PHASE TRANSFORMATIONS

 Nucleation & Growth

 TTT and CCT Diagrams
 APPLICATIONS
 Transformations in Steel
 Precipitation
 Solidification & crystallization
 Glass transition
 Recovery, Recrystallization & Grain growth

Phase Transformations in Metals and Alloys

David Porter & Kenneth Esterling
Van Nostrand Reinhold Co. Ltd., New York (1981)
Phase Transformations: an overview
 When one phase transforms to another phase it is called phase transformation.
 Often the word phase transition is used to describe transformations where there is no
change in composition.
 In a phase transformation we could be concerned about phases defined based on:
 Structure → e.g. cubic to tetragonal phase
 Property → e.g. ferromagnetic to paramagnetic phase
 Phase transformations could be classified based on (pictorial view in next page):
 Kinetic: Mass transport → Diffusional or Diffusionless
 Thermodynamic: Order (of the transformation) → 1st order, 2nd order, higher order.
 Often subtler aspects are considered under the preview of transformations.
E.g. (i) roughening transition of surfaces, (ii) coherent to semi-coherent transition of
interfaces.
 PHASE TRANSFORMATIONS
Based on
Mass
transport
Diffusional Diffusionless
Involves long range mass transport E.g. Martensitic

PHASE TRANSFORMATIONS
Based on
order

1nd order 2nd (& higher) order

nucleation & growth Entire volume transforms
 Transformations in Materials
 Phase transformations are associated with change in one or more properties.
 Hence for microstructure dependent properties we would like to additionally ‘worry about’
‘subtler’ transformations, which involve defect structure and stress state (apart from
phases).
 Therefore the broader subject of interest is Microstructural Transformations.

Phases Defects Residual stress

Phases can transform Defect structures can change Stress state can be altered

Phase Defect Structure Stress-State

Transformation Transformation Transformation

Geometrical Physical
Structural Property

Phases Microstructure

Phases Transformations Microstructural Transformations

 Some of the questions we would like to have an answer for…

 What is a Phase?
 What kind of phases exist?
 What constitutes a transformation?
 How can we cause a phase transformation to occur?
 The stimuli: P, T, Magnetic field, Electric field etc.
 What kind of phase transformations are there?
 Why does a phase transformation occur?
 Energy considerations of the system?
 Thermodynamic potentials (G, A…)

Answers for some these questions may be found in other chapters

 Revise concepts of surface and interface energy before starting on these topics

 When a volume of material (V) transforms three energies have to be considered :

(i) reduction in G (assume we are working at constant T & P),
(ii) increase in  (interface free-energy),
(iii) increase in strain energy.
 In a liquid to solid phase transformation the strain energy term can be neglected (as the
liquid can flow and accommodate the volume/shape change involved in the transformation-
assume we are working at constant T & P).

Bulk Gibbs free energy ↓

New interface created

Energies involved Interfacial energy ↑

Strain energy ↑
Important in solid to solid transformations
Volume of transformed material
  The origin of the strain energy can be understood using the schematics as below. Eshelby
construction is used for this purpose.
 In general a solid state phase transformation can involve a change in both volume and
shape. I.e. both dilatational and shear strains may be involved. For simplicity we consider
only change in volume of the material, leading to an increase in the strain energy of the
system (in future considerations).
Bulk Gibbs free energy ↓

Energies involved Interfacial energy ↑

Strain energy ↑

(a)
Considering only
Only volume change
(c) volume change (d)
(b)

