Recovery and Recrystallization Kinetics in AA1050 and AA3003 Aluminium Alloys

Recovery and Recrystallization Kinetics in
AA1050 and AA3003 Aluminium Alloys

Recovery and Recrystallization Kinetics in
AA1050 and AA3003 Aluminium Alloys
PROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus Prof. dr. ir. J.T. Fokkema,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op maandag 31 maart 2003 te 13.30 uur
door
Shangping CHEN
Master of Science
Xian Jiaotong University, China
geboren te Shaaxi province, China
Dit proefschrift is goedgekeurd door de promotor:
Prof. dr.ir. S. van der Zwaag
Samenstelling promotiecommissie:
Rector Magnificus, voorzitter
Prof. dr.ir. S. van der Zwaag, Technische Universiteit Delft, promotor
Prof. dr.ir. J.TH.M. de Hosson, University of Groningen
Prof. dr.ir. H.J. Huetink, University of Twente
Prof. dr.ir. M.J.L. van Tooren, Technische Universiteit Delft
Prof. dr.ir. A. Bakker, Technische Universiteit Delft
Prof. ir. L. Katgerman, Technische Universiteit Delft
Prof. dr. I.M. Richardson, Technische Universiteit Delft
This research was carried out in the framework of the Strategic Research Program of the
Netherlands Institute for Metals Research in the Netherlands (www.nimr.nl).
Published and distributed by: DUP Science
DUP Science is an imprint of
Delft University Press
P.O.Box 98
2600 MG Delft
The Netherlands
Telephone: +31 15 2785678
E-mail: Info@Library.Tudelft.nl
ISBN 90-407-2382-6
Keywords: Aluminum alloys; Recrystallization kinetics; Modeling
Copyright 2003 by Shangping Chen
All rights reserved. No part of the material protected by this copyright notice may be
reproduced or utilized in any form or by any means, electronic or mechanical, including
photocopying, recording or by any information storage and retrieval system, without
permission from the publisher: Delft University Press.
Printed in the Netherlands
v
Contents
1. Introduction ............................................................................................................. 1
1.1. Current status of recrystallization kinetics models....................................... 1
1.2. Scope of dissertation......................................................................................... 3
References ................................................................................................................ 5
2. Quantification of the recrystallization behavior in Al-alloy AA1050................. 7
2.1. Introduction ...................................................................................................... 7
2.2. Experimental details......................................................................................... 8
2.2.1. Material preparation ................................................................................. 8
2.2.2. Microstructural characterization............................................................... 9
2.2.3. Composite image method....................................................................... 10
2.3. Results.............................................................................................................. 11
2.3.1. Microstructural evolution ....................................................................... 11
2.3.2. Composite imaging analysis................................................................... 12
2.3.3. OIM observation..................................................................................... 14
2.3.4. Micro-hardness measurements ............................................................... 16
2.4. Discussion........................................................................................................ 18
2.4.1. Composite image method....................................................................... 18
2.4.2. Comparison of optical and OIM techniques........................................... 18
2.4.3. Effect of recovery................................................................................... 18
2.5. Conclusions ..................................................................................................... 20
3. Modeling the kinetics of grain boundary nucleated recrystallization processes
after cold deformation using a single grain approach ....................................... 21
3.1. Introduction .................................................................................................... 21
3.2. The Kinetic Model .......................................................................................... 23
3.2.1. Grain Geometry...................................................................................... 23
3.2.2. The deformed microstructure ................................................................. 25
vi
3.2.3. Nucleation............................................................................................... 27
3.2.4. Growth.................................................................................................... 28
3.3. Results and discussion.................................................................................... 31
3.3.1. The effect of grain geometry on the recrystallization kinetics ............... 31
3.3.2. Driving pressure for recrystallization..................................................... 35
3.3.3. Simulation of recrystallization kinetics .................................................. 38
3.4. Conclusions ..................................................................................................... 41
4. A refined single grain approach applied to the modeling of recrystallization
kinetics for cold rolled single-phase metals......................................................... 45
4.1. Introduction .................................................................................................... 45
4.2. Experiments .................................................................................................... 47
4.3. Modeling the recrystallization kinetics......................................................... 50
4.3.1. The orientation dependent microstructure in the deformed state ........... 50
4.3.2. Nucleation and Growth........................................................................... 51
4.3.3. Kinetics approach ................................................................................... 53
4.4. Results.............................................................................................................. 54
4.4.1. Effect of the grain geometry and the nucleation site density on the
recrystallization kinetics......................................................................... 54
4.4.2. Effect of the strain on the recrystallization kinetics ............................... 57
4.4.3. Effect of the initial grain size prior to deformation on the
recrystallization kinetics......................................................................... 58
4.4.4. Effect of the orientation on the recrystallization kinetics....................... 59
4.5. Application of the model to the experiment data ........................................ 60
4.6. Discussions....................................................................................................... 61
4.7. Conclusions ..................................................................................................... 64
5. A FE comparison study of hot rolling operation and PSC testing.................... 67
5.1. Introduction .................................................................................................... 67
5.2. Mathematical approach................................................................................. 68
5.2.1. Constitutive model.................................................................................. 68
5.2.2. Hot rolling operation .............................................................................. 69
5.2.3. Plain strain compression......................................................................... 73
5.2.4. Friction condition ................................................................................... 73
5.3. Results.............................................................................................................. 73
5.3.1. Simulation of hot rolling process ........................................................... 73
vii
5.3.2. Simulation of PSC test............................................................................ 79
5.4. Discussions....................................................................................................... 80
5.4.1. Characteristics of hot rolling deformation.............................................. 80
5.4.2. Comparison of the hot rolling operation and PSC testing...................... 82
5.5. Conclusions ..................................................................................................... 83
6. Modeling recrystallization kinetics in AA1050 following simulated break
down rolling.85
6.1. Introduction .................................................................................................... 85
6.2. The recrystallization kinetics model ............................................................. 87
6.2.1. The Single grain approach...................................................................... 87
6.2.2. Kinetics approach ................................................................................... 91
6.3. Experimental................................................................................................... 91
6.3.1. Material and experimental detail ............................................................ 91
6.3.2. Finite element analysis ........................................................................... 93
6.4. Results.............................................................................................................. 95
6.4.1. The deformation structure ...................................................................... 95
6.4.2. Softening behavior by micro-hardness indentation................................ 95
6.4.3. Recrystallization kinetics by optical microscopy................................... 96
6.4.4. The relationship between static softening and recrystallization............. 98
6.4.5. The grain size of the fully recrystallized structure ................................. 99
6.4.6. JMAK and the S-F analysis of the recrystallization kinetics ............... 100
6.5. Application of the single grain model to predict the recrystallization
kinetics ........................................................................................................... 102
6.5.1. The subgrain size.................................................................................. 102
6.5.2. The calibration constant
d
C ................................................................. 102
6.5.3. Parameters for recovery........................................................................ 104
6.5.4. Activation energy for recrystallization................................................. 105
6.5.5. Textural components after hot deformation ......................................... 105
6.5.6. The critical size of a viable nucleus...106
6.5.7. Prediction of recrystallization kinetics ................................................. 106
6.6. Conclusions ................................................................................................... 110
7. Effect of microsegregation and dislocations on the nucleation kinetics of
precipitation in aluminium alloy AA3003 ......................................................... 113
7.1. Introduction .................................................................................................. 113
viii
7.2. Mathematical description of the C-curve................................................... 114
7.2.1. Activation energy for nucleation.......................................................... 115
7.2.2. Nucleation site density ......................................................................... 117
7.2.3. The evolution of dislocation density .................................................... 118
7.2.4. Effect of micro-segregation of Mn....................................................... 118
7.3. Experimental................................................................................................. 119
7.4. Results............................................................................................................ 120
7.4.1. The evolution of microstructure during homogenization..................... 120
7.4.2. The evolution of the conductivity and precipitation during isothermal
annealing............................................................................................... 120
7.4.3. The isothermal precipitation kinetics ................................................... 122
7.4.4. The softening kinetics during isothermal annealing............................. 125
7.5. Discussion...................................................................................................... 126
7.5.1. Analysis of the experimental results..................................................... 126
7.5.2. Application the model to the experiment data...................................... 127
7.6. Conclusions ................................................................................................... 133
8. On the precipitation and recrystallization behavior in an AA3003 following
hot deformation ................................................................................................... 135
8.1. Introduction .................................................................................................. 136
8.2. Experimental................................................................................................. 136
8.3. Results............................................................................................................ 138
8.3.1. The evolution of microstructure during preheat treatment ................... 138
8.3.2. High temperature mechanical behavior................................................ 139
8.3.3. The decomposition kinetics of the supersaturated matrix during
isothermal annealing............................................................................. 142
8.3.4. The softening kinetics........................................................................... 145
8.3.5. Nucleation mechanisms and recrystallized grain structure .................. 151
8.4. Discussion...................................................................................................... 153
8.4.1. Effect of the deformation, recovery and recrystallization on
the precipitation kinetics....................................................................... 153
8.4.2. Effect of precipitation on the recrystallization kinetic ......................... 157
8.4.3. Interaction bewteen precipitation and recrystallization........................ 166
8.5. Conclusions ................................................................................................... 168
Appendix.170
ix
Summary .................................................................................................................. 173
Samenvatting............................................................................................................ 177
Publications .............................................................................................................. 181
Acknowledgements .................................................................................................. 183
About the author...................................................................................................... 185
S.P. Chen
Chapter 1
Introduction

Recrystallization and related annealing phenomena which occur during the
thermomechanical processing of metals have long been recognized as being of great
importance in technological and scientific interest [1-3]. Recovery is the annealing
processes occurring in deformed metals without the migration of a high angle grain
boundary. It contributes to a balance of high strength and high toughness.
Recrystallization is the formation of a new grain structure in a deformed material owing
to the formation and movement of high angle grain boundaries driven by the stored
energy of deformation. It controls the grain structure of the final product. The
combination of recovery and recrystallization is largely responsible for the
microstructural evolution in metals during hot rolling and during annealing after cold
rolling. Metallurgical research in this field is mainly motivated by the requirements of
industry, and currently, a strong need exists for quantitative, physically-based models
which can be applied to metal-forming processes so as to control, improve and optimize
the microstructure and the texture of the finished products. Such models require a more
detailed understanding of both the deformation and annealing processes than we have at
present.
There are three major issues of significance with regard to recrystallization and related
phenomena, namely kinetics, microstructure and texture. Knowledge of the recovery
and recrystallization kinetics is essential in the aluminium industry for predicting rolling
loads and final rolled grain size. This dissertation aims to develop a physical model to
predict the kinetics of recovery and recrystallization in aluminium alloys, as a function
of deformation treatment, thermal history, alloy composition and precipitate content.
1.1. Current status of recrystallization kinetics models
The appropriate model would provide equations which describe the recrystallization
process in sufficient detail. The model should include the following parameters:
Chapter 1 2
Deformed microstructure: This determines the number, spatial distribution, and
viability of nucleation sites, and provides the driving force for growth of nuclei.
Microstructure: Grain boundaries and second phase particles will influence the
nucleation mechanisms and may also affect the growth.
Recovery: Recovery before and during recrystallization will affect the growth rate
and in certain cases the nucleation rate.
Up to now, this goal is far from being reached although models for recrystallization
have been developed over a very long time.
The basis for most analytical models to describe the kinetics of recrystallization is the
approach developed by Johnson, Mehl, Avrami and Kolomogorov (JMAK) [3]. In
JMAK theory it is assumed that the recrystallized nuclei form randomly in the pre-
existing microstructure and that the growth rate of these nuclei is constant and isotropic.
The ideal JMAK behavior is rarely exhibited by real materials. There have been several
attempts to improve the JMAK model, one of which is the Microstructural Path
Methodology (MPM) developed by Vandermeer and Rath [4]. MPM allows more
detailed information about nucleation and growth rates to be extracted from
experimental measurements than the original JMAK analysis permits. However, the
methodology is still based on the assumptions of a random nuclei distribution and
isotropic growth of the recrystallization boundaries. A theoretical method of treating
impingement caused by clustered nucleation at grain boundaries and grain edges was
presented a number of years ago by Cahn [5] and was further developed recently for
time dependent growth rates by Vandermeer and Masumura [6]. These types of
expressions have been shown to be successful in calculating the fraction of
recrystallization and the recrystallized grain size for specific circumstances. However,
one obvious problem concerning these models is the large number of empirical
parameters that have to be determined specifically for each material and pre-processing
history. Another disadvantage of these models is the lack of concern about the origin of
the recrystallized grains. Without the knowledge of the origin of the recrystallized
grains, the models remain at an empirical level and this limits the ability to understand
and control the process. Furthermore, none of the above analytical models describes the
effect of concurrent recovery on the recrystallization kinetics.
A recent development in modeling recrystallization is on the basis of an improved
understanding of the nucleation mechanism and how it is affected by microstructural
heterogeneities resulting from deformation. The topic has been reviewed by Nes and
Hutchinson [7] and Humphreys [8]. In a model developed by Nes and coworkers [9]
and a model by Sellars [10], the nucleation sites for recrystallized grains of different
Global introduction 3
crystallographic orientations, particle stimulated nucleation (PSN) and nucleation from
grain boundary regions have been incorporated. The microstructural parameters of the
deformed structure, subgrain size, sub-boundary misorientation, are used as transient
variables to correlate the recrystallization kinetics to the mechanical properties, such as
strain, stress, and strain rate. However, these models still apply the assumption that the
recrystallization kinetics follows the JMAK equation.
In the last decade, several microstructural models have been developed for simulating
the temporal evolution of the recrystallization microstructures, as well as for predicting
the recrystallization kinetics. These models can be grouped as cellular models [11-13],
computer Avrami models [14,15] and the models based on the Monte Carlo (MC) [16-
18], and Cellular Automaton (CA) [19] techniques. All these models involve inevitably
some assumptions about the nucleation of recrystallization. Although they may yield
realistic predictions of the recrystallization kinetics and the grain size distribution as a
function of various material and processing parameters, they provide little insight into
the mechanism of recrystallization and therefore lack predictive capability.
The points made here emphasize the need for a model that can predict the
recrystallization kinetics and has a firmer foundation in the physics of the nucleation
and growth of the recrystallization.
1.2. Scope of this dissertation
The overall goal of the project is to develop a physical model to predict the recovery
and recrystallization kinetics after cold and/or hot working of aluminium alloys when
the processing parameters such as the deformation strain, strain rate, deformation
temperature and the annealing temperature are known. The basic theory of this model is
described in Chapter 3. The model can be employed to describe the recrystallization
kinetics of polycrystalline metals, based on a single grain representation of the
deformed microstructure (characterized by a mean subgrain size and mean
misorientation of subgrain boundaries). A salient feature of the model is that the initial
grain size, the deformation geometry, the texture components and concurrent recovery
can be taken into account. The grain geometry is shown to have a large effect on the
recrystallization kinetics through a change in the impingement space. The
recrystallization kinetics in grains with different orientations are different because of the
inhomogenity of the deformed microstructure. The validation of the model is done on
the basis of dedicated experiments and using information from the literature. Chapter 4
is an application of the model to the recrystallization kinetics in commercial purity alloy
AA1050 after cold deformation. In Chapter 6 the model is adapted to treat the
recrystallization kinetics following hot deformation and relevant experiments to validate
the model are conducted on a commercial purity alloy AA1050.
Chapter 1 4
In aluminium alloys recovery and recrystallization processes are closely entangled,
since they not only often take place simultaneously, but also they have a distinct
influence on each other. As a consequence, the softening kinetics determined by means
of different methodologies may be different. In Chapter 2 several techniques are used to
quantitatively determine the softening and recrystallization kinetics. A mathematical
method is proposed to separate the effect of recovery and recrystallization.
In the literature, laboratory plane strain compression testing (PSC) is assumed to
adequately simulate hot rolling in terms of microstructure development. However, the
correctness of this assumption remains questionable. In Chapter 5 a comprehensive
study on the comparison of hot rolling and PSC by FEM calculation is reported. The
evolution of the key physical variables in both processes is found to be quite different. It
indicates that a further examination of the deformation history should be made when
applying the relationships established from PSC testing to real industrial hot rolling
operations.
Many aluminium alloys contain a mixed particle structure, which consists of a coarse
distribution of large inclusions resulting from the casting and a fine particle dispersion
due to subsequent hot rolling and annealing treatment. It has been well established that a
fine particle inhibits both the grain nucleation and growth rates while the coarse particle
(>1 m) stimulates the nucleation rates by acting as nucleation sites. However, the
mutual interaction between the softening and decomposition processes is still poorly
understood. Quantitative description of these processes is far from being complete.
Chapter 7 and Chapter 8 deal with such phenomena taking place during
thermomechanical processes in a technological important alloy AA3003. The strain
induced precipitation kinetics are experimentally studied and modeled in Chapter 7. A
comprehensive experimental study on the softening kinetics following hot deformation
is described in Chapter 8, taking into account of the effect of the second phase particles
and the mutual interaction between the precipitation and softening. This thesis ends with
a summary in which the major results and conclusions are grouped together.
Global introduction 5
References
1. P. Cotterill and P.R. Mould, in 'Recrystallization and Grain Growth in Metals',
Surrey Univ. Press. London, 1976.
2. R.D. Doherty, D.A. Hughes, F.J. Humphreys, J.J. Jonas, D.J. Jensen, M.E. Kassner,
W.E. King, H.J. McQueen and A.D. Rollett, Mater. Sci. Eng. A, 238, 1997, 219-
274.
3. F.J. Humphreys and M. Haltherly, in 'Recrystallization and Related Annealing
Phenomena', London, Pergamon, 1996.
4. R.A. Vandermeer and B.B. Rath, Metall. Trans. A, 20A, 1989, 391-401.
5. J.W. Cahn, Acta Metall., 4, 1956, 449-459.
6. R.A. Vandermeer and R.A. Masumura, Acta Metall., 40, 1992, 877-886.
7. E. Nes and W.B. Hutchinson, in 'Proc. 10th Riso Int. Symp. on 'Materials
Architechure: the Scientific Basis for Engineering Materials'', J.Bilde-Sorensen
(Eds), Roskilde, Denmark, Riso National Labotatory, 1989, 233-245.
8. F.J. Humphreys, in 'Proc. Int. Conf. on 'Recrystallization' 90', T. Chandra (Eds),
Warrendale, PA, TMS, 1990, 113-122.
9. H.E. Vatne, T. Furu, R. Orsund and E. Nes, Acta Mater., 44, 1996, 4463-4473.
10. C.M. Sellars, in 'Thermomechanical Processing in Theory, Modeling & Practice
[TMP]
2
', B. Hutchinson et.al (Eds), Swedish Society for Metals Technology, 1996,
35-51.
11. H.V. Atkinson, Acta Metall., 36, 2001, 469-485.
12. V. Marx, F.R. Reher, and G. Gottstein, Acta Mater., 47, 1999, 1219-1230.
13. F.J. Humphreys, Scr. Met. Mat., 27, 1992, 1557-1562.
14. T. Furu, K. Marthinsen and E. Nes, Mater. Sci. Techn., 6, 1990, 1093-1102.
15. T. Saetre, O. Hunderi and E. Nes, Acta Metall., 34, 1986, 981-992.
16. A.D. Rollett, Prog. Mat. Sci., 42, 1997, 79-99.
17. A.D. Rollett, D.J. Srolovitz, M.P. Anderson and R.D. Doherty, Acta Metall., 40,
1992, 3475-3482.
18. D.J. Srolovitz, G.S. Grest and M.P. Anderson, Acta Metall., 34, 1986, 1833-1842.
19. H.W. Hesselbarth and I.R. Gobel, Acta Metall., 39, 1991, 2135-2152.
Chapter 1 6
S.P.Chen
Chapter 2
Quantification of the recrystallization
behavior in Al-alloy AA1050
A new methodology for the determination of the recrystallized volume fraction from
anodically etched aluminium alloys using optical microscopy is described. The method
involves the creation of a composite image from multiple micrographs taken at a series
of orientations. The results of quantitative analysis of images obtained by this new
method are compared with those obtained using the traditional single image optical
microscopy technique, orientation imaging microscopy (OIM) and microhardness
indentation. The multiple orientation image method is shown to consistently yield a
recrystallized volume fraction which is significantly higher than that determined from a
single image while multiple orientation imaging and OIM results are found to be in
good agreement. Furthermore it is shown that, after the subtraction of the effect of
concurrent recovery using the rule of mixtures, microhardness indentation can also be
used to determine the recrystallized volume fraction.
2.1. Introduction
In the study of the kinetics of recrystallization, it is important to determine the volume
fraction of recrystallized material as accurately as possible. Several techniques are
currently in use to quantify the recrystallization behavior of deformed metals. The most
widely used method is that of optical microscopy performed on a series of samples
recrystallized to different extents [1-4]. In addition to yielding quantitative information
regarding the extent of recrystallization, this technique provides some insight into other
microstructural features such as grain size as well as patterns of nucleation and growth.
However, when this method is applied to aluminium alloys such as AA1050, it turns out
Chapter 2 8
to be rather difficult to obtain precise data on the recrystallized fraction. Traditionally,
an anodic etching technique is used to reveal the grain structure and a number of
micrographs are taken of randomly selected areas. A point-counting technique is then
applied to obtain average values of the recrystallized fraction [1,3]. However, when
observed directly at a single orientation with respect to the polarisors the grain structure
of a partially recrystallized aluminium sample prepared in this manner is rarely clearly
and unambiguously visible in its entirety [5]. When the sample is rotated under
polarized light, the microstructure in a field of view seemingly undergoes a
metamorphosis in which apparently unrecrystallized regions begin to appear
recrystallized (as previously hidden boundaries become visible) and apparently
recrystallized regions begin to appear unrecrystallized (as internal substructural details
become visible). This situation leads to extra complications in the identification of
recrystallized grains with only a single micrograph. In order to eliminate such
uncertainties in the determination of the fraction recrystallized, in this paper, a new
method has been designed which is based on the construction of a composite image of a
single region produced from a set of single micrographs taken at a series of stage
rotations. This method enables the structure to be faithfully revealed and thus enables
the recrystallized fraction to be more accurately determined.
Recently, orientation imaging microscopy (OIM) has been employed to determine the
recrystallized fraction [6-9] in Al-base alloys. Although so far OIM has been
predominately applied to the determination of texture, grain boundary structure and
phase determination, an important potential application of OIM lies in the field of
recrystallization, in particular the determination of recrystallization kinetics and the
crystallographic relationships between recrystallized and unrecrystallized grains. By
OIM, it is possible to determine accurately whether an area is recrystallized. Other,
indirect methods (hardness indentations, x-ray diffraction, neutron diffraction and
electrical resistivity) have also been employed [1,10-12] to determine the
recrystallization fraction. These methods measure certain effects of microstructural
changes on the properties and provide only average values including both recovery and
recrystallization effects. Nevertheless, if the effects of recovery and recrystallization can
be separated such techniques may provide valuable additional information regarding
recrystallization behavior. In this study OIM and microhardness measurement have
been employed along with the optical microscopy techniques described above to
determine the recrystallization fraction in the commercially pure AA1050 alloy. The
results so obtained are compared and critically discussed.
Qualification of the recrystallization behavior 9
2.2. Experimental details
2.2.1. Material preparation
The chemical composition of the commercially pure aluminium alloy AA 1050 used is:
0.185 wt.% Fe, 0.109 wt.% Si and Al in balance. The material contains plate-shaped
intermetallic compounds FeAl
3
, which have an aspect ratio in the range of 1 to 6. The
size of FeAl
3
particles ranges between 0.2 and 7 m.
In order to produce material in a suitable form for further experimentation a cast ingot
of AA1050 was hot rolled in 19 passes, resulting in a reduction in thickness from 500
mm to 4 mm. The hot rolling processes started at 520C and finished at 305C. The hot-
rolled material was annealed at 600C for 2 h and then quenched into water. The
material was again heated to 400C and held at this temperature for 2 h to reduce the
content of iron in solid solution. The average grain size after this treatment was 90 m.
The materials were finally cold rolled to a reduction in thickness of 50% (from 4 mm to
2 mm). The rolled sheet was cut into small samples of 20152 mm. The samples were
annealed in a salt bath at 340C for times ranging from 30 s to 3 h and then quenched
into water to obtain different extents of recrystallization. The temperature of the salt
bath was controlled to within an accuracy of 2C. The time taken for a specimen to
reach the set temperature was approximately 5 seconds.
2.2.2. Microstructural characterization
The optical microscopy and OIM as well as micro-hardness examinations were carried
out on the section parallel to the RN plane (R rolling direction and N surface normal) to
encounter as many grain boundaries as possible. After standard sectioning and
polishing, specimens for optical metallography were etched anodically with Barkers
reagent (1% HBF
4
aqueous solution) [13] at 20V for approximately 120 s depending on
the annealing time. For OIM scanning, the specimens were etched with Kellers reagent
for 30 seconds. In order to account for inhomogeneities in the microstructure in the
through thickness direction of the sheet, all the metallography and hardness
measurements were performed at locations along the center line of the sheet, in a band
covering approximately one third of the thickness of the sheet.
The optical microscopy examinations were performed using a NEOPHOT inverted
stage metallurgical microscope. Both a standard point-counting technique and a LEICA
QUANTIMET digital image analysis facility were used to evaluate the recrystallized
fraction. In all cases measurements were conducted on five separate areas. A Buehler
OMNIMET MHT automatic micro hardness tester was used for the microhardness
measurements. The micro hardness of each specimen was determined, using 50 g load
Chapter 2 10
and 15 s loading time. Hardness tests were made after polishing and 18 measurements
were taken on each specimen.
Orientation imaging microscopy (OIM), developed by TexSEM Laboratories Inc., was
integrated with a Philips XL-30s FEG scanning electron microscope (SEM) and
employed for the combined microstructural and crystallographic analysis. By a fast
procedure of capturing and processing electron back-scatter diffraction patterns
(EBSPs), the OIM system produces thousands of orientation measurements, linking
local lattice orientation with grain morphology. Each measurement is represented by a
pixel in the orientation micrographs, to which a color or gray scale value is assigned on
the basis of the local details of lattice orientation or the quality of the corresponding
EBSP image quality (IQ). Recrystallized grains are distinguished according to IQ and
orientation spread. The volume fraction of recrystallized material was determined from
the ratio of the area of the recrystallized grains to the area of the whole image. In this
work the area of a typical OIM scan was about 1800600 m, containing
approximately 130 original grains.
2.2.3. Composite image method
In order to construct composite images a total of four micrographs at 20 stage rotation
intervals were taken from each area. The rotation center of the stage should be aligned
perfectly with the objective lens so that no misalignment of the micrographs results. The
following step was to trace all visible recrystallized grains from each of the micrographs
onto an acetate sheet in order to construct a single representation containing all
recrystallized regions. To reveal the contribution to the total volume fraction made by
each rotation, a separate color was used to trace the visible recrystallized structure from
each individual micrograph. The border of each micrograph was also traced in order to
define a common overlaid area. Quantitative metallography (a standard point counting
technique) was then performed using the composite image. In this study the criteria used
to determine whether a grain is recrystallized were the following:
Size: Grains with an area considerably smaller (<10%) than the original deformed
average grains were taken to be potentially recrystallized grains (when the fraction
recrystallized is small).
Polygonality: Polygonal grains with sharp triple points and straight boundary edges
were counted as potentially recrystallized grains.
Equiaxiallity: Grains with low aspect ratios were taken to be potentially recrystallized
grains. In this study those grains having an arbitrarily chosen aspect ratio up to and
including 1.75 are considered to be recrystallized, the reminder being classed as
unrecrystallized.
Absence of a visible internal structure: Absence of deformation bands was taken as
an additional indication of recrystallized grain character.
A further development of the technique is to use color micrographs in an attempt to
reduce the apparent metamorphosis. By adding a 1/4 filter, grains with different
orientation will exhibit different colors under polarized light. This facilitates a better
identification of the recrystallized grains. In this investigation, similar composite images
were composed from color digital micrographs according to the procedure described
above. The unrecrystallized material can be extracted from the digital image. The
fraction of the recrystallized area, the recrystallized grain sizes (defined as square root
of grain area divided by / 4 ) of individual grains and grain size distribution were then
calculated using automated areal analysis.
2.3. Results
2.3.1. Microstructural evolution
The deformed structures show that the grains in the central part of the rolled sheet
follow the same shape change as the whole specimen. The grains are elongated in the
rolling direction and exhibit a pancake shape with a thickness of 45 m on average, i.e.
half of the initial grain size. The triple point angles among the grains are distorted and
not equal to 120. The as-deformed grain structure showed non-uniform colors and
somewhat unclear grain boundaries. The non-uniform colors reflect the local variation
in the disturbance of the crystal created during deformation.
After annealing the deformed sheet for 5 minutes at 340C, the triple point angles
tended to return towards 120. A few nuclei were observed at the pre-existing grain
corners and along the grain boundaries. The newly formed recrystallized grains
exhibited a uniform color within the grains and sharp grain boundaries. There were
some indications of strain-induced migration of pre-existing grain boundaries (SIBM).
As the annealing time increased to 10 min, the pre-existing grain boundaries became
corrugated. More nuclei could be seen at the pre-existing grain corners or along the
grain boundaries. In some cases clusters of nuclei could be observed. Fig. 1 shows the
nuclei along the boundaries of the pre-existing deformed grains and the SIBM in
regions close to grain edges. After 20 minute annealing, evidence of strain-induced
migration of pre-existing grain boundaries is obvious. The number of nuclei is still
increasing and at the same time nuclei coalescence is observed. Some of the deformed
grains were found to have many more nuclei than others whilst some of them were still
Chapter 2 12
nuclei free. When annealed for 30 minutes or longer, the density of the nuclei remained
constant. However, the size of the recrystallized grains increased with annealing time.
No nuclei were found around FeAl
3
particles.
Fig. 1. Optical micrographs showing the microstructure of AA 1050 recrystallized at
340C for (a) 10 min and (b) 60 min. Arrows indicate the SIBM of region close to the
pre-existing grain boundaries.
2.3.2. Composite imaging analysis
Fig. 2 shows a series of color micrographs taken at 20 rotation intervals of a sample
after annealing 50 min at 340C. As can be seen, the area labeled A in Fig. 2a seems to
be one grain but it is clear in Figs. 2b and 2c that it actually consists of two grains. One
may easily identify the grain boundary at position B on Figs. 2b and 2c, but it is hardly
visible in Fig. 2a. The difference among the individual micrographs is a clear
illustration that it is difficult to determine an accurate recrystallized fraction from a
single micrograph.
The areal fractions recrystallized obtained from single and composite images using the
point counting technique are displayed in Fig. 3. It can be seen that consideration of a
single image leads to a significant underestimate of the volume fraction of recrystallized
material. For all samples studied the composite images constructed from three
orientated micrographs yield a volume fraction about 50% higher than estimated from a
single micrograph. For the samples under investigation here, more than three
micrographs do not show a significant modification to the composite images and
consequently do not lead to a further increase of the measured fraction. The points in
Fig. 3 represent the mean values, and the error bars (shown only for the measurements
on the composite images composed from three micrographs) indicate the spread of the
individual measurements.
Fig. 2. Multiple micrographs showing the microstructural metamorphosis at place A
and B when rotating the sample annealed at 340C for 50 min.
Chapter 2 14
Fig. 3. Recrystallized fraction determined from single and composite images using a
point counting technique, showing the effect of multiple orientated micrographs.
Fig. 4 shows a histogram of the recrystallized grain size at various times obtained by
composite images combined with the automated areal analysis technique. It seems that
the recrystallized grain sizes in the partly recrystallized material follow an
approximately log-normal distribution. The maximum equivalent diameter of the
recrystallized grain, d
max
, and the mean grain size, d
mean
, are employed to describe the
grain size evolution, which is shown in Fig. 4d. Both curves demonstrate that the
growth rate of the recrystallized grains decreases with annealing time and the difference
between the two curves becomes larger. As can be seen in Fig. 1, the impingement
between the clustered grains may prevent them from growing further. This explains the
big difference between d
max
and d
mean
at the later stages of recrystallization.
2.3.3. OIM observation
Fig. 5 shows an OIM image of the same sample as in Fig. 2. The left part in Fig. 5 is the
same region as in Fig. 2 in which there is a nearly perfect grain match with the
conventional microstructure (a mirror image). The black lines indicate high angle grain
boundaries with a misorientation larger than 15, and the thin blue and gray lines in a
deformed grain represent sub-grain boundaries with a misorientation angle between 10-
15 and 4-9, respectively. From the combined information of Inverse Pole Figure (IPF)
maps and IQ maps, recrystallized grains can easily be distinguished from non-
recrystallized grains. The recrystallized grains are uniform in color, which means that
the spread in orientation within the grains is quite small, and free from low angle sub-
grain boundaries. In contrast, the non-recrystallized grains contain deformation
substructures as revealed in slightly different colors and low angle sub-grain boundaries.
0
0.2
0.4
0.6
0.8
1
0 2000 4000 6000 8000
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
Single micrograph
Two micrographs
Three micrographs
Correspondingly, the recrystallized grains are much brighter than the non-recrystallized
grains on the IQ map. This is indicative of a high degree of crystal perfection.
Fig. 4. Size evolution of recrystallized grains vs annealing time.
Fig. 5. OIM map (IPF) showing the microstructure and boundaries with different
misorientation levels in the same sample as in Fig. 2.
20 min
0
0.05
0.1
0.15
0.2
0.25
0.3
3 10 17 25 35 40 More
Grain size (m)
F
r
e
q
u
e
n
c
y
a
90 min
0
0.05
0.1
0.15
0.2
0.25
0.3
8 26 48 70 92 More
Grain size (m)
F
r
e
q
u
e
n
c
y
c
50 min
0
0.05
0.1
0.15
0.2
0.25
0.3
4 16 30 44 58 More
Grain size (m)
F
r
e
q
u
e
n
c
y
b
0
20
40
60
80
100
120
0 1000 2000 3000 4000 5000 6000
Annealing time (s)
d
m
a
x
,

d
m
e
a
n

(
m
)
dmean
dmax
d
Chapter 2 16
Fig. 6 shows the recrystallized fraction determined from OIM for several annealing
times. The results obtained with the optical microscopy technique (the manual point
counting and automated area analysis) are also shown for comparison. Here the attached
error bars are only shown for the area analysis (error is within 5% ). The values from
the point counting technique are lower than those from the areal analysis and both
results are slightly lower than the data from OIM but still within the experimental
scatter of areal analysis data.
Fig. 6. Comparison of the fraction of recrystallized material determined by composite
image method and by OIM.
2.3.4. Micro-hardness measurements
The micro hardness (HV) measurements after isothermal annealing at 340C for various
times are shown in Fig. 7a. From the error bars it can be seen that there is a large scatter
between the individual hardness measurements. However, a clear overall pattern of
annealing behavior can be observed. Similar to the optical microscopy determination,
the sizes of the error bars are larger in the intermediate time range, which demonstrates
the inhomogeneous character of the partially recrystallized material. It can be seen that
the hardness curve from the onset of annealing to the point that represents 2%
recrystallized material (according to optical microscopy examination) can be
represented by a straight line. The slope of the curve increases when the
recrystallization process starts. It indicates a rapid decrease of hardness with increasing
fraction of recrystallized material. The microhardness of a sample directly after cold
work (H
m
) is 46.8 and that of the fully recrystallized material (
0
H ) is 24.5. The
0
0.2
0.4
0.6
0.8
1
0 2000 4000 6000 8000
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
s
e
d
By point counting
By area analysis
By EBSD
softening fraction, R , of an annealed sample could be calculated from these two values
as follows [1,15]
0
m
m
H H
R
H H
(1)
where H is the measured hardness. R calculated according to formula (1) is plotted in
Fig. 7b (diamond symbol). Obviously these data include the softening effects
contributed by both recovery and recrystallization, as will be discussed later.
Fig. 7. Microhardness measurement (a) and the derived fraction of recrystallized
material before and after subtraction of recovery effects (b).
0
0.2
0.4
0.6
0.8
1
10 100 1000 10000 100000
Time (s)
F
r
a
c
t
i
o
n

s
o
f
t
e
n
e
d
Including prior and
concurrent recovery
Excluding prior recovery
Excluding prior recovery
and concurrent recovery
b
10
2
10
3
10
5
10
4
10
1
20
25
30
35
40
45
50
10 100 1000 10000 100000
Time (s)

