Phase field

modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Phase field modelling

Basics: Microstructure, materials thermodynamics and
transformation kinetics
Mogadalai P Gururajan1
Department of Metallurgical Engineering and Materials Science
Indian Institute of TechnologyBombay
Mumbai 400076 INDIA

8 December, 2011

1 guru.mp@iitb.ac.in,gururajan.mp@gmail.com

Outline
Phase field
modelling
M P Gururajan,
IIT-B

1 Preamble

Preamble
Introduction
Classical diffusion

2 Introduction

Non-classical
diffusion
Phase field
modelling

3 Classical diffusion

Summary

4 Non-classical diffusion
5 Phase field modelling
6 Summary

Acknowledgements
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Thanks to the organisers for inviting me to give this talk!

Thanks to my:
Teachers: Abi, Ferdi, Peter
Colleagues: Shankara, Ram, Deep, Saswata, Rajdip, Chirru,
Kuo-An, Ian, Vichu, Prita, and Phani
Students: Amol, Yash, Arpit, Jayatheerdh, Kedar, Arijit, Vivek,
...
Funding agencies: Volkswagen, DAAD-UGC, CSIR, ONR, IRCC
at IIT-B, DRDO ER&IPR,SASE

A disclaimer
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction

This is a pedagogical presentation;

Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Hence, some (but, hopefully, not all :-) of what I am saying is known
to you;

Please indulge me!

Modelling: Time and length scales

Phase field
modelling
M P Gururajan,
IIT-B

Metres,Months

Preamble
Introduction
Classical diffusion

Finite Element
Process

Non-classical
diffusion

Phase field Micron, Minutes

Summary

Time

Phase field
modelling

MD

Nano, Nano

Ab initio Angstrom, Femto

Length
Figure: Time and length scales for some modelling techniques

Microns and Minutes

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Figure: Microstructure in Udimet 520, Xu et al, Met. Mat. Trans. A, 29,

1998

Formation and evolution of microstructures

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Microstructures: typically at the micron length scale

Evolution of microstructures: typically, in the course of a few
seconds to several hours
Phase field model: the ideal tool for formation and evolution of
microstructures considering the time and length scales

Microstructure: a closer look

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Figure: Microstructure in Udimet 520, Xu et al, Met. Mat. Trans. A, 29,

1998

What do we want to model?

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion

Elastic stress effects

Phase field
modelling

Grain growth

Summary

Growth and coarsening of precipitates

Another example!
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Figure: Dendritic microstructure in a nickel-base superalloy single crystal

weld, David et al, JOM, June, 2003.

Solidification: phase field modelling

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Microstructures: typically at the micron length scale

Evolution of microstructures: typically, in the course of a few
seconds to several hours
Phase field model: the ideal tool for formation and evolution of
microstructures

What is phase field modelling?

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion

Short answer: Partial differential equations that describe diffusion (of

atoms and heat) as well as phase transformations

Phase field
modelling
Summary

Long answer: The rest of this talk!

Some reading material

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Spinodal decomposition, J E Hilliard, Chapter 12, Phase

transformations, ASM, 1970.
My favourite: Derivations of the important expressions are given
in full, on the premise that it is easier for a reader to skip a step
than it is for another to bridge the algebraic gap between it is
easily shown that and the ensuing equation.
Original papers from Cahn and co-workers: Cahn and Hilliard in
J. of Chem. Phys., 31, p. 688 and 28, p.258; Cahn, Trans. Met.
Soc. AIME, 242. p. 166; Cahn, Acta Met., 9, p 795.
Few recent reviews: K Thornton et al, Acta Mat., 51, p. 5675;
L-Q Chen, Ann. Rev. of Mat. Res., 32, p. 113; W J Boettinger
et al, Ann. Rev. of Mat. Res., 32, p. 163.

Ficks first law

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion

Constitutive law: the flux is proportional to concentration gradient

Phase field
modelling
Summary

J = Dc
Why?

(1)

Ficks first law: atomistic picture

Phase field
modelling
M P Gururajan,
IIT-B

If n > n ,
2
1

Preamble
Introduction

J
1

Classical diffusion
Non-classical
diffusion
Phase field
modelling

J>J
1 2

J
2
n n
1 2

Summary

Figure: An atomistic model to explain why Ficks first law is reasonable!