Schematic of the Eshelby construction to understand the origin of the stresses due to phase transformation
of a volume (V): (a) region V before transformation, (b) the region V is cut out of the matrix and allowed to
transform (the transformation could involve both shape and volume changes), (c) the transformed volume
(V‘- shown to be larger in the figure) is inserted into the hole (here only volume change is shown for
simplicity), (c) the system is allowed to equilibrate. The continuity of the system is maintained during the
transformation. The system is strained as a larger volume V’ is inserted into the hole of volume V.
 Let us start understanding phase transformations using the example of the solidification of
a pure metal. (This process is a first order transformation*. First order transformations
involve nucleation and growth**).
 There is no change in composition involved as we are considering a pure metal. If we
solidify an alloy this will involve long range diffusion.
 Strain energy term can be neglected as the liquid melt can flow to accommodate the
volume change (assume we are working at constant T & P).
 The process can start only below the melting point of the liquid (as only below the melting
point the GLiquid < GSolid). I.e. we need to Undercool the system. As we shall note, under
suitable conditions (e.g. container-less solidification in zero gravity conditions), melts can
be undercooled to a large extent without solidification taking place.

Bulk Gibbs free energy ↓

Energies involved Interfacial energy ↑
Strain energy ↑ Solid-solid transformation

Nucleation Growth till

1nd order Trasformation
** = of +  is
nucleation & growth
→  phase exhausted

* Click here to know more about order of a phase transformation

 Video snap shots of
1 2 solidification of stearic acid

Caution: here we are seeing an

Liquid increase time experiment and
Liquid → Solid phase transformation: Solidification

soon we will be ‘talking of’

increasing undercooling
Solid experiments
Growth of Crystal

3 4

Two crystal going to join to

form grain boundary

Crude schematic!
Grain boundary ↑t
5 6
Growth of nucleated crystal

See video here Solidification complete “For sufficient Undercooling”

Liquid → Solid phase transformation
 On cooling just below Tm solid becomes stable, i.e. GLiquid < GSolid.
 But even when we are just below Tm solidification does not ‘start’.
 E.g. liquid Ni can be undercooled 250 K below Tm.
 We will try to understand Why?
 The figure below shows G vs T curves for melt and a crystal.
 The undercooling is marked as T and the ‘G’ difference between the liquid and the solid
(which will be released on solidification) is marked as Gv (the subscript indicates that the
quantity G is per unit volume). Hence, Gv is a function of undercooling (T)

Solid stable Liquid stable

Gv
Solid (GS)

G → ve
G →

T Liquid (GL)
T - Undercooling G → +ve
Tm T →
TABLE: Degree of supercooling (ΔT) Values (Homogeneous
Nucleation) for Several Metals
 As pointed out before solidification is a first order phase transformation involving
nucleation (of crystal from melt) and growth (of crystals such that the entire liquid is
exhausted).
 Nucleation is a ‘technical term’ and we will try to understand that soon.
 In solid  solid phase transformation, which involve strain energy, heterogeneous
nucleation (defined below) is highly preferred. Even in liquid  solid transformations
heterogeneous nucleation plays an very important role.

Solidification = Nucleation + Growth

of crystals from melt of nucleated crystals till liquid is exhausted

Homogenous
Nucleation Heterogenous nucleation sites
Heterogeneous  Liquid → solid  walls of container, inclusions
 Solid → solid  inclusions, grain boundaries,
dislocations, stacking faults

 In Homogenous nucleation the probability of nucleation occurring at point in the parent

phase is same throughout the parent phase.
 In heterogeneous nucleation there are some preferred sites in the parent phase where
nucleation can occur
 Homogenous nucleation Let us start with a ‘text-book’ description of nucleation before taking up an alternate
perspective

 Let us consider LS transformation taking place by homogenous nucleation. Let the
system be undercooled to a fixed temperature T. Let us consider the formation of a
spherical crystal of radius ‘r’ from the melt. We can neglect the strain energy contribution.
 Let the change in ‘G’ during the process be G. This is equal to the decrease in bulk free
energy + the increase in surface free energy. This can be computed for a spherical nucleus
as below.