M
i
c
r
o

H
V
ln
rec m
H H B t
a
10
4
10
1
10
3
10
5
10
2
Chapter 2 18
2.4. Discussion
2.4.1. Composite image method
The work was aimed at accurately quantifying the fraction of recrystallized material
using optical microscopy. As demonstrated in Fig. 2, taking only a single image can not
provide complete information of the partially recrystallized microstructure. If two
neighboring grains are oriented in such a way that certain lattice planes of both lie
roughly perpendicular to the direction of the polarized light, they will exhibit a similar
color. Therefore, some recrystallized grains may not be distinguished from the
deformed parent grain in a single micrograph. As a result, the measured fraction may be
underestimated. Here it is shown that this problem can be easily overcome by using
composite micrographs produced by repeatedly rotating the sample (by e.g. 20) around
the optical axis. Those grain boundaries that are not shown in a single micrograph can
be revealed in others. By this new method the characterization of recrystallized grains is
more reliable and a more accurate recrystallization fraction is obtained.
2.4.2. Comparison of optical and OIM techniques
The recrystallization fractions determined by both optical microscopy and OIM show
the same course of transformation. Measurements by OIM yield slightly higher values
of the fraction, especially at shorter annealing times. This is due to the resolution
difference between the two measuring systems. The smallest recrystallized grain size
that can be clearly observed by the optical microscopy under polarized light is of the
order of 2-5 m. However, the spatial resolution of EBSP is less than 1 m. Recovery
will not affect the results obtained by OIM and optical microscopy, as (1) recovered
regions still contain fairly small subgrains; (2) no significant lattice reorientation occurs
(no new high angle grain boundaries are created); (3) only a misorientation angle larger
than 15 is classified as recrystallized. Recovered regions will therefore be registered as
deformed material.
2.4.3. Effect of recovery
The hardness measurements show a slightly different annealing behavior than derived
from the optical and OIM experiments. The hardness value is proportional to the flow
stress of the material, which is strongly dependent on the dislocation density. This
measurement should reflect both recovery and recrystallization effects. Determination
of recrystallization kinetics from hardness measurements requires separation of the
effects associated with concurrent recovery and recrystallization. Assuming that the
isothermal recovery kinetics follows a logarithmic decay relationship, the time
dependence of micro hardness is then given by the following expression [14,15]
ln
rec
H A B t (2)
Where H
rec
is the hardness of an annealed sample if only recovery takes place,
1
A H ,
the hardness at 1 s annealing, and B, the slope of the straight line, as seen in Fig. 7a.
If the measured hardness, H, of the partially recrystallized materials follows the rule of
mixtures it holds that
0
(1 )
rec
H f H f H (3)
Where f is the volume fraction of recrystallized grains. Assuming that the kinetics of
concurrent recovery follow a straight line when recrystallization starts, by substituting
equation (2) into (3), the recrystallization fraction excluding prior and concurrent
recovery can be obtained as shown in triangle labels in Fig. 7b. If it is assumed that
there is no concurrent recovery taking place after recrystallization starts, by setting
H
rec
=40.5 the hardness at the start point of recrystallization (doted line in Fig. 7a), one
may exclude the prior recovery effect and obtain the recrystallized fraction only, as
shown in Fig. 7b (square label). The recrystallized fraction with exclusion of only the
prior recovery effect is about 10% higher than the value with exclusion of both prior
and concurrent recovery effects. This also indicates that the concurrent recovery effects
are significant during the early stages of recrystallization. By comparing the three
curves in Fig. 7b, one can see that taking into account recovery is indeed necessary to
properly determine recrystallization fraction. After subtraction of the recovery effect the
resulting fraction recrystallized is in good agreement with the metallographic
measurements, as shown in Fig. 8. The difference is within 5%.
Fig. 8. Comparison of fraction of recrystallized material quantified by micro-hardness
tests and optical microscopy on the base of a composite image method.
0
0.2
0.4
0.6
0.8
1
10 100 1000 10000 100000
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d

By Micro Hardness
excluding prior and
concurrent recovery
By Optical Microscopy
10
2
10
5
10
4
10
3
10
1
Chapter 2 20
2.5. Conclusions
1. A new method for characterizing the microstructure in AA1050 by optical
microscopy is presented. Combining this method with an areal analysis technique yields
a more accurate measurement of the fraction of recrystallized material, which is in
excellent agreement with OIM measurements.
2. Microhardness measurements demonstrate another course of annealing behavior,
which includes both effects associated with concurrent recovery and recrystallization
processes. After subtraction of the effect of the concurrent recovery, the micro hardness
test can provide a reasonable quantification of the recrystallized fraction developed with
annealing time.
References
1. E.C.W. Perryman, Trans. AIME, J. Metals, Sept. 1955, 1053-1061.
2. P.L. Orsetti Rossi and C.M. Sellars, Acta Mater., 1997, 45, 137-148.
3. R.A. Vandermeer, Metal. Trans., April 1970, 1, 819-826.
4. P. Faivry, R.D. Doherty, J. of Mater. Sci. 1979, 14, 897-919.
5. C.N. Sparks, PhD thesis, University of Sheffield, 1993.
6. M.P. Black and R.L. Higginson, Scr. Mater. 41, (2), 1999, 125-129.
7. E. Woldt, D. Juul Jensen, Metall. Trans., 26A, July 1995, 1717-1724.
8. O. Engler, G.Gottstein, Steel Research, 63, 9, 1992, 413-418.
9. D.J. Dingley and K. Baba-Kishi, Scanning Electron Microsc., 1986, 27, 383-395.
10. T. Furu, R. Orsund and E. Nes, Acta Met. Mater., 43, 1995, 2209-2232.
11. D. Bowen, R.R. Eggleston and R.H. Kropschot, J. Applied Physics. 1952, 23 (6),
630-635.
12. K. Mukunthan and E.B. Hawbolt, Metall. Mat. Trans. A 27, Nov. 1996, 3410-3423.
13. F. Li and P.S. Bate, Acta Metall. Mater. 39, 11, 1991, 2639-2650.
14. J.G. Byrne: "Recovery, Recrystallization and Grain Growth", MacMillan Co., New
York, NY, 1965, 37-59.
15. F.J. Humphreys and M. Hatherly, Recrystallisation and Related Annealing
Phenomena, Pergamon, London, 1996.
S.P. Chen
Chapter 3
Modeling the kinetics of grain boundary
nucleated recrystallization processes after
cold deformation using a single grain
approach
A physical model to predict the recrystallization kinetics of single phase polycrystalline
metals, based on a single grain representation of deformed microstructure (characterized
by a mean subgrain size and mean misorientation of subgrain boundaries), is presented.
The model takes into account the grain geometry, the position and the density of the
nucleation sites. The selected geometry is a regular tetrakaidecahedron, combining
topological features of a random Voronoi distribution characteristic for polycrystalline
material with the advantages of a single-grain calculation. The model employs
empirically determined relationships from existing literature to describe the deformed
microstructure and in so doing, facilitates the prediction of the recrystallization behavior
when only the deformation strain and the recrystallization temperature are known. The
boundary mobility and the driving force as well as the nucleation density are related to
the true plastic strain of deformation through the microstructure. The model also
describes the effects of concurrent recovery on the overall recrystallization kinetics.
3.1. Introduction
Recrystallization has long been an important subject for metallurgical research [1]. The
microstructural evolution during recrystallization can be described phenomenologically
as a nucleation and a growth process. The basis for the theoretical treatment of the
recrystallization transformation kinetics is the JMAK equation which is based on the
Chapter 3 22
assumptions that the recrystallized nuclei form randomly in the pre-existing
microstructure and that the growth rate of these nuclei is constant and isotropic. The
modeling process is implemented by posing the nucleation and growth characteristics of
the product phase based on the relationship between the volume fraction transformed,
X , and the extended volume fraction,
ex
X :
(1 )
ex
dX X dX = (1)
The ideal JMAK behavior is rarely exhibited by real materials [1-3].
It is widely recognized that the nucleation sites in recrystallization are non-randomly
distributed [4,5]. Nucleation tends to occur at preferred sites such as prior grain
boundaries, edges and corners or inhomogeneities as transition bands and shear bands.
Indeed, grain boundary nucleation has been frequently observed [1,6,7] and it plays a
dominant role in recrystallization kinetics when the initial grain size is small, or at lower
strains. Another important issue [8-10] that arises from the analysis of experimentally
determined recrystallization kinetics is that the average interface migration rate is often
found to decrease with time.
For cases where the critical assumption of randomness is violated, the JMAK equation
can not be employed directly and an alternative approach to the impingement problem is
necessary. A theoretical method of treating impingement caused by clustered nucleation
at grain boundaries and grain edges was presented a number of years ago by Cahn [11]
and was developed recently for time dependent growth rate by Vandermeer and
Masumura [12]. In his analysis, which was based on Eq. (1), Cahn considered two
separate types of impingement: first, impingement among nodules originating from the
same grain boundary (or edge), and second, the impingement between nodules
originating from different boundaries or edges. Eq. (1) is, however, only an
approximation which applies when nucleation occurs at the heterogeneities. A more
accurate description of the kinetics and the resulting microstructure in these cases can
be obtained by computer simulation [1,13].
The goal of the present program is to develop a physical model to predict the
recrystallization kinetics when only the deformation strain and annealing temperature
are known. Since deformation is a necessary precursor for both nucleation of
recrystallized grains and sustained growth during recrystallization, the distribution of
nucleation sites is strongly dependent upon both the deformed grain geometry and the
defect structure induced by deformation. As a first step, this chapter deals with
recrystallization kinetics of the grain boundary nucleation with a simplified model. The
influence of the grain geometry, nucleus site density and the relative positions of the
Modeling the recrystallization kinetics of the grain boundary nucleation 23
nuclei inside the original grain on the kinetics of the recrystallization will be considered
in detail. The boundary mobility and the driving force as well as the nucleation density
will be related to the true plastic strain of deformation through the microstructure. The
effects of concurrent recovery and the starting grain size on recrystallization kinetics
will be also explored.
3.2. The Kinetic Model
3.2.1. Grain Geometry
The mathematical concept that models the recrystallized microstructure (equiaxed
grains) most ideally consists of a construction of randomly chosen Voronoi cells. The
construction of a Voronoi tessellation of space shows a strong resemblance to the
evolution of a real microstructure. It has been demonstrated that a reasonable
correspondence exists between the geometrical parameters for the average values for the
Voronoi tessellation and the tetrakaidecahedron single grain [14]. Therefore, the
microstructure of an well-annealed material can be described easily by a distribution of
tetrakaidecahedra of various sizes. A number of studies has shown that this single grain
representation with discrete nucleation sites offers a new approach to modeling the
phase transformations in steels and ferrous alloys and a model for treating growth of a
ferrite into austenite was developed [14-16]. It will now be briefly summarized and
adapted to recrystallization.
In this phase transformation model, the austenite grain is described as a regular
tetrakaidecahedra. The ferrite grains are assumed to nucleate from the austenite grain
boundary. After nucleation the growth of each nucleus is modeled as an expanding
sphere and is controlled by interface velocity, which depends on the driving force and
interface mobility. In the case of recrystallization, there is no phase transformation.
However, since the deformed metal stores some energy in the form of various types of
imperfections, the deformed state can be considered as a phase that has a higher energy
than the fully recrystallized phase. In this sense, the recrystallization can be treated as if
a new phase would form. For a phase transformation, the driving force is the difference
in the Gibbs energy between two phases, and for recrystallization, this comes from the
stored energy. On the other hand, the recrystallization process is recognized to be the
motion of a boundary. This is true not only for the growth stages but also for the
nucleation stage because the formation of a mobile recrystallized front, i.e. nucleation,
results from subgrain growth, subgrain coalescence or strain-induced grain boundary
migration. Therefore, that the concept of the transformation is controlled by boundary
velocity is also employed in the case of recrystallization. Furthermore, it is well known
that the new recrystallized grains are frequently nucleated from the original grain
boundaries, the area around the second phase particles and deformation inhomogeneities
Chapter 3 24
such as deformation bands [1]. For pure aluminium without second phase particles
under intermediate deformation strains the grain boundary nucleation should be the
dominant mechanism.
Based on the above analysis, the concept of the previous phase transformation model is,
in principle, also applicable to the recrystallization process. To investigate the kinetics
of recrystallization, we first study the behavior of a single deformed tetrakaidecahedron.
Homogenous deformation not only changes the shape of the tetrakaidecahedron but also
leads to an increase in the grain surface area per unit volume, while there is no change
in the number of grain corners per unit volume. The deformation of a
tetrakaidecahedron can be described by a 33 deformation matrix [17], which operates
on each vector in turn to generate a set of new vectors defining the new shape. Thus a
vector, u, becomes a new vector, v, as a consequence of a homogenous deformation S:
|
|
|
.
|
\
|
|
|
|
.
|
\
|
= =
3
2
1
0 0
0 0
0 0
u
u
u
c
b
a
Su v (2)
where
i
u are the components of u; and a , b , and c are the principle distortions (ratios
of the final to initial lengths of the unit vectors along the principal axes ( , , x y z )).
Therefore, a ln , b ln and c ln are the true strains along the three principal axes of the
deformation and 1 = abc . Rolling involves plane strain deformation with 1 = b and
1 = ac .
The three-dimensional construction, as shown in Fig. 1, is used in the simulation. This
has a reference frame of rolling direction (RD), transverse direction (TD), normal
direction (ND) with RD and ND along x and z axes, respectively. The length of the
undeformed cube side, d , the distance between two opposite square faces of the
tetrakaidecahedron, is defined as the grain size prior to the deformation. In the
numerical calculations the cube is divided into volume elements that are mapped onto a
three dimensional matrix. The number of elements used to describe the cube is
( N N N , , ). After a deformation the number of the elements takes the integer of
( cN bN aN , , ). The aspect ratio of grains in the direction RD/ND is c a / depending on
the strain of the rolling deformation. The recrystallized nuclei can be randomly
distributed in the deformed grain or be arranged on the grain boundary or corners. In the
model it is assumed that grains grow independently of one another. After each
nucleation event, the nucleus grows in a spherical manner with a growth velocity V
perpendicular to the interface, which is a constant in space but may vary with time. The
simulated geometric grain is permitted to grow beyond the faces of the
tetrakaidecahedron, but the volume of the new grain is computed only for the portion of
the grain contained within the tetrakaidecahedron. An approach, which is comparable
with the time cone method of Cahn [18], is used to calculate the volume fraction
transformed. We calculate the probability that the element ( , , ) x y z X in the volume is
untransformed at time t . At each time step, the radius of each nuclei, ( , ) R t t , and the
distance between each element and the center of every nuclei, X' , are calculated and
compared. If the magnitude of the grain radius ( , ) R t t exceeds the distance between X'
and X:
2
2
( , ) X X' 0 R t t > (3)
the element X will have transformed before time, t . The number of untransformed sites
is then counted to calculate the transformed fraction and from that the transformation
rate.
Fig. 1. A deformed tetrakaidecahedron in a cuboid used as computer specimen.
3.2.2. The deformed microstructure
In a high stacking fault energy metal such as aluminium it is assumed that the
application of a strain of c ~ 0.2 results in a uniform equiaxed cellular structure [19,20].
These cells or subgrains are slightly misoriented with respect to each other. For
intermediate rolling strains ( 0.2 1.5 c < < ), the substructure is inhomogeneous, with the
size and the shape of the cells/subgrains varying from region to region in the material.
x
z
y
RD
ND
TD
Chapter 3 26
At larger strains ( 1.5 c > ), the substructure becomes more complex but more spatially
homogeneous.
It has been demonstrated that mean subgrain size and misorientation depend on the true
deformation strain [19-21]. The variation in the average subgrain/cell size with strain
follows a relationship that is of a similar type for a range of materials, and independent
of the deformation mode. The relationship between the mean cell/subgrain size, o , and
deformation strain in aluminium is given by [21]:
0.35 0.17 / o c = + (4)
where c is the true strain, for rolling deformation ln a c = , and o is the mean subgrain
diameter in m.
In the case of aluminium the average subgrain boundary misorientation, u , is reported
to increase rapidly with strain, reaching about 2-3 at a strain of about 1, after which it
remains constant up to rolling strain as high as 4 [20]. The following relationship
between mean subgrain diameter and misorientation has been established for pure
aluminium deformed by cold rolling [22]:
/ b C ou = (5)
where b is the Burgers vector and C is a constant between 60~80 rad.
For aluminium, the mean subgrain size and mean misorientation of the subgrain
structure as a function of the true strain are shown in Fig. 2. It can be seen that the
subgrain size decreases while the mean misorientation increases as strain increases.
Both saturate at higher strains.
Fig. 2. The variation of the mean subgrain size and misorientation with deformation
strain.
3.2.3. Nucleation
In this chapter, nuclei are assumed to be present at pre-existing high angle grain
boundaries after deformation and only the kinetics of grain boundary nucleated
recrystallization is considered. Other potential nucleation sites such as deformation
bands and shear bands increase in significance when the deformation true strain
becomes greater than 1.3 [18]. The present study on strictly grain boundary nucleated
recrystallization is limited, therefore, to the range of small to intermediate deformation
strains.
The probability of finding a critical-sized subgrain on the grain boundary depends on
the average subgrain size, o , and the grain boundary area per unit volume,
v
S . The
nucleation site density
v
N in the grain boundary can be estimated using the expression
[23,24]:
2
/
V d v
N C S o = (6)
where
d
C is a calibration constant which determines the potency of the grain boundary
as a nucleation site (
d
C , in the present work, is taken as
4
2.5 10
).
Deformation can increase the nucleation site density for recrystallization through an
increase in grain boundary area per unit volume,
v
S , of the deformed structure. The
0
0.5
1
1.5
2
2.5
3
3.5
4
0 0.5 1 1.5 2 2.5
True strain

M
i
s
o
r
i
e
n
t
a
t
i
o
n

(

)
0
0.5
1
1.5
2
2.5
S
u
b
g
r
a
i
n

s
i
z
e

(
m
)
o
u
Chapter 3 28
value of
v
S is dependent on the rolling reduction strain and the initial grain size, and for
a tetrakaidecahedron, varies with rolling strain in a manner described by [17]:
( )
2 2 2 2
1
3 1 2/ 3 2/ 2 2 /
2
v
S a a a a a a a
d
= + + + + + + (7)
where a e
c
= is the distortion along the rolling direction.
Fig. 3 shows the variation of the nucleation density with true strain for three different
initial grain sizes. As the strain increases the nucleation site density
v
N increases
significantly as a result of both an increase in
v
S and a concurrent decrease in o .
Fig. 3. The nucleation site density as a function of true strain.
3.2.4. Growth
The migration of low and high angle boundaries is the most important atomic-scale
mechanism that occurs during recovery and recrystallization of the deformed materials.
The relation between the rate of migration of the interface, V , and the driving pressure
for boundaries with a specific energy and mobility,
n
M , moving into a uniformly
deformed matrix is given by the simple expression:
n
V M P = (8)
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2 2.5
True strain
N
u
c
l
e
a
t
i
o
n

s
i
t
e

d
e
n
s
i
t
y

(
a
.
u
)
d=50m
100
150
where P is the driving pressure which is assumed to arise from the stored energy of the
dislocations in the subgrain boundaries and in the subgrain interior.

While a real deformed microstructure contains subgrains of a variety of sizes and
misorientations, in the present model, following the approach of Humphreys [25], we
simplify the analysis by considering the substructure to be adequately described using
two components. The major component is considered as an assembly of equiaxed
subgrains of mean equivalent radius, R , mean misorientation, u , and with boundaries
of mean energy and mobility, and M respectively. The minor component we
consider as particular subgrains (effectively sub-critically sized recrystallization
nuclei) which have a larger size (
n
R ) and different boundary characteristics ( , ,
n n n
M u )
to that of the average subgrain assembly.
The total energy of a particular subgrain as a function of radius R is given by:
3
4
3
n n v
E A R E t = (9)
where
2
~ 4
n
A R t is surface area of the particular subgrain and
v
E is stored energy in the
average assembly estimated by [1]:

2
/ / 2
v i
E R Gb o = + (10)
The driving pressure for this particular subgrain to grow is given by the relative change
of the total energy when the dislocation density inside cell,
i
, can be ignored.
2 1
n
n n
dE
P
A dR R R
o
= = + (11)
where o is a geometrical constant and has a value of ~1.5 [1].
If the main mechanism of recovery in the subgrain assembly is subgrain coarsening, the
growth rate of a uniform subgrain assembly may be expressed in the form [1,26]:
dR M
dt R
o
= (12)
As a simple approach, the boundary energy is assumed to be dependent only on the
average misorientation angle and is given by [1]:
Chapter 3 30
1 ln
m
m m
u u

u u
( | |
=
( |
( \ .
(13)
where
m
and
m
u are the values of boundary energy and misorientation for high-angle
boundaries and which are commonly taken as 0.324 J/m [1] and 15 respectively.
Subgrain boundary mobility is assumed to be related to u and the limited data in the
literature shows that the M vs u curve is sigmoidal [25,29,30]. In this analysis, the
following empirical form is taken:
( )
1
n
m
B
n
M M e
u
u
| |
| =
|
\ .
(14)
where
n
M is the mobility of a high angle boundary at an annealing temperature given
by:
( )
0
exp /
n g
M M Q R T = (15)
where
0
M =1.78 10
2
m
4
/Js and Q=147 kJ /mol [27,28],
g
R , gas constant, 4 = n and
5 = B for aluminum [25]. It is suggested [25] that Eq. (14) should be applicable over the
temperature range 250 - 400C. It should be noted that Eq. (14) is an empirical
relationship and M is very sensitive to the values of B and n .
Up to now the model has dealt with the recrystallization kinetics of a single deformed
grain. As long as the physical parameters,
d
C ,
0
M , Q, B and n could be experimentally
determined, the recrystallization kinetics in an as-deformed grain with size, d , which is
given as a data array of the fraction recrystallized, f, vs time, t, could be predicted
numerically provided the deformation strain, c , and the annealing temperature, T , are
known.
Actually, the starting grain size distribution is also an important parameter for the
recrystallization kinetics. Experimental measurements [31,32] have shown the grain size
distribution in annealed materials to be approximately log-normal, where the maximum
grain size is typically 2.5~3 times mean size. To determine the overall recrystallization
kinetics in a log-normal polycrystalline aggregate, we treat it in the following way: the
distribution function of the starting grain size is divided into j equal size classes in the
interval
max
0 / 2.5 ~ 3 d d s s (depending on the standard deviation). For each of the
size classes, the average grain size
i
d and frequency
i
are calculated. A data array of
the fraction
i
f vs time t for each of the j size classes is computed from the single grain
model. Each class is operating in a monolithic body without reference to other size
classes. The overall fraction recrystallized f is obtained by weighted summation over the
j classes through Eq. (16):
( )
i i i
f f = _ _ (16)
where
i
is the frequency of the initial grain with diameter of
i
d ,
i
f is the fraction
recrystallized of a class of the grains with diameter of
i
d at annealing time, t .
3.3. Results and discussion
3.3.1. The effect of grain geometry on the recrystallization kinetics
In real materials, the space into which the recrystallization nuclei can grow is effectively
defined by both initial grain size and the deformation geometry. Ideally, the starting
grain structure prior to deformation is isotropic and essentially uniformly sized. As
deformation ensues, the grain structure is constrained to conform to the deformation
geometry. In this section a number of simulations of the recrystallization kinetics will be
presented, in which all the nucleation takes place at time, 0 t = , and the interface
velocity is assumed to remain constant throughout the annealing cycle and to be
independent of the applied deformation strain. Thus, only the influence of the grain
geometry resulting from various degrees of deformation, the position of the nuclei and
the nucleation site density on the recrystallization kinetics will be considered and
discussed. The results are presented as a function of the dimensionless reduced time
*
t ,
which is defined as:
*
V
t t
d
= (17)
where d is the starting grain size. The reduced time parameter indicates that the actual
transformation time depends on the ratio d V / rather than solely on V .
Let us first consider an ideal case, assuming six nuclei either each located at one of the
corners of the six quadrilateral faces or at the center of the quadrilateral faces of a
tetrakaidecahedron, to see the effect of the grain geometry on the recrystallization
kinetics. The calculated kinetics curves, giving the recrystallized fraction as a function
of reduced time after various strains, are shown in Fig. 4. Curves in thin line represent
the cases where each of the six nuclei is at a corner of the quadrilateral faces while the
Chapter 3 32
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Reduced time t*
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
1
2
3 4
Fig.4. S.P.Chen
curves in thick line are the cases where each of six nuclei is at the center of a
quadrilateral face.
Fig. 4. Calculated recrystallization kinetics for six nuclei each at one of quadrilateral
faces after various rolling strains. 1. 0 c = , 2. 0.69 c = , 3. 1.1 c = , 4. 1.6 c = (for
details, see text).
It is clear from these simulations that the geometrical shape change caused by the
deformation exerts a large effect on the overall recrystallization kinetics with the rate
becoming slower as deformation strain increases. The reduced rate is a result of the
elongation of the tetrakaidecahedron in the rolling direction and the concomitant change
in the relative positions and distances between corners and faces. The distance between
two opposite quadrilateral faces of the tetrakaidecahedron along the rolling direction
increases whilst along the normal direction it decreases. Early impingement between
growing grains along the normal direction in the deformed structure, therefore, occurs
and this is intensified with increasing deformation strain.
The shape of the curves in Fig. 4 changes with increasing deformation strain. An
evolution of the shape of the curves from the typical sigmoidal form ( 0 c = ) to the
truncated sigmoidal shape (after a small deformation) and to a straight line with a kink
(at large strains) with increasing deformation is evident. In the case where there is no
deformation the nuclei are symmetrically distributed in space on the six quadrilateral
faces of the tetrakaidecahedron. It is this uniform distribution of nuclei which gives rise
to the sigmoidal shape of the kinetics curves. After deformation the distribution of
nuclei is no longer uniform in space because of relative change of the nucleation site
positions. Early impingement in the normal direction but late in the rolling direction,
therefore, is responsible for the transition from a three dimensional, to a two-
dimensional and finally to a one-dimensional growth, and gives rise to the observed
shape change of the recrystallization kinetics curve.
In Fig. 4 the recrystallization kinetics for random nucleation using the JMAK equation
assuming an exponent 3 k = are also shown. For the cases of six nuclei, when each
nucleus is on one of the quadrilateral faces or at one corner of the quadrilateral faces the
number of nuclei per tetrakaidecahedron,
n
N , is 3 and 1.5 respectively. The nucleus
density
n
n , is, therefore, given by
3
2 /
n n
n N d = . The thick dotted line in Fig. 4
corresponds to 3
n
N = while the thin dotted line applies to 1.5
n
N = . Clearly,
impingement geometry resulting from deformation is quite complex and the JMAK
model is not capable of taking this into account.
Fig. 5. Calculated recrystallization kinetics for cases of 1 and/or 2 nuclei each at one of
quadrilateral faces in different directions after a rolling strain of 0.69 c = .
The dependence of the kinetic curves on the relative positions of the nuclei can be seen
in Fig. 5, which is calculated for a deformed grain geometry corresponding to a strain of
0.69. The number on each curve corresponds to the number of nuclei. The symbols R, N
and T refer to the nuclei located at the quadrilateral faces perpendicular, normal or
transverse to the rolling direction respectively. For example, when each of the two
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2
Reduced time t*
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
2N 2T
1N
2R
1T 1R
Chapter 3 34
nuclei is at one of the two quadrilateral faces perpendicular to normal direction the rate
(curve marked 2N in Fig. 5) is steeper than those of the other two cases (2T and 2R).
The balance between early impingement and larger interface area of recrystallized /
unrecrystallized regions explains this relative order. When two nuclei start from two
opposing quadrilateral faces of the deformed tetrakaidecahedron in the direction of the
rolling deformation, the impingement occurs later but the interface area inside the
deformed grain during growth is comparatively small.
Fig. 6. Calculated recrystallization kinetics as a function of the number of nuclei after a
particular rolling strain. (a). 0 c = , (b). 0.69 c = .
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2
Reduced time t*
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d

24 6 2N 1N
2R 1R
(b)
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2
Reduced time t*
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d

