Microstructure From

Processing: Evaluation and

Modelling
Grain size control: Lecture 3

Martin Strangwood,
Phase Transformations and Microstructural Modelling,
School of Metallurgy and Materials
Introduction
• Thermal exposure generally increases grain size and
the control is by (i) transformation, (ii) solute drag or
(iii) Zener drag

• Added deformation leads to an increase in stored

energy that can be used to form ‘new’ grains
 Starting state
• Annealing converts a high energy state (deformed) into a lower
energy one and so the processes are driven by the energy of the
deformed starting state

• Deformation introduces a number of defects:

• Point defects - vacancies and interstitials (latter only really

for irradiated material); vacancies often arise from
dislocation motion, e.g. jogs
Vacancies
Dislocations

Jog
 Starting state
• Annealing converts a high energy state (deformed) into a lower
energy one and so the processes are driven by the energy of the
deformed starting state

• Deformation introduces a number of defects:

• Line defects - dislocations

• Planar defects - stacking faults and boundaries

• Defects store excess energy in elastic strain fields around them and
annealing removes them at a rate dependent on the type and number
of defects
Defect behaviour
Energy Number Mobility Overall

Vacancies Low High High

Dislocations

Stacking
faults
Grain
High Low Low
boundaries
Defect behaviour
Energy Number Mobility Overall

Vacancies Low High High Small

Dislocations Dominant

Stacking Secondary
faults
Grain Tertiary
High Low Low
boundaries
Stages in annealing
• The dominant nature of dislocations means that these determine
the stages of annealing

• Recovery - re-arrangement of dislocations, but little

reduction in numbers
Stages in annealing
• Primary recrystallisation - removal of excess dislocations
reducing stored energy as new dislocation-free grains
form
• Uniform grain growth - reduction in grain boundary area

• Secondary recrystallisation - non-uniform grain growth

• Tertiary recrystallisation
Deformed structure - Cu
• FCC structure with high stacking fault energy

• Structures imaged on normal-transverse plane after cold

deformation

Rolling direction
Deformed structure - Cu

Cells
Up to 10 %  ~ 0.5 - 1.0 m
diameter

{111}

Microbands ~ 40 x 0.2 m
> 10 % 
Deformed structure - Cu

Microbands rotate
Increasing  parallel to rolling
plane

35° Macroscopic shear bands

Higher  form - thin sheets of
elongated narrow cells

• These features are associated with nucleation

of recrystallised grains
Deformed structure - 70:30 brass
• FCC structure with low stacking fault energy

• Initially just a random arrangement of dislocations with no clear

defined cell structure

• 40 - 60 % cold reduction

• Very thin deformation twins form on planes of

high shear stress, ~ 0.02 m thick

• 50 - 80 % cold reduction
• Twins rotate to become parallel to sheet surfaces
Deformed structure - 70:30 brass
• 60 - 90 % cold reduction

• Macroscopic shear bands form at lower strains

than in high SFE materials

Random arrangement of
dislocations
 Recrystallisation and softening
• High SFE, e.g. Cu

Primary recrystallisation

Hv Work hardened

50 % softening = 50
% primary
recrystallisation

Annealing temperature
 Recrystallisation and softening
• Low SFE, e.g. brass

Recovery

Primary recrystallisation
Hv Work hardened

50 % softening Excessive recovery

leads to retardation of
50 % primary
recrystallisation recrystallisation

Annealing temperature
Laws of recrystallisation
• Minimum deformation needed to initiate recrystallisation
Activation barrier

• Smaller amounts of deformation need higher recrystallisation

temperatures
Thermally activated

• Increasing annealing times reduce temperature needed for

recrystallisation
Diffusional growth
Laws of recrystallisation
• Recrystallised grain size depends on on degree of deformation
(chiefly) and annealing temperature (to a lesser extent); smaller
grains result from larger degrees of deformation and lower
annealing temperatures
Nucleation rate effects
• As original grain size increases, greater amounts of
deformation are required for equivalent recrystallisation time
and temperature
Driving force and nucleation site reduction

• The amount of ‘cold’ work needed for equivalent hardening

increases as deformation temperature increases

Rearrangement of defects during

deformation and reduced driving force
Laws of recrystallisation
• New grains do not grow into deformed grains of similar orientation

Texture affects recrystallisation

• Continued heating can cause the recrystallised grain size to

increase (not always the case)

Grain boundaries can drive some

annealing processes
Dislocation stored energy
• Dislocations store the greatest amount of excess energy
in a cold deformed metallic alloy and primary
recrystallisation is driven by the removal of excess
dislocations

• The size of this driving force can be estimated in two

ways
Stress-strain curve
• Al-based alloys can have a fairly flat work hardening curve
200