Free energy change on nucleation  Neglected in L → S transformations

Reduction in bulk free energy  increase in surface energy  increase in strain energy
ΔG  (Volume).(GV )  (Surface).( )

Gv  f (T ) 4 
ΔG   r 3 .(Gv )  4r 2 .( )  
3 
 Note that GV is negative

f (r ) r3  Note that below a value of ‘1’ the lower power of ‘r’ dominates;
while above ‘1’ the higher power of ‘r’ dominates.
 In the above equation these powers are weighed with other
r2 ‘factors/parameters’, but the essential logic remains.

r 1
 4 
 
ΔG   r 3 .(Gv )  4r 2 .( )
Funda Check A note on minimization versus criticality conditions. 3 

 In the above equation, the r3 term is +ve and the r2 term is ve. Such kinds of equations
are often encountered in materials science, where one term is opposing the process and the
other is supporting it. Example of such processes are crack growth (where surface energy
opposes the process and the strain energy stored in the material supports crack growth).
 In the current case it is the higher power is supporting the phase transformation. Since the
higher power dominates above ‘1’, the function will go through a maximum as in fig.
below. This implies the G function will go through a maximum. I.e. if the process just
even starts it will lead to an increase in G! (more about this soon).
 On the other hand the function with ve contribution from the lower power (to G) will
go through a minimum (fig. below) and such a process will take place down-hill in G and
stop.
1
4
0.8 (x - x^2)
x^2
3.5 (x^2 - x)
x^3 0.6 Goes through a maximum
3
0.4

2.5 0.2
x
2 0
x^n

f(x)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

1.5 -0.2

1 -0.4

0.5
-0.6 Goes through a minimum
-0.8
0
0 0.2 0.4 0.6 0.8 x 1 1.2 1.4 1.6 -1
 Homogeneous Nucleation & Energy Effects
Surface Free Energy- destabilizes
the nuclei (it takes energy to make
an interface)
DGS = 4r 2 
 = surface tension

DGT = Total Free Energy

= DGS + DGV

Volume (Bulk) Free Energy –

stabilizes the nuclei (releases energy)
4
DGV = r 3 DG
3
volume free energy
DG =
unit volume
r* = critical nucleus: for r < r* nuclei shrink; for r >r* nuclei grow (to reduce energy)
 4 
 
ΔG   r 3 .(Gv )  4r 2 .( ) Note that G is a function of T, r & 
3 

 As we have noted previously G vs r plot will go through a maximum (implying that as a

small crystal forms ‘G’ will increase and hence it will tend to dissolve).
 The maximum of G vs r plot is obtained by by setting dG/dr = 0. The maximum value of
G corresponds to a value of ‘r’ called the critical radius (denoted by superscript *).
 If by some ‘accident’ (technically a ‘statistical random fluctuation’) a crystal (of ‘preferred’
crystal structure) size > r* (called supercritical nuclei) forms then it can grow down-hill in
‘G’. Crystals smaller than r* (called embryos) will tend to shrink to reduce ‘G’. The critical
value of G at r* is called G*.
 Reduction in G (below the liquid state) is obtained only after r0 is obtained (which can be
obtained by setting G = 0). dG
0
dr
r1*  0 Trivial solution
dG
0 2
dr r2*  
Gv
As Gv is ve, r*is +ve G  0
2
r*  
Gv

16  3 Note that we are r* r0

G  
G →

*
at a constant T
3 Gv2
Embryos Supercritical nuclei
3
r0  
G  0 Gv
r →
What is the effect of undercooling (T) on r* and G*?
 We have noted that GV is a fucntion of undercooling (T). At larger undercoolings GV
increases and hence r* and G* decrease. This is evident from the equations for r* and G*
as below (derived before).
 At Tm GV is zero and r* is infinity!
  That the melting point is not the same as the freezing point!!
 This energy (G) barrier to nucleation is called the ‘nucleation barrier’.