%
24 12 6 2 1
(a)
It may also be inferred that nucleation density has a significant influence on the kinetics
of recrystallization. In Fig. 6 the calculated recrystallization curves for different number
of nuclei positioned within the deformed tetrakaidecahedron after strains of 0 c =
(Fig. 6a) and 0.7 c = (Fig. 6b) are shown. The curves marked 24 represent the condition
that each of the 24 nuclei is positioned at a corner of the tetrakaidecahedron. Curves
marked 6 are the cases that each of the 6 nuclei is arranged at one of the quadrilateral
faces. For the case of 2 nuclei, each is located at one of the two parallel quadrilateral
faces. R, N, T have the same meaning as in Fig. 5. The greater the number of nuclei, the
faster the transformation rate becomes. Fig. 6b also suggests that, when the density of
nuclei are low, the relative positions of the nuclei in the deformed grain are more
significant for the overall recrystallization time than the number of nuclei present.
We must emphasize that the above simulations are based on the assumptions that the
growth rate is constant in time and equal for all conditions, and that the nuclei start from
the same equivalent position irrespective of the applied strain. These results, therefore,
just indicate that the grain geometry change resulting from deformation is of importance
in deciding the kinetics of recrystallization. In reality, larger deformation results in a
higher stored energy and a larger surface area per unit volume that make growth rate
and the nucleation site density increase with increasing deformation, hence the true
trend in the recrystallization kinetics is the opposite to that shown here.
In the following sections we will first establish the relationships between the driving
pressure and the deformation strain and then present the simulations of the
recrystallization kinetics in aluminium, in which the effect of plastic deformation is
taken into account properly. The positions of nuclei on the faces of the
tetrakaidecahedron are determined randomly assuming that there is an equal probability
that each of the 14 faces may act as a nucleation site. Each simulation is therefore
repeated a number of times and the average of the simulated kinetics taken as
representative.
3.3.2. Driving pressure for recrystallization
1. The influence of deformation on driving pressure and critical nucleus size
As demonstrated by Eq. (9), the driving pressure for recrystallized grain growth depends
on two terms: the boundary surface energy of the nuclei and the stored energy in the
assembly. The surface energy term depends on the misorientation,
n
u , (which is
assumed to be 10-15) between the nucleus and its surroundings and radius of the
nucleus. The stored energy in the assembly is determined by the mean subgrain size and
misorientation in the assembly. When it is assumed that there is no concurrent recovery
Chapter 3 36
then the stored energy in the unrecrystallized material remains constant. The sum of the
two terms yields the driving pressure that increases nucleus size and approaches
asymptotically the value of the stored energy in the assembly as recrystallization
proceeds. When the driving pressure is positive, the "nucleus" may begin to grow. The
critical nucleus size for recrystallization is dependent on , R u and
n
u and is given by:
2 ( )
( )
n
c
R R
u
o u
= (18)
Fig. 7 shows the variation of the driving pressure with the radius of the recrystallized
grain after various deformation strains. As the deformation increases, the subgrain size
decreases and misorientation increases, the driving pressure increases whilst the critical
nucleus size decreases.
On the basis of an earlier assumption,
n
M is a constant for high angle grain boundary
(u =15) at a given temperature [1]. The growth rate will, therefore, vary in the same
manner as the driving pressure as recrystallization proceeds. That is to say, it will
increase rapidly during the early stages of recrystallization and then, as the process
continues, approach asymptotically a value equal to the product of the stored energy in
the assembly with the mobility.
Fig. 7. The variation of driving pressure with recrystallized grain radius after various
deformation strains.
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 2 4 6 8 10
Recrystallized grain radius (m)
D
r
i
v
i
n
g

p
r
e
s
s
u
r
e

(
M
P
a
)
c=1.95
0.69
0.22
0.41
Rc=3.5m
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 1 2 3 4 5
Timex10
4
(s)
D
r
i
v
i
n
g

p
r
e
s
s
u
r
e

(
M
P
a
)
No recovery
Recovery n=4
Recovery n=3.5
c=0.69
0.22
0.41
Fig.8, S.P.Chen
2. Effect of concurrent recovery on the driving pressure
The above results apply when there is no recovery in the assembly. When concurrent
recovery takes place the driving pressure may decrease as the recrystallizing grain
grows. The magnitude of this recovery effect on recrystallization will depend on the rate
of the recovery reaction as compared with migration rate of the interface. As proposed
previously (section 2.4) the dominant recovery mechanism is considered to be subgrain
growth, a situation expected during annealing of aluminium at higher temperature
[33,34]. If u , and thus , are assumed to remain constant, which is reasonable if it is
assumed that no orientation gradient is present [1], the kinetics of subgrain growth will
be parabolic. As might be expected, the increase in deformation strain results in a higher
u and smaller R , and so, should increase the kinetics of subgrain growth. The effect of
concurrent recovery on the driving pressure for recrystallization after various degrees of
deformation is shown in Fig. 8. In contrast to the case of no recovery, where the driving
pressure continues to increase to a stable level, in the presence of concurrent recovery it
will first increase due to the decrease of the surface energy term as the recrystallized
grain grows and then decreases because of the decrease in the stored energy in the
assembly. The driving pressure is lower when concurrent recovery is active and this
effect increases with increasing strain.
Fig. 8. The effect of concurrent recovery on the driving pressure for recrystallization.
Chapter 3 38
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5
Timex10
4
(s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
E
v
/
E
0
No recovery
Recovery n=4
Recovery n=3.5
c=0.22
0.41
0.69
Fig.9, S.P.Chen
3.3.3. Simulation of recrystallization kinetics
1. The recrystallization kinetics by single grain approach
After considering the grain shape change, the microstructure (subgrain size,
misorientation) evolution and the nucleation site density change as a function of
deformation strain we are now in a position to simulate the kinetics of the
recrystallization process in aluminium by specifying various microstructures after
different degrees of rolling deformation and inserting appropriate values of driving
force and interface mobility from Eqs. (9) and (12) into Eq. (6). In this simulation we
consider a single grain, 100 m in diameter. The number of nuclei after true strains of
0.22, 0.41 and 0.69 are 4, 10 and 20, respectively.
Fig. 9 shows the predicted recrystallization kinetics in aluminium after various strains
for the cases of no recovery and with recovery and also the variation in the stored
energy
0
/ E E
v
(
v
E and
0
E being the stored energy in the assembly at time t and 0 t =
respectively) when concurrent recovery is active. The rate of recrystallization increases
with the amount of deformation. While the stored energy increases with deformation,
the increase in nucleation site density resulting from
v
S as deformation increases could
account for a significant portion of the corresponding increase in recrystallization
kinetics. It can be noted that the effect of concurrent recovery on the recrystallization
kinetics increases as deformation strain increases. The magnitude of the effect of the
recovery on the recrystallization kinetics is strongly dependent on the value of the
parameter n employed in Eq. (14).
Fig. 9. Simulation of recrystallization kinetics in aluminium after various degrees of
deformation by the single grain model.
0
10
20
30
40
50
60
70
80
90
0 1 2 3 4 5
Time x10
4
(s)
R
r
e
x

(
m
)
No recovery
Recovery n=4
Recovery n=3.5
0.22
0.41
c=0.69
Fig.10, S.P.Chen
Fig. 10. Prediction of the change in the recrystallizing grain radius with time.
The predictions of the recrystallized grain radius as a function of annealing time are
illustrated in Fig. 10 for the case of no-recovery and the case of recovery. When the
condition of no recovery is assumed linear growth is predicted. This is a result of the
stored energy and hence driving pressure increasing throughout the nucleation stage and
then attaining a constant value during recrystallization. In the cases where recovery is
taken into account a departure from this linear growth occurs for all strains. As the
deformation strain increases the deviation from the linear growth condition becomes
more obvious with the extent of the deviation depending on the magnitude of the
concurrent recovery assumed.
2. The recrystallization kinetics when there is a grain size distribution
A more realistic simulation of recrystallization kinetics should consider the range of the
grain sizes before deformation. The initial grain size affects the recrystallization kinetics
in three ways: by a change in the nucleation site density as a consequence of
v
S , which
decreases as grain size increases, by changing the relative distances among nuclei and
by a variation in stored energy resulting from variation in dislocation density after
deformation. The last factor is most complicated as it also depends on texture and
orientation with respect to deformation field and is not considered here.
Fig. 11 shows simulations of recrystallization kinetics for an assembly with a grain size
distribution (of mean size d =100 m and a standard deviation =1.5) in comparison
with the simulated kinetics from a single grain of mean size. In this case, the
Chapter 3 40
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5
Timex10
4
(s)
F
r
a
c
t
i
o
n

r
e
c
r
y
t
a
l
l
i
z
e
d

Single grain
Multigrains =1.5
c=0.22
0.41
0.69
Fig.11, S.P.Chen
distribution function is divided into 12 equal size classes. The overall kinetics curves are
derived according to Eq. (16) by assuming that the subgrain size does not change with
grain size. The temperature in the simulation is kept at 340C and it is assumed that
there is no concurrent recovery present. Fig. 12 shows the effect of the standard
deviation of the grain size distribution on the recrystallization kinetics after a strain of
0.69 for both the case of concurrent recovery and that of no recovery. As can be seen,
the rate of recrystallization of the assembly of grains (thick line) is slower than that of
the single grain of the mean size (thin line). The rate difference between the two cases
decreases with increasing deformation strain in the strain range studied. These can be
explained as follows. The recrystallization kinetics is accelerated by an increase in the
nucleation site density that is proportional to the grain boundary area per unit volume S
v
.
Fine-grained materials recrystallize faster because of a larger S
v
. Deformation also leads
to an increase in S
v
. The increase in S
v
by grain size will have a less effect as strain
increases. On the other hand, deformation makes the geometrical shape change of the
grain and leads to a change in the relative positions of the nuclei. As we have analyzed
in section 3.1, the relative positions of the nuclei have a stronger effect on the overall
recrystallization kinetics than the change in nucleus density with increasing strain.
Hence, recrystallization kinetics becomes less dependent on the original grain size as
strain increases. A notable and interesting fact is that the
0.68
t of recrystallization for
both single grain and multigrain approaches are the same for all the grain size
distributions considered.
Fig. 11. Simulation of recrystallization kinetics in aluminum after various degrees of
deformation assuming a grain size distribution.
0
0.2
0.4
0.6
0.8
1
0 0.4 0.8 1.2 1.6
Timex10
4
(s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d

Single grain
Multigrains =1.38
Multigrains =1.5
Recovery n=4
No recovery
Fig.12, S.P.Chen
Fig. 12. The effect of the width of the grain size distribution on the overall
recrystallization kinetics after a strain of 0.69.
3.4. Conclusions
A physical model to predict the grain boundary nucleated recrystallization kinetics of
single-phase metals based on the microstructure of the deformed state is proposed. The
model, notwithstanding its simplicity, provides a new insight into the kinetics of
recrystallization. The model includes the grain geometry of the deformed structure, the
nucleation density and the relative positions of the nuclei and these factors have been
shown to have a large effect on the overall recrystallization kinetics. By employing data
from the open literature, in conjunction with the deformed grain geometry, the
recrystallization kinetics could be simulated for a range of deformation and heat
treatment conditions. The influence of concurrent recovery on recrystallization kinetics
is shown to increase with increasing deformation strain. The potential of the model is to
be explored further for recrystallization kinetics following warm deformation.
Chapter 3 42
References
1. F. J. Humphreys and M. Hatherly, Recrystallization and Related Annealing
Phenomena, Pergamon, London, 1996.
2. D. Juul Jensen, N. Hansen and F. J. Humphreys, Acta Metall., 1985, 33, 2155-2255.
3. N. Hassen, T. Leffers and J. K. Kems, Acta Metall., 1981, 29, 1523-1622.
4. P. A. Beck and P. R. Sperry, J. Appl. Phys., 1950, 21, 150-157.
5. J. Hjelen, R. Orsund and E. Nes, Acta Metall., 1991, 39, 1377.
6. P.L. Orsettirossi and C.M. Sellars, Mater. Sci. and Techn.,1999, 15, 185-192.
7. P.L. Orsettirossi and C.M. Sellars, Mater. Sci. and Techn.,1999, 15, 193-201.
8. T. Furu and E. Nes, Materials Science Forum, 1993,113-115, 311-316.
9. R. A.Vandermeer and B. B. Rath, Metall. Trans. A, 1989, 20, 391-401.
10. R. A.Vandermeer and D. Juul Jensen, Acta Metall. 1994, 42, 2427-2436
11. J. W. Cahn, Acta Metall., 1956, 4, 449-459.
12. R. A. Vandermeer and R. A. Masumura, Acta Metall., 1992, 40, 877-886.
13. A. D. Rollett, Progress in Materials Science, 1997, 42, 79-99.
14. Y. Van Leeuwen, S. Vooijs, J. Sietsma and S. van der Zwaag, Metall. and Mat.
Trans. A, 1998, 29, 2925-2931.
15. S. I. Vooijs, Y. Van Leeuwen, J. Sietsma and S. van der Zwaag, Metall. and Mat.
Trans. A, 2000, 31, 379-385.
16. T.A. Kop, Y. Van Leeuwen, J. Sietsma and S. van der Zwaag, ISIJ International,
2000, 40, 713-718.
17. H. K. D. H. Bhadeshia: Worked examples in the geometry of crystals, 1987, The
Institute of Metals, London.
18. J. W. Cahn, Trans. Indian Inst.Metall.,1997, 50 (6), 573-580.
19. T. Furu, R. Orsund and E. Nes, Acta Metall. Mater. 1995, 43 (6), 2209-2232.
20. J. Gil Sevillano, P. van Houtte and E. A. D. Aernoudt, Progress in Materials
Science, 1980, 25, 69-412.
21. H. E. Vatne, T. Furu and E. Nes, Proceedings of the International Conference on
Recrystallization and Related Topics, Monterey, California, 21-24 Oct., 1996.
22. N. Hansen and D. A. Hughes, Phys. Stat. Solidii (b), 1995, 149, 155-172.
[TMP]
2
', B.Hutchinson et.al (Eds), Swedish Society for Metals Technology, 1996,
35-51.
24. H. E. Vante, T. Furu, R. Orsund and E. Nes, Acta Mater. 1996, 44, 4463-4473.
25. F. J. Humphreys, Acta Mater., 1997, 45, 4231-4240.
26. R. Orsund and E. Nes, Scripta Metall. 1989, 23, 1187-1192.
27. Y. Huang and F .J. Humphreys, Acta Mater., 1999, 47, 2259-2268.
28. Y. Huang and F .J. Humphreys, Acta Mater., 2000, 48, 2017-2030.
29. G. Gottstein and L. S. Shvindlerman, Scripta Metall., 1992, 27, 1515-1520.
30. A. D. Rollett, E. A. Holm, Proc. of the int. Conf. on Recrystallization and Related
Topics, Monterey, California, 21-24 Oct., 1996.
31. F. N. Rhines and B. R. Patterson, Metall. Trans. A, 1982, 13, 985-993.
32. A. Thorvaldsen, Acta Mater., 1997, 45, 587-597.
33. R. Sandstrom, Acta Metall., 1977, 25, 897-904.
34. E. Nes, Acta Metall., 1995, 43, 2189-2207.
Chapter 3 44
S.P. Chen
Chapter 4
A refined single grain approach applied to
the modeling of recrystallization kinetics for
cold rolled single-phase metals
A comprehensive model for the recrystallization kinetics is proposed which incorporates
both microstructure and the textural components in the deformed state. The model is
based on the single grain approach proposed previously. The influence of the as-
deformed grain orientation, which affects the stored energy via subgrain size and sub-
boundary misorientation, is taken into account. The effects of the deformed grain
geometry, the nucleation site density and the initial grain size prior to deformation on
the recrystallization kinetics is assessed. The model is applied to the recrystallization
kinetics of cold-rolled AA1050 alloy.
4.1. Introduction
Recrystallization kinetics in simple metallic systems is frequently described by JMAK
type models [1]. The basis of the JMAK analysis, which yields the volume fraction
transformed, X, against time t, is:
1 exp( )
n
X kt = (1)
The constant k involves the nucleation rate, or the nucleation site density, and the
growth rate. The exponent n relates to the time dependencies of nucleation rate, growth
and the dimensionality of the growth fronts. The crucial assumption made in deriving
Chapter 4 46
the various forms of JMAK is that the nucleation sites are randomly distributed in space
leading to n values to be 3 or 4. However, in almost all experimental studies of
recrystallization kinetics in aluminium alloys the exponent n is less than 2. Such a
deviation from ideal JMAK behavior is attributed to be the occurrence of concurrent
recovery [2], or a non-uniform distribution of stored energy [3,4].
The occurrence of the simultaneous recovery during recrystallization will reduce the
dislocation density and hence the stored energy. Consequently, the concurrent recovery
should retard recrystallization kinetics and cause a negative deviation from linear
JMAK behavior. However, both experiments [5,6] and theoretical predictions [3,7] have
shown that concurrent recovery can not fully explain the observed lower value for n.
Hence a more detailed analysis of the uniformity of stored energy in deformed metals
and its effect on the recrystallization behavior seems appropriate.
It is well known that the deformation in a polycrystal material is inhomogeneous [8,9].
The inhomogenity is of a dual nature. Firstly, the area along the original grain
boundaries undergoes a more severe plastic deformation than the average macroscopic
level. As a result, a deformation gradient will form in the grain; the original grain
boundaries are of highest dislocation density, and therefore, the main source of
recrystallization nuclei, especially for single-phase alloys. Like the effect of concurrent
recovery, this kind of the inhomogenity also leads to a time dependent growth rate of
recrystallizing grains and therefore reduces the slope of JMAK plot. However, it was
found that continuous reductions in growth rate by a factor of 10 were required to cause
the JMAK exponent to drop significantly below a value of 3. The variation in growth
rate by a factor of 2 in real experiments caused no detectable change in the simulated
exponent [10,11]. Furthermore, this explanation can not be applied to cases where the
nucleation is on a coarser scale than the distribution of the stored energy.
The second inhomogenity refers to the fact that the deformation microstructures varies
from grain to grain because of the initial crystallographic texture. As a result, the
nucleation sites for recrystallization in deformed grains with different orientations are
different [12,13]. Each as-deformed grain will recrystallize at a rate that depends on its
size as well as initial orientation with respect to the deformation field depending on the
accumulation of the stored energy.
Hence, the non-uniformity of the recrystallization behavior is most likely due to the
variation of the stored energy as a function of the grain orientation. Dillamore et al [14]
has pointed out that the variation of stored energy as a function of the texture is related
to the variation of the Taylor factor. The order of the stored energy of the deformation
textures in low carbon steel is [15],
{110} 110 {111} {211} 011 {100} 011 uvw
E E E E
< > < > < > < >
> > > . The
A single grain approach applied to the modeling of recrystallization kinetics 47
dislocation structures and density in aluminium alloys were observed to be dependent
on the crystallographic orientation too although there was a wide scatter in the
experimental data [16,17].
In addition, it has been realized that the recrystallization kinetics and the resulting grain
structures are determined by the ratio of the nucleation to growth rate, the density and
the spatial distribution of nucleation sites as well as the impingement space. All these
factors are strongly dependent on the microstructure, the substructures in individual
grains and the local orientations between grains, developed during plastic deformation.
Therefore, a more realistic recrystallization model should take into account all these
details of deformed microstructure.
The advanced techniques such as EBSD and hard x-rays from a synchrotron source
make it possible to examine the recrystallization behaviors at the level of an individual
grain [8,18,19]. However, theoretical models at this level are still lacking. In [7] we
have proposed a single grain approach to predict the recrystallization kinetics of a
polycrystal aggregate. The structure of the deformed metal is taken into account in the
model, which contains the information about the main features of rolled material, i. e.
on the degree of cold rolling, the grain size, and the grain shape. An as-deformed
tetrakaidecahedron is applied to describe the grain geometry. The grain shape change
with strain is introduced through a parameter of aspect ratio according to the
macroscopic change in the shape of the specimen (principle of the Taylor model).
Additionally, the subgrain size and the misorientation between subgrains belonging to
the same grain are related to the degree of deformation.
In the present Chapter, the physical model of the recrystallization kinetics, based on the
single grain approach, is further developed to incorporate both microstructural
inhomogeneities and the textural components in the as-deformed state. The influence of
the as-deformed grain orientation, which affects the stored energy via subgrain size and
sub-boundary misorientation, is taken into account. The effect of the initial grain size,
the grain geometry and the nucleation site density on the recrystallization kinetics is
assessed. The model is applied to the recrystallization kinetics in cold rolled AA1050.
4.2. Experiments
The AA1050 studied has a chemical composition of 0.185 wt.% Fe, 0.109 % Si, balance
Al. Cold-rolled plate with thickness of 5 mm was annealed at 600C for 2 h and
quenched into water. The material was then subjected to a precipitation treatment at
400C for 2 h to reduce and stabilize the content of iron in solid solution. The average
grain size after this treatment was 985 m. Metallographic examination revealed
Chapter 4 48
coarse (2-4 m) iron-rich constituent produced during casting and fine (0.1-0.3 m)
precipitates resulting from annealing. Following a cold rolling treatment leading to a
reduction of 50% in thickness ( 0.69 c = ), the samples were annealed in a salt bath at
340C for a range of times and then quenched into water to get different degrees of
recrystallization. All the metallographical and OIM observations were confined to the
center layer of the sample. The OIM facility performed a scan over an area of
1200600 m
2
, containing approximately 120 original grains. The degree of
recrystallization was determined quantitatively and accurately using the method
described in [20].
Five typical orientations of grains in the deformed state, namely, Cube {100}<001>,
Goss {110}<100>, Brass {110}<112>, S {124}<211> and Copper {112}<110> are
observed in OIM analysis. The volume fractions of the main texture components in as-
deformed state were calculated by allowing a 15 spread around the ideal orientations,
as listed in Table 1. The average fully recrystallized grain size after deformation to
strain of 0.69 and annealed at 340C was 695 m.
Figure 1 shows a partially recrystallized structure in a sample annealed at 340C for 50
min, showing the characteristic of the inhomogeneous nucleation. Recrystallization was
found to occur faster in the grains with Copper and/or S orientations. Grains with a
Cube orientation are slowest in the recrystallization.
The overall recrystallization kinetics as measured by optical microscopy is shown in
Figure 2 by the square doted line. Also in this Figure we illustrate schematically the
approximate recrystallization kinetics examined in grains with different orientations by
OIM.
Table 1. The Talyor factor of the main texture components and related substructural
parameter values after cold rolling 50%
Texture
component
{hkl}<uvw> M value o
(m)
u Frequency
%
RG {011}<110> 4.90 0.46 4.68 0
Copper {112}<111> 3.67 0.61 3.53 2
S {123}<364> 3.52 0.64 3.34 7
Brass {011}<211> 3.27 0.68 3.15 9
Goss {011}<100> 2.75 0.82 2.63 20
Cube {001}<100> 2.45 0.94 2.29 13
Random 3.06
average
0.73 2.92 49
Fig. 1. OIM of an AA1050 sample annealed at 340C for 50 min, showing the
inhomogeneous nature of the recrystallization process.
Fig. 2. The overall softening kinetics measured by optical microscopy (square marks)
and approximate recrystallization kinetics of 3 texture components at 340C after a
strain of 0.69.
0
0.2
0.4
0.6
0.8
1
0 2000 4000 6000 8000
Time (s)
F
r
a
c
t
i
o
n
Brass
Copper
S
Cube
Random
Chapter 4 50
4.3. Modeling the recrystallization kinetics
4.3.1. The orientation dependent microstructure in the deformed state
The dislocation density within a given grain after a plastic deformation is primarily
related to the amount of crystallographic slip I, which is defined as [21]:
Mc I = (2)
where M is the orientation factor of the as-deformed grain (Taylor factor), c is the
macroscopic shape change (true strain) of the grain, which is assumed to be the same as
that of the specimen.
As an approximation, we assume that aluminium follows a linear strain hardening
behavior, i.e.:
0 c
t t | = I (3)
where
c
t is the resolved shear stress,
0
t , the frictional part of
c
t , |, a constant.
The relationship between the internal stress and dislocation density can be written as:
'
0 c
t t | = (4)
where |' is another constant. Combining equations (3) and (4) one obtains:
2
I (5)

In most cases, and certainly for this aluminium grade, the dislocations are arranged in a
cellular substructure or into subgrain. If the substructure in a specific deformed grain
can be described by a subgrain size o and a sub-boundary misorientation u then [1]:
u
o
(6)
The relationship between the deformation features ( , o u ) of an individual grain and the
average values ( , o u ) of the polycrystalline aggregate can be expressed as:
2
av
M
M
u u
o o
| |
=
|
\ .
(7)
where
av
M is the average value of the Taylor factor. In addition, an empirical
relationship between the average cell/subgrain size, o , and deformation strain [22,23]
is employed in the present work:
0.35 0.17 / o c = + (8)
and
/ b C ou = (9)
where o is in m, b is the Burgers vector and C is a constant between 60~80 rad.
Eq. (7) indicates that the stored energy is proportional to the square of the Taylor factor.
4.3.2. Nucleation and Growth
Based on the experimental data, nucleation is assumed to start from the original grain
boundaries separating grains of different orientations. In a deformed grain with a
specific orientation, the microstructure will have a variety of cell sizes and a distribution
of misorientations. We assume that the microstructure can be described using two
components [24]: an assembly of equiaxed subgrains and specific subgrains, effectively
sub-critically sized recrystallization nuclei. The former can be characterized by a mean
equivalent radius, R , a mean misorientation, u , and with boundaries of mean energy
and mobility, and
sb
M , respectively. The latter have a larger size (
n
R ) and different
boundary characteristics ( , ,
n n n
M u ). In the present study we assume that the
misorientation angle between the particular subgrain and surroundings is 15.
The effective nucleation site density for grain boundary nucleation is expressed by
[25,26]:
2
/
V d v
N C S o = (10)
where
d
C is a calibration constant which determines the potential of the grain boundary
as a nucleation site. S
v
is the grain boundary area per unit volume, which is related to the
original grain size d, rolling strain c , and the grain shape tessellation in the space. For a
tetrakaidecahedron, S
v
varies with rolling strain in a manner described by [7]:
( ) a a a a a a a
d
S
v
/ 2 2 / 2 3 / 2 1 3
2
1
2 2 2 2
+ + + + + + = (11)
Chapter 4 52
where
c
e a = is the distortion along the rolling direction.
The driving pressure acting on the recrystallized front is:
2
n
n
P
R R
o
= + (12)
where o is a geometrical constant and has a value of ~1.5 [1].
The growth rate of a recrystallized grain is expressed as:
2
n n
n n
n
dR
M P M
dt R R
o | |
= =
|
\ .
(13)
where
n
M is the mobility of a high angle boundary at the annealing temperature and
which is given by:
( )
0
exp /
n g
M M Q R T = (14)
where
0
M is a pre-exponential factor, Q, activation energy for boundary migration,
g
R ,
gas constant and T, the annealing temperature.
The boundary energy is assumed to depend on the average misorientation angle only
and is given by [1]:
1 ln
m
m m
u u

u u
( | |
=
( |
\ .
(15)
where
m
and
m
u are the values of boundary energy and misorientation for high-angle
boundaries and which are commonly taken as 0.324 J/m [1] and 15 respectively.
Combining Eqs (9), (13) and (15), the relationship between the growth rate of a
recrystallizing grain and the subgrain size of the assembly is approximately given by:
2 2
1
n m n
m
dR bC M K
V
dt
o
u o o
= = (16)
with K being a temperature dependent constant.
4.3.3. Kinetics approach
The single grain model described in Chapter 3 and [7] is expanded along the line
described above and is used to calculate the recrystallization kinetics in individual
grains with a specific orientation. Nucleation event in an as-deformed grain is assumed
to be site-saturated. A growing recrystallized grain is modeled as an expanding semi-
sphere nucleated on grain boundaries. New grains grow independently of one another
until hard impingement occurs.
As the position of the nuclei has a significant effect on the recrystallization kinetics [7],
the position of nuclei on the faces of the tetrakaidecahedron are determined randomly
assuming that there is an equal probability that each coordinate on the 14 faces
(including grain edges and corners) may act as a nucleation site. Each simulation is
therefore repeated a number of times (40) in the model calculation and the average of
the simulated kinetics is taken as the representative, i.e. the time to obtain a given
fraction of recrystallization, f, is given by the average over the number of the
simulation, j.
, ,
/
av i i j
t t j =
_
(17)
where
, i j
t is the time at a given fraction in the j th simulation for a grain with a specific
orientation labeled i.
Generally, a deformed polycrystal contains many grains of different orientations. Each
grain has a different set of surrounding grains. In this attempt at modeling
recrystallization kinetics in a polycrystal material we assume that there is no correlation
between the orientations of neighboring grains and all the grains of one crystallographic
orientation (within a certain small range) are represented by a single grain of the mean
size. (The simulation results in the next section will show that this assumption is
reasonable).
The overall kinetics of the polycrystalline aggregate can then be obtained by a weighted
summation of the kinetics on the main textural components in the as-deformed state.
The fractional recrystallization f of the assembly is given by:
( )
i i i
f f = _ _ (18)
where
i
is the fraction of a main textural component.
Chapter 4 54
4.4. Results
4.4.1. Effect of the grain geometry and the nucleation site density on
the recrystallization kinetics
Before we present the simulation results, two new concepts need to be introduced, i.e.,
the effective nucleation site density and the numerical nucleation site density. As long
as C
d
can be experimentally determined the nucleation site density, N
v
, can be calculated
using Eq. (10), which is called the effective nucleation site density. In reality, a
recrystallized grain at the original grain boundary may invade into two (or three)
adjacent grains, at which the growth of the recrystallizing grain in one side is faster than
in the other. However, in the single grain kinetics model the new grain is assumed to
start from the original grain boundaries and only the portion of the new grain inside the
objective grain is taken into account when fractional recrystallization is calculated. As a
consequence, only one half of a physical nucleus is considered. Therefore the numerical
nucleation site density, taken to be twice the effective nucleation site density, is
employed in calculating the nucleation sites in a single grain in the present model. In the
model the number of the nuclei in a single grain calculated might be a non-integer. For
such a case, the representative of the simulated recrystallization kinetics is treated as a
weighted summation on the simulation kinetics obtained from simulations for the two
nearest integer nuclei per grain values.
In the single grain kinetics model, the recrystallization fraction is proportional to
3
( / ) N Vt d , in which N is the number of the nuclei in an as-deformed grain of diameter
d, t is real time,
2
/ V K o = is the growth rate. If the interface mobility is assumed to
remain constant throughout the recrystallization process and to be independent of the
applied deformation strain, and for a certain number of N, we can present the fraction of
the recrystallized material as a function of the dimensionless reduced time
*
t , which is
defined as:
*
2
K
t t
d o
= (19)
The real time can be calculated provided the K is known at a given annealing
temperature.
Figure 3 shows the simulated recrystallization kinetics after a strain of 0.69 assuming
that six nuclei are located randomly on the surface of a tetrakaidecahedron. Each curve
represents one simulation. As can be seen, the simulated recrystallization kinetics
differs from each other because of the difference in the nucleus positions. The
difference in the kinetics reflects the characteristics of the nucleation nature. The six
nuclei can either be present as clusters or as being far apart. When they form as a
cluster, a slow recrystallization rate is obtained due to early impingement. Contrarily,
when nuclei form far apart from each other, a faster rate will result. In order to obtain
the representative of the recrystallization kinetics in a single deformed grain simulations
were run a number of times with the same number of nuclei but different nucleation
positions. 40 simulations were found to be an adequate number to yield a representative
with a stable kinetics curve. In Figure 3 this average behavior is indicated by a square
dotted curve. The representative curve shows a typical sigmoidal shape. The JMAK
equation is only applied to the representative of simulation kinetics to obtain least
square best fit of the JMAK exponent n, to determine the effects of underlying
assumption and parameter values on n.
Fig. 3. The determination of the representative of the simulated recrystallization
kinetics for six nuclei located randomly on the surfaces of a tetrakaidecahedron at a
strain of 0.69.
Figure 4 shows the simulated recrystallization kinetics for a tetrakaidecahedron of size
100 d m = at different strains, 0.41, 0.69, 1.10, 1.37, 1.61 and 2.3 with the number of
the nuclei per grain fixed at 6 and 1 m o = . Because in this subset of the simulations
we assume that the subgrain size, o , does not change with the strain, then the kinetics
after various degree of rolling strain will only reflect the effect of the grain geometry
resulting from deformation on the recrystallization kinetics. As deformation increases,
the aspect ratio of the grain increases, the resulting n decreases.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2
t*=Kt/d
F
r
a
c
t
i
o
n
n = 2.68
-5
-4
-3
-2
-1
0
1
2
-2.5 -2 -1.5 -1 -0.5 0
Lnt*
L
n
(
L
n
(
1
/
(
1
-
x
)
)
)
Chapter 4 56
Fig. 4. The simulated recrystallization kinetics for the case of six nuclei located
randomly on the surfaces of a tetrakaidecahedron at various strains. From leftmost to
rightmost, c =0.2, 0.41, 0.69, 1.10, 1.37, 1.61, 2.30.
Fig. 5. Effect of the grain geometry resulting from deformation and the number of the
nuclei on the JMAK exponent.
1.6
1.8
2
2.2
2.4
2.6
2.8
3
3.2
0 0.5 1 1.5 2 2.5
Strain
J
M
A
K