Assume random jumps

Assume jump frequency to be independent of composition
Higher the number of atoms in a plane, higher the flux
Concentration gradient net flux

Ficks second law

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

The conservation of mass coupled with the constitutive law: the

classical diffusion equation!
c
= J = Dc
t

(2)

The surprise!
Phase field
modelling
M P Gururajan,
IIT-B
Preamble

Classical diffusion law of the form

Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling

c
= Dc
t

(3)

fails in certain cases! It needs modification!

Summary

Why failure? Remember, it was a reasonable constitutive law plus a

sure conservation law.
What is the needed modification?

Thermodynamics of phase separating systems

Phase field
modelling
M P Gururajan,
IIT-B
Preamble

Regular solution model

Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling

Configurational entropy assuming random arrangement

Enthalpy assuming pair interactions and negligible pressure term
(i.e., H E )

Summary

(1 xB )GA + xB GB + xB (1 xB )
+RT [xB ln xB + (1 xB ) ln (1 xB )]

(4)

Phase separating systems

Phase field
modelling
M P Gururajan,
IIT-B

( > 0) and lower temperatures

Preamble
Introduction
Classical diffusion
Non-classical
diffusion

Phase field
modelling
Summary

x
B

x
B

x
B

Figure: Free energy of a phase separating system

Mechanisms of phase separation

Phase field
modelling
M P Gururajan,
IIT-B

G<0

Preamble

G = 0

Introduction
Classical diffusion

G = 0

Non-classical
diffusion
Phase field
modelling

G> 0

Summary

x
B

x
B

x
B

Figure: Positive curvature (nucleation) and negative curvature (spinodal)

of the free energy versus composition diagram; phase separation mechanism
changes from nucleation to spinodal at the point of zero curvature.

Spinodal: driving force

Phase field
modelling

M P Gururajan,
IIT-B
Preamble
Introduction

Classical diffusion

Non-classical
diffusion
Phase field
modelling

Driving force: decrease in G

Summary

Y
A

Y
B

Y
C

Figure: In regions of negative curvature, A-rich (and B-rich) regions

spontaneously become richer in A (and B), because, the free energy
decreases in such a process.

Classical diffusion
Phase field
modelling
M P Gururajan,
IIT-B

2 xB
xB
=D
t
x 2

Preamble

(5)

Introduction
Classical diffusion

Summary

Composition

Phase field
modelling

Composition

Non-classical
diffusion

Position

Position

Figure: Evolution of composition during homogenisation and phase

separation. While homogenisation obviously follows from the diffusion
equation above, phase separation implies D < 0.

Chemical potential
Phase field
modelling
M P Gururajan,
IIT-B
Preamble

Of course you knew!

2

Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling

1 2

Summary

Figure: Flux should be driven to even out chemical potential differences.

Diffusion equation
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

J = NV M where = (B A ) =

G
nB

xB
= NV M
t
G
Noting NV (2 1 ) = x
, it can be shown
B

 2 


xB
M
G
M
2
=

x
=
G 00 2 xB
B
t
NV
NV
xB2

(6)
(7)

(8)

Are we done?
Phase field
modelling
M P Gururajan,
IIT-B

xB
=
t

Preamble

M
NV

G 00 2 xB

(9)

Introduction

Phase field
modelling
Summary

Composition

Non-classical
diffusion

Composition

Classical diffusion

G > 0

G < 0
G

G
Position
Composition

Position
Composition

Figure: Evolution of composition during homogenisation and phase

separation. By combining the signs of the curvature of the composition
profile and free energy versus composition curve, we can explain both.

Why we are not yet done!

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

xB
=
t

M
NV

G 00 2 xB

(10)

A solution to the above equation (when it is linearised):

xB xB0 = A(, t) exp (ix)

(11)

where, xB0 is the average composition, A(, t) is the amplitude of

Fourier component of wavenumber ( = 2/) at time t ( being
the wavelenth)
By substituting the solution in the equation, one can show
 


M
A(, t) = A(, 0) exp
G 00 2 t = A(, 0) exp [R()t] (12)
NV

Failure of classical diffusion equation

Phase field
modelling
M P Gururajan,
IIT-B
Preamble

 


M
00 2
A(, t) = A(, 0) exp
G t = A(, 0) exp [R()t] (13)
NV

Introduction

Solution to
classical equation

Classical diffusion
Non-classical
diffusion
Phase field
modelling

Solution
that we want

Critical
Wavenumber

Summary

Maximally growing wavenumber

Figure: Predicted growth rate from the classical diffusion equation and
what we want it to be.

As 0, the A(, t)
Cu-Ni-Fe, 100
A and not ordered BCC as predicted here!