Tm Gv  f (T )
Decreasing G*

The bulk free energy reduction is a function of undercooling

2
r*  
Gv
r →
16  3
G  
*

3 Gv2
Decreasing r*
G →

Using the Turnbull approximation (linearizing the G-T curve

close to Tm), we can get the value of G interms of the enthalpy
of solidification. 16 3 Tm2
G  


3 T 2 H 2
Turnbull’s approximation

Tm  T T
G  H f  H f
Tm Tm

Solid (GS) ΔH f  heat of fusion

G 2
16 3  Tm 
G   
*

3  H f T
 
G →

T
Liquid (GL)
Tm
T →
 Jump
Quantum How are atoms assembled to form a nucleus of r*  “Statistical Random Fluctuation”

To cause nucleation (or even to form an embryo) atoms of the liquid (which are randomly moving
about) have to come together in a order, which resembles the crystalline order, at a given instant of
time.
Typically, this crystalline order is very different from the order (local order), which exists in the liquid.
This ‘coming together’ is a random process, which is statistical in nature  i.e. the liquid is exploring
‘locally’ many different possible configurations and randomly (by chance), in some location in the
liquid, this order may resemble the preferred crystalline order.
Since this process is random (& statistical) in nature, the probability that a larger sized crystalline
order is assembled is lower than that to assemble a smaller sized ‘crystal’.
Hence, at smaller undercoolings (where the value of r* is large) the chance of the formation of a
supercritical nucleus is smaller and so is the probability of solidification (as at least one nucleus is
needed which can grow to cause solidification). At larger undercoolings, where r* value is relatively
smaller, the chance of solidification is higher.

Tm
Chances of nucleation increases

↑ T
 r*
 What is meant by the ‘Nucleation Barrier’  an alternate perspective Funda Check

 Here we try to understand: “What exactly is meant by the nucleation barrier?”.

 It is sometime difficult to fathom out as to the surface energy can make freezing of a small
‘embryo’ energetically ‘infeasible’ (as we have already noted that unless the crystallite size is >
r0 the energy of the system is higher). Agreed that for the surface the energy lowering is not as
much as that for the bulk*, but even the surface (with some ‘unsaturated bonds’) is expected to
have a lower energy than the liquid state (where the crystal is energetically favoured). I.e. the
specific concern being: “can state-1 in figure below be above the zero level (now considered for
the liquid state)?”  “Is the surface so bad that it even negates the effect of the bulk lowering?”
 We will approach this mystery from a different angle by first asking the question: “what is
meant by melting point?” & “what is meant by undercooling?”.

* refer to surface energy and surface tension slides.

Melting point, undercooling, freezing point (in the realm of homogenous nucleation)
 The plot below shows melting point of Au nanoparticles, plotted as a function of the particle radius. It is to
be noted that the melting point of nanoparticles decreases below the ‘bulk melting point’ (a 5nm particle
melts more than 100C below Tmbulk). This is due to surface effects (surface is expected to have a lower
melting point than bulk!?*)  actually, the current understanding is that the whole nanoparticle melts
simultaneously (not surface layer by layer).
 Let us continue to use the example of Au. Suppose we are below Tmbulk (1337K=1064C, i.e. system is undercooled
w.r.t the bulk melting point) at T1 (=1300K  T = 37K) and suppose a small crystal of r2 = 5nm forms in the liquid.
Now the melting point of this crystal is ~1200K  this crystal will ‘melt-away’. Now we have to assemble a
crystal of size of about 15nm (= r1) for it ‘not to melt’. This needless to say is much less probable (and it is
better to undercool even further so that the value of r* decreases) . Thus the mystery of ‘nucleation barrier’
vanishes and we can ‘think of’ melting point  freezing point (for a given size of particle)!
 Tm is in heating for the bulk material and in cooling if we take into account the size dependence of melting
point everything ‘sort-of’ falls into place .

T1

Other materials like Pb, Cu, Bi,

r1 Si show similar trend lines

* Surface atoms are loosely bound as compared to the bulk atoms.