n
Real
N=2
N=6
N=12
N=24
n = 2.82
n = 1.78
-5
-4
-3
-2
-1
0
1
2
-3 -2 -1 0 1
Ln t* (=Kt/d)
L
n
(
l
n
(
1
/
(
1
-
x
)
)
)
The preceding simulations are repeated to study the effect of the number of the nuclei in
an objective grain on the recrystallization kinetics. The resulting variations of n against
strain for various numbers of the nuclei are shown in Figure 5. It can be seen that the
JMAK exponent depends on the number of the nucleation sites. In a deformed
tetrakaidecahedron an increase in the number of the nuclei gives a rise in n. However, in
the case of the undeformed one, an opposite trend is found.
4.4.2. Effect of the strain on the recrystallization kinetics
In reality, as strain increases the sub-grain size decreases, S
v
increases. Therefore, the
nucleation site density and the growth rate increases with strain. These factors favor the
recrystallization kinetics. Figure 6 shows the simulated recrystallization kinetics in a
tetrakaidecahedron with an initial size of 100 m and average substructures after
various rolling strains. In these simulations, the subgrain size and the nucleation site
density at each strain are calculated according to Eq. (8) and Eq. (10), respectively. The
d
C is taken as
4
1.5 10
(which is calculated from the current experiment and this value

will be used in the following simulations). The fractional recrystallization is plotted
against
2
1 100
( / ) Kt d o with
1
o being 1 m. As can be seen, as the strain increases, the
kinetics curves shift to shorter times while the JMAK exponent n decreases (which is
shown in Figure 5 by thick line with legend real). The final variation of n with strain
reflects the combined effect of the grain geometry and the nucleation sites. On the one
hand, the increase in the number of the nuclei will increase n. On the other hand, the
grain geometry change resulting from increasing strain leads to a lower value of n.
Fig. 6. The simulated recrystallization kinetics in a single grain after various strains.
From leftmost to rightmost, =2.3, 1.61, 0.69, 0.41.
n= 2.76
n = 2.65
n= 2.37
n= 2.22
-5
-4
-3
-2
-1
0
1
2
-6 -5 -4 -3 -2 -1 0
ln t*(=Kt/
1
d
100
)
L
n
(
l
n
(
1
/
(
1
-
x
)
)
)
Chapter 4 58
Fig. 7. The effect of the initial grain size on the recrystallization kinetics. (a), =0.2, (b),
=0.69, (c), =1.61, (d) =2.30.
4.4.3. Effect of the initial grain size prior to deformation on the
recrystallization kinetics
Apparently, the initial grain size affects the nucleation site density as it affects
v
S ,
which decreases as initial grain size increases. In addition, it can lead to a larger
variation in stored energy resulting from variation in dislocation density after
deformation. The second factor is most complicated as it also depends on texture and
orientation with respect to the deformation field and will be considered later. In order to
determine to which extent the initial grain size affects the recrystallization kinetics
through
v
S , we assume that the subgrain size does not change with strain and fixes at
1m, then the nucleation site density at a given strain is only a function of the initial
grain size. Figure 7 shows the simulated recrystallization kinetics, which is presented
against
2
1 100
( / ) Kt d o , in the grains with 3 different initial diameters, i.e. 58 m, 100 m
and 200 m for 4 strain levels. As can be seen, the effect of the initial grain size on the
recrystallization kinetics depends on the deformation strain applied. In an undeformed
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
d=58m
d=100
d=200
a
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
d=58m
d=100
d=200
b
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
d=58m
d=100
d=200
c
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2 2.5 3
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
d=58m
d=100
d=200
d
tetrakaidecahedron, smaller grain does recrystallize faster than coarser one. However,
this effect decreases as the deformation strain increases. For example, at larger strains
the recrystallization kinetics in a coarser grain is faster than that in the finer grains.
These results indicate that the increase in
v
S resulting from deformation strain plays a
dominant role in the increase of the nucleation density.
4.4.4. Effect of the orientation on the recrystallization kinetics
Table 1 lists the Taylor factors of the main deformation textural components found in
metals with f.c.c. crystal structure [22,27]. The Taylor factor of the Goss component is
taken as an average of the values corresponding to the inclination of the shear band with
respect to the rolling direction being 30 deg. and 35 deg. respectively. The related
substructure parameters in grains with different orientations at a strain of 0.69 are
calculated using Eqs. (7)-(9) and listed in Table 1. We assume that
d
C has the same
value for all grain boundaries and equals
4
1.5 10
when calculating the nucleation site

density in grains with different textural components. Figure 8 shows the simulated
recrystallization kinetics in grains with different orientations but a fixed initial grain size
of 100 m at a strain of 0.69. As can be seen, a grain with a larger Taylor factor
recrystallizes earlier and at a higher rate.
Fig. 8. The simulated recrystallization kinetics in grains with different orientations and
overall recrystallization kinetics in AA1050. Thin solid line: from leftmost to rightmost,
RG, Copper, S, Brass, Random, Goss, Cube. Square dotted line: the overall
recrystallization kinetics.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
Overall
Chapter 4 60
4.5. Application of the model to the experiment data
As indicated above, the overall recrystallization kinetics should include the effect of the
grain geometry, the nucleation site density, the starting grain size and the strain
variation from grain to grain as a consequence of the grain orientation. The grain size
distribution in annealed materials is approximately log-normal, where the maximum
grain size is typically 2.5~3 times mean size. To determine the overall recrystallization
kinetics of the grains with different sizes, we can apply the method specified in [7].
However, from the proceeding simulation results we have realized that the grain
orientation has a much stronger effect on the recrystallization kinetics than the initial
grain size. Hence, for the sake of simplicity, we assume in the overall recrystallization
kinetics that all the grains with a particular crystallographic orientation can be
represented by a single grain of mean size.
Now the single grain model can be applied to predict the recrystallization kinetics if the
input parameters, deformation strain, , the initial grain size, d, the calibration constant,
C
d
,

the mobility at a given annealing temperature, M
n
, and the volume fraction of the
main texture components in the as-deformed state are determined.
In the current experiment on AA1050, the initial grain size prior to deformation as well
as the fully recrystallized grain size after deformation to strain of 0.69, annealed at
340C are taken to be 100 m and 69 m, respectively. From these values the number
of the effective nuclei in a single grain of mean size is estimated to be 3, leading to a
calibration constant C
d
is of
4
1.5 10
. The corresponding nucleation site density and

the subgrain size in a grain with a given orientation are determined according to the
section 4.4. Since M
n
could not be determined from the present experiment we leave it
as a free parameter in the model. The simulated recrystallization kinetics in grains with
different orientations is shown in Figure 8.
The overall recrystallization kinetics can be obtained now using Eq. (18) with
considering the frequency of the texture components in the deformed state listed in
Table 1. The resulting curve is shown in Figure 8 by the square dotted line. The
Avrami-plots of the experimental data and the simulation is shown in Figure 9. The
JMAK exponents measured and simulated are 1.99 and 1.94, respectively. Obviously,
the present model could yield a much more practical prediction of the JMAK n, by
taking into account the deformation inhomogenity from grain to grain due to variation
of the Taylor factor.
Fig. 9. The comparison of the experiment with the prediction from the single grain
model by considering orientation dependent kinetics.
4.6. Discussions
The simulation results from the present model shows that the JMAK exponent of a
apparent recrystallization kinetics curve depends on the grain geometry, nucleation site
density, the initial grain size prior to deformation and the main textural components
after deformation. This result, which is one of the important features that the present
model reveals, implies that the recrystallization space (the space that the recrystallizing
grains grow into) is of importance in deciding the recrystallization kinetics. In real
materials, recrystallization space is effectively defined by both the initial grain size and
the deformation geometry. The change of the grain geometry due to deformation leads
to a change in the impingement space of the recrystallized grains. An increase in the
number of the nuclei will also cause a change in the impingement space.
Another salient feature that the present refined model shows is that the effect of the
texture components in the as deformed state on the recrystallization kinetics is
significant. This point will be elaborated further by simulations. Figure 10 shows the
simulated orientation dependent recrystallization kinetics at two strain levels, 0.69 and
1.61, respectively. We use the two starting structures with different fraction of texture
components as listed in Table 2. The overall kinetics curves of the two examples are
also indicated in Figure 10. In the same figure in the smaller box the Avrami plot for
these two examples are plotted. It can be seen that the shape and the slope of the JMAK
plot are related to the mixture of the texture components. When there are Goss and Cube
n = 1.94
n = 1.99
-4
-3
-2
-1
0
1
2
3
-5 -3 -1 1 3 5 7 9 11
Ln t* for simulation, Ln t for experiment
L
n
(
l
n
(
1
/
(
1
-
x
)
)
)
Simulation
Experiment
Chapter 4 62
texture components present in the as-deformed state, the JMAK plot shows a non-linear
behavior, i.e. two-stage n, which has been frequently reported in the literature.
Fig. 10. The simulated recrystallization kinetics in grains with different orientations and
overall recrystallization kinetics for the two examples. Thin solid line: from leftmost to
rightmost, RG, Copper, S, Brass, Random, Goss, Cube. (a), =0.69, (b), =1.61.
Table 2. The frequency of textural components in the as-deformed state for two
assumed structures
Texture
components
Copper S Brass Goss Cube Random
Example 1 16 20 23 0 0 41
Example 2 3 8 12 22 8 47
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
Example 1
Example 2
n = 1.88
n= 2.33
-4
-3
-2
-1
0
1
2
3
-3 -2 -1 0
Ln t*
L
n
(
l
n
(
1
/
(
1
-
x
)
)
)
0
0.2
0.4
0.6
0.8
1
x
a
0
0.2
0.4
0.6
0.8
1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
t*=Kt/
1
d
100
F
r
a
c
t
i
o
n
Example 1
Example 2
n = 2.10
n= 1.53
-4
-3
-2
-1
0
1
2
3
-5 -4 -3 -2 -1
Ln t*
L
n
(
l
n
(
1
/
(
1
-
x
)
)
)
0
0.2
0.4
0.6
0.8
1
x
b
Many factors that control recrystallization are not specifically treated in JMAK model
but are casually lumped into the constants in Eq. (1). Two important factors that are
included in this category are the initial grain size prior to deformation and the amount of
deformation. These points can be clarified with the help of the present model.
Theoretical studies in the literature suggested that the finer grained material
recrystallizes after a shorter incubation time and within a shorter relative time period
than the coarser grained material although the experiments yielded controversial results
[28,29]. Generally, since recrystallization nucleates primarily along prior grain
boundaries, the rate of nucleation is directly proportional to S
v
, which increases as the
grain size decreases. Therefore, after similar amounts of deformation and for similar
growth rates the rate of recrystallization should increase as the prior grain size
decreases. However, if it is assumed that the substructure in a fine grain and a coarse
one is equal (this is true for aluminum alloys at strain larger than 0.5 [30]) and the
fraction of the texture components is the same in both cases, present model shows that it
depends on the strain level whether small grains recrystallize faster or slower than large
grains. At lower strains, small grains recrystallize faster. However, at larger strains, the
recrystallization rate in coarser grain becomes higher. This is attributed to the change in
the impingement space due to the deformation geometry and the relatively large number
of nucleation sites in the coarser grains as strain increases. This prediction is in
excellent agreement with experimental study in Sellarss group [31]. It should be
pointed out explicitly that the mechanism of intra-granular nucleation in larger strains
need not to be involved to explain this result, as in the current simulations nucleation
always occurs at the grain boundaries.
Fig. 11. The volume fractions of the texture components in the cold-rolled AA1050 as a
function of the degree of deformation. (a) fine initial grain size (50 m), (b), coarse
initial grain size (350 m) [32].
0
20
40
60
80
0 20 40 60 80 100
Degree of cold rolling (%)
T
e
x
t
u
r
e

c
o
m
p
o
n
e
n
t
s

(
%
)
{110}<112>+{123}<634>+{112}<111>
{100}<001>
{110}<001>
a
0
20
40
60
80
0 20 40 60 80 100
Degree of cold rolling (%)
T
e
x
t
u
r
e

c
o
m
p
o
n
e
n
t
s

(
%
)
{110}<112>+{123}<634>+{112}<111>
{100}<001>
{110}<001>
b
Chapter 4 64
Finally, the initial grain size and the amount of strain are both known to have a large
effect on the texture components in the as deformed state. As can be seen from
Figure 11, which shows the textural component evolution with cold rolling strain in a
commercial aluminum alloy [32], the rolling texture components intensify while the
fraction of the soft components, Goss and Cube, decreases with increasing strain. The
variation of the textural concentration with strain in the fine grain material is different
from that in the coarse grain one. Therefore the initial grain size and the degree of the
deformation can affect the recrystallization kinetics through their effect on the texture
components. The model correctly predicts this changeovers in relative behavior.
4.7. Conclusions
The following conclusions are obtained from the refined recrystallization model which
incorporates both the deformed microstructure and the textural components.
(1). The kinetics model based on the single approach provides a new insight to gain a
better understanding of annealing phenomena in deformed polycrystals, particularly the
effect of non-random nucleation sites and the geometry by which recrystallized grains
fill space on recrystallization kinetics.
(2). The simulation shows that the JMAK exponent depends on the grain geometry,
nucleation site density, the initial grain size prior to deformation and the main textural
components after deformation. The effect of grain size on recrystallization kinetics
depends on the amount of the prior strain applied.
(3). After introducing the deformation inhomogenity from grain to grain due to
difference in the Taylor factor, the single grain approach could predict correctly the
JMAK exponent.
References
2. R.A. Vandermeer and P. Gordon, in 'Recovery and Recrystallization of Metals',
New York, Interscience, 1962.
5. E.C.W. Perryman, Trans. AIME, J. Metals, 203, 1955, 1053-1061.
6. R.A. Vandermeer and B.B. Rath, in 'Simulation and theory of evolving
microstructures', M.P. Anderson and A.D. Rollett (Eds), TMS, 1990, 119-126.
7. S.P. Chen, I. Todd and S. van der Zwaag, Metall. Mater. Trans. A, 33A, 2002, 529-
537.
8. L. Delannay, O.V. Mishin, D.J. Jensen and P. van Houtte, Acta Mater., 49, 2001,
2441-2451.
9. V. Randle, H. Hansen and D.J. Jensen, Phil. Mag. A, 73, 1996, 265-282.
11. A.D. Rollett, D.J. Srolovitz, R.D. Doherty and M.P. Anderson, Acta Metall., 37,
1989, 627-639.
12. E. Nes and J.K. Solberg, Mater. Sci. Techn., 2, 1986, 19-23.
14. I.L. Dillamore and H. Katoh, Met. Sci., 8, 1974, 73-83.
15. W.B. Hutchinson, Met. Sci., 8, 1974, 185-196.
16. D.P. Field and H. Weiland, Mater. Sci. For., 157-162, 1994, 1181-1188.
17. X.Y. Wen and W.B. Lee, in 'Sheet Metal Forming Technology', M.Y. Demeri (Eds),
The minerals, Metals & Materials Society, 1999, 233-243.
18. E.M. Lauridsen, D.J. Jensen, H.F. Poulsen and U. Lienert, Scr. Mat., 43, 2000, 561-
566.
19. R.D. Doherty, D.A. Hughes, F.J. Humphreys, J.J. Jonas, D.J. Jensen, M.E. Kassner,
W.E. King, H.J. McQueen and A.D. Rollett, Mater. Sci. Eng. A, 238, 1997, 219-
274.
20. S.P. Chen, D.N. Hanlon, S. van der Zwaag, Y.T. Pei and J.Th.M. de Hosson, J. Mat.
Sci., 37, 2002, 989-995.
21. J. Gil Sevillano, P. van Houtte and E. Aernoudt, Scr. Met., 10, 1976, 775-778.
22. J. Gil Sevillano, P. van Houtte and E.A.D. Aernoudt, Pro. Mat. Sci., 25, 1980, 69-
412.
23. N. Hansen and D.A. Hughes, Phys. Sta. Sol. (b), 149, 1995, 155-172.
24. F.J. Humphreys, Acta Mater., 45, 1997, 4231-4240.
Chapter 4 66
[TMP]
2
35-51.
27. H.J. Bunge, in 'Texture Analysis in Materials Science', London, Butterworths, 1982.
28. J.C. Blade and P.L. Morris, in 'Proc. 4th Int. Conf. on Textures', Cambridge, 1975,
171-178.
29. L. Ryde, W.B. Hutchinson and S. Jonsson, in 'Recrystallisation '90', T. Chandra
(Eds), Warrendale, PA, TMS, 1990, 313-318.
30. P. Cotterill and P.R. Mould, in 'Recrystallization and Grain Growth in Metals',
Surrey Univ. Press. London, 1976.
31. P.L. Orsetti Rossi and C.M. Sellars, in 'Aluminium Alloys, Their Physical and
Mechanical Properties, Proc.of 6th Int. Conf. on Aluminium Alloys, ICAA-6', T.
Sato, S. Kumai, T. Kobayashi, and Y. Murakami (Eds), The Japan Institute of Light
Metals, 2,1998, 1227-1232.
32. N. Hansen and D.J. Jensen, Metall. Trans. A, 17, 1986, 253-259.
S.P. Chen
Chapter 5
A FE comparison study of hot rolling
operation and PSC testing
A comprehensive study on the comparison of hot rolling and Plane Strain Compression
testing (PSC) is reported in the present chapter. The evolution of the key physical
variables during hot rolling and PSC testing are quantified and compared, using Finite
Element Method (FEM) calculation. Calculated results show that the process variables,
i.e. local stress, strain, strain rate and temperature, vary from entry to exit and
throughout the thickness of work piece during hot rolling, while the evolution patterns
of these parameters during PSC testing are quite uniform under constant nominal strain
rate. Therefore, for a given amount of nominal deformation (reduction), the loading
history, the actual distributions of these process variables in time, is quite different. It is
indicated that a further examination of the deformation history should be made when
applying the relationships established from PSC testing to real industrial hot rolling
operation.
5.1. Introduction
Current studies in the literature assumed that PSC adequately simulates hot rolling in
terms of microstructure development and hence has the potential to allow laboratory
scale experiments to be performed rather than carrying out expensive, full scale plant
trials [1,2]. The principal reason is that both deformation processes approach plain strain
compressive deformation. However, application of the relationship of the basic physical
metallurgy of hot working obtained from PSC testing to industrial rolling operation
Chapter 5 68
requires accurate knowledge of the time variation of the temperature and the strain rate.
Although many investigators have developed mathematical models to predict the
temperature distribution and the strain distribution of the slab during hot rolling [2-4],
very few publications paid attention to the deformation history and the variation of the
strain rate pattern during hot rolling, which have been recognized as two important
parameters for hot deformation. In practical hot rolling conditions, where the nominal
strain rate ranges from 0.1 to 100 /s, the yield stress characteristics of the material are
strain rate rather than strain dependent. In addition, significant differences in
microstructure occur depending on the deformation history, which are reflected in the
subsequent recrystallization kinetics and the final grain size [5].

The variation in local strain and strain path in samples deformed in PSC testing has
been studied in detail using analytical slip-line-field theory [6], grid analysis [7], FEM
[8-10] with the aim of determining the optimum sample geometry and friction
conditions for as uniform deformation conditions in as large a part of the sample as
possible. Although the shape similarity is not maintained in PSC testing as it is in
rolling it was found that an acceptable uniform deformation pattern was obtained for
cases in which the tool width to specimen thickness ( / w h ) ratio is larger than 1.5.

In this chapter, the deformation behavior and heat transfer phenomena in AA1050
alloys during hot rolling and in PSC testing at elevated temperature were studied using
FEM simulation to obtain the simultaneous variations of the field quantities. The
detailed strain, stress, strain rate and the temperature history of elements throughout the
rolling strip and inside the specimen in the PSC testing are quantified and analyzed.
Such a study provides a better understanding of the hot rolling and the PSC test, and
therefore, gives guidance to procedures to determine appropriate values for the
experimental parameters in PSC test to simulate hot rolling operation more accurately.
5.2. Mathematical approach
5.2.1. Constitutive model
A simple AA1050 commercial aluminium alloy is used as the model material in this
study. The physical properties of this alloy as a function of temperature are listed in
Table 1. The constitutive equation for the material used in the calculation was derived
from hot, plain strain compression tests on a Gleeble 3500, with the ratio of the tool
width to sample thickness being 2. The shape of stress strain curve at elevated
temperatures is well described by a generalized Voce type of equation [11]:
0
( ) exp( / )
ss ss ss
C o o o o c o = (1)
A FE comparison study of hot rolling and PSC testing 69
with ( )
0 1 1
ln A Z B o = , ( )
2 2
ln
ss
A Z B o = ,
( )
1
3 3
sinh ( )
m
C A Z B
= , exp( / )
def g def
Z Q R T c = .
where c is nominal strain, c is nominal strain rate;
0
o ,
ss
o and o are yield stress, the
steady-state stress and the flow stress, respectively; Z is the Zener-Hollomon
parameter, 156 kJ/mol
def
Q = , the apparent activation energy of the deformation process
during plastic flow,
g
R , the universal gas constant, T , the absolute temperature;
i
A ,
i
B ,
m are fitting parameters whose values are given in Table 2.
Table 1. Physical properties of AA1050 used to run FEM simulation in Marc
Temperature
(C)
Coefficient of
thermal
expansion
(x10
-5
) K
-1
Specific heat
(N/mm
2
K
-1
)
Thermal
conductivity
(W/m K
-1
)
E
(GPa)
20 2.20 900 230 71.8
100 2.39 938 232 68.8
200 2.43 984 234 65.1
300 2.53 1030 236 61.3
400 2.65 1076 237 57.6
500 2.82 1116 238 53.8
Table 2. Fitting parameters for the constitutive model used
1
A
2
A
3
A
1
B
2
B
3
B m
3.4 4.92 111.6 1.47x10
9
5.58x10
8
1.91x10
7
0.28
5.2.2. Hot rolling operation
1. Analytical analysis
The nominal strain during rolling deformation, when assumed to be homogeneously
distributed through the thickness of the strip, is given by:
0
2
ln
3
f
h
h
c = (2)
Chapter 5 70
while the instantaneous strain rate is defined as:
2 1
3
dh
h dt
c = (3)
Using the geometric relation as shown in Fig. 1, Eq. (3) can be reformulated into:
2 2 sin
2 (1 cos ) 3
R
f
V
h R
u
c
u
=
+
(4)

Eq. (4) indicates that the instantaneous strain rate depends on both the roll gap geometry
and the roll speed. The strain rate reaches a maximum value at the entry and decreases
to zero at the exit. The corresponding average value of the nominal strain rate is given
by integrating Eq. (4):
0
( )
R
f
V
R h h
c
c ~
(5)

where R is the radius of the roll,
0
h and
f
h are the entry and exit thickness of the work-
strip, respectively, h is current height and t is time,
R
V R e = with e the rotation rate
of the roll, and u is defined in Fig. 1.
Fig. 1. Roll gap geometry and sampling plane evolution during hot rolling.
2. Finite element analysis
As the rolling operation is performed at elevated temperature, a heat-transfer analysis is
included and coupled with the deformation analysis. The temperature profile in the slab
during hot rolling was calculated under the following assumptions. 1). Since the length
of the piece is much greater than either its width or thickness, modeling of this
particular process can be reduced to a 2D approach. 2). Oxide formation does not
significantly affect the heat transfer between the strip and the work roll. 3). The work
rolls are assumed not to change shape or size during the deformation process, and
remain perfectly spherical in cross section.
The mathematical model for a nonlinear thermomechanical process is given by the
following coupled equation [12]:
( , ) K T = u u f (6)
p
T T
U C k q
x y y
| | c c c
= +
|
c c c
\ .
(7)
1
r r
r pr r
T T
C rk
t r r r
c c c | |
=
|
c c c
\ .
(8)
Eqs. (7) and (8) are, respectively, the governing equation for heat flow in the slab and in
the work roll. K is the material stiffness matrix, u is the nodal displacement vector, T
is the nodal temperature, f is the prescribed force vector, k ,
p
C , are the thermal
conductivity, specific heat and the density of the slab respectively.
r
k ,
pr
C ,
r
are the
thermal conductivity, specific heat and the density of the roll respectively. U is the strip
entry velocity in mm per second ( x Ut = ). q is a heat-generation term representing the
heat released due to plastic work and is given by:
q qoc = (9)
where o is the equivalent flow stress, c is the equivalent strain rate, and q is the
efficiency of conversion of strain energy to heat which is taken to be 0.95, as has been
determined empirically for aluminium alloys [1].
The finite element formulation of the mathematical model given by Eqs. (6), (7) and (8)
leads to a set of non-linear equations which in MARC [13] are solved through the full
Newton-Raphson method. The flow formulation method and an Arbitrary Eulerian
Langragian Technique are used. Continuum mechanics is employed as the deformation
behavior is on a macro scale.
Chapter 5 72
Due to symmetrical geometry, computations are carried out only in top half after
assuming equal conditions on top and bottom. Initial conditions are listed in Table 3.
Furthermore, the following boundary conditions are imposed:
(1). Symmetrical cooling occurs at the centerline of the strip, the temperature change in
the work roll is confined to a thin layer (o =25 mm):
0, 0 0
T
t y k
y
c
> = =
c
(10)
0, 25 0
r
T
t r R k
r
c
> = =
c
(11)
(2). heat transfer between the strip and the work roll
0, ( ) / 2, ( )
con strip roll
T
t y Y t k h T T
y
c
> = =
c
(12)
(3). heat transfer to the environment at surface boundaries exposed to air
0, ( ) / 2, ( )
env strip room
T
t y Y t k h T T
y
c
> = =
c
(13)
where ( ) Y t is the strip thickness at each zone.
(4). the velocity at the center of the strip in the y direction is zero due to the geometric
symmetry.
0, 0, 0 t y V > = = (14)
Table 3. Variables used for simulation
Work roll Initial temperature C 50
Strip initial temperature C 390
Environment sink temperature C 20
Work roll thermal conductivity (W/m K) 36
Work roll heat capacity (N/mm
2
K) 3.77
Film coefficient to environment (strip and roll)
env
h
0.01
Contact heat transfer coefficient (roll) kW/m
2
K
con
h
20
radius of the rolls (mm) 100
Shear friction for rolling
0.6 =
5.2.3. Plain strain compression
Plain strain compression is simulated under an imposed constant nominal strain rate
which is controlled by the velocity of the upper tool. The displacement boundary
conditions can be summarized in terms of the applied strain in the y direction:
0 0
x
U at x = = (symmetry)
0
(1 exp( ))
y
U h t at y h c = = (15)
0 0
y
U at y = =
where
x
U and
y
U are, respectively, the displacement along x and y direction. For
geometric symmetry reasons, simulation is only done in one half of the specimen to
save computational time.
5.2.4. Friction condition
Friction conditions at the roll-workpiece interface greatly influence the metal flow and
the formation of surface. In the present study, the method used to model the interface
friction is the interface shear factor approach, defined by [14]:
0
2
[ arctan ]
s
s s
u
f k
u t
= (16)
where
s
f is the frictional stress,
s
k is the shear strength of the work strip,
s
u is the
relative velocity between roll and work slab,
0
u is a constant, usually takes a value of
(0.01 0.1)
s
u . The interface shear factor, , represents the ratio of the interface shear
stress of the object being deformed to the actual material shear strength ( /
i s
k t = ),
where
i
t is the interface shear stress.
5.3. Results
5.3.1. Simulation of hot rolling process
In the hot rolling deformation study it is assumed that the radius of the rolls is 100 mm.
Simulations are performed under 4 different conditions, i.e., two different reductions,
50% (
0
60 h = , 30
f
h = ) and 37.5% (
0
48 h = , 30
f
h = ) with nominal strain rates of
2.5 /s and 0.25 /s. The interface friction coefficient and the relative velocity are selected
by sensitivity analysis. It is found that an increase in friction coefficient from 0.4 to 1.0
does not significantly affect the pattern of the equivalent strain and equivalent strain rate
distribution during hot rolling. Variation of the parameter
0
u in Eq. (16) in the range of
Chapter 5 74
(0.01 0.1)
s
u has little effect on the strain distribution and the roll force. In the
following hot rolling simulations, =0.6 and
0
0.1
s
u u = are applied.
1. The deformation history during hot rolling
We first consider the case of 50% nominal reduction (
0
60 h = , 30
f
h = ) rolled at a
nominal strain rate of 2.5 /s. Let us follow a perpendicular plane of nodes through
thickness (C-S) pass through the roll gap (see Fig. 1). When it enters the roll gap, the
plane bends gradually. The temperature profile, equivalent strain, equivalent stress, and
equivalent strain rate along this row at 6 increments are shown in Fig. 2a-d.
Fig. 2. The equivalent strain, strain rate, stress and temperature profiles of the
sampling plane from surface to center at various positions for the case of a nominal
reduction of 50% and a nominal strain rate of 2.5 /s.
310
320
330
340
350
360
370
380
390
400
410
0 0.2 0.4 0.6 0.8 1
Relative distance from surface
T
e
m
p
e
r
a
t
u
r
e

C
S1-C1
S2-C2
S3-C3
S4-C4
S5-C5
S6-C6
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

S1-C1 S2-C2
S3-C3 S4-C4
S5-C5 S6-C6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 0.2 0.4 0.6 0.8 1
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e

(
1
/
s
)
S1-C1 S2-C2
S3-C3 S4-C4
S5-C5
0
10
20
30
40
50
60
0 0.2 0.4 0.6 0.8 1
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
S1-C1 S2-C2
S3-C3 S4-C4
S5-C5 S6-C6
At position C1-S1, about 5 mm ahead of the entry plane, the surface node S1 is not yet
in contact with the roll. Due to surrounding material being sucked into the roll gap, the
equivalent strain, stress, and strain rate begin to increase while the temperature shows a
little parabolic fall towards the surface due to air cooling of the stock on leaving the
reheating furnace.
At C2-S2, 5 mm behind the entry plane, the test plane is in the roll gap. The temperature
begins to drop quickly near the surface because of the chilling effect of the cold roll.
The equivalent strain rate, equivalent strain and equivalent stress now increase
significantly throughout the thickness.
At C3-S3, 15 mm from the entry, the test plane is at about one third of the contact arc
length, the temperature near the surface continues to drop but at center the temperature
increases due to heat generation from deformation. This temperature rise is obliterated
nearer the surface by the chilling effect of the rolls. The strain rate at the surface
decreases but reaches a maximum level at the center. Correspondingly, the stress
reaches its highest value in the center.
At C4-S4, 30 mm from the entry, the test plane is at about two third of the contact arc
length, the temperature begin to increases both at the surface and in the center. The
strain rate drops to a lower level, which is rather uniformly from surface to center. The
stress also decreases.
At C5-S5, 49 mm from entry, the test plane comes to the exit of the roll gap, the strain
rate drops to zero. The stress continues to decrease. The temperature near the surface
keeps increasing but decreases in the center.
At longer times (C6-S6, 16 mm away from the exit plane) the strain distribution does
not change. The stress drops down to a lower level, which may be considered as
residual stress. The temperature rebounds rapidly near surface but drops monotonically
in the center due to high thermal conductivity of aluminum alloy. The temperature
difference between surface and center of the slab becomes smaller.
2. The effect of the nominal strain rate and nominal reduction
In order to demonstrate the deformation history more clearly, the traces of two sampling
points C (at center of the strip) and S (on the surface) are followed. Fig. 3 and Fig. 4
show the evolution of equivalent strain, strain rate and temperature and resulting
equivalent stress during rolling operation under the 4 deformation conditions considered
(see section 3.1).
Chapter 5 76
The equivalent strain increases as sampling points pass through the roll gap (Fig. 3a and
Fig. 4a). The strain at point S is larger than that in point C, which is approximately same
value as the prediction by Eq. (2). However, the equivalent strain is also affected by the
nominal strain rate (see Fig. 3a). The same level of strain at point C occurs earlier in
case of a higher nominal strain rate. This may be a general characteristic of the rolling
operation.
Fig. 3. The effect of the nominal strain rate on the deformation history of the sampling
points C and S with a nominal reduction of 50%, (a). equivalent strain, (b). equivalent
strain rate, (c). temperature, (d). equivalent stress.
0
10
20
30
40
50
60
70
-20 0 20 40 60 80 100
Distance from entrance (mm)
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
Point C,2.5/s
Point S,2.5/s
Point C,0.25/s
Point S,0.25/s d
0
0.2
0.4
0.6
0.8
1
1.2
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n
Point C,2.5/s
Point S, 2.5/s
Point C,0.25/s
Point S,0.25/s
eq.(2)
a
230
270
310
350
390
430
-20 0 20 40 60 80 100
T
e
m
p
e
r
a
t
u
r
e

