
as
FBB hF
where FBB is the fraction of broken bonds at the interface and is the atom
surface density. Using hF = HF VA and
(1/VA )2/3
and taking FBB 1/2 as found for liquid/solid interfaces in our chapter on
surfaces, we can nd
1/3 Tm
r VA
T
172 CHAPTER 6. SOLIDIFICATION

for a fcc with lattice parameter a this becomes

a Tm Tm
r 1/3
= 0.63a
(4) T T

For Cu with a lattice parameter of a 3.6 A and a melting temperature of

Tm 1358 K, we nd a critical radius of about 23 A for an undercooling of
10% which is about 136 K.
Now lets turn to the nucleation rate. Starting with the nucleation event
being dened as the process of adding an atom to a critically sized nucleus
we found:
Z
I = RA
VA

where A is the area of a critically sized nucleus, R is the attachment rate,
is the number of atomic sites per area, Z is the number fraction of criti-
cally sized clusters and VA is the atomic volume. Using The Becker-Doring
expression for Z we nd

1/2   
A 0 G G + GM
I= 5/3
exp
VA N 3kB T kB T

where N is the number of atoms in a critically-sized nucleus and we have

again used 1/VA . G is the nucleation barrier given (for spherical
2/3

nuclei) by
16 3 16FBB3
hF
G = 2
=
3GV 3(T /Tm )2
Taking
4
A = 4(r )2 and N = (r )3
3VA
and again using the above expression for the interface energy we nd
   
20 FBB hF G + GM
I= exp
VA kB T kB T
Further, we recognize that nucleation takes place close to the melting point
so hF /(kB T ) hF /(kB Tm ). Earlier in the course we noted that the
enthalpy of fusion scaled with the melting point, and that for a large number
of materials
hF
1.1
kB Tm
6.1. NUCLEATION 173

Using again FBB 1/2 we nd

  
1.5 0 G + GM
I exp
VA kB T

To make further progress we must estimate the migration energy. This is

dierent from the activation barrier for diusion in that it need not contain
the cost of forming a defect. Hence our earlier scaling relations wont be
much help. We instead just assume that the energy cost for migration from
liquid to solid of the order of the dierence in bond strength between solid
and liquid. Hence we nd
2
GM
hF
z
Hence    
GM 2Tm hF
exp exp 0.8
kB T zT kB Tm
where we have assumed we are near T = Tm . Hence it appears that this
migration energy cost does not play a big role in nucleation at small under-
coolings2 .
Lastly we must make some estimate of 0 . This is likely close to but less
than a typical atomic vibration, and in absence of other information we take
0 1012 /s
Putting this all together we nd


1.2 1012 G
I exp
VA kB T

 
1.2 1012
Tm 2 Tm
exp 2.3
VA T T

Using the volume per atom of Cu we nd a prefactor of 1 1041 /m3 s, which

is a factor of 100 times that obtained by Turnbull, who assumed a a factor
of 10 higher frequency, larger atomic volume and a larger migration energy
cost (taken to be that of viscous ow so exp(GM /kB T ) 102 ). This
uncertainty in prefactor is endemic to nucleation theory and and variations
2
As the temperature decreases, nucleation will be curtailed by this migration energy
cost, and with rapid cooling, can sometimes be suppressed completely, resulting in an
amorphous glass.
174 CHAPTER 6. SOLIDIFICATION

by a factor 2 in the exponent are common. As discussed below, this is not

as bad as it seems as the undercooling where copious nucleation is observed
is relatively insensitive to the prefactor. In fact to a good approximation the
nucleation rate is just given by


0 G
I const. exp
VA kB T
where the constant is in the range of 102 1.
We can plot the nucleation rate per atom site as a function of undercool-
ing as shown in Figure 6.9. The sudden onset of copious nucleation occurs
somewhere between 20% and 25% undercooling. For example, a nucleation
rate of one per cubic micron per second corresponds to IVA 1.2 1011 /s,
which occurs at an undercooling of about 23%. This gure also shows that
the undercooling at which copious nucleation occurs is relatively insensitive
to the prefactor, since prefactors which dier by a factor of 100 only change
the onset of copious nucleation by a fraction of a percent.
The onset of copious nucleation at 23% is at a rather higher undercooling
than is usually achieved in practice3 . In most practical situations nucleation
occurs at undercoolings below 10%. This is due to the dominance of hetero-
geneous nucleation on the mold wall or on refractory particles in the melt.
Heterogeneous nucleation takes place on the interface between the melt
and the solid surfaces in which the melt is in contact (the mold walls or
inclusions in the melt). We can apply the same analysis to nd the nucleation
rate per area for this case with only a slight modication of our starting
expression