Work done
Stress

Strain 0.5
Work done  200 x 0.5
~10 % is stored; P1 ~ 107 J m-3
 100 MPa
 100x106 J m3
Hot working?
Dislocation density
• Excess dislocations store energy is strain energy around the
extra half plane

• Dislocation energy per unit length can be multiplied by dislocation

density (length / unit volume) to give total energy

Line energy   b 2

•  = shear modulus; b = Burger’s vector


 Dislocation density in Al
• For a dislocation density of N dislocations per unit area of slip plane,
then:

Energy   N  b 2
  const.  1
N 1016 m2
  100 GPa
b  2x1010 m
Energy  40 x 106 J m3
• Recovery reduces the value of 


Measuring stored energy?
Measuring stored energy
DSC

XRD

EBSD

TEM
SIBM
• Grains with different Schmid factors will start yielding under
different applied stresses

• During deformation the Schmid factors change as the sample gets

longer and thinner so that the slip planes and directions rotate

• During cold rolling adjacent grains will undergo different levels of

strain and so have different stored energies (dislocation densities)

• Dislocation density of A < dislocation density of B

Boundary bulging
• If the stabler A grain is allowed to grow into the less stable grain
B by bulging of the A/B grain boundary then there will be a
decrease in free energy (Gv) but an increase in interfacial
energy
Original A/B boundary

B Bulged A/B boundary

2 b
A L
Gv 
L
• Bulge is stable and can grow if: L  10 m
 b  0.3 mJ m2
• Schmid factor can give up to Gv  60000 J m3
10 % variation, hence viable i.e.~ 6%
Intragranular nucleation
• Grain refinement would require greater nucleation and the
formation of new recrystallised grains within the deformed grains
would also block the growth of those from grain boundaries

• During deformation (depending on SFE) sub-grains can form

separated by boundaries (SGB) with a few degrees misorientation

• Under the correct processing conditions transition bands can form

across grains and act as intragranular nucleation sites for
recrystallised grains
 Transition bands
• Transition bands consist of a large number density of fine sub-grains,
which achieve a large orientation change rapidly

{111}

Transition bands

• The high number density of fine sub-grains gives

conditions for SIBM with the misorientation
established allowing growth into the surrounding
sub-grains
Transition band formation
• As plastic deformation proceeds the slip planes and directions rotate
towards the tensile axis (i.e. towards or away from <110> depending
on sense of stress and crystal structure)

• Rotation away from <110> can be either to <100> or <111> so that

parts of a grain can rotate one way and other parts the other way;
these two parts are separated by a transition band
 Transition band formation
BCC
FCC

Tension
Transition bands

Compression
 Dutta-Sellars Equations
300,000  2.75105  
Rs  6.751020
D  exp
2
0
4
exp  185Nb 
 RT   T  
2
 Rf 
0.85  1  expln 0.95 
  Rs 

270,000   2.51010 
P5  3106 Nb   Z 0.5 exp
1 1
exp


2 
  RT  T ln K s 
3

 Q   12 N 
Nb C
Z  Ýexp def   14 
RTdef  Ks  6770
 2.26 
10 T
Dutta-Sellars Equations
• The D-S equations have a relatively simple form and are easy to apply
with clearly identified parts, that have been shown to fit data well

Original fit data

Dutta-Sellars Equations
• The original fitting data used was limited to 0.3 strain; 0.03 – 0.04 wt %
Nb; and a mean prior austenite grain sizes in the range 12 – 100 μm

• Application of the
D-S equations to
more recent
datasets has
revealed less
good agreement

• Improvement of
the D-S approach
is needed along
with expansion to
more realistic
aspects

Subsequent data
 Re-crystallisation Ranges
• Initially a 0.046 wt % Nb steel was used that had been homogenised
at 1225 °C for four days

• Re-heating at 1225 °C for one hour prior to Gleeble testing gave a

uni-modal large grain size (250 - 280 μm) for which the original D-S
equations should apply
 Grain Size Distributions
• Relationships of the type

 1
0.67
drex  D'D 0
were replaced by a simple halving of the original grain size size on
recrystallisation – this effectively limits each prior austenite grain to
the formation of two recrystallised grains


• This allowed each class in the starting grain size distribution to be
treated separately to build up a recrystallised grain size distribution
(a similar approach has been used for TSDR material by the CEIT
group).
 Grain Size Distributions
• This approach gives good quantification with the percentage
recrystallised and for grain size distributions both unimodal and
bimodal
Grain Size Distributions
 Literature Values
• A grain class-based approach improves the fit for our experimental data
and for literature data. The outliers generally have high (0.09 wt %) and
low (0.02 wt %) [Nb] or higher [N].