C
Point C,2.5/s Point S,2.5/s
Point C,0.25/s Point S,0.25/s
c
0
1
2
3
4
5
6
7
8
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e

(
1
/
s
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e

(
1
/
s
)
eq.(4)
Point S,2.5/s
Point C,2.5/s
Point S,0.25/s
Point C,0.25/s
b
Fig. 4. The effect of the nominal reduction on the deformation history of the sampling
points C and S at a nominal strain rate of 2.5/s, (a).equivalent strain, (b). equivalent
strain rate, (c). temperature (d). equivalent stress.
FEM simulation shows that the strain rate evolution is much more complicated (Fig. 3b
and Fig. 4b). A very high strain rate (~3 times of the nominal strain rate) exists just
beneath the surface at the entry into the roll gap. However at the center position the
strain rate increases first and then decreases as the strip moves forwards from the entry
to exit. The peak value is slightly larger than the nominal strain rate. The peak position
appears at one third to the two third of the contact arc length and is affected by the
nominal strain rate and the nominal reduction. A lower nominal strain rate and a larger
nominal reduction shift the peak towards the exit. At the exit of the roll gap the strain
rate decreases to zero. There is a large difference between the average strain rate
derived from the analytical Eq. (4) and the values from the FEM simulation over much
of the contact distance while the shape of the two curves is more or less comparable.
0
0.2
0.4
0.6
0.8
1
1.2
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n
eq.(2),37%
Point C,37%
Point S,37%
eq.(2),50%
Point C,50%
Point S,50%
a
0
1
2
3
4
5
6
7
8
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e

(
1
/
s
)
eq.(4),37%
Point C,37%
Point S,37%
eq.(4),50%
Point C,50%
Point S,50%
b
290
310
330
350
370
390
410
-20 0 20 40 60 80 100
T
e
m
p
e
r
a
t
u
r
e

C
Point C,37%
Point S,37%
Point C,50%
Point S,50%
c
0
10
20
30
40
50
60
70
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
Point C,50%
Point S,50%
Point C,37%
Point S,37%
d
Chapter 5 78
Temperature profiles demonstrate the strong cooling effect of the rolls, the rotation rate
of the roll (nominal strain rate) and the nominal reduction. The surface temperature of
the strip experiences large variations as it passes through the roll gap. The larger
reduction, the greater is the drop in slab surface temperature. The temperature in the
center point C is more sensitive to the nominal strain rate as can be seen by comparing
the Fig. 3c and Fig. 4c. At a higher strain rate the process in the point C becomes almost
adiabatic and the heat generated raises the temperature whereas at low strain rate the
temperature in point C will decreases since there is enough time for the internal heat
generated by inelastic deformation to be dissipated towards surroundings.
Fig. 5. The effect of the friction coefficient on the deformation history of the sampling
points C and S with a nominal reduction of 50%, (a). equivalent strain, (b). equivalent
strain rate, (c). temperature, (d). equivalent stress.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n
Point S,=0.8
Point C,=0.8
Point C,=0.5
Point S, =0.5
eq.(2)
a
290
310
330
350
370
390
410
-20 0 20 40 60 80 100
T
e
m
p
e
r
a
t
u
r
e

C
Point C,=0.8
Point S,=0.8
Point C,=0.5
Point S,=0.5
c
0
10
20
30
40
50
60
70
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
Point C,=0.8
Point S,=0.8
Point C,=0.5
Point S,=0.5
d
0
1
2
3
4
5
6
7
8
-20 0 20 40 60 80 100
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e

(
1
/
s
)
Point S,=0.8
Point C,=0.8
Point S,=0.5
Point C,=0.5
eq.(4)
b
As a combined effect of the strain rate, strain, and the temperature, the equivalent
stresses at each point also presents a peak evolution during the rolling process. The
position of the maximum values for the Point C corresponds to that of the equivalent
strain rate, and is affected by nominal strain rate and nominal reduction. At the entry of
the roll gap, although the strain rate is very high the equivalent stress at the surface is
low, because the strain is small and the steady state has not been reached yet.
3. Effect of the friction between the roll and strip
In hot rolling it is usually assumed that the friction is sufficiently high to give 'sticking'
friction over the whole arc of contact. As shown in Fig. 5, an increase in the friction
coefficient leads to a rise of the strain level near surface but has less effect on the strain
in the center. The strain gradient throughout the thickness increases significantly.
However, the friction has little effect on the strain rate history and no effect on the roll
strip temperature in the center of the strip.
5.3.2. Simulation of PSC test
For plain strain compression, the initial temperature is set at 390C for both sample and
tools. We first ignore the heat generation due to the deformation which is the actual
experimental condition on Gleeble, where the temperature is controlled by a sensitive
temperature controlling unit involving a direct current heater and a water cooling
circuit. Because the heat loss by radiation and convection from the specimens to the
atmosphere is negligible, an isothermal deformation can be assumed. Then a
comparison was run for specimens with heat generation from plastic deformation. We
apply this simulation for the case / 2 w h = , where the geometrical effect of the tool to
specimen dimensions can be avoided. In an ideal frictionless compression test the
distribution of the local strain, strain rate and stress in all parts of the specimen is
considered to be uniform. The applied strain rate is constant as the nominal reduction
increases. However, the presence of friction will decrease the strain, strain rate at the
specimen surface while increasing these values in the center. Fig. 6 shows the
distribution of the equivalent strain rate along the y-y symmetrical plane for the cases of
a nominal strain rate of 2.5 /s with an interface fraction coefficient =0.2. As the
friction coefficient increases, the equivalent strain and strain rate gradient between the
surface and the center increases. Obviously, in PSC testing and hot rolling friction has a
different effect on the evolution of the process parameters.
When heat generation by deformation is considered the temperature will increase during
deformation. The temperature gradually rises as the test proceeds and reaches its highest
value at the end of the test, i.e. when the maximum deformation is attained. The
Chapter 5 80
temperature increase depends to a large extent on the nominal strain rate applied. For
example, at a nominal strain rate of 0.25 /s, the temperature increases only 1.7C after
deforming to a nominal reduction of 65%, while for a higher strain rate of 2.5 /s, the
temperature increase is 10C. Fig. 7 shows the effect of the nominal strain rate on the
temperature in the specimen center and its effect on the equivalent stress. The
temperature distribution is more or less uniform in the deformation area. This
temperature increase slightly lowers the equivalent stress.
5.4. Discussions
5.4.1. Characteristics of hot rolling deformation
The above results show the equivalent strain experienced by the stock during the rolling
pass is completely non-uniform. It increases with time to a maximum at the exit of roll
gap and varies through thickness. The strain level at the slab center corresponds closely
to the values of nominal true strain for each reduction and it is affected by the nominal
strain rate applied. The excess strain near the surface represents the redundant shear
brought about by constraining the metal to flow through the roll gap.
The variation of the local equivalent strain rate is more complicated and varies from
entry to exit and from point to point through the thickness. The values in both the
surface and center of the sheet are significantly different from the analytic prediction
given by Eq. (4). The strain rate pattern in the center of the rolling strip experiences a
slow increase at first and then decreases gradually to zero.
The importance of a knowledge of the accurate temperature of the slab within the
deformation zone of the rolling operation has been recognized [15,16]. Most of the
literature documented only pays attention to the temperature at the exit of the roll gap.
Of course this temperature is important because the following softening procedure is
strongly dependent on it. However, use of this exit temperature in the constitutive
equations for whole deformation process may not be so reasonable, especially for lower
nominal strain rates. In the condition of a higher nominal strain rate the temperature at
which the peak stress is reached is approximately the exit temperature. However, for
lower nominal strain rates, the temperature at which the equivalent stress reach its
maximum differs considerably from the temperature at the exit.
As can be seen, a severe deformation (lower temperature, larger equivalent strain and
higher initial strain rate) occurs near the surface of the rolling strip. So the
characterization of the deformation and the following softening behavior should be
applied to the center area.
Fig. 6. The distributions of the equivalent strain, strain rate and stress along the
symmetrical line in the tool direction at several nominal reductions for the case of a
nominal strain rate 2.5 /s, no heat generation, =0.2.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 2 4 6 8 10
Distance from upper tool (mm)
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n
0.71
0.63
0.5
0.39
0.3
0.16
a
30
35
40
45
50
55
0 2 4 6 8 10
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
0.71 0.63
0.5 0.39
0.3 0.16
c
0
0.5
1
1.5
2
2.5
3
3.5
4
0 2 4 6 8 10
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e
0.71 0.63 0.5
0.39 0.3 0.16
b
Chapter 5 82
Fig. 7. The effect of the nominal strain rate on the temperature and the equivalent stress
(=0.2).
It should be noted that a considerable strain, strain rate and stress are present before the
material enters into the roll gap. The equivalent stress reaches a maximum stress at
about one third to two third of the contact arc length from the entry. The fact that the
equivalent stress reaches a maximum value early in the roll gap could have important
practical implications. This will influence the following processes such as recovery and
recrystallization, as they both depend on the stored energy from the prior deformation
and the recrystallization temperature. This result is in good agreement with experiments
by Sheppard et. al [17]. They observed an imperfect cellular structure with ragged walls
and considerable dislocation activity on the spot which is still outside the quasi-static
deformation zone. At the onset of observable macroscopic plastic deformation the
electron micrograph indicated a structure consisting of well developed subgrain walls
but containing a multiplicity of microcells within each subgrain. Within one third of the
contact arc length from entry, the subgrain cell decreases. Beyond this point the
subgrain size is stabilized and no further change in size could be observed.
5.4.2. Comparison of the hot rolling operation and PSC testing
It can be concluded from the present simulation that only the history of equivalent strain
in the center of the rolling slab is comparable with that in the PSC testing, especially for
lower strain rates. However, as a key parameter for hot deformation, the evolution of
equivalent strain rate during deformation in both cases is quite different. In the PSC
testing a constant equivalent strain rate during deformation is obtained under a constant
30
35
40
45
50
0 0.2 0.4 0.6 0.8
Nominal reduction
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
385
390
395
400
405
410
T
e
m
p
e
r
a
t
u
r
e

C
0.25/s 1/s 2.5/s
nominal strain rate experiment. In real industrial rolling operation the equivalent strain
rate in the center of the slab presents a peak variation.
It is assumed that the experimental PSC testing is carried out in an isothermal way. In
practice this is true at low strain rates since the internal heat generated by friction and
inelastic deformation is dissipated towards the surroundings, and therefore the specimen
temperature is not raised. However, at a high strain rate the process becomes almost
adiabatic and the heat generated raises the specimen temperature. For hot rolling
operation under higher strain rates the temperature in the sampling point in the center of
the strip will increase, but under lower strain rate the temperature drops significantly.
The friction in both cases also presents a different effect. In the hot rolling condition,
the increasing friction coefficient between the roll and the strip has little effect on the
distribution and the evolution of the deformation parameters in the center point.
However, increasing the friction coefficient results in an increase in the equivalent
strain, strain rate in the center sampling point in the case of PSC testing.
5.5. Conclusions
The evolution of the equivalent strain rate and temperature during hot rolling operation
and in constant nominal strain rate PSC testing is quite different. The highest strain rate
distribution is concentrated just beneath the surface of the slab around the entry point.
The strain rate in the center of the rolling slab experiences a peak variation, i.e. first
increases and then decreases as the slab passes the roll gap.
The equivalent stress, which is a combined effect of the strain rate, strain and the
temperature, reaches a maximum values at about one third to two third of the arc contact
length from the entry, the position of the peak corresponding to that of the equivalent
strain rate.
It is suggested that in order to simulate the rolling operation more accurately in a PSC
experiment the strain rate should be designed to impose a peak variation instead of a
constant value. The peak value is the equivalent value of the nominal strain rate. In PSC
testing the friction between the sample and the tools should be as small as possible. The
comparison should be made in a confined area which is the center of the strip or the
PSC specimen. For lower strain rates, the temperature drop in the hot rolling operation
must be taken into account.
Chapter 5 84
References
1. G.E. Dieter, in 'Mechanical Metallurgy', New York, McGraw-Hill, 1986.
2. M.A. Wells, D.J. Lloyd, I.V. Samarasekera, J.K. Brimacombe and E.B. Hawbolt,
Metall. Mater. Trans. B, 29, 1998, 709-719.
3. D.Q. Jin, V.H. Hernandez-Avila, I.V. Samarasekera and J.K. Brimacombe, in
'Modeling of Metal Rolling Progresses', J.H. Beynon, P. Ingham, H. Teichert, and
K. Waterson (Eds), The Institute of Materials, 1996, 36-58.
4. R. Colas, Model. Simul. Mater. Sci. Eng., 3, 1995, 437-453.
5. T. Furu, H.R. Shercliff, C.M. Sellars and M.F. Ashby, Mater. Sci. For., 217222,
1996, 453-458.
6. R. Hill, in 'The Mathematical Theory of Plasticity', Clarendon Press, Oxford, 1950.
7. J.H. Beynon and C.M. Sellars, J. Test. and Eval., 13, 1985, 28-38.
8. R.J. Hand, S.R. Foster and C.M. Sellars, Mater. Sci. Techn., 16, 2000, 442-450.
9. C.H. Lee and S. Kobayashi, Int. J. Mech. Sci., 12, 1970, 349-370.
10. S.I. Oh, S.L. Semiatin and J.J. Jonas, Metall. Trans. A, 23A, 1992, 963-975.
11. C.M. Sellars, F.J. Humphreys, E. Nes and D.J. Jensen, in 'Numerical Predictions of
Deformation Processes and the Behaviour of Real Materials', S.I. Andersen, J.B.
Bilde-Soresen, T. Lorentzen, O.B. Pedersen, and N.J. Sorensen (Eds), Riso National
Laboratory Roskilde, Denmark, 1994, 109-134.
12. O.C. Zienkiewicz, in 'The Finite Element Method', McGraw-Hill, London, 1977.
13. Marc, in 'Users' manual, Marc Analysis Research Corporation', Palo Alto, CA,
1996.
14. S. Kobayashi, in 'Numerical Analysis of Forming Processes', J.J. Pitman, O.C.
Zienkiewicz, R.D. Wood, and J.M. Alexander (Eds), Wiley, 1984, 45-71.
15. C.M. Sellars, Mater. Sci. Techn., 1, 1985, 325-332.
16. C. Devadas, I.V. Samarasekera and E.B. Hawbolt, Metall. Trans. A, 22, 1991, 307-
319.
17. T. Sheppard, M.A. Zaidi, P.A. Hollinshead and N. Raghunathan, in 'Microstructural
Control in Alminium Alloys: Deformation, Recovery and Recrystallization', E.
Henry Chia and H.J. McQueen (Eds), The Metallugical Society, Inc., 1985, 19-43.
S.P. Chen
Chapter 6
Modeling recrystallization kinetics in
AA1050 following simulated break down
rolling
A physical model based on the single grain approach is refined and developed to predict
the recrystallization kinetics of a single-phase metal following hot deformation. The
model involves the original and deformed grain geometry, the characteristics of the
dislocation networks formed and the average mobility of the moving interfaces. The
model accounts properly the effect of the concurrent recovery and textural components
in the deformed microstructure on the recrystallization kinetics. Experimental work is
conducted on the recovery and recrystallization kinetics in AA1050 following plastic
deformation at elevated temperatures as encountered during break down rolling to
extract the physical input parameters and to validate the model. The Plane strain
compression (PSC) test is used to simulate the hot rolling deformation. The predictions
are in good agreement with the experimental results.
6.1. Introduction
Hot working is a process of changing the shape and the microstructure of metals and
alloys at elevated temperatures. The important microstructural changes occurring during
hot working are recovery and recrystallization, which are the restoration processes that
result from the release of the internal stored energy during deformation [1-3]. Modeling
of thermomechanical processing is well established as a valuable procedure for
optimizing processing conditions and is currently the subject of major research efforts in
the aluminum industry.
Chapter 6 86
Several attempts have been proposed to model the recrystallization kinetics physically
[3,4]. These models consider the nucleation positions and growth behavior of the
recrystallized grains by linking them to the deformation microstructure through internal
state variables. The results have shown that such models describe the experimental
results for specific alloys well, though only for uniform processing conditions.
However, these models could not take into account of the effect of concurrent recovery
on the recrystallization kinetics properly. The contribution of the recovery on the
softening kinetics could be separated using different techniques [5] and in some cases
this effect is quite large and could not be ignored [6]. Moreover, recent experimental
investigations have revealed that the recrystallization is inhomogeneous from grain to
grain [7,8]. Such a general characteristic is believed to be caused by an inhomogeneous
distribution of stored energy. The deformed grains with different orientations have
different microstructures and stored energy. In this case not only will the growth rate
decrease during recrystallization, but growth of the grains will be restricted by
impingement with their neighbors within the original grains [1]. It appears that a more
sophisticated and fundamental description of the recrystallization should start from
considering the microstructure evolution in a single grain. An approach based on such a
principle has been proposed by the present authors to predict the recrystallization
kinetics after cold deformation in Chapters 3 and 4 [9,10]. In the single grain model, the
structure of the deformed metal is taken into account, which contains the information
about the main features of rolled material, i. e. on the degree of cold rolling, the grain
size, and the grain shape. An as-deformed tetrakaidecahedron is applied to describe the
grain geometry. The subgrain size and the misorientation between subgrains are related
to the degree of deformation. Additionally, the effect of recovery on the recrystallization
kinetics can be nicely predicted. The model has been further developed in Chapter 4 to
incorporate both microstructural inhomogeneities and the textural components in the as-
deformed state [10].
In this Chapter the model based on the single grain approach is adapted and elaborated
to predict the recrystallization kinetics following hot deformation. As a physical model,
all the input parameters are related to the microstructure and can be determined from the
experiment only leaving the pre-exponent factor of the mobility as an adjustable
parameter. Experimental study is carried out on the recovery and recrystallization
kinetics in AA1050 following simulating hot deformation in PSC to extract the input
parameters and to validate the model. The actual strain, strain rate and temperature
distribution in the PSC specimens were analyzed using FEM simulation. The model
predictions show a good agreement with the experimental results in a wide range of
annealing conditions.
Modeling recrystallization kinetics following hot deformation 87
6.2. The recrystallization kinetics model
6.2.1. The Single grain approach
A physical model to predict the recrystallization kinetics of single-phase metals, based
on deformed microstructure, after cold deformation, has been developed in Chapters 3
and 4 [9,10]. The model can be used to predict the recrystallization kinetics when only
the deformation strain and the annealing temperature are known. It takes into account
the grain geometry, the position and the density of the nucleation sites as well as stored
energy evolution. The effect of the concurrent recovery and the microstructural
inhomogeneous and textural component in the deformed state can also be incorporated.
With some adaptation we will apply it to modeling recrystallization kinetics after hot
deformation.
In the single grain approach, it is assumed that the microstructure of an well-annealed
material can be described by a distribution of regular tetrakaidecahedra of various sizes.
After rolling deformation, the grains elongate along the rolling direction. If it is further
assumed that each grain under the same strain as the macroscopic one, then a deformed
tetrakaidecahedron with a size distribution will represent the main features of rolling
deformation, degree of rolling, the grain size and grain shape.

For aluminum alloys the microstructure after hot deformation can be adequately
described by subgrain size and the average misorientation between the subgrains. In a
deformed grain with a specific orientation, the microstructure will have a variety of cell
sizes and a distribution of misorientations. We assume that the microstructure can be
described using two components [11]: an assembly of equiaxed subgrains and specific
subgrains, effectively sub-critically sized recrystallization nuclei. The former can be
characterized by a mean equivalent diameter, o , a mean misorientation, u , and with
boundaries of mean energy and mobility, and
sb
M , respectively. The latter have a
larger size (
n
o ) and different boundary characteristics ( , ,
n n n
M u ). In the present study
we assume that the misorientation angle between the particular subgrain and
surroundings is 15.
The nucleation site density in a single-phase material is proposed to be given by [3]:
1 2 3 4 2 3
0 0 0
V V
v V
L S P
N c P c c c
u
o o o
= + + + (1)
where
V
P is the number of grain corners, where four grains meet per unit volume,
V
L is
the line length per unit volume where three grains meet, and
v
S is the grain boundary
Chapter 6 88
area per unit volume between two grains, P
u
depending on finding a mobile boundary
inside an original grain, a deformation band boundary of 10 u > , 0 o is the average
subgrain size in the as-deformed state.
1
c to
4
c are constants. From the present
experimental observation and literature studies it can be concluded that nucleation at
grain boundary is the dominant term for a large regime of deformation.
Assuming grain boundary nucleation being dominant mechanism, the nucleation site
density, N
v
, in a grain with a specific orientation, can be expressed as:
2
0 0
/
v d v
N C S N o = + (2)
where
d
C is a calibration constant,
0
N , the nucleation site density contributed from
other mechanisms,
0
o , the average subgrain size in the as-deformed state in a deformed
grain. The value of
v
S is dependent on the rolling strain and the initial grain size, and
for a tetrakaidecahedron, varies with rolling strain in a manner described by [12]:
( )
2 2 2 2
1
3 1 2/ 3 2/ 2 2 /
2
v
S a a a a a a a
d
= + + + + + + (3)
where a e
c
= is the distortion along the rolling direction, d is the grain size of the
starting material.
The growth velocity of a recrystallizing grain front is described by:
n
n
dR
V M P
dt
= = (4)
where V is the migration rate,
n
M is mobility, P is the overall driving pressure.
The mobility of a high angle boundary at an annealing temperature given by:
0
exp( )
n rex g
M M Q R T = (5)
where
rex
Q is the activation energy for grain boundary motion,
0
M is the pre-exponent
factor,
g
R is gas constant, T is the annealing temperature.
The overall driving pressure for a viable nucleus to grow is given by the sum of the
driving pressure from stored energy and the retarding pressure from boundary
curvature:
4
n
n
P
o
o o
= + (6)
Where o is a geometrical constant and has a value of ~3 [1], is the subgrain
boundary energy, which is assumed to depend on u as follows:
(1 ln )
m
m m
u u

u u
= (7)
where
m
u is a value of misorientation at which the boundary approaches high-angle
character (taken to be 15 [1]) and
m
is high angle boundary energy per unit surface
(taken to be 0.324 J/m
2
for Al [1]).
A deformed polycrystal contains many grains of different orientations. In this model,
based on the experimental fact, we assume such a deformation pattern that
microstructure may change from grain to grain and it can be related to the Taylor factor
of the individual grain. The relationship between the deformation features ( , o u ) of an
individual grain and the average values ( , o u ) of the polycrystalline aggregate can be
related to Taylor factor as [10]:
2
av
M
M
u u
o o
| |
=
|
\ .
(8)
where M and
av
M are the Taylor factors of a grain with a specific orientation and the
average value of all the orientations, respectively.
We need to know the evolution of microstructural parameters: o and u with the flow
stress. Generally u is considered to be a constant value of about 4 [2] when the steady
state has been reached and this value is assumed not to change during recovery.
A relationship exists between equivalent flow stress o and o ,
i
for a variety of
deformation conditions and during recovery process, i.e. [13,14]:
1 2
/
i av i av
M Gb M Gb o o o o o = + + (9)
where
i
o is the initial flow stress, G is the shear modulus,
av
M is the Taylor factor of
the polycrystal, b is the value of Burgers vector, and
1
o and
2
o are constants. During
steady state deformation the principle of similitude applies [2,3]:
Chapter 6 90
/
i
C
o
o = (10)
where C
o
is a constant of typical value of the order 5 [2]. Now the average subgrain
size in the steady state is predicted to be a function of flow stress:
0 3
( )
av s i
M Gb o o o o = (11)
where
3 1 2
C
o
o o o = + , with
3
o being of the order 3 [2],
s
o is the flow stress in the
steady state. The subscript in
0
o denotes the subgrain size in the as deformed state.
The subgrain size will increase during recovery stage. A widely used equation to
describe the recovery in Al alloys is expressed as [13]:
( )
1 ln(1 )
g
i
rec
s i
R T
t t
R
A
o o
o o t
= = +
(12)
with relaxation time parameter
0
exp
rec
g
Q
R T
t t
| |
= |
|
\ .
(13)
where
rec
R is the fraction residual strain hardening at time t , ( ) t o , the instantaneous
flow stress during recovery,
rec
Q , an activation energy for recovery. A and
0
t are
constants both dependent on the composition and deformation microstructure of the
material.
Combining Eqs. (11) and (12), the subgrain size evolution in a deformed grain with a
average Taylor during recovery is now given by:
0
/
rec
R o o = (14)
By analogy to Eq. (14), the subgrain size evolution in a specific deformed grain during
recovery can be written as:
0
/
rec
R o o = (15)
6.2.2. Kinetics approach
The single grain model described in Chapters 3 and 4 is elaborated along the line
described above and is used to calculate the recrystallization kinetics in individual
grains with a specific orientation. Nucleation event in an as-deformed grain is assumed
to be site-saturated. A growing recrystallized grain is modeled as an expanding semi-
sphere nucleated on grain boundaries. New grains grow independently of one another
until hard impingement occurs.
The position of nuclei on the faces of the tetrakaidecahedron are determined randomly
assuming that there is an equal probability that each coordinate on the 14 faces
(including grain edges and corners) may act as a nucleation site. Each simulation is
therefore repeated a number of times (40) and the average of the simulated kinetics is
taken as the representative, i.e. the time to obtain a given fraction of recrystallization, f,
is given by the average over the number of the simulation, j.
, ,
/
av i i j
t t j =
_
(16)
where
, i j
t is the time at a given fraction in the j th simulation for a grain with a specific
orientation labeled i.
In the attempt at modeling recrystallization kinetics in a polycrystal material we assume
that there is no correlation between the orientations of neighboring grains and all the
grains of one crystallographic orientation (within a certain small range) are represented
by a single grain of the mean size.
The overall kinetics of the polycrystalline aggregate can be obtained by weighted
summation of the kinetics on the main textural components in the as-deformed state.
The fractional recrystallization f of the assembly is given by:
( )
i i i
f f = _ _ (17)
where
i
is the fraction of a main textural component.
6.3. Experimental
6.3.1. Material and experimental detail
The material used in the study is a commercial purity aluminium alloy AA 1050. The
as-received material is a sheet of transfer gauge of 25 mm in thickness that had been DC
Chapter 6 92
casted, homogenized, break-down-rolled and annealed. It has a fully recrystallized
microstructure with a grain size 104.4 m in the center of the sheet and 95.2 m in the
surface layer. Rectangular PSC specimens, the dimensions 102015 mm, were
machined from the middle thickness of the sheet with the deformation in PSC being
parallel to the rolling direction. The deformation is applied on the sample thickness (of
10 mm) direction and the tool width is 5 mm. The PSC tests were carried out on a
Gleeble 3500. All tests were conducted in an atmosphere of vacuum. A water based
graphite lubricant was used to reduce the friction between the specimen and the tools.
A series of tests were conducted at the following combinations of temperature and strain
rate: 340C and 0.25 /s; 390C and 2.5 /s and 450C and 26.2 /s, all to a strain of 0.62.
These process parameters, leading to an equal Zener-Hollomon parameter, were
calculated using an activation energy for deformation 156 kJ/mol. After deformation to
the defined strain the samples were unloaded and held at the annealing temperature,
which is the same as the deformation temperature, for various time intervals, followed
by jet water quenching.
Taking into account the deformation inhomogeneities in PSC samples, the
microstructural and micro-hardness analyses were carefully confined in three prescribed
areas. Sampling points are shown in Fig. 1. Their co-ordinates are, A (0, h/4, 0), B
(1.25, 0, 0), C (0, 0, 0) and D (0, h/8, 0), with h being the instantaneous thickness of the
specimen, for current deformation strain 5.8 h = . The actual distribution of the strain,
strain rate pattern is estimated from FEM simulation.
Fig. 1. Schematic illustration of the deformed specimen geometry and the sampling
points.
A composite image method [5] was applied to determine the volume fraction of the
recrystallized material. A Buehler OMNIMET MHT automatic micro hardness tester
was used for the micro-hardness measurements, using 100 g load and 15 s loading time.
Hardness tests were made after polishing and 18 measurements were taken for each
experimental point.
6.3.2. Finite element analysis
To decide the actual local strain stress distribution, the FEM program MARC [15], was
utilized. The deformation patterns (strain, stress and strain rate) are strongly dependent
on two factors: the ratio ( / w h ) of the tool width to the sample thickness and the
fraction coefficient between specimen and the tool surface. The extent of the non-
uniformity increases as the / w h decreases and/or the friction factor increases.
However, when the ( / w h ) is less than 1, which is the present case, the simulation
results do not show very much difference in the distribution of the strain, strain rate and
stress for various assumed frictional conditions at the sample-tool interface. Therefore
the deformation pattern is considered to be only dependent on the sample geometry.
Fig. 2 shows the simulated local equivalent strain, strain rate and stress distribution
along the vertical symmetrical line (yy in Fig. 1) for the specimens deformed to a
nominal reduction of 42%. As can be seen, the deformation patterns are identical for the
case of the deformation at 390C, nominal strain rate 2.5 /s and the case of 340C,
nominal strain rate 0.25 /s.
The local equivalent strain in the center of the specimen is about 1.18, which is almost
two times as that of the nominal strain (0.62). The strain rate gradient is even larger, for
example, for the case of 390C2.5 /s, equivalent strain rate in the center is 7.8 /s,
which is 3 times more than the nominal strain rate.
Chapter 6 94
Fig. 2. The distributions of a). the equivalent strain, b). strain rate, and c). stress along
the sampling line yy (Fig. 1) at a nominal strain of 0.62.
0
1
2
3
4
5
6
7
8
9
0 1 2 3 4 5 6
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n

r
a
t
e
390C,Nominal
strain rate 2.5/s
340C,Nominal
strain rate 0.25/s
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1 2 3 4 5 6
E
q
u
i
v
a
l
e
n
t

s
t
r
a
i
n
390C,Nominal
strain rate 2.5/s
340C,Nominal
strain rate 0.25/s
30
35
40
45
50
55
60
0 1 2 3 4 5 6
E
q
u
i
v
a
l
e
n
t

s
t
r
e
s
s

(
M
P
a
)
390C,Nominal
strain rate 2.5/s
340C,Nominal
strain rate 0.25/s
6.4. Results
6.4.1. The deformation structure
The microstructure in the deformed region of a PSC sample is not homogenous. A
single Xs shaped deformation zone along the diagonals between the tool edges can be
observed. The grains become elongated along the rolling direction of the sample.
Obviously, the grains in the center of the specimen experience a larger deformation. An
area in contact with each tool over a certain length near the center undergoes a much
lower strain. In the deformed grains one or two sets of deformation bands are observed
to form within 5 of an angle of 35 to the rolling plane. Fewer deformation bands
were observed in a distance away from the center (for example, point A in Fig. 1).
Fig. 3 shows the true stress versus the true strain plots for specimens tested at the 3
conditions to a strain of 0.62, with the data corrected using a friction coefficient =0.08.
At a strain of 0.4 the stress has reached the steady state level of about 44 MPa. The
difference between the steady state stresses of 3 sets of testing conditions with
combined temperatures and strain rates is within 3 MPa. The overlap in the strain stress
curves indicates that the condition of equal Z value has indeed been obtained.
Fig. 3. The strain stress curves of the PSC tested at three different conditions.
6.4.2. Softening behavior by micro-hardness indentation
The softening kinetics as measured by micro-hardness testing is presented in Fig. 4. We
confined our HV measurement to an area not more than 100 m of the point C and point
A. Clearly, the softening kinetics exhibits two stages: the first stage corresponding to
recovery, followed by an accelerated softening regime associated with recrystallization.
0
10
20
30
40
50
0 0.2 0.4 0.6 0.8
Strain
S
t
r
e
s
s

(
M
P
a
)
340Cx0.03/s
390Cx2.5/s
450Cx26.2/s
Chapter 6 96
In addition, the softening kinetics due to recovery is faster in the point C where the
deformation is largest.
The percentage of softening is defined by:
def t
rec
def rex
H H
f
H H
(18)
where
def
H is the hardness of the material in as-hot-deformed state,
t
H is the
instantaneous hardness after annealing time t , and
rex
H is the hardness of completely
recrystallized material. As the microhardness value is linearly proportional to stress
value, 1
rec rec
f R = . The fraction softened in point C and A as determined by Micro-
hardness testing is shown in Fig. 4 by dotted lines.
6.4.3. Recrystallization kinetics by optical microscopy
The recrystallization kinetics in point C, annealed at various temperatures, as monitored
by the metallographic measurements, is shown in Fig. 5, which is similarly measured in
the central region within a size of 500500 m. Fig. 6 shows partially recrystallized
structures obtained after an anneal for 6000 s at 340C (Fig. 6a) and after an anneal for
100 s at 390C (Fig. 6b). As can be seen, recrystallization was initiated primarily at
prior grain boundaries; recrystallization is not homogenous, in some deformed grains
recrystallized grains appear rather large, but in some other deformed grains there are
still no nuclei to be found.
Fig. 4. Microhardness measurements of the specimens annealed at different
temperatures for various times. Solid symbols refer to position C; open symbols refer to
position A; dotted lines is the best fit of Eq. (12).
23
25
27
29
31
33
35
1.E-01 1.E+01 1.E+03 1.E+05
Time (s)
M
i
c
r
o
H
V
HV in point C
HV in point A
Recovery line
450C 390C 340C
10
-1
10
5
10
3
10
1
Fig. 5. The recrystallization kinetics measured by quantitative optical microscopy and
the softening kinetics by Microhardness test in point C. Solid symbols indicate the
optical microscopy data; open symbols indicate microhardness data.