Atomic perspective of nucleation: Nucleation Rate

 The process of nucleation (of a crystal from a liquid melt, below Tmbulk) we have described so
far is a dynamic one. Various atomic configurations are being explored in the liquid state some
of which resemble the stable crystalline order. Some of these ‘crystallites’ are of a critical size
r*T for a given undercooling (T). These crystallites can grow to transform the melt to a
solid by becoming supercritical. Crystallites smaller than r* (embryos) tend to ‘dissolve’.
 As the whole process is dynamic, we need to describe the process in terms of ‘rate’  the
nucleation rate [dN/dt  number of nucleation events/time].
 Also, true nucleation is the rate at which crystallites become supercritical. To find the
nucleation rate we have to find the number of critical sized crystallites (N*) and multiply it by
the frequency/rate at which they become supercritical.
 If the total number of particles (which can act like potential nucleation sites in homogenous
nucleation for now) is Nt , then the number of critical sized particles given by an Arrhenius type
function with a activation barrier of G*.

 G * 
 
 kT 
N *  Nt e  
 The number of potential atoms, which can jump to make the critical nucleus supercritical are
the atoms which are ‘adjacent’ to the liquid let this number be s*.
 If the lattice vibration frequency is  and the activation barrier for an atom facing the nucleus
(i.e. atom belonging to s*) to jump into the nucleus (to make in supercritical) is Hd , the
frequency with which nuclei become supercritical due atomic jumps into the nucleus is given
by:  H 
d

 '  s  e kT 
*

No. of critical sized Frequency with which they

Rate of nucleation = 
particles become supercritical
dN
I  G * 

 kT 

 H d 
 
dt N *  Nt e    '  s* e  kT 

No. of particles/volume in L  → lattice vibration frequency (~1013 /s)

s* atoms of the liquid facing the nucleus

Critical sized nucleus (r*)

Jump taking particle to supercriticality

→ nucleated (enthalpy of activation = Hd)

Outline of critical sized nucleus

  The nucleation rate (I = dN/dt) can be written as a product of the two terms as in the equation
below.
 How does the plot of this function look with temperature?
 At Tm , G* is   I = 0 (as expected if there is no undercooling there is no nucleation).
 At T = 0K again I = 0
 This implies that the function should reach a maximum between T = Tm and T = 0.
 A schematic plot of I(T) (or I(T)) is given in the figure below.
 An important point to note is that the nucleation rate is not a monotonic function of
undercooling.

Tm
T = Tm → G* =  → I = 0
Increasing T

 G* ↑  I ↓
 G *  H d 
  Note: G* is a function of T

I  Nt s*  e
T (K) →

 kT 
T↑ I ↑

T=0→I=0
0 I →
Heterogenous nucleation
 We have already talked about the ‘nucleation barrier’ and the difficulty in the nucleation
process. This is all the more so for fully solid state phase transformations, where the strain
energy term is also involved (which opposes the transformation).
 The nucleation process is often made ‘easier’ by the presence of ‘defects’ in the system.
 In the solidification of a liquid this could be the mold walls.
 For solid state transformation suitable nucleation sites are: non-equilibrium defects such
as excess vacancies, dislocations, grain boundaries, stacking faults, inclusions and
surfaces.
 One way to visualize the ease of heterogeneous nucleation
heterogeneous nucleation at a defect will lead to destruction/modification of the defect
(make it less “‘defective’”). This will lead to some free energy Gd being released → thus
reducing the activation barrier (equation below).

ΔG hetro,defect  (V)  Gv  Gs   A   (Gd )

 Increasing Gd (i.e. decreasing G*)
 Homogenous sites
 Vacancies
 Dislocations
 Stacking Faults
 Grain boundaries (triple junction…), Interphase boundaries
 Free Surface
Heterogenous nucleation
 Consider the nucleation of  from  on a planar surface of inclusion .
 The nucleus will have the shape of a lens (as in the figure below).
 Surface tension force balance equation can be written as in equation (1) below. The contact angle
can be calculated from this equation (as in equation (3)).
 Keeping in view the interface areas created and lost we can write the G equation as below (2).