Birmingham data Literature data

• The grain-size class approach expand the original
D-S equations and should be used as standard for
future applications
 Segregation Effects during Hot Rolling
• The presence of solute-rich and solute-depleted regions will alter:
• Degree of dissolution during re-heating (already mentioned)
• Driving force for re-precipitation of strain-induced precipitates
during deformation and holding
• Solute drag

• The differing Nb levels (currently experimental but to be

predicted) can be used as input for the D-S equations with the
appropriate grain size distribution
 Segregation Effects during Hot Rolling
• For an average grain size this leads to a widening of the partial
recrystallisation regime
 Segregation Effects during Hot Rolling
• For the use of a class-based prediction then assumptions need to be
made regarding the fraction of the each class in solute-rich and solute-
depleted material

• The D-S equations are then applied for each class and composition
 Recrystallisation Prediction
• The use of different compositions for different regions gives good
agreement with Gleeble simulations (single hit)
 Grain Size Distributions
• The class-based segregated approach gives reasonable agreement
for grain size distribution

1225/1025 1225/975

• The extent of recrystallisation for different grain

size classes changes with deformation
temperature
Grain size distributions
 Grain size distributions were evaluated in terms of D5% (grain size class
constituting the first 5 area% of the total area measured), Dmode (the
mode grain size class) and Dmax (the largest grain size class in the
distribution).
Dmode

Dmax
D D5%
5%
 Effect of strain and initial grain size on recrystallised grain
size
 The recrystallised grain size decreases with an increase in strain for all
initial grain sizes during single hot deformation – shown here for mode
grain size:

 Whilst the mode grain size decreases as the applied strain increases, the
rate of refinement decreases and the initially different grain size values
converge at high strain.
Predicting the recrystallised grain size distribution

 The modified equation (exponents derived from the classical theory for rate
of nucleation) was used to predict the grain size distributions
 Grain boundary nucleation is assumed in deriving the equation
Original equation

Modified equation

 D’ values were fitted to the measured grain size distribution in terms of

D5%, Dmode and Dmax for Fe-30Ni using the original equation in order to
establish the best fit D’ values
 Best fitted D’ values for predicting the D5% Dmode & Dmax D5%, Dmode & Dmax in
the 50-60 µm, 110-120 µm & 160-180 µm distributions for different applied strains

 D’ represents the ‘efficiency’ of nucleation of new grains and is dimensionless.

 For D5% Dmode & Dmax D’ values increase steadily with increasing strain for the smallest initial
grain size distribution
 A flattish region is observed with low strain for the largest initial grain size distribution
 A lower D’ value for the Dmode and D5% might suggest that fine grains recrystallise first due to
higher stored energy.
 Best fitted D’ values for predicting the D5% & Dmax in the 50-60 µm, 110-120
µm & 160-180 µm distributions for different applied strains

 D5% & Dmax have similar trends to those observed for the Dmode D’ values.

 D’ values increase steadily with increasing strain for the smallest initial grain size distribution

 A flattish region is observed with low strain for the largest initial grain size distribution
 Variation of stored energy with initial grain size

 The deformation curves shown are for the 50 - 60 µm,110 -120µm and
160 -180µm samples deformed to a strain of 0.3 cold strain

 The finest initial grain size distribution (50-60 µm) material shows a greater
work hardening rate than the other samples, which is consistent with a
higher stored energy for the same plastic strain
 Example distribution fit - 110-120 micron mode, 0.7 strain
 D’ = 0.83 reported in literature for predicting Carbon Manganese (C-Mn) steels does not
give a good fit to the measured grain size distribution Carbon Manganese (C-Mn) steels
 D’ = 1.1 reported in literature for predicting Nb - containing steels was used to show that
the literature equations can predict the mode grain size
 The variable D’ approach gives a good fit to the measured grain size distribution
 Summary of RMS errors for predicted grain size
distributions in highly alloyed steels

RMS error
Alloy type Initial mode grain size / µm Strain D' variation Constant D’ approach (1.1)
Strip steel 0.15 3.14 7.71
160-180 0.3 7.61 16.00
0.45 8.00 21.87
High Aluminium
steel 25-30 0.7 8.00 27.92
0.15 11.70 7.00
9Cr forging steel 170 - 180 0.3 10.83 12.50
0.45 4.42 18.80
360 - 380 0.3 6.90 17.00

 The fit between measured and predicted grain size distributions for the
variable D’ approach deteriorates as the complexity of the alloy increases
and at lower strain values, although in almost all cases the fit is better than
for a constant D’ approach
 Control of grain size

• The coverage of distributions has been

extended to consider
• strain and grain size (including the order
of recrystallisation)
• rates of recrystallisation
• growth rates (and particle
distribution)