I het = RA Z het
where the only dierence is that now Z het is the number fraction of critically
sized particles, so Z het is the number of critically sized particles per unit
interface area. We modify the Becker-Doring expression for the homogeneous
case to nd:  1/2  
het 1 Ghet Ghet
Z = exp
N 3kB T kB T
where Ghet is the heterogeneous nucleation barrier given by
Ghet = G S()
3
in fact one could even question the validity of using the Turnbull free energy extrap-
olation down this far
6.1. NUCLEATION 175

-12
100x10 5000

80 4000
FBB
I VA (1/s)

60 0.5 3000

N*
0.45
40 0.40 2000
0.35
20 1000

0 0
0 5 10 15 20 25
T/Tm (%)

Figure 6.9: Nucleation rate per atom volume as a function of fractional un-
dercooling calculated using the assumptions detailed above. The dashed line
corresponds to the nucleation rate of 1 per second per cubic micron. The
dotted line is the same expression with just a factor of 100 smaller prefactor.
Also shown is the number of atoms in the critical nucleus. Numbers corre-
spond to values of FBB . We see that a change of 10% in the estimation of
the interface energy has a change in predicted nucleation rate comparable to
a change of a factor of 100 in the prefactor.
176 CHAPTER 6. SOLIDIFICATION

Hence the heterogeneous nucleation rate is given by:

 1/2   
A 0 Ghet Ghet + GM
I het
= 4/3
exp
VA N 3kB T kB T
2/3
where we have again made the substitution = 1/VA . Using
2 4(r )3

a = 2(r ) (1 cos ) and N = S()
3VA
we nd
   
20 FBB hF Ghet + GM
I het = 2/3
exp
VA kB T (2 + cos ) kB T
It is interesting to compare our estimations for homogeneous and het-
erogeneous nucleation rates. To do so we consider a collection of N atoms
with volume V = N VA . Homogeneous nucleation takes place throughout the
sample volume so the total number of homogeneous nucleation events per
unit time is
N hom = V I hom
However heterogeneous nucleation takes place only on the area of the sample
in contact with the mold. Hence the total number of heterogeneous nucle-
ation events per unit time is
N het = AI het
A is the area of the mold wall where heterogeneous nucleation will occur. If,
for simplicity, we assume that nucleation takes place at the container walls,
we can say A = V 2/3 . To compare these we can multiply each by VA /V to
put it on a per atomic site basis, so the quantities we compare are
N hom VA
= I hom VA
V    
FBB hF G + GM
= 20 exp
kB T kB T
and
N het VA VA
= I het 1/3
V V    


VA 1/3 FBB hF Ghet + GM
= 20 exp
V kB T (2 + cos ) kB T
6.2. GROWTH 177

We can see right away that the prefactor for the heterogeneous case is smaller
by the factor4


VA 1/3
V
Taking a typical volume of 1 liter = 1103 m3 , and using the atomic volume
of Cu, we nd


VA 1/3
= 2.3 109
V
Hence the prefactor for the heterogeneous nucleation rate is much smaller.
However, the lowering of the nucleation barrier by a factor of S() more
than compensates. We can estimate the undercooling range over which het-
erogeneous nucleation dominates. Setting N het = N hom and solving for the
undercooling gives:
   
1/2
T  3
16FBB [1 S()] hF Tm

Tm c ln(V /VA ) kB Tm T
Using numbers as above, assuming a volume of 1 liter, and taken = 60
corresponding to the case where SL LM 2SM , so that S() 0.16, we
can nd the crossover undercooling or about 40%. Hence, we predict that
heterogeneous nucleation dominates for the most undercoolings of practical
interest. This is shown in Figure 6.10, where we show both the homogeneous
and heterogeneous nucleation rate for a volume of 1 liter.

6.2 Growth
6.2.1 Atom Attachment Rate
Once a nucleus has formed, the transformation proceeds with the process of
growth, where atoms are transported across the interface from one phase to
the other. The undercooling which has paid the thermodynamic price for
the formation of the interface surrounding the nucleus, provides the ther-
modynamic driving force for this process as well. The process of an atom
jumping from the liquid to the solid is a diusive jump with a driving force,
just as we saw in diusion with a chemical concentration gradient. (Sec. 4.3)
The energy as a function of position during this jump is shown in Fig. 6.11
4

Ignoring the factor 1/ 2 + cos .