(a) (b)
Fig. 6. Optical microstructures in the center of PSC specimens (a) deformed at 340C,
0.25/ s c = to a strain of 0.62, annealing at 340C for 6000 s, (b) deformed at 390C,
2.5/ s c = to a strain of 0.62, annealing at 390C for 100 s, showing grain boundary
nucleation and the deformation bands in the deformed grains.
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
(
s
o
f
t
e
n
e
d
)
BY HV
BY OPM
450C 390C 340C
10
-1
10
4
10
3
10
2
10
1
10
0
10
5
Chapter 6 98
6.4.4. The relationship between static softening and recrystallization
Fig. 7 shows a plot of the fractional recrystallization determined by optical microscopy
against the softening fraction measured in point C. As can be seen, the fractional
recrystallization was non-linearly related to the softening (after subtraction of the
recovery the softening follow a straight line as the open marks shown in the Fig. 7).
This phenomenon was previously reported by Sellars group [16]. A large amount of
softening was involved even before the recrystallization start, which is identified by
optical microscopy, and this nonlinearity becomes insignificant as the annealing
temperature increases. Because both of the recovery and the recrystallization processes
are thermally activated, these processes are strongly temperature-dependent. Since, in
general, recovery proceed with lower activation energies than those which characterize
the recrystallization process, low temperatures will favor recovery over
recrystallization. At higher temperatures, thermal fluctuations assist both the initiation
and the progress of recrystallization. The earlier start of the recrystallization process
will reduce the relative softening due to recovery. Therefore, the extent of recovery is
more significant at lower isothermal annealing temperatures.
Fig. 7. The relationship between recrystallization fraction (measured by OPM) and
static softening (by HV). Filled marks: the static softening from HV before subtraction
of recovery. Open marks: after subtraction of recovery.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Fraction recrystallized
F
r
a
c
t
i
o
n

s
o
f
t
e
n
e
d
340C
390C
450C
6.4.5. The grain size of the fully recrystallized structure
In a space-filled assembly of fully-recrystallized grains, the relationship between the
number of grains per unit volume,
v
N , and the average grain size, d , is given by [17]:
( )
1/ 3
1/
v
d N = (19)
where is a constant having a value of unity for the case of site saturation [17]. Fig. 8
shows the grain size distribution near the different locations in a fully recrystallized
specimen annealed at 390C. it shows the expected log-normal behavior. As the
distance away from the center point C increases the range of the grain size becomes
larger as a consequence of the deformation inhomogenity. Table 1 lists the quantitative
measurement of mean grain size (which are given by both arithmetic mean ( d ) and
square root mean (
s
d ) and calculated
v
N (from d ) in the fully recrystallized samples
after annealing at different temperatures. The data in the table indicate that the grain
size, hence the nucleation site density, is a strong function of the location (strain and
stress vary with location) in the sample but is not much affected by the annealing
temperature. There is surprisingly little difference in arithmetic mean grain size after
annealing at different temperatures while the square root mean grain size increases as
the annealing temperature decreases. We will use the arithmetic mean d to characterize
the grain size, and consequently, assume that the final grain size does not depend on the
annealing temperature in the range studied.
Fig. 8. The grain size distribution in the different sampling points in a fully
recrystallized specimen annealed at 390C.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
10 30 60 90 120 150 180 210 240 More
Equivalent diameter (m)
F
r
e
q
u
e
n
c
y
Point C,d=75.4m
Point D,d=85.3m
Point A,d=101m
Chapter 6 100
Table 1. The mean grain size and the number of the grains per unit volume in the
sampling points
Annealing
temperature
Point C Point D Point A
d (m) 76.8 86.5 103.6
s
d (m)
96.5 107.5 129.4 340C
v
N
2.210
-6
1.510
-6
9.010
-7
d (m) 75.4 84.3 101.5
s
d (m)
89.4 97.8 121.4 390C
v
N
2.310
-6
1.710
-6
9.710
-7
d (m) 74 81.5 102.2
s
d (m)
83.6 93.7 113.5 450C
v
N (m
-3
)
2.410
-6
1.810
-6
9.410
-7
Average ( , )
V
N c o (m
-3
)
2.310
-6
1.710
-6
9.310
-7
6.4.6. JMAK and the S-F analysis of the recrystallization kinetics
The JMAK equation and the Speicl-Fisher (SF) equation are used to characterize the
recrystallization kinetics. JMAK equation is [1,18]:
1 exp( )
n
X bt = (20)
and SF equation [1,19]:
1
m
X
kt
X
=
(21)
where X is the volume fraction recrystallized in time, t, and n, b m, k are constants.
The JMAK description of isothermal kinetics used a least-squares analysis to obtain a
best-fit line for ln ln(1/(1 ) ln X vs t . The SF description of same kinetics used for
ln( /(1 ) ln X X vs t . Both equations describe the experimental data reasonably well in
the range of 10 to 90% of recrystallization with
2
R values in the range of 0.93 to 0.98.
The resulting kinetic parameters are listed in Table 2 and Table 3.
Table 2. The fitting parameters of JMAK and SF analysis in Point C
Temp C n lnb m lnk
340 2.138 -18.949 3.145 -27.090
390 2.222 -12.390 3.339 -7.343
450 2.313 -5.239 3.130 -6.146
Table 3. The fitting parameters of JMAK and SF analysis in Point A
Temp C n lnb m lnk
340 1.983 -18.872 3.030 -28.017
390 2.124 -12.645 2.747 -15.749
450 2.297 -6.232 2.888 -7.2569
As can be seen, the time exponents, n and m, remain relatively constant over the test
temperatures, while the parameters lnb and lnk, being reflective of the temperature
dependence of the nucleation and the growth rates, were strongly dependent on
temperature. In the present work the kinetics data are analyzed in terms of the average
values of n and m (listed in Table 4), the corresponding lnb and lnk values being
recalculated at each test temperature to characterize the isothermal recrystallization
kinetics. The kinetic curves resulting from this analysis, together with the experimental
measurements are shown in Fig. 9.
Fig. 9. The JMAK and S-F analysis of the experimental data.
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
Point C
Point A
JMAK fit
S-F fit
450C 390C 340C
10
-1
10
2
10
0
10
1
10
4
10
3
10
5
Chapter 6 102
Table 4. The average values of n and m and the appropriate lnb and lnk recalculated
n lnb m lnk
Point C 2.22 -59345/T+77.276 3.20 -85782/T+112.5
Point A 2.13 -57871/T+74.326 2.89 -78408/T+101.4
6.5. Application of the single grain model to predict the
recrystallization kinetics
We now turn to predict the recrystallization kinetics after hot deformation using the
single grain approach. To use the model we firstly specify the determinations of the
physical parameters from the experimental results.
6.5.1. The subgrain size
From Eq. (11) the average subgrain size could be decided if we know the value of the
initial flow stress
i
o and the steady state stress
s
o at each sampling point. In this study
the 16
i
o = MPa determined from strain stress curve is used and taking
3
2.6 o = in
Eq. (11) the resulting average subgrain sizes in two sampling points are listed in
Table 5, which are in a very good agreement with the literature data [20].
The subgrain size in a grain with a specific orientation is obtained from Eq. (8) with an
assumption that the subboundary misorientations are equal in the grains with different
orientations and take to be 4.
Table 5. The local equivalent strain, stress, subgrain size and
v
S in the sampling points
Strain
s
o (MPa)
0 o (m) v
S
lnZ
Point C 1.16 54.4 1.56 0.062 30.34
Point A 0.44 46.2 1.99 0.037 28.96
6.5.2. The calibration constant
d
C
From Eqs. (3) and (11),
v
S and
0
o depend on, respectively, the equivalent strain and
the equivalent stress at the sampling points and both are a function of the positions in
the sample. As shown in Fig. 10, along the y-y sampling line, the calculated
2
0 v
S o
increases linearly with local strain and varies with local stress in a manner of power
function. The values of ( , )
v
N c o measured at 3 sampling points against strain and stress
present an identical trend, which conforms that the
2
0 v
S o vs ( , )
v
N c o presents a linear
relationship, as shown in Fig. 11. The slope and the intercept of the plot ( , )
v
N c o
against
2
0 v
S o yield the calibration constant
5
9 10
d
C

= and
9
0
7.85 10 N

= /m
3
,
respectively, which indicates that the nucleation sites from other mechanisms such as
grain corners and deformation bands are neglectable compared with that of the grain
boundary nucleation.
Fig. 10. The variation of
2
0 v
S o against a). strain and b) stress.
0
2
4
6
8
10
12
14
35 40 45 50 55
Stress (MPa)
S
v
/
x
1
0
3
0
2
4
6
8
10
12
14
N
v
(
)
x
1
0
7
Sv/0
Nv(,)
0
2
4
6
8
10
12
14
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Strain
N
v
(
)
x
1
0
7
0
2
4
6
8
10
12
14
S
v
/
x
1
0
3
Nv(,)
Sv/
Chapter 6 104
The nucleation site density can now be calculated from Eq. (2). The effective number of
the nuclei in a deformed single grain can then be decided. Since the nuclei are assumed
to initiate from the original grain boundaries only half of the new grain grows inwards
the objective grain. So we need to modify the number of the nucleation site calculated
by a factor of 2, hereafter, this is referred as the numerical number of the nucleation
sites and used in simulation.
Fig. 11. A linear relationship between ( , )
v
N c o and
2
0 v
S o .
6.5.3. Parameters for recovery
By applying Eqs. (12) and (18) to the present experimental data, the best fit solutions
are shown in Fig. 4, and the fitting parameters are listed in Table 6. The reaction is
associated with an activation energy of 192 kJ/mol, which is a typical value of
activation energy for Fe diffusion in aluminium [21]. It suggests that the very small
amount of iron in this alloy control the recovery process.
0
t is temperature dependent
and it is given by:
0 1 2
bT b t = (22)
Table 6. Fitting parameters for recovery
A (kJ/mol) b
1
b
2
Q
rec
( kJ/mol)
Point C 95.8 3.6374x10
-18
2.1304x10
-15
192
Point A 95.8 2.4033x10
-17
1.4144x10
-14
192
0
2
4
6
8
10
12
0.002 0.004 0.006 0.008 0.01 0.012 0.014
S
v
/
0
N
v
(
)
x
1
0
-
7

(
m
-
3
)
6.5.4. Activation energy for recrystallization
An apparent activation energy can be found by assuming that recrystallization process
follows an Arrhenius relationship with respect to the temperature, i.e., that
exp
rex
R
Q
t
RT
| |
|
\ .
(23)
where
rex
Q is the activation energy related to the recrystallization processes.
The time
R
t , required for 25%, 50%, 68% and 75% recrystallization was calculated for
each annealing temperature using the JMAK analysis; the
R
Q value was estimated from
linear plots of ln(1/ )
R
t vs
1
1/ ( ) T K
, resulting in an apparent activation energy of

230 kJ/mol in the center point C and 225 kJ/mol in the point A. These values are in
good agreement with that of 234 kJ/mol reported by Perryman's study [22] and that of
230 kJ/mol by Sellars at al [23].
6.5.5. Textural components after hot deformation
A wide matrix of experimental plane strain compression data on the microstructure and
texture in hot worked AA1050 has been generated in the literature. The main texture
components found in AA1050 after hot deformation are: cube {100}<001>, Goss
{110}<001>, brass {110}<112>, copper (C) {112}<111> and S {123}<634>. The
concentrations of the textures are a function of the strain but not a function of Zener-
Holloman parameter. The amount of Cube component after hot working decrease as the
strain increases. The overall amount of rolling texture components increases, while the
ratio of the four rolling textural components Goss, brass, copper and S in AA1050 was
found not to alter in a range of hot working conditions. In this study we take the data
from Vernon-Pary et al [24], as listed in Table 7, as inputs for the modeling and
assuming that the grain with an orientation which has the larger Taylor factor
recrystallizes faster. The Taylor factors of the different texture components are
determined from H.J. Bunge [25].
Table 7. The input for the volume concentrations of texture components in two
sampling points
Sampling points Volume fraction (%)
Cube Goss Copper brass S Random
Point A 5.2 9 8.1 11.6 20.4 45.7
Point C 3 8.9 8.9 12.8 24.3 42.1
Chapter 6 106
6.5.6. The critical size of a viable nucleus
Another input parameter is the critical size of a viable nucleus. According to
Thompson [26], the condition for instability, leading to the growth of a potential
nucleus is:
0
n
n
dR dR
R R
dt dt
> (24)
where R and R
n
are the average subgrain radius in the assembly and radius of the
potential nuclei, respectively.
Using Eqs. (4), (6), (12) and (15) and taking 2R o = , Eq. (24) is solved for the
condition that the left-hand side equals zero and the critical diameter of the potential
nuclei is given as:
2
( ) 2
criticl
n n n n
d d
M M
dt dt
o o
o o o o
| |
=
|
\ .
(25)
where
0
2
( )
g
rex
R T
d
dt A t R
o
o
t
=
+
(26)
In the present modeling, it is assumed that the nucleation is site saturated and the
boundary between the potential nuclei and surroundings is high angle boundary, the size
of the critical nuclei can be calculated by using Eq. (25) taking 0 t = .
6.5.7. Prediction of recrystallization kinetics
Until now we have determined all the input parameters except the pre-exponent factor
of the mobility, which is used as a tuning parameter. This constant is determined to be
6
0.68 10
m
4
/Js by adjusting it to produce coincidence between the prediction and the
experimental points for the case of point C annealing at 390C.
The predicted recrystallization kinetics in point C and point A for the case of 390C
annealing by single grain model is shown in Fig. 12a and Fig. 12b. The recrystallization
kinetics in a grain with different orientations is taken as average of 40 simulations with
different nucleation site positions. The overall recrystallization kinetics is the weighted
summation of the kinetics from various textural components.
The overall recrystallization fraction against time in logarithmic scale depicted in the
small box of these figure shows a two-stage JMAK n behavior. The apparent derivation
occurs as the recrystallization increases up to 90%. The derivation is more obviously
seen in recrystallization rate in point A (Fig. 12b) where is subjected to a smaller
deformation. This is the contribution of the Cube texture component, which has been
predicted to have a sluggish recrystallization kinetics.
(a)
(b)
Fig. 12. The predicted recrystallization kinetics in point C (a) and in point A (b) at
annealing temperature 390C, thin curves: from leftmost to rightmost, the
recrystallization kinetics in grains with Copper, S, B, Random, Goss and Cube. Square
pointed curve: the overall recrystallization kinetics.
0
0.2
0.4
0.6
0.8
1
0 400 800 1200 1600
Annealing time (s)
F
r
a
c
t
i
o
n
-4
-3
-2
-1
0
1
2
3.5 4.5 5.5 6.5
ln time (s)
ln
ln
(
1
/
(
1
-
x
)
)
0
0.2
0.4
0.6
0.8
1
0 400 800 1200 1600 2000 2400
Annealing time (s)
F
r
a
c
t
i
o
n
-5
-4
-3
-2
-1
0
1
2
4 5 6 7 8
ln time (s)
ln
ln
(
1
/
(
1
-
x
)
)
Chapter 6 108
Fig. 13 shows the effect of the concurrent recovery on the recrystallization kinetics. The
effect of concurrent recovery on the recrystallization kinetics decreases as annealing
temperature increases and is slightly affected by the degree of the deformation.
(a)
(b)
Fig. 13. The predicted recrystallization kinetics in point C (a) and in point A (b) at 3
different annealing temperatures, cross-dotted curves: no concurrent recovery, circle
solid curves: with concurrent recovery.
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
Annealing time (s)
F
r
a
c
t
i
o
n
450C 390C 340C
10
0
10
5
10
4
10
3
10
2
10
1
10
-1
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
Annealing time (s)
F
r
a
c
t
i
o
n
450C 390C 340C
10
-1
10
4
10
3
10
2
10
1
10
0
10
5
Fig. 14 is a comparison of the prediction by the single grain model with the
experimental data at three annealing temperatures, together with JMAK fit. It can be
seen that the single grain model yields a good prediction of the experimental data in the
whole range of the recrystallization.
Fig. 14. The comparison of the experiment with the predicted recrystallization kinetics
at 3 different annealing temperatures, solid points: experiment in point C, empty points:
experiment in Point A. thin solid curve: JMAK fit, Point dotted curves: prediction by
single grain model.
6.6. Conclusions
1. The static softening kinetics of a commercial aluminum alloy AA1050 following hot
deformation includes both recovery and recrystallization. The rate of the static
restoration increases with increasing local equivalent strain and strain rate, i.e. with
increasing local equivalent stress.
2. The recovery kinetics could be modeled by the logarithmic decay relationship. The
contribution of the recovery to the softening kinetics decreases as the annealing
temperature increases.
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
Time (s)
F
r
a
c
t
i
o
n

r
e
c
r
y
s
t
a
l
l
i
z
e
d
450C 390C 340C
10
-1
10
5
10
4
10
3
10
2
10
1
10
0
Chapter 6 110
3. The nucleation site density was found to be a strong function of the local strain and
stress but weakly affected by the annealing temperature.
4. A relatively novel approach of the single grain model was applied to model the
kinetic aspects of recrystallization. Following this method, the recrystallization kinetics
was related to the evolution of microstructure and the grain geometry. The effect of the
concurrent recovery and the textural components on the recrystallization can be
included.
References
2. E. Nes, H.E. Vatne, O. Daaland, T. Furu, R. Orsund and K. Marthinsen, in 'The 4th
International Conference on Aluminium Alloys', 2000, 18-49.
[TMP]
2
35-51.
5. S.P. Chen, D.N. Hanlon, S. van der Zwaag, Y.T. Pei and J.Th.M. de Hosson, J. Mat.
Sci., 37, 2002, 989-995.
7. E.M. Lauridsen, D.J. Jensen, H.F. Poulsen and U. Lienert, Scr. Mat., 43, 2000, 561-
566.
8. L. Delannay, O.V. Mishin, D.J. Jensen and P. van Houtte, Acta Mater., 49, 2001,
2441-2451.
9. S.P. Chen, I. Todd and S. van der Zwaag, Metall. Mater. Trans. A, 33A, 2002, 529-
537.
10. S.P. Chen and S. van der Zwaag, Metall. Mater. Trans. A, submitted, 2002.
11. F.J. Humphreys, Acta Mater., 45, 1997, 4231-4240.
12. H.K.D.H. Bhadeshia, In 'Worked examples in the geometry of crystals', London,
The institute of metal, 1987.
13. E. Nes, Acta Metall. Mater., 43, 1995, 2189-2207.
14. C.M. Sellars, Iron. and Steelm., 22, 1995, 459-464.
15. Marc, in 'Users' manual, Marc Analysis Research Corporation', Palo Alto, CA,
1996.
16. D.R. Barraclough and C.M. Sellars, Met. Sci., 13, 1979, 257-268.
17. E.E. Underwood, in 'Quantitative Microscopy', McGraw-Hill Book Co.,N.Y., 1968.
18. W.A. Johnson and R.F. Mehl, Trans. Am. Inst. Min (Metall. ) Eng., 135, 1939, 416-
443.
19. G.R. Speich and R.M. Fisher, in 'Recrystallization, Grain Growth and Textures', H.
Margolin (Eds), ASM, Cleveland, OH, 1966, 563-598.
20. E. Nes and T. Furu, Scr. Met. Mat., 33, 1995, 87-92.
21. G.M. Hood, Phil. Mag. A, 21, 1969, 305-324.
23. C.M. Sellars, A.M. Irisarri and E.S. Puchi, in 'Microstructural Control in Alminium
Alloys: Deformation, Recovery and Recrystallization', E. Henry Chia and H.J.
McQueen (Eds), The Metallurgical Society, Inc., 1985, 179-196.
Chapter 6 112
24. K.D. Vernon-Parry, T. Furu, D.J. Jensen and F.J. Humphreys, Mater. Sci. Techn.,
12, 1996, 889-896.
25. H.J. Bunge, in 'Texture Analysis in Materials Science', London, Butterworths, 1982.
26. C.V. Thompson, H.J. Frost and F. Spaepen, Acta Metall., 35, 1987, 887-890.
S.P. Chen
Chapter 7
Effect of microsegregation and dislocations
on the nucleation kinetics of precipitation in
aluminium alloy AA3003
Nucleation kinetics of the precipitation of MnAl
6
and MnAl
12
in the aluminum alloy
AA3003 alloy has been investigated experimentally and theoretically. The results show
that cold rolling enhances the rate of precipitation and this effect increases as the
annealing temperature decreases. Micro-segregation of the Mn solute atom at the
dislocation network during cold deformation is found to have a significant effect on the
nucleation kinetics of the precipitation of MnAl
6
and MnAl
12
in AA3003 in addition to
the effect of dislocations, which increase the nucleation site density and reduce the
nucleation barrier. A model to predict the start times of the precipitation during
isothermal holding is constructed by considering the effects of dislocations and recovery
as well as micro-segregation of Mn on the nucleation kinetics of precipitates. The
predictions of the model are in good agreement with the experiment data.
7.1. Introduction
Generally, AA3003 alloys contain a mixed particle structure consisting of a coarse
distribution of large intermetallic inclusions due to the casting and a fine dispersion of
small particles precipitated during subsequent hot rolling and annealing treatments.
Unlike in the precipitation-hardening alloys such as Al-Cu, where homogenous
precipitation can take place, the precipitation in the dispersoid-containing non-heat-
Chapter 7 114
treatable Al-Mn alloys only occurs heterogeneously. Therefore, the precipitation
processes will be influenced by defects related to deformation and softening processes.
While the decomposition of the supersaturated solid solution of the undeformed binary
alloy Al-Mn [1,2] and AA3003 [3,4] has been studied extensively, relatively few reports
document the decomposition kinetics from the deformed state. Limited data show that
the precipitation rate in deformed material is appreciably higher and the MnAl
6
phase is
formed at lower temperature [5]. However, since recovery, recrystallization and
precipitation overlap and interact strongly, the general qualification of the effect of the
deformation and softening processes on the precipitation kinetics, especially on the
initial stage of the process, is complicated.
The effect of deformation on precipitation kinetics has often been attributed to the
increase of dislocation density, which increases both the nucleation site density and
diffusivities of precipitation-forming elements in the material [6,7]. However, as will be
seen in the current experiment, the precipitation process can still be promoted even
when the dislocations formed during deformation have largely been eliminated before
the precipitation starts. Consequently, the effect of deformation on the precipitation can
not be explained from dislocation density consideration only.
Many experimental observations have suggested that deformation can cause the
redistribution of the precipitate-forming elements between the dislocation cell walls and
the cell interiors [8-10]. Therefore, segregation of the solute atoms to the potential
nucleation sites may contribute to precipitation kinetics and provide an explanation for
the observed acceleration of the precipitation in a recovered structure.
In this chapter the early stages of the precipitation kinetics in an AA3003 alloy has been
investigated. The decomposition kinetics is monitored by conductivity measurements
and microstructural analysis. An Avrami type equation is applied to analyze the
precipitation kinetics. Thereafter, the precipitation-time-temperature relationship (C-
curve) is established. The contributions from the dislocations, recovery and micro-
segregation of solute to the nucleation kinetics of the precipitation are assessed by using
a model on the basis of classical nucleation theory. The model predictions and the
experimental observations are compared and a good agreement is obtained.
7.2. Mathematical description of the C-curve
In AA3003, precipitation only occurs heterogeneously by nucleation at high energy
defect sites such as grain boundaries, inclusions and dislocations. Thus, the precipitation
sequence is strongly influenced by deformation and recovery. In this paper, classical
nucleation theory is developed to take these effects into account. Assuming that the
Effect of microsegregation and dislocations on the nucleation kinetics 115
grain boundary and dislocation nucleations are the main mechanisms, the total steady
state nucleation rate can be written as:
I I I = +
gb dis
(1)
where I
gb
and I
dis
are grain boundary nucleation rate and dislocation nucleation rate
respectively. In the spirit of classical nucleation theory the steady state nucleation rate
per unit volume I can be expressed as [11]:
2 *
( / ) exp( / ) I N C a D G RT ~ A
v
0 eff
(2)
where G
-
A is the energy barrier for nucleation (J/mol). N
v
, the number of available sites
for the nucleation. a, the lattice parameter of the matrix phase.
0
C , the solute
concentration, T, the temperature, R, the gas constant, and D
eff
, the diffusion coefficient
of the rate controlling element,
eff 0 d
exp( / ) D D Q RT = , with D
0
, the pre-exponential
factor, Q
d
, the activation energy for diffusion.
During the early stages of a precipitation reaction, the reaction rate is controlled by the
nucleation rate. At that stage, the number of the nuclei per unit volume as a function of
annealing time t is given by:
0 0
t t
N I dt I dt = +
) )
gb dis
(3)
The time, t
*
, taken to precipitate a certain number
*
N of particles of the new phase for
nucleation to be detected can be determined from Eq. (3).
Application of the above nucleation theory to predict the precipitation-time-temperature
relationship requires the physical parameters,
v
N ,
*
G A and D
eff
under two different
nucleation mechanisms. In the following we will formulate a model which takes into
account the effect of deformation and softening. The determination of the relevant
parameters will be specified in section 5.2.1.
7.2.1. Activation energy for nucleation
For the case of grain boundary nucleation, the nucleation barrier for a grain boundary
nucleus with two abutted spherical caps is [12]:
* *
hom
( ) G G S u A = A
gb
(4)
where
2
( ) (2 cos )(1 cos ) / 2 S u u u = + and cos / 2
oo o|
u = with
o|
being the specific
energy of the o | interface,
oo
, matrix grain boundary energy,
*
hom
G A , the nucleation
barrier for homogenous nucleation.
The ratio of the nucleation energy on a dislocation to the homogenous nucleation energy
is [13]:
Chapter 7 116
*
hom
/ ( ) G G f o A A =
*
dis
(5)
where
2
2 / G
o| o|
o t = A , with
2
/[4 (1 )] Gb t u = for edge dislocations
2
/ 4 Gb t =
for screw dislocations, G is the elastic shear modulus, b the Burgers vector and u the
Poisson ratio. The value of ( ) f o , which was numerically shown in Fig. 3 in reference
[13], can be approximately expressed as:
0.58
( ) 1 f o o = (6)
The nucleation barrier,
*
hom
G A , for a spherical nucleus in an undeformed matrix, is given
as [11,12]:
3 3
*
hom 2
16
3 ( )
G
G G
o|
o|
t
A =
A + A
E
(7)
where G
o|
A is the chemical driving force, G A
E
, strain energy resulting from lattice
distortion, is a modifier of
o|
, a measure of the interaction energy between the
surface of the nucleus with the surrounding dislocations, which falls between 0 and 1.
For the precipitation of an intermetallic phase such as MnAl
6
(|) from aluminium with
manganese in solid solution , the chemical driving force per unit volume of precipitate,
G
o|
A (J/mol), is given by [14]:
0
ln( / )
1
C C
RT
G C C
C V
| o
o| o
o
A =
m
(8)
where R is the universal gas constant, T is the isothermal annealing temperature, V
m
the
molar volume of the | -phase, C
0
, the initial concentration of Mn in the solid solution,
C
, the equilibrium concentrations of Mn at the annealing temperature. C
, the
composition of the intermetallic | (MnAl
6
).
The interfaces of
6
/ MnAl o and
12
/ MnAl o are of a semi-coherent nature, with the
orientation relationship being listed in Table 1 [15]. The misfit between the two lattices
is defined as ( ) / a a a
| o o
A = , with a
o
and a
|
being the lattice parameters of the
matrix and precipitates respectively. The interfacial energy of a semi-coherent interface
can be written as [14]:
c s o|
= + (9)
where
c
is a chemical contribution from the coherent portion of the boundary,
s
, a
structural part due to the dislocations which depends on the misfit between the two
lattices and for a small A:
1 s
C = A (10)
where C
1
is a constant. In this study
c
and
12
MnAl
are assumed to be 0.2 J/m
2
and
0.22 J/m
2
[12], respectively, and
6
MnAl o
is calculated from Eqs. (9) and (10) by applying
data from Table 1, to be 0.26 J/m
2
.
The value of strain energy
E
G A associated with the semi-coherent interface can be
estimated directly from G
o|
A as [16]:
1.5
0
( / ) 1 /
E
G G
o| o|
= + A A (11)
where
0
is the specific energy of an incoherent interface, which is taken as 0.5 J/m
2
[12].
Table 1. The orientation relationship between the precipitates and Al matrix and the
misfit between the two lattices
Precipitate Orientation relationship disregistry
12 MnAl Al
(310) //(111)
1.88% MnAl
12
12 MnAl Al
[111] //[310]
1.88%
6 MnAl Al
(001) //(315)
5.69% MnAl
6
6 MnAl Al
[100] //[130]
1.68%
7.2.2. Nucleation site density
Assuming that the heterogeneous nucleation sites at grain boundaries are positioned
within a plate of thickness, o
b
, around the grain boundary, the number of heterogeneous
nucleation sites at grain boundaries, N
gb
, relative to the total number of nucleation sites
per unit volume for homogenous nucleation within the matrix,
hom
N , is [12]:
hom
N
N d
o
=
gb
b
g
(12)
where d
g
is the mean grain diameter and o
b
is the thickness of the grain boundary.
Assuming that the number of available sites for nucleation along dislocations is equal to
the number of atoms in the core of the dislocations, then heterogeneous nucleation
density on dislocations, i. e. the number of atoms per unit volume, N
dis
associated with
dislocations can be written as [17]:
Chapter 7 118
2
hom
N
r
N
t =
dis
c
(13)
where is dislocation density, r
c
is the core radius of the dislocation.
7.2.3. The evolution of dislocation density
Dislocations in a deformed material are assumed to arrange into cellular structures. The
cell size after deformation, o , can be estimated as [18]:
0.35 0.17 / o c = + (14)
where c is the true strain. The dislocation density in the deformed state, , can be
obtained by applying the principle of similitude [19]:
0.5
c o
=