  Interfacial Energies



  Created Alens 

Created Acircle 
 is the contact angle
    
Cos  (3) Lost Acircle 
 
Surface tension force balance   Cos       (1)

ΔG  (Vlens )Gv  (Alens )   ( Acircle )    ( Acircle )   (2)

Vlens = h2(3r-h)/3 Alens = 2rh h = (1-Cos)r rcircle = r Sin

 Using the procedure as before (for the case of the homogenous nucleation) we can find r* for
heterogeneous nucleation. Using the surface tension balance equation we can write the
formulae for r*and G* using a single interfacial energy  (and contact angle ).
 Further we can write down Ghetero
*
in terms of Ghomo
*
and contact angle .

dG 2  
 
3
0 *
r  Ghetero
* 4
 

2  3Cos  Cos 3
Gv 3 Gv
hetero 2
dr
2  3Cos  Cos 3 
Ghetero
*

 
1 * 1 Just a function of 
G *
hetero  Ghomo  2  3Cos  Cos 3
4 *
Ghomo 4  the contact angle

Decreasing tendency to wet the substrate

Increasing
contact angle

Complete wetting
 = 0  f() = 0
Ghetero
*

*
Ghomo

1
4

2  3Cos  Cos3  f ( )   = 90  f() = ½ Partial wetting

 = 180  f() = 1
No wetting
The plot of Ghetero
*
/ Ghomo
*
is shown in the next page.
 Plot of G*hetero/G*homo is shown below. This brings out the benefit of heterogeneous nucleation vs homogenous nucleation.
 If the  phase nucleus (lens shaped) completely wets the substrate/inclusion (-phase) (i.e.  = 0)
 then G*hetero = 0  there is no barrier to nucleation.
 On the other extreme if -phase does not we the substrate (i.e.  = 180)
 then G*hetero = G*homo  there is no benefit of the substrate.
 In reality the wetting angle  is somewhere between 0-180
 Hence, we have to chose a heterogeneous nucleating agent with a minimum ‘’ value.

1
G*hetero (180o) = G*homo
no benefit
0.75
G*hetero (0o) = 0
G*hetero / G*homo →

no barrier to nucleation G*hetero (90o) = G*homo/2

0.5
    
Cos 
 
0.25

Complete wetting Partial wetting No wetting

0
0 30 60 90 120 150 180
 (degrees) →
 Choice of heterogeneous nucleating agent  

 Heterogeneous nucleation has many practical applications. 


 During the solidification of a melt if only a few nuclei form and
 
these nuclei grow, we will have a coarse grained material (which
will have a lower strength as compared to a fine grained
material- due to Hall-Petch effect).      
Cos 
 Hence, nucleating agents are added to the melt (e.g. Ti for Al  
alloys, Zr for Mg alloys) for grain refinement.

How to get a small value of ? (so that ‘easy’ heterogeneous nucleation).

 Choosing a nucleating agent with a low value of  (low energy  interface)
 (Actually the value of (  ) will determine the effectiveness of the heterogeneous
nucleating agent → high  or low )
How to get a low value of ?
 We can get a low value of  if:
(i) crystal structure of  and  are similar and
(ii) lattice parameters are as close as possible
 Examples of such choices:
 In seeding rain-bearing clouds → AgI or NaCl are used for nucleation of ice crystals
 Ni (FCC, a = 3.52 Å) is used a heterogeneous nucleating agent in the production of
artificial diamonds (FCC, a = 3.57 Å) from graphite.
 Why does heterogeneous nucleation dominate? (aren’t there more number of homogenous nucleation sites?)

 To understand the above questions, let us write the nucleation rate for both cases as a pre-
exponential term and an exponential term. The pre-exponential term is a function of the
number of nucleation sites.
 However, the term that dominates is the exponential term and due to a lower G* the
heterogeneous nucleation rate is typically higher.

 Ghomo
*   Ghetero
* 

   
 
I homo  I homo
0
e  kT 
I hetero  I hetero
0
e  kT 

= f(number of nucleation sites) = f(number of nucleation sites)

~ 1042 ~ 1026

BUT
the exponential term dominates

Ihetero > Ihomo

Heterogeneous nucleation in AlMgZn alloy