What is the physics that we have missed?

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Interfacial energy!
At the interfaces, there are more unfavourable (AB) bonds; they cost
the system energy. This energy is, obviously, a function of
concentration gradient!
In other words, free energy (G ) should not just be a polynomial in
composition (xB ); it should instead be a functional in xB .
G
If so, = n
, where {/nB } is the variational derivative.
B

The free energy

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

G=

1
NV

Z
dV



g0 (xB ) + |xB |2

(14)

g0 (xB ): free energy per atom; a polynomial in xB ; has at least two

minima and a maxima.
: gradient energy coefficient
xB : gradient

The chemical potential

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling

Euler-Lagrange equation:
NV =

xB

G
xB


(15)

Summary

NV =

G
22 xB
xB

(16)

The non-classical diffusion equation

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion

xB
=
t

M
NV

 00 2

G xB 24 xB

The fourth order term is what makes sure that for larger , the
solution remains finite!

Phase field
modelling
Summary

Cahn-Hilliard equation
Gibbs
van der Waals
Time Dependent Ginzburg Landau
Allen-Cahn
Hohenberg and Halperin: Models A, B, C, ...

(17)

The non-classical diffusion equation

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

xB
=
t

M
NV

 00 2

G xB 24 xB

(18)

Physically, the interfacial energy sets the lower limit on the

wavelength of composition modulations; phase separation reduces
overall free energy; however, new interfaces are formed as phase
transformation proceeds; these interfaces cost the system energy; so,
unless the gain in energy can more than compensate for the
interfacial energy increase, phase separation will not proceed.
How about coherency?
Miscibility gap, chemical spinodal and coherent spinodal

A movie!
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

A solution to the Cahn-Hilliard equation:



M
xB
=
G 00 2 xB 24 xB
t
NV
Obtained using Fourier transforms!

Spinodal decomposition

(19)

The recipe
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Order parameter: A field variable continuous and has values

for all positions and time; could be conserved or non-conserved;
describes the topology (Microstructure)
Free energy functional: To describe the free energy for a given
microstructure; a functional of the order parameters
(Thermodynamics)
Variational derivative equated to the chemical potential,
which in turn, determines how the order parameters
change: Microstructure evolves in such a way that the free
energy keeps decreasing (Kinetics)
Note that there is no the phase field model; any model, which is
formulated according to the principles above is a phase field model;
and, for the same phenomena, depending on the free energy
functional that one writes, there could be several phase field models!

Characteristics of a phase field model

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Interfaces are not sharp; diffuse interface model

No tracking of interface: numerical solutions are easier
Gradient energy coefficient: interfacial energy contributions
(Gibbs-Thomson, for example) are automatically accounted for
Topological singularities (splitting or disappearance of
interfaces): naturally taken care of
Elastic stress, magnetic and electric field: can be coupled by
adding the relevant free energy term!

A couple of more movies

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Solidification: four-fold and six-fold dendrites

Order parameters
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

: non-conserved order parameter; can be thought of as solid fraction

or as crystallanity
T : temperature; another non-conserved order parameter

Thermodynamics
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Free energy minimization: isothermal systems

Entropy maximization: Non-isothermal
Thermodynamically consistent phase-field models: entropy is a
functional of and T

Kinetics
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction

Non-classical Fourier law of heat conduction

Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

T
= 2 T + Lsl
t
t
Equation of phase transformation

=  + g 0 (, T )
t

(20)

(21)

What can phase field models not do?

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Necessity of non-dimensionalization
Advantage of non-dimensionalization
Disadvantage of non-dimensionalization
Measuring quantities that enter phase field model

Ideal problems for phase field modelling

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Microstructural evolution: qualitative

Parametric space exploration: example of elastic stress effect.
Given that elastic moduli difference and elastic anisotropy affect thin
film instability, which one affects it more for a given eigenstrain and
interfacial properties?
Complex interactions between different phenomena where controlled
experiments or analytical explorations are difficult

A word about numerical methods

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

FFTW: why the best

Why Finite Element techniques are not that successful
Finite volume/ finte difference techniques

Summary
Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

Phase field model: non-classical partial differential equations

Ideal for the study of microstructural evolution
Ideal for discerning trends and parameter ranges that can give
rise to specific phenomena
Huge scope for exploration!

Phase field
modelling
M P Gururajan,
IIT-B
Preamble
Introduction
Classical diffusion
Non-classical
diffusion
Phase field
modelling
Summary

THANK YOU!