(15)
where c
o
is a constant about ~5. On the early stages of the annealing process the
fraction residual strain hardening can be described by [20]:
1 ln(1 )
RT t
R
A t
= +
rec
(16)
where
0
exp
Q
RT
t t
| |
=
|
\ .
rec
, A and
0
t , constants, Q
rec
, the activation energy for recovery.
The softening fraction, which is mainly from recovery decay on the early stage of the
annealing process, can be obtained from the microhardness measurement as:
H H
R
H H
rex
rec
def rex
(17)
where H
def
is the hardness of the material in the as-deformed state, H , the instantaneous
hardness after annealing time t , H
rex
, the hardness of completely recrystallized
material.
By assuming that the value of microhardness is proportional to
1/ 2
, the dislocation
density change during softening can be determined from Eqs. (16) and (17).
7.2.4. Effect of micro-segregation of Mn
The redistribution of the precipitate-forming elements between the dislocation cell walls
and the cell interiors during deformation may result in an increase of local
supersaturation, which will affect the thermodynamics of precipitation. As the
mechanism for the segregation is not completely understood yet, in this paper we apply
the solute atmosphere theory [21,22] to quantitatively describe the solute distribution
around dislocations in order to discuss the effect of the segregation on the precipitation
process.
The extent of solute enrichment around a moving dislocation depends on the velocity of
solute atom diffusion relative to the dislocation velocity at the deformation temperature
and the difference between the molar volume of the solute atom and that of the matrix
atom [21,22]. As an approximation, for a dilute solution the concentration of solute
around a dislocation, C
dis
, when the dislocation velocity approaches zero, is given by:
( )
0
exp ( ) / C C p V V RT =
dis s m
(18)
where
(1 ) sin
3 (1 )
Gb
p
r
u |
t u
+
=
, V
m
and V
s
are the partial molar volumes of matrix and
solute elements, respectively, G, the shear modulus, b, Burgers vector, u , Poisson's
ratio, r and | are the polar coordinates with respect to the core of the moving
dislocations,
0
C , the concentration of the solute far away from a dislocation.
7.3. Experimental
The actual chemical composition of AA3003 has been determined using spectroscopic
analysis, and is given in Table 2. The starting material was provided in the transfer state
condition (26 mm thick hot-rolled plate). Two pieces of the plate were solution treated
at 630 for 15 h and water quenched to room temperature. Samples I were directly
annealed isothermally between 550 and 350C in a salt bath with the temperature
controlled to within 1C. Samples II were cold rolled with a 50% reduction in thickness
before applying the same isothermal annealing conditions. These treatments produced
samples with essentially two different microstructures: sample I was highly
supersaturated with Mn (~0.68 wt. %) and sample II was highly supersaturated as well
as being strain hardened. The precipitation kinetics was monitored using conductivity
measurements. Sigmatest D2.068 was used to measure the conductivity at 252C. All
the specimens were carefully prepared to have the same thickness. The surface of the
specimen was polished to make good contact with the conductivity detector. 20
measurements were taken for each experiment to give an averaged value. The accuracy
of the measurement is within 1% MS/m. The distribution of the undissolved dispersoids
and the grain structures of all the samples were examined by quantitative optical
metallography. The recovery and recrystallization processes were followed by
microhardness measurements and optical metallography. The micro-hardness was
determined using 50 g load and 15 s loading time on a Buehler OMNIMET MHT
automatic micro hardness tester. An experimental point represents an average value of
18 measurements.
Chapter 7 120
Table 2. The chemical composition of the material studied
Si Fe Cu Mn other Al
AA3003 <0.30 <0.56 <0.25 1.10 <0.15(total) bal
7.4. Results
7.4.1 The evolution of microstructure during homogenization
The optical microstructure of the transfer gauge shows an incompletely recrystallized
structure. A fully recrystallized structure was obtained after homogenization at 630C
for 10 min. However, the recrystallized grains were still elongated along the rolling
direction with an aspect ratio l/w~3-5. The average grain size (equivalent diameter from
area measurement) was determined as 150 m. The grain size seems not to increase
with increasing homogenization time up to 15 h owing to the pinning effect of the
second phase particles.
There are two kinds of intermetallic particles in the transfer gauge material,
| -(Fe,Mn)Al
6
or MnAl
6
particles (rod-like with a aspect ratio >5, size >2 m) and
o -(FeMn)SiAl (spherical, size between ~1 m and 4 m). The rods of (Fe,Mn)Al
6
are
oriented in the rolling direction.
The small particles (less than 1.0 m) dissolve with increasing time during
homogenization. However, the dispersion of insoluble particles of (Fe,Mn)Al
6
and
(FeMn)SiAl did not change during annealing at 630C while the rod-like | particles
became more or less spherical. Application of 50% cold rolling deformation to the
homogenized samples did not change the size of the undissolved particles and their
distribution.
The electrical conductivity of AA3003 in the transfer state is measured as 24.8 MS/m. It
decreases rapidly with homogenisation time (t=30 min) and then approaches a constant
value of 19.96 MS/m, which indicates that the solid solutions of (Mn, Fe, and/or Si) in
the matrix reach the equilibrium value. After applying cold rolling to 50% thickness
reduction to the homogenized samples, the final conductivity of the material became
20.15 MS/m.
7.4.2 The evolution of the conductivity and precipitation during
isothermal annealing
Fig. 1 shows the conductivity evolution of the samples for both treatments (no
deformation and with deformation) during isothermal annealing at temperatures
between 350C and 550C. As can be seen, the conductivity increases rapidly at the
very beginning and the rate slows down gradually with annealing time. The
conductivity of the cold rolled samples increases faster than that of the non-strained
sample. This indicates that strain promotes the precipitation processes. The effect of the
cold rolling on the precipitation increases as the annealing temperature decreases.
(a) (b)
Fig. 1. The conductivity evolution of the samples after homogenization at 630C, (a)
without strain; (b) 50% cold rolling, then annealed at various temperatures.
(a) (b)
Fig. 2. The microstructure of the precipitates after annealing at 400C for 35 min, (a)
without strain; (b) 50% cold rolling. (Coarse particles-undissolved, fine particles-
precipitates)
19
20
21
22
23
24
25
26
1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
Time (min)
C
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
350C
400C
450C
500C
550C
10
-2
10
4
10
3
10
2
10
1
10
0
10
-1
20
21
22
23
24
25
26
27
28
29
30
1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
Time (min)
C
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
350C
400C
450C
500C
550C
10
-2
10
-1
10
1
10
4
10
3
10
2
10
0
Chapter 7 122
The optical microstructures after annealing at 400C for 30 min are shown in Fig. 2.
The difference between the two types of samples is obvious. In the samples without
strain, the precipitation starts at the original grain boundaries, as shown in Fig. 2(a). As
annealing time increases, more and more precipitation takes place within the original
grains and the number and size of precipitates increase with annealing time. However,
in the deformed samples, fine precipitates are more or less uniformly distributed in the
matrix (Fig. 2(b)). This indicates that the heterogeneous nucleation sites provided by
dislocations now play an important role.
7.4.3. The isothermal precipitation kinetics
The electrical conductivity is the reciprocal of the resistivity. It is well established that,
for low solute concentration, the resistivity of an alloy, _ , is given by the following
relation [23],
' '
( )
i i i i
C C _ _ o o = + +
_ Al
(19)
where
6
2.64 10 _

=
Al
cm is the electrical resistivity of the pure matrix.
i
C and
'
i
C
the respective concentration of solute atoms.
i
o and
'
i
o are the specific resistivities of
the ith elements in solid solution and in precipitated form, respectively, which are listed
in Table 3. Metals in solid solution promote the resistivity to a greater extent than when
out of solution.
Using Eq. (19) and Table 3 the contribution that each alloying element makes to the
resistivity can be predicted. Because of the low solid solubility of Fe in aluminium and
the relatively low total concentration of Si, their contribution to the resistivity change
during heat treatment can be neglected in comparison with the one due to the change of
the Mn concentration. Therefore, the analysis of the conductivity curves can be based
on an analysis of the quasi binary Al-Mn system, taking into account that the addition of
the elements Fe and Si to the pure Al-Mn alloy reduces the solid solubility of Mn in Al
[4,24]. The solubility of Mn in the solid solution corresponding to the equilibrium phase
MnAl
6
in the current Al-Mn-Si alloy as a function of temperature is shown in Fig. 3
[25,26].
Table 3. Effect of the additions on the resistivity of pure Al at 20C
Element Mn Fe Si
i
o (cm/wt.%)
2.94 2.56 1.02
'
i
o (cm/wt.%) 0.34 0.058 0.088
Max solubility,wt.% 1.55 (658C) 0.052 (655C) 1.65 (580C)
Fig. 3. The solubility of Mn in Al and the saturation conductivity as a function of
temperature.
The precipitation fraction can now be determined using:
0 0
0 0
t t
t
x
_ _
_ _

= =

(20)
where
t
is the conductivity at annealing time t,
0
, the conductivity of the samples
before annealing and
, the saturation level of the conductivity of the fully

precipitated samples at long time, which corresponds to the equilibrium content of Mn
in solid solution and increases with decreasing temperature.
i
_ is the corresponding
resistivity.
can be calculated from equation (19) by inputting the solubility of Mn

from Fig. 3 and other data from Table 2 and Table 3. The graphical presentation of the
against temperature is shown in Fig. 3, too. The precipitation fraction as a function

of the annealing time is shown in Fig. 4.
It is supposed that the precipitation process can be described by the JMAK equation, i. e
[12].
( )
1 exp
n
x kt = (21)
By applying Eq. (21) to the experimental data, the parameters, k and n are determined
by iteration on the method of least squares and are listed in Table 4. n is close to 0.5
indicating that the precipitation is diffusion controlled process. Application of the
deformation results in a decrease in n but a increase in k. Increase in k indicates that
0
0.2
0.4
0.6
0.8
300 350 400 450 500 550 600 650 700
Temperature (C)
S
o
l
u
b
i
l
i
t
y

M
n

(
w
t
%
)
19
21
23
25
27
29
31
33
S
a
t
u
r
a
t
i
o
n

c
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
Chapter 7 124
cold deformation reduces the nucleation barrier of the precipitation. Decrease in n is due
to the concurrent recovery and recrystallization processes, which weakens the effect of
the dislocations. The relation between ln k and (1/T) is not linear but shows a kink,
which indicates two precipitation reactions taking place in the temperature range
studied.
Table 4. Fitting parameters n and k for the precipitation kinetics using JMAK equation
undeformed 50% Cold rolling Temperature
n k n k
350 0.40 0.008 0.39 0.058
400 0.42 0.018 0.35 0.091
450 0.39 0.036 0.37 0.065
500 0.42 0.030 0.34 0.107
550 0.37 0.055 0.30 0.159
(a) (b)
Fig. 4. The precipitation fraction as a function of the time, annealed at various
temperatures, (a) without strain; (b) 50% cold rolling.
If the beginning of the precipitation is defined as 0.05 x = transformed, the time-
temperature-transformation representing nucleation kinetics of the precipitation reaction
can be constructed by solving Eq. (21) at 0.05 x = for each temperature. The resulting
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100 1000 10000
Time (min)
F
r
a
c
t
i
o
n
550C
500C
450C
400C
350C
10
2
10
-1
10
-2
10
0
10
1
10
4
10
3
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100 1000 10000
Time (min)
F
r
a
c
t
i
o
n
550C
500C
450C
400C
350C
10
-2
10
-1
10
4
10
3
10
0
10
2
10
1
C-curves are shown in Fig. 5. In the TTT diagram of the non-strained samples, there is a
nose toward 450C and the other at higher temperatures, indicating two precipitation
reactions, which is in a good agreement with the literature [4,15]. At temperatures
higher than 500C, the equilibrium MnAl
6
precipitate forms directly. At lower
temperatures (T<500C), initially the metastable phase G
1
(MnAl
12
) is precipitated
which then transforms into MnAl
6
for long annealing times. In the cold deformed
material the precipitation starting time is appreciably shorter and MnAl
6
phase forms
even at lower temperatures. The character of the double nose is more noticeable.
Fig. 5. The temperature dependence of the start (t
0.05
) of the precipitation (C-curve).
7.4.4. The softening kinetics during isothermal annealing
The microhardness evolution after annealing at various temperatures is shown in Fig. 6.
In theory, the microhardness evolution should include the contributions from both
softening and precipitation. The microhardness is measured to be 352 HV for the
sample after 630C solution treatment for 15 h and quenching. After annealing at 400C
for various times such as 30 min, 15 h, 96 h, the microhardness values are 362, 352
and 352 respectively. They indicate that the microhardness is not sensitive to the
precipitation process. Therefore the microhardness measurement can be considered only
to reflect the softening behavior. The best fit of the equation (16) and (17) to the
experimental data yields the activation energy for recovery Q
rec
=220 kJ/mol, which is
approximately the activation energy of Mn diffusion in Al [24]. The other two
parameters, A and
0
t , are listed in Table 5.
300
350
400
450
500
550
600
0.1 1 10 100 1000
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
50%CR
No strain
MnAl
6
MnAl
12
10
1
10
3
10
2
10
0
10
-1
Chapter 7 126
Fig. 6. The microhardness as a function of annealing time in the samples with 50% cold
rolling.
Table 5. The fitting parameters A and
0
t for recovery
350C 400C 450 500C
A (kJ/mol) 210 204 198 195
0
t (1/s)
21
9.86 10
25
1.05 10
26
1.20 10
27
1.96 10
7.5. Discussion
7.5.1. Analysis of the experimental results
The enhancement of precipitation kinetics due to prior plastic deformation can be
attributed to the following reasons. Deformation increases the dislocation density, which
increases both the nucleation site density and diffusivities of precipitate-forming
elements in the material. Furthermore, the nucleation barrier on dislocations is much
lower than that for homogenous nucleation. In addition, the solute atmosphere around
dislocations enhances the rate of precipitation via an increase in the local chemical
driving force. The last point can be indirectly seen from the current experimental
results, and forms the basis of the new analysis of the precipitation reaction in AA3003.
As it is well known, applying deformation to the uniform solution will increase its
resistivity, i.e. decrease electric conductivity. However, in the present experiment the
30
35
40
45
50
55
60
65
1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
Time (min)
H
a
r
d
n
e
s
s

(
V
H
N
)
450
400
350C
500
10
-2
10
2
10
1
10
0
10
-1
10
4
10
3
conductivity of the deformed material shows a slight increase (compare Fig. 3a and
Fig. 3b), which corresponds to a decrease of 0.025 wt. % Mn in solute solution. This
indicates that solute atom redistribution has occurred during cold rolling. The Mn
segregation at dislocations results in a decrease of Mn concentration in the matrix,
leading to an increase of the conductivity.
In the deformed samples, the precipitation of MnAl
6
or MnAl
12
and recovery are
coupled phenomena during annealing at the temperature range studied here. Obviously,
when annealing at higher temperatures, some of dislocations formed during cold rolling
can be eliminated before the precipitation starts as a result of concurrent restoration
process. Therefore, the effect of the cold rolling on the precipitation kinetics is likely to
decrease with annealing temperature, as is indeed observed. In the sample annealed at
550C, the deformed samples fully recrystallize within about 12 s, while the
precipitation starts at 15 s. As the time to the complete recrystallization is less than the
start time of precipitation, dislocations formed during cold rolling have been eliminated
completely before precipitation starts. It would be expected that there would be no
enhancing effect of strain-induced precipitation in this case in the point of view that the
presence of the dislocation will increase the nucleation sites and diffusivites of the
precipitation-forming elements. However, it is of interest that the precipitation kinetics
is still promoted (Fig. 5). This, again, can be explained as being a result of the Mn
segregation effect. Precipitate-forming elements are enriched at dislocation cell walls
during deformation or after deformation as a result of segregation. While the dislocation
cell walls are swept out by moving boundaries of new recrystallized grains, the already
established concentration distribution during deformation is unlikely to be altered. Thus,
precipitation may still take place at regions of prior dislocation cell walls where the
higher local supersaturation of precipitate-forming elements is retained for a sufficiently
long time.
In the following section the effect of the initial and transient dislocation density and Mn
segregation on the nucleation kinetics of precipitation will be further analyzed by means
of the theoretical predictions.
7.5.2. Application the model to the experiment data
1. Parameters determination
The physical parameters used in the prediction from literature and other constants
determined from the tuning are listed in Table 6. The diffusion coefficient of Mn in Al
reported in the literature shows a very large scatter [24,27,28]. Therefore, in this study
the activation energy Q
d
is firstly evaluated from the experimental C-curve obtained in
this work according to the procedure described by Ryum quoted in [29]. The Q
d
is
determined from Fig. 5, by dividing the slope of the lower asymptote by gas constant R,
Chapter 7 128
respectively, as 223 kJ/mol and 198 kJ/mol from the non-strained C-curve and strained
C-curve. Hence the diffusion coefficient of Mn in Al in the literature closest to these
value,
d
211 Q = kJ/mol [27] is applied in the prediction.
The solvus boundary for metastable phase MnAl
12
in aluminium needs to be
determined. A mathematical description of the solvus boundaries of the equilibrium
precipitate MnAl
6
is given as [29]:
eq 6
ln[% ] [% ]
H
T
S R Al Mn
A
=
A
(22)
where S A and H A are the standard entropy and enthalpy of the reaction. [%Al] and
[%Mn] are the matrix concentrations of elements Al and Mn in wt%. T
eq.
is

the solvus
temperature of MnAl
6
. By fitting Eq. (22) to the literature data (solid line in Fig. 3) the
values of S A and H A are determined as 25 JK/mol and 65 kJ/mol, respectively.
Assuming that the thermodynamic properties of the metastable MnAl
12
phase are
similar to those of the equilibrium MnAl
6
phase and taking into account the extra
pressure on the particle from the curvature of the interface of the precipitate interface in
the solvus enthalpy, the solvus temperature of metastable equilibrium, T
met
, can be
expressed as:
met. 6
ln[% ] [% ]
H
T
S R Al Mn
A O
=
A
(23)
where O is the contribution of the interface curvature to the reaction enthalpy, given by:
m
2 / V r O = (24)
where
m
V is the molar volume of the precipitate, r the radius of the curvature, the
interfacial energy of the particle. The ratio between Eqs. (22) and (24) is:
met.
eq
T H
T H
A O
=
A
(25)
which is independent of temperature and composition, and only dependent on the type
and shape of precipitate. By setting the average radius 1 r = nm and 0.26 = Jm
-2
,
then:
met.
eq
0.90
T
T
= (26)
The temperature dependent chemical driving force, G
o|
A , and the strain energy term,
E
G A can be determined from Eqs. (8) and (11), respectively, and from these values the
nucleation barriers for grain boundary nucleation and for dislocation nucleation can then
be calculated from Eqs. (4) to (7). The total nucleation site density takes the sum of the
gb
N and
dis
N which could obtained from Eqs. (12) and (13), respectively. All these
values can be put into Eq. (3), which could be solved numerically, to obtain the
0.5
t .
Table 6. Parameters and variables employed in the model prediction
Symbol value definition
a 0.40496 nm Lattice parameter of Al
V
Al
5
10
m
3
/mol Molar volume of Al
V
6
MnAl
6
9.68 10
m
3
/mol Molar volume of MnAl
6
V
12
MnAl
6
9.10 10
m
3
/mol Molar volume of MnAl
12
d
g
150m The original grain size
b
o
0.8 nm [12] Grain boundary thickness
6
MnAl o
0.26 J/m
2
Specific energy of
6
/ MnAl o interface
12
MnAl o
0.22 J/m
2
[12] Specific energy of
12
/ MnAl o interface
oo
0.324 J/m
2
[14] Matrix grain boundary energy
r
c
0.286 nm The core radius of dislocation
Q
d
211 kJ/mol [27] Activation energy for Mn diffusion
in Al
D
0
3
4.934 10
m
2
/min Pre-exponential factor of Mn
diffusivity in Al
0
10
6
/mm
2
Dislocation density in the fully
recrystallized state
Q
rec
220 kJ/mol Activation energy for recovery
G 25.4 GPa Shear modulus
b 0.286 nm Burgers vector
u 0.33 Possions ratio
0.16
modifier
Chapter 7 130
The C-curve of the unstrained case is now used as a reference to adjust the constants,
(in Eq. (7)) and
0
D in the model. As shown in Fig. 7, in the non-strained case the grain
boundary nucleation is dominant mechanism because of lower nucleation barrier and
relatively higher nucleation site density on the grain boundaries. The nose in the C-
curves of dislocation nucleation appears at lower temperature than that of grain
boundary nucleation. This is because that the nucleation barrier at a grain boundary is
lower than that on dislocations. On the other hand, since o is proportional to G
o|
A ,
which decreases with increasing temperature, the effectiveness of a dislocation to
catalyze nucleation decreases as temperature increases.
Fig. 7. Application of the model to the experimental C-curve of the case without
deformation to determine parameters, and
0
D .
2. Effect of the dislocation density, recovery and Mn segregation on the C-Curve
The effect of the dislocation density on the precipitation kinetics is shown in Fig. 8. As
the dislocation density increases, the nose of the C-curve shifts to the left. This is
because of the strong increase in the nucleation site density produced by dislocations
and as a result, the dislocation nucleation gradually becomes dominant.
The effect of the softening reaction on C-curve can be also seen from the dotted line in
Fig. 8. Here we assume that the softening follows Eq. (16) and when it is fully
recrystallized the dislocation density is fixed at 10
6
/mm
2
. As can be expected, softening
reaction reduces the dislocation density, and therefore, shifts the upper part of the C-
300
350
400
450
500
550
600
650
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
MnAl6 MnAl12 Experimental
GB =10
6
/mm
2
10
-1
10
5
10
3
10
1
10
7
curve to the right and this effect decreases as annealing temperature decreases because
of the sluggish recovery process in this alloy.
The segregation increases the Mn concentration near the dislocations and/or grain
boundaries, and therefore, increases the driving force for precipitation. Fig. 9 shows that
the effect of the different segregation level of local Mn content on the C-curves. The
bulk concentration of Mn is 0.68 wt. %. As can be seen, Mn segregation displaces the
nose of the C-curve towards higher temperatures and shorter times. The enhancing
effect of the segregation is stronger at higher temperatures.
Fig. 8. The effect of the dislocation density and softening restoration on the C-Curve of
the precipitation. Solid lines: dislocation effect, dotted lines: dislocation effect and
softening effect, dash-dotted lines: grain boundary nucleation.
250
300
350
400
450
500
550
600
650
1.E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
=10
13
/mm
2
10
11
10
9
10
6
GB
GB
MnAl
6
MnAl
12
10
-4
10
4
10
2
10
0
10
-2
10
6
Chapter 7 132
(a) (b)
Fig. 9. The effect of the strain-induced microsegregation of Mn on the C-Curve of the
precipitation.
3. Prediction of the C-Curve after the current deformation
The model was applied to the experimental data after cold rolling. In the as-rolled state
the dislocation density is calculated to be
9
5 10 = /mm
2
and the relative
concentration ratio of
0
/ C C
dis
Mn Mn
is ~1.35 from Eq. (18). Fig. 10 shows predictions of C-
Curves together with experimental data. The dotted curve only considers the
contribution from the dislocations. The double dotted curve is the one assuming
9
5 10 = /mm
2
and with Mn segregation from 0.69 to 0.90 (
0
/ C C
dis
Mn Mn
=1.35). The
solid black curve is the one in which all the effects, the dislocation density, Mn
segregation and softening are taken into account. As can be seen, the presence of
dislocations promotes precipitation reaction. However, only considering the increase in
the nucleation site density from dislocations does not lead to a satisfactory description
of the experimental data. The effect of the microsegregation of Mn is significant. The
prediction of the present model gives a good agreement with the experiments when all
the effects from dislocation density, Mn microsegregation and recovery are considered.
300
350
400
450
500
550
600
650
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
=10
11
/mm
2
MnAl
6
0.7wt.%Mn
0.8
0.9
1.0
1.2
GB
10
-3
10
3
10
1
10
-1
300
350
400
450
500
550
600
0.001 0.01 0.1 1 10 100 1000
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
MnAl
12
=10
11
/mm
2
0.7wt.%Mn
0.9
1.2
0.8
1.0
GB
10
-3
10
1
10
0
10
-1
10
-2
10
3
10
2
Fig. 10. The comparison of the prediction with the experimental C-curve of the case
with 50% cold rolling (CR). Dotted curve:
9
5 10 = /mm
2
, double dotted curve:
9
5 10 = /mm
2
and a segregation of Mn from 0.69 to 0.90 %wt. Solid curve: including
the effect of dislocation density, Mn segregation and recovery. Circle points and square
points: experimental C-curves of 50% CR and without strain, respectively.
7.6. Conclusions
1. Micro-segregation of the solute atom Mn at the dislocation network during cold
deformation has an apparent effect on the nucleation kinetics of the precipitation of
MnAl
6
and MnAl
12
in AA3003 alloy.
2. The present model can account for the effects of Mn micro-segregation on the
nucleation kinetics of the precipitation in addition to the contributions from the
dislocations and concurrent softening reaction.
300
350
400
450
500
550
600
650
0.01 0.1 1 10 100 1000
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
MnAl
6
MnAl
12
c=0.67, =5x10
9
/mm
2
10
-1
10
3
10
2
10
1
10
0
10
-2
Chapter 7 134
References
1. E. Nes, J.D. Embury, Z. Metallkde. 66 (1975) 589-593.
2. A.K. Jena, D.P. Lahiri, T.R. Ramachandran, M.C. Chaturvedi, J. Mat. Sci. 16 (1981)
2544-2550.
3. G. Hausch, P. Furrer, H. Warlimont, Z. Metallkde. 69 (1978) 174-180.
4. D.B. Goel, U.P. Roorkee, P. Furrer, Neuhausen, H. Warlimont, Aluminium 50
(1974) 511-516.
5. N.J. Luiggi, Metall. Mater. Trans. B 28 (1997) 125-133.
6. B. Dutta, E.J. Pamiere, C.M. Sellars, Acta Mater. 49 (2001) 785-794.
7. E.J. Pamiere, C.I. Garcia, A.J. DeArdo, Metall. Mater. Trans. A 25 (1994) 277-287.
8. T. Abe, H. Onodera, J. of Phase Equilibria 22 (2001) 491-497.
9. A.H. Cottrell, B.A. Bilby, Proc. Phys. Soc. (London) A62 (1949) 49-61.
10. H. Yoshinaga, S. Morozumi, Phil. Mag. A23 (1971) 1351-1356.
11. K.C. Russell, Adv. Colloid Interface Sci. 13 (1980) 205-318.
12. D.A. Porter, K.E. Easterling, Phase Transformations in Metals and Alloys, Stanley
Thornes (Publishers) Ltd, 1992.
13. J.W. Cahn, Acta Metall. 5 (1957) 169-172.
14. A.K. Jena, M.C. Chaturvedi, Phase Transformations in Materials, Englewood Cliffs,
New Jersey 07632, Prentice Hall, 1992.
15. L.F. Mondolfo, Aluminium Alloys: Structure and Properties, Butterworth and Co.,
London, 1976.
16. W.J. Liu, J.J. Jonas, Mater. Sci. Techn. 5 (1989) 8-12.
17. R. Gomez-Ramirez, G.M. Pound, Metall. Trans. 4 (1973) 1563-1570.
18. J. Gil Sevillano, P. van Houtte, E.A.D. Aernoudt, Pro. Mat. Sci. 25 (1980) 69-412.
19. C.M. Sellars, in: B. Hutchinson et.al (Eds), Thermomechanical Processing in
Theory, Modeling & Practice [TMP]
2
(1996) 35-51.
20. E. Nes, Acta Metall. Mater. 43 (1995) 2189-2207.
21. R. Fuentes-Samaniego, R. Gasca-Neri, J.P. Hirth, Phil. Mag. A49 (1984) 31-43.
22. H. Yoshinaga, S. Morozumi, Phil. Mag. A23 (1971) 1367-1385.
23. J.E. Hatch, Aluminium: Properties and Physical Metallurgy, American Society for
Metals, Metals Park, Ohio, 1988.
24. J.J. Theler, P. Furrer, Aluminium 50 (1974) 467-472.
25. M. Hasen, Constitution of Binary Alloys, Mc Graw Hill, New York, 1958.
26. F.A. Shunk, Costitution of Binary Alloys, Second supplement, Mc Graw Hill, New
York, 1969.
27. G.M. Hood, R.J. Schultz, Phil. Mag. A23 (1971) 1479-1489.
28. T.S. Lundy, J.F. Murdock, J. Appl. Phys. 33 (1962) 1671-1673.
29. O. Grong, Metallurgical Modeling of Welding, The Institute of Materials, 1994.
S.P. Chen
Chapter 8
On the precipitation and recrystallization
behavior in an AA3003 alloy following hot
deformation
An AA3003 alloy, pre-heat treated to produce two different starting microstructures,
was deformed in plane strain compression at strain rates in the range of 0.1 to 10 /s at
temperatures varying between 350C to 550C. As-deformed specimens were
subsequently annealed at different temperatures to study the effect of the hot
deformation parameters on the precipitation and recrystallization behavior. The solute
content of Mn in the matrix was found to have a large effect on the deformation,
decomposition and softening behavior of the alloy. The precipitation kinetics was
enhanced by deformation but weakened by recovery and recrystallization. Dynamic
precipitation during concurrent deformation depends on the strain rate and deformation
temperature. The softening kinetics is either fast or sluggish depending on the pre-heat
treatments, deformation conditions and the annealing temperature. When
recrystallization is slow the contribution of the recovery to the softening can be as high
as 70%. When recrystallization is fast recovery is responsible for only ~25% of the
softening. Recrystallization kinetics is not simply proportional to the value of the Zener-
Hollomon parameter. The fully recrystallized grain size increases as annealing
temperature decreases as a result of precipitation effect. Although particle stimulated
nucleation (PSN) is observed in the cold deformed samples annealed at higher
temperatures (>500C), nucleation by subgrain growth and strain induced boundary
migration (SIBM) are the dominant nucleation mechanisms in the hot deformed samples
annealed in the temperature range 450-540C.
Chapter 8 136
8.1. Introduction
It has been recognized that Mn is one of the most efficient elements to affect the
deformation and recrystallization behavior of Al [1]. This influence strongly depends on
the supersaturation of the alloy and on the size and distribution of the precipitate
particles present in the matrix [2-4].
In AA3003 alloy the precipitation of MnAl
6
and the recrystallization reaction during
thermo-mechanical treatments are coupled phenomena [5,6]. The competition of these
two processes determines the final microstructure, and therefore, the properties. The
presence of the second phase particles is well known to affect recrystallization kinetics
and grain size and shape. The coarse particles speed up recrystallization while fine
particles slow it down.
Numerous investigators have studied the precipitation kinetics and recrystallization
behavior in 3 alloys after cold working [2,7,8]. However, relatively few workers
have studied the concurrent Mn precipitation, or examined the potential interaction
between this precipitation and the matrix recrystallization during and after hot
deformation such as break down rolling. Hence, systematical studies on the
recrystallization kinetics in AA3003 after hot deformation are limited in number. As the
deformation temperature and strain rate clearly affect both the dislocation substructure
and precipitation kinetics, microstructure evolution due to warm deformation may be
quite different from that due to cold deformation. As concurrent precipitation strongly
affects the kinetics of the softening process, the content of the Mn in the matrix and
particle size distribution just prior to the warm deformation will, for a given
deformation condition, play an important, but as yet unclear, role.
The objectives of the present investigation are therefore: (1) to elucidate the effect of the
thermo-mechanical processing on precipitation and softening behavior of AA3003
following hot deformation, with particular attention being paid to the roles of: pre-heat
treatment; deformation temperature; strain rate; and annealing temperature. (2) To
provide basic experimental data for further modeling study on how the hot deformation
affects MnAl
6
precipitation and how the latter, in turn, influences the recrystallization
process.
8.2. Experimental
The chemical composition of AA3003 (in wt.%), determined using spectroscopic
analysis, was 0.30 Si, 0.56 Fe, 0.25 Cu, 1.10 Mn and bal. Al. The starting material was
provided in the transfer gauge state (26 mm thick hot-rolled plate) that had been DC
cast, homogenized, break-down-rolled and annealed. The transfer gauge material
On the precipitation and recrystallization behavior in an AA3003 137
(AA3003) was thermally pre-treated to produce two different initial levels of
manganese supersaturation and dispersoid distribution. One piece of the plate was
solutionised at 630 for 16h and water quenched to room temperature (designated by A:
630C16h). The other piece was directly annealed at 450C in a salt bath for 24 h to
produce a fully recrystallized structure and to reduce the Mn content in the matrix
(designated by B: 450C24h). These treatments produced samples with essentially two
different microstructures: sample A was highly supersaturated with Mn and sample B
was highly precipitated with a low level of Mn dissolved in the matrix.
To simulate the deformation that occurs during conventional industrial processing such
as hot rolling, specimens of both treatments were deformed in plane strain compression
using a Gleeble 3500 over the temperature range 350 to 550C. Rectangular PSC
specimens, dimensions 53010 mm
3
, were machined from the middle thickness of the
sheet. The elongation direction of the deformation in PSC samples is parallel to the
original rolling direction. A tool with width of 10 mm is used so that the ratio of the tool
width to the sample thickness is 2 to ensure almost uniform deformation in the sample.
Each sample was heated to the test temperature at a heating rate of 5 /s, held for 30 s at
the test temperature, deformed at a given strain rate to a strain of 0.70, then quenched by
water within 1 s to room temperature. The deformation temperatures were 350C,
450C and 550C. The strain rates used were 0.1 /s, 1 /s and 10 /s. All tests were
conducted in vacuum. A water based graphite lubricant was used to reduce the friction
between the specimen and the tools. As the Gleeble uses resistance heating, the contact
between the two platens and the specimen becomes crucial for obtaining a uniform
distribution of the temperature within the specimen. Hence, it is very important to
machine specimens precisely to the exact shape. The specimens were ground on
sandpaper to ensure the difference in thickness at both ends being less than 0.02 mm.
The temperature difference between the two ends of a sample is less than 5C. To show
the effect of the deformation temperature material heat treated according to schedule A
was also cold rolled to the same amount of strain of 0.7. Samples are labeled by
referring preheat treatment and deformation conditions. For example, A350C0.1/s
refers a case of treatment A, deformed at 350C with a strain rate of 0.1/s, CR refers
samples after cold rolling deformation.
Following the hot or cold deformation, specimens were isothermally annealed in a salt
bath for times ranging from 6 s to 500 h at temperatures of 450, 500 and 550C to study
the softening kinetics and decomposition kinetics.
For all samples the distribution of the undissolved dispersoids and the grain structures
were determined by quantitative optical metallography. The recovery and
recrystallization processes were followed by microhardness measurements and optical
Chapter 8 138
metallography. Specimens for optical metallography were anodized with an aqueous
solution 3% Barker agent at a potential of 15 V for about 100 s. the microstructure was
studied using polarized light. The micro-hardness was determined on a Buehler
OMNIMET MHT automatic micro hardness tester using 300 g load and 15 s loading
time. An experimental point in the graphs represents the average value of 18
measurements. Sigmatest D2.068 eddy current conductivity tester was used to measure
the conductivity at 202C. All specimens were carefully prepared to have the same
thickness. The surface of the specimen was polished to make a good contact with the
conductivity detector. 20 measurements were taken for each experiment. The accuracy
of the measurement is within 0.05 MS/m.
8.3. Results
8.3.1. The microstructural evolution during preheat treatment
The optical microstructure of the starting material showed an incompletely
recrystallized structure. A fully recrystallized structure was obtained after
homogenization at 630C for 2 min. In the fully recrystallized material, the grains are
still elongated along the rolling direction with an aspect ratio l/w~4-6. The average grain
size (equivalent diameter from area measurement) was determined as 125 m. The grain
size seems not to increase with increasing homogenization time up to 168 h due to the
pinning effect of the second phase particles.
There are two kinds of intermetallic particles present in the transfer gauge material,
| -(Fe,Mn)Al
6
or MnAl
6
particles (rod-like with a aspect ratio >5, size >2 m) and
o -(FeMn)SiAl (spherical, size between ~1 m and 4 m). The (Fe,Mn)Al
6
particles are
oriented in the rolling direction.
The small particles dissolve with increasing time during homogenization at 630C.
However, the size distribution of insoluble particles of (Fe,Mn)Al
6
and (FeMn)SiAl did
not change during annealing while the rod-like | particles became more or less
spherical. Therefore material with treatment A has a distribution of coarse particles.
Table 1 shows the effect of homogenization time on the number of particles, their size
and aspect ratio and the area fraction. Particles with a size smaller than 1m are not
included in these statistics.
The average grain size and the aspect ratio of the grain after treatment B are almost the
same as those after treatment A. There are two kinds of particles in the samples with
treatment B. The coarse particles (>0.5 m) are those already present in the transfer
state which have undergone a further growth process during subsequent treatments. The
fine dispersoids are those produced by precipitation and growth during further
isothermal annealing at 450C. The average size of the coarse particles is 1.95 m. The
average size of the fine dispersoids after a similar treatment is about 0.12 m according
to Hansen at al [9].
For the modeling purpose of the PSN effect it is necessary to characterize the large
particle size distribution. The larger particle size distribution observed in a two-
dimensional section is generally in the form of [10]
( ) exp( ) f H L q q q = (1)
whereq is the particle diameter measured in 2-D, H and L are experimental constants
which can be determined from particle size distribution on a plane.
2
2/
d
L q = ,
2
0
H L N = . Where
2d
q is the average diameter of the particles with diameter larger than
c
q in the 2-D measurement,
0
N the total number of particles in the plane. The number
of particles per unit volume, which is larger than q in 3-dimensions, is given by
Sandstroms analysis [10]:
2 3
( ) exp( ) ( )
4
v
H L
N L erfc L
L
q
q q q
t
| |
= +
|
|
\ .
(2)
Application of cold rolling and hot deformation to a strain of 0.7 does not change the
size of the undissolved particles and their distribution in the samples A. However, the
fine dispersoids in the samples B tend to form bands parallel to the rolling direction, as
shown in Fig. 2d.
Table 1. Particle evolution during homogenization at 630C and annealing at 450C
Treatment
conditions
Average size
(m)
Aspect ratio Number/mm
2
10
3
%Area
transfer 1.82 1.78 23.24 9.9
450C24h 1.95 1.76 25.12 10.5
630C16h 2.25 1.72 13.96 6.2
630C78h 2.94 1.70 9.46 6.0
630C168h 3.42 1.63 5.06 5.6
8.3.2. High temperature mechanical behavior
The flow stress-strain curves under different deformation conditions are shown in
Fig. 1. The steady state stress of the sample A is always higher than that of the sample B
Chapter 8 140
under same deformation conditions. This indicates that the level of Mn in solute
solution and the presence of the fine second phase particles have a complicated effect on
the flow stress. The mechanical behavior is well described by Sellars-Tagart constitutive
equation [11]:
(sinh ) exp( / )
m
g
A Q R T c oo = (3)
where A, o , m are material constants, c , true strain rate, R
g
, the universal gas constant,
T, absolute temperature, Q, the activation energy of deformation and o , the flow stress.
Conditions of c and T are often incorporated into a single parameter, the Zener-
Hollomon parameter, Z [11]
exp( / )
g
Z Q R T c = (4)
Fig. 1. The equivalent stress-equivalent strain curves of samples A and B, a). Deformed
at three temperatures for a fixed strain rate of 1 /s, b) deformed at 450C for three
strain rates.
The parameters obtained by fitting Eq. (3) to the steady state flow stresses are listed in
Table 2 in the first two rows. Although the exact physical significance of each constant
in Eq. (3) is rather vague, it is interesting to compare the obtained values with those of
commercial purity aluminum and with those of conventional Al-Mn series alloys [12-
14]. A very wide variation has been reported on these parameters of Al-Mn alloys in
differing states, a few examples are listed in Table 2. If the parameter A is regarded as a
frequency factor (proportional to the density of activateable sites), the higher value in
the treatment A reflects that more dislocations overcome obstacles to move in unit time.
If the parametero is a measure of activation volume, the lower value in treatment B
0
10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8
Strain
S
t
r
e
s
s

(
M
P
a
)
A350C
B550C
A550C
B450C
A450C
B350C
1/ s c =
a
0
10
20
30
40
50
0 0.2 0.4 0.6 0.8
Strain
S
t
r
e
s
s

(
M
P
a
)
Td=450C
A1/s
A0.1/s
A10/s
B1/s
B10/s
b
suggests a smaller volume of material participating in a single deformation event. And if
the parameter m is associated with climb-controlled processes, irrespective of dispersed
particles, the lower value in the treatment A indicates a larger solute drag effect during
deformation. The activation energy basically reflects the energy barriers over which
deformation can take place. For pure aluminum the activation energy is an indicator of
the energy barriers for self-diffusion, which largely governs dislocation motion. For
aluminum alloys, the activation energy is affected by solute concentration and the
dispersion of second-phase particles in a complicated way, which usually act as addition
barriers to the free motion of dislocations. Thus, as the high temperature mechanical
behavior of the materials depends largely on the content of Mn in the matrix and the
amount of dispersoids, the effect of the pre-deformation state should not be ignored.
Table 2. The fitting parameters of the material constants in the constitutive equations for
treatments A and B and some related data in literature.
A sec
-1
o MPa
-1
m Q (kJ/mol) deformation
state
Treatment A
11
1.0 10 0.056 4.0 186
Treatment B
10
1.2 10 0.052 4.5 160
Al-1Mn [12]
10
1.21 10 0.039 5.0 156 transfer
Al-1Mn [14] ~ 0.052 2.6 152 homogenized
AA3003 [13] ~ ~ 3.1 180 homogenized
AA1050 [12]
11
1.96 10 0.036 5.0 155 transfer
The subgrain size, o , directly after hot deformation was estimated from the following
relationship [15]:
1
/( )
av s i
M Gb o o o o = (5)
where
1
o is a constant of a order 3,
av
M , the average value of Taylor factor, G, the
shear modulus, b, the burgers vector and
i
o ,
s
o , the initial flow stress and the steady
state flow stress, respectively.
As the stored energy is inversely proportional to the subgrain size, the driving force for
recrystallization in the samples of the treatment A having a higher flow stress is higher
than that of the treatment B under the same deformation conditions.
Chapter 8 142
8.3.3. The decomposition kinetics of the supersaturated matrix during
deformation and isothermal annealing
The electrical conductivity of AA3003 in the transfer state is measured as 24.8 MS/m
which corresponds to 0.34 wt.% Mn in solution. For the samples with treatment A,
during homogenizing at 630C the conductivity decreases rapidly in the first (within 30
minutes) and then approaches a constant value of 19.96 MS/m, which indicates that the
solid solution (of Mn, Fe, and/or Si) in the matrix reaches the equilibrium level. The
solubility of Mn in solute solution in this temperature is 0.68 wt.% and this value is
maintained during down quenching. For the samples after treatment B, the conductivity
is 26.98 MS/m. The corresponding Mn content in the matrix is 0.22 wt.%.
Table 3. The conductivity measurement in the samples A after deformation and the
fraction decomposed during deformation with respect to the equilibrium concentration
at different annealing temperatures
The fraction decomposed during
deformation in the annealing temperature
Deformation
conditions
LnZ Conductivity
(MS/m) 450C 500C 540C
As-
homogenized
19.96 0.0 0.0 0.0
CR 20.15 3.6 4.3 5.7
550C0.1/s 24.9 20.46 8.0 9.5 12.6
550C1/s 27.2 20.35 6.5 7.6 10.1
450C0.1/s 28.6 20.36 6.6 7.8 10.4
550C10/s 29.5 20.27 5.3 6.3 8.4
450C1/s 30.9 20.23 4.8 5.6 7.5
450C10/s 33.2 20.18 4.1 4.8 6.4
350C0.1/s 33.6 20.42 7.5 8.8 11.7
350C1/s 35.9 20.20 4.4 5.1 6.8
It is interesting to note that the conductivity measured in the deformed samples of the
treatment A is higher than that before deformation, as listed in Table 3. Fig. 2 shows the
distribution of the second phase particles in the as-deformed samples for 4 different
deformation conditions. In Fig. 2a, for material of treatment A after cold rolling, the
matrix is clean except for coarse primary particles. However, fine precipitates are found
along the grain boundaries and subgrain boundaries in the samples after hot deformation
(Fig. 2b and Fig. 2c). More precipitates can be observed in the sample A350C0.1/s
than in A350C1/s. Fig. 2d shows the particle distribution in B450C1/s in which the
smaller particles are formed before deformation. The conductivity in the deformed
samples B also increases when the deformation temperature is below 450C. The
increment in conductivity is related to the deformation conditions in the same manner as
in the samples A but is smaller. However, when the sample B is deformed at 550C the
conductivity decreases, in line with resolutionising of Mn.
Fig. 2. The distribution of the precipitates in the as deformed samples, (a) Cold Rolling,
(b) A350C0.1/s, (c) A350C1/s and (d) B450C1/s
That the conductivity of the as-deformed samples is higher than that in the undeformed
samples indicates that the solute redistribution of the solute atoms (Mn, Fe, Si) has
occurred during deformation. The fine dispersoids observed along grain boundaries or
subgrain boundaries in some of the hot deformed samples give a further indication that
strain-induced precipitation could happen during hot deformation.
Fig. 3 shows the conductivity variation with annealing time in the samples of treatment
A after different deformation conditions. In Fig. 3a, material annealed at 450C, the rate
of the precipitation in the sample after cold deformation is highest at the very beginning
of the annealing (~1 min.) and then decreases. The effect of the hot deformation
Chapter 8 144
conditions on the precipitation kinetics depends on the Zener-Hollomon parameter, Z
(lnZ is listed in Table 4) and deformation temperature. In Fig. 3b, material annealed at
500C, the enhancing effect of the deformation on the precipitation at first sight bears
no relation with Z. However, when the effect of the recrystallization on the precipitation
kinetics is considered these results could be fully understood. This point will be
elaborated in detail in the discussion.
Fig. 3. The decomposition kinetics in the samples A measured by means of conductivity
( 0.70 c = for both cold and hot deformation), (a) annealed at 450C, (b) annealed at
500C.
19
20
21
22
23
24
25
26
27
28
29
30
0.01 0.1 1 10 100 1000 10000
Annealing time (min)
C
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
no strain 50% CR
350Cx0.1/s 350Cx1/s
450Cx0.1/s 450Cx1/s
450Cx10/s 550Cx0.1/s
550Cx1/s 550Cx10/s
a
19
20
21
22
23
24
25
26
27
28
0.01 0.1 1 10 100 1000 10000
C
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
no strain 50% CR
350Cx0.1/s 350Cx1/s
450Cx0.1/s 450Cx1/s
450Cx10/s 550Cx0.1/s
550Cx1/s 550Cx10/s
b
Fig. 4 is the evolution of the conductivity measured in the samples of B450C1/s
annealed at different temperatures. During annealing at 540C the conductivity
decreases rapidly and then approaches a constant value. This indicates that the second
phase particles are dissolving during early stage of annealing. Deformation not only
promotes the precipitation process but also enhances the particle dissolution process.
Fig. 4. The conductivity in the samples of B450C1/s as a function of annealing time
at different temperatures.
8.3.4. The softening kinetics
The softening kinetics is determined from the microhardness measurement and it is
given by:
def t
def rex
H H
y
H H
(6)
where
def
H ,
rex
H and
t
H are the values of the microhardness measured in the as
deformed state, fully recrystallized state and at annealing time t respectively. The
softening kinetics after different deformation conditions is listed in Table 4 and is
shown in Fig. 5 to Fig. 7 for materials after hot deformation and Fig. 8 for material after
cold rolling. OPM in Fig. 5a labels the recrystallization kinetics measured from optical
microscopy. As can be seen, the softening kinetics is either fast or slow, strongly
25
26
27
28
29
30
31
1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
C
o
n
d
u
c
t
i
v
i
t
y

(
M
S
/
m
)
450C
500C
540C
10
-2
10
4
10
3
10
1
10
-1
10
2
10
0
Chapter 8 146
depending on the pre-heat treatment, deformation temperature, strain rate and annealing
temperature.
For a given deformation condition, the rate of softening increases as the annealing
temperature increases. For a given strain rate, an increase in the deformation
temperature results in a decrease in the rate of recrystallization. For a given deformation
temperature, an increase in the strain rate leads to an increase in the rate of
recrystallization kinetics. Generally, for a given annealing temperature, the softening
kinetics increases as the Z value increases. However, attention should paid to special
cases. Although the Z values for the deformation conditions of A350C0.1/s and
A450C10/s are almost the same, the softening kinetics in the samples of
A450C10/s is much faster than that in A350C0.1/s (see Table 4, Fig. 6 and Fig. 7).
This seemingly unexpected result was verified in a new set of experiments.
When recrystallization is fast, recovery is responsible for only ~25% of the softening.
However, when the softening kinetics is sluggish the contribution of the recovery to the
softening can be as high as 70%, as shown in Fig. 5 to Fig. 7, in which arrows indicate
the start of recrystallization.
The softening kinetics in the cold rolling samples is faster than that in the hot deformed
samples under the same annealing temperatures. This is in line with a finer dislocation
network due to cold deformation. The contribution of the recovery to the softening is
about 60% and it slightly increases as at annealing temperature increases.
It should be noted that when the recrystallization is fast, the softening rates in samples
of the treatment B is higher than that in samples of the treatment A (see Fig. 5a, Fig. 6a
and Fig. 7a). Contrarily, when the recrystallization is slow, the softening rates in
samples of the treatment A is higher than that in samples of the treatment B (see Fig. 5b,
Fig. 6b and Fig. 7d). This indicates that the precipitates existing prior to the plastic
deformation play a less effective role in retarding recrystallization even though the
particle size may be in the range of 50~150 nm.
The times required for 5% and 95% recrystallization are now plotted against reciprocal
temperature for several deformation conditions, as shown in Fig. 9. A striking feature of
Fig. 9 is the pronounced upward curvature of the data at low temperature for both the
start and finish of the recrystallization except for the case of B350C1/s where there is
no curvature for the start of the recrystallization. The recrystallization time is longer
than would be predicted by extrapolation from high temperature using Arrhenius
relation. The departure of the curves from the extrapolation to low temperature suggests
that both the nucleation and growth processes are retarded by the concurrent
decomposition process.
Fig. 5. The softening kinetics annealed at 450C after different deformation conditions.
0
0.2
0.4
0.6
0.8
1
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
F
r
a
c
t
i
o
n
A350Cx1/s,OPM
B350Cx1/s,OPM
A350Cx1/s
B350Cx1/s
Ta=450C
a
10
-3
10
4
10
3
10
2
10
1
10
0
10
-1
10
-2
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05
F
r
a
c
t
i
o
n
A450Cx1/s
A450Cx10/s
B450Cx1/s
B450Cx10/s
Ta=450C
b
10
-1
10
5
10
4
10
3
10
2
10
1
10
0
0
0.2
0.4
0.6
0.8
1
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
F
r
a
c
t
i
o
n
A450Cx0.1/s
A450Cx1/s
B450Cx1/s
A450Cx10/s
B450Cx10/s
Ta=500C
b
10
-3
10
4
10
3
10
2
10
1
10
0
10
-1
10
-2
0
0.2
0.4
0.6
0.8
1
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03
F
r
a
c
t
i
o
n
A350Cx0.1/s
A350Cx1/s
B350Cx1/s
Ta=500C
a
10
-3
10
3
10
2
10
1
10
0
10
-1
10
-2
Chapter 8 148
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1
F
r
a
c
t
i
o
n
B350Cx1/s
A350Cx1/s
A350Cx0.1/s
Ta=540C
a
10
-2
10
0
10
-1
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100
F
r
a
c
t
i
o
n
A450Cx0.1/s
A450Cx1/s
A450Cx10/s
Ta=540C
b
10
-2
10
2
10
1
10
0
10
-1
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10
F
r
a
c
t
i
o
n
B450Cx1/s
B450Cx10/s
A450Cx1/s
A450Cx10/s
Ta=540C
c
10
-2
10
0
10
1
10
-1 0
0.2
0.4
0.6
0.8
1
0.001 0.01 0.1 1 10 100 1000 10000
F
r
a
c
t
i
o
n
B550Cx1/s
A550Cx0.1/s
A550Cx1/s
A550Cx10/s
Ta=540C
d
10
-3
10
-2
10
4
10
3
10
2
10
1
10
0
10
1
Fig. 8. The softening kinetics annealed at different temperatures after cold deformation.
Fig. 9. Times to 5% and 95% recrystallization for several deformation conditions.
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100 1000 10000
S
o
f
t
e
n
i
n
g

f
r
a
c
t
i
o
n
350C
500
450
400
10
-2
10
1
10
0
10
3
10
4
10
2
10
-1
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
1.2 1.25 1.3 1.35 1.4 1.45 1.5
1000/T (K)
t0.05 A350Cx1/s
t0.95 A350Cx1/s
t0.05 B350Cx1/s
t0.95 B350Cx1/s
t0.05 CR
t0.95 CR
T
i
m
e
s

t
o

5
%

o
r

9
5
%

r
e
x

(
m
i
n
)
10
5
10
0
10
1
10
2
10
3
10
4
10
-2
10
-1
a
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.2 1.25 1.3 1.35 1.4 1.45 1.5
1000/T (K)
t0.05 A450Cx10/s
t0.95 A450Cx10/s
t0.05 B450Cx10/s
t0.95 B450Cx10/s
t0.05 CR
t0.95 CR
10
6
10
-1
10
0
10
1
10
2
10
3
10
4
10
5
10
-2
T
i
m
e
s

t
o

5
%

o
r

9
5
%

r
e
x

(
m
i
n
)
b
Chapter 8 150
Table 4. The start and finish times and recrystallized grain size of the two treatments
after different deformations and different annealing temperatures
Defor-
mation
TC
Strain
rate
(1/s)
LnZ Anneal
T C
Starting
time
(min)
Complete
time
(min)
Grain size
(m) RD/ND
Aspect
ratio
540 0.15 1 1438/362 4
500 60 10
3
2857/180 16
0.1 33.6
450 No recrystallized within 7620 min
540 0.2 496/175 2.8
500 0.3 2 562/150 3.3
A350
1 35.9
450 1080 10
4
1088/145 7.5
540 12 40 1875/358 5.2
0.1 28.6
540 0.2 2 1312/375 3.5
500 960 4200 2000/313 6.5
1 30.9
540 0.1 0.5 637/262 2.4
500 0.5 10 725/212 3.4
A450
10 33.2
450 4800 510
4
1375/195 7.2
0.1 24.9 540 No recrystallized within 5040 min
1 27.2 540 240 4320 6500/900 7.2
A550
10 29.5 540 0.2 2 1687/412 4
540 0.15 536/112 4.5
500 0.12 1 612/124 4.9
B350 1 30.9
450 0.3 60 1466/156 9.4
540 0.5 5 3875/295 13-18
500 7620 ~
1 26.6
540 0.1 0.4 1125/150 7.5
500 0.3 5 1450/175 8.2
B450
10 28.9
450 23280 410
5
2140/196 11
B550 1 23.4 540 No recrystallized within 5040 min
500 0.5 26.8 1.36
450 0.2 30 28.5 1.42
400 30 6000 280/123 2.28
A
Cold rolling
8.3.5. Nucleation mechanisms and recrystallized grain structure
After hot PSC deformation the grains are further elongated along the rolling direction.
The average aspect ratio of the grains now is 14-18. The grain structure after the same
amount of cold rolling is comparable to that after hot deformation. In addition to the
pancaking of the grains, deformation bands can be found in the as-deformed
microstructure. These deformation bands are not uniformly distributed. In the heavily
dynamic precipitated samples, the subgrains are observed to be elongated along the
deformation direction because precipitation occurs on the subgrain boundaries (as
shown in Fig. 3b).
Fig. 10. The partially recrystallized structures in the samples (a) A450C10/s,
annealed at 450C for 128 h, (b) B350C1/s, annealed at 450C for 40 s, (c)
A450C1/s, annealed at 540C for 18 s and (d) CR, annealed at 400C for 1 h.
Particle stimulated nucleation (PSN) was only found in the cold rolled samples
annealed at temperature above 500C. In this circumstance, the initial stage of
recrystallization is more or less uniform and the early site saturation can be inferred.
However, in all other cases, the initial stage of the recrystallization appears to be highly
Chapter 8 152
localized. Nucleation by subgrain growth (Fig. 10a) and strain induced boundary
migration (SIBM) (Fig. 10b) are often observed. The growing recrystallized grains have
an irregular shape and jagged grain boundaries which is typical for particle impeded
grain boundary migration. No PSN effect was found in the current hot deformed
samples annealed in the temperature range studied under optical microscopy. Evidently,
hot deformation leads to a reduced tendency for nucleation. The recrystallized grain
size in the sample after hot deformation are typically 5~10 times larger than those at the
same stage of transformation after cold deformation (compare Fig. 10c and Fig. 10d).
The deformation inhomogenity, e.g. near original grain boundaries, is prerequisite for
the nucleation of recrystallization (Fig. 10c).
The final recrystallized grain size resulting from annealing at different temperatures
were measured by means of optical light microscopy. The final grain size varies
considerably within the specimens. The average values are listed in Table 4, where the
grain size is given by two dimensions, RD (grain dimension in the elongation direction
in PSC specimens) and ND (grain dimension in the compress direction). Measured grain
dimensions for fully recrystallized material for several cases are given as a function of
annealing temperature in Fig. 11 for both RD and ND. It is clearly shown that for a
given deformation condition, the fully recrystallized grain size and its aspect ratio
decrease as annealing temperature increases. This relation of grain size with annealing
temperature is opposite to that commonly cited as recrystallization law which does
not involve dispersoids effects. The size increase at lower temperature is primarily in
the RD direction.
Fig. 11. Recrystallized grain size as a function of annealing temperature for several
deformation conditions.
0
200
400
600
800
1000
1200
1400
1600
440 460 480 500 520 540 560
Annealing temperature (C)
R
e
c
r
y
s
t
a
l
l
i
z
e
d

g
r
a
i
n

s
i
z
e

(
m
)
A350Cx1/s RD
A350Cx1/s ND
B350Cx1/s RD
B350Cx1/s ND
a
0
500
1000
1500
2000
2500
440 460 480 500 520 540 560
Annealing temperature (C)
R
e
c
r
y
s
t
a
l
l
i
z
e
d

g
r
a
i
n

s
i
z
e

(
m
)
A450Cx10/s RD
A450Cx10/s ND
B450Cx10/s RD
B450Cx10/s ND
b
For a given deformation temperature, an increase in the strain rate leads to a decrease in
the recrystallized grain size. The grain size is larger for the B preheat practice. The
aspect ratio of the recrystallized grains in the samples of the treatment B remains larger
than that in the samples of the treatment A. The fully recrystallized grain size and its
aspect ratio are much smaller in the cold deformed samples.
The lower the annealing temperature, the larger the recrystallized grain size and the
larger aspect ratio of the grain and these phenomena can be associated with the sluggish
recrystallization process.
8.4. Discussion
8.4.1. Effect of the deformation, recovery and recrystallization on the
precipitation kinetics of supersaturated alloy
The microstructural evolution during annealing is complex and needs to be interpreted
in terms of the combination of precipitation and recrystallization. In Al-Mn alloys,
precipitation occurs only on high or low angle grain boundaries. Thus, the precipitation
kinetics is influenced by the deformed, recovered and recrystallized microstructure.
That deformation of the supersaturated solution increases the precipitation rate is
attributed to the following reasons [16]: deformation increases the dislocation density,
which increases both the nucleation sites and diffusivities of precipitate-forming
elements in the material. Furthermore, the nucleation barrier for precipitation on
dislocations is much lower than that for homogenous nucleation. In addition, the solute
segregation to the dislocation network during deformation enhances the rate of
precipitation via an increase in the local chemical driving force. Therefore, the more
dislocations introduced by deformation, the larger effect of the enhancement. Any
recovery or recrystallization will result in a decrease in the dislocation density, and
therefore, weaken the effect of deformation.
Fig. 12 shows schematically a time dependence of the dislocation density for various
experimental conditions for samples of treatment A deformed at 450C. The hot
deformation is done at a temperature range where precipitation and dynamic recovery
occur concurrently. The dislocation density reaches a saturation value at a strain of 0.7.
The higher the strain rate, the higher the dislocation density. However, the loading time
increases as strain rate decreases. In the following annealing processes, the dislocation
density decreases because of static recovery and recrystallization. The reduction of the
dislocation density is faster in the samples deformed at a higher strain rate and annealed
at a higher temperature. Here we need to analyze the effect of the deformation on the
Chapter 8 154
dynamic precipitation (during deformation) and static precipitation (during annealing
after deformation) separately.
Fig. 12. The dislocation density as a function of time for various experimental
conditions deformed at 450C.
The solute segregation or dynamic precipitation during hot deformation is a complicated
problem. Data in Table 3 show that the conductivity increases after deformation, which
indicates that the decomposition of the supersaturated matrix has happened during
deformation. Solute atmosphere formation around a moving dislocation leads to the
redistribution of solute elements between the dislocation cell walls and the cell interiors.
When the concentration of the solute atoms at dislocations reaches the composition of
the precipitate nucleus, a precipitate can form. The concentration of the solute atoms
around a moving dislocation is determined by the velocity of solute atom diffusion
relative to the dislocation velocity which is proportional to the macro strain rate. The
solute atom diffusion may be enhanced by vacancy production during hot deformation.
Therefore, this process depends strongly on the strain rate and the deformation
temperature. However, no apparent correlation between dynamic precipitation and the Z
value can be deduced.
The contribution of the dynamic precipitation during deformation to the overall
precipitation kinetics is different at different annealing temperatures thereafter. The
fraction decomposed during deformation is given by:
time
D
i
s
l
o
c
a
t
i
o
n

d
e
n
s
i
t
y
deformation annealing
0.1/ s c =
10 / s c =
1/ s c =
450C
540C
500C
Td=450C
1
h d
h
x
_ _
_ _
(7)
where
h
_ and
d
_ are the resistivities in the as-homogenized state (before deformation)
and in the as-deformed state, respectively. _
is the saturation level of the resistivity of

the fully precipitated sample at long time, which corresponds to the equilibrium content
of Mn in solid solution at a given annealing temperature. _
is calculated in Chapter 7.
The fraction decomposed during deformation can be up to 12% of overall precipitation,
as listed in Table 3.
By referring to the recrystallization kinetics listed in Table 4 the experimental data in
Fig. 3 can be understood as follows. When annealed at 450C, the recrystallization start
and finish time after cold deformation is 0.1 and 10 min, respectively. The precipitation
rate in the cold deformed sample is relatively high at the very beginning of annealing
(Fig. 3a) and then decreases because of the annihilation of the dislocation density as a
result of recovery and recrystallization. On the contrary, softening in all the other hot
deformed samples is still in the slow recovery stage and the effect of deformation
conditions on the precipitation rate will follow the rule of the higher the Z value, the
higher dislocation density, the higher precipitation rate. However, loading time also
plays an important role. For example, although the Z value of the sample 350C1/s is
higher than that of sample 350C0.1/s, the precipitation rates of the two samples are
almost identical. This is due to the stronger segregation effect caused during
deformation in the sample 350C0.1/s. In the case of Fig. 3b, material annealed at
500C, precipitation rate at deformation conditions of higher Z becomes lower because
of the early recrystallization. Thus, the experimental data in Fig. 3 show, in fact, a
combined effect of deformation and softening on the precipitation.
The JMAK equation is now applied to describe the static precipitation kinetics during
annealing at a given temperature. The precipitation kinetics is written as:
2
1 exp( )
n d t
d
x kt
_ _
_ _
= =
(8)
where
t
_ is the resistivity at annealing time t.
d
_ and _
are the same as above. k and

n are constants. The best-fitting parameters for k and n under different deformation
conditions are listed in Table 5. The value of n decreases as Z value increases but k
varies in the opposite manner except the cases where the deformation is carried out at
lower strain rate (0.1 /s).
Chapter 8 156
Table 5. Fitting parameters for k and n under different deformation conditions using
JMAK equation
Annealing temperature C
450 500 540
Deformation
conditions
lnZ
n k n k n k
No strain 0.39 0.037 0.42 0.031 0.37 0.057
CR 0.37 0.066 0.33 0.108 0.30 0.160
A5500.1/s 24.9 0.54 0.016 0.43 0.056 0.34 0.109
A5501/s 27.2 0.51 0.025 0.42 0.067 0.34 0.125
A4500.1/s 28.6 0.49 0.032 0.38 0.104 0.33 0.132
A55010/s 29.5 0.50 0.029 0.41 0.082 0.33 0.135
A4501/s 30.9 0.43 0.070 0.36 0.135 0.32 0.146
A45010/s 33.2 0.43 0.076 0.34 0.106 0.30 0.162
A3500.1/s 33.6 0.38 0.122 0.29 0.212 0.29 0.175
A3501/s 35.9 0.38 0.132 0.29 0.161 0.29 0.178
The overall precipitation kinetics includes dynamic precipitation and static precipitation
and is given by:
1 ( ) exp( )
n d
h
x kt
_ _
_ _
(9)
The times to 15% decomposition and 95% decomposition at different annealing
temperatures are calculated and represented by the C-curves, as shown in Fig. 13. The
introduction of deformation has the effect of shifting the C-curve to shorter times. The
start-up of precipitation is strongly affected by deformation condition, i.e. the strain rate
and deformation temperature. The complete of precipitation is not only promoted by
deformation but also affected by recovery and recrystallization. For example, in
Fig. 13e, the times to 95% precipitation for both cases of A45010/s and cold
deformation are almost the same when the samples are annealed at 540C. However,
when materials are annealed at 450C the time to 95% precipitation of A45010/s is
much shorter than that of samples after cold deformation. This is due to the fact that the
enhancing effect of deformation is weakened by different degree of recovery and
recrystallization at different annealing temperatures. The recrystallization process in
both cases completes within 0.5 minutes during annealing at 540C. When dislocations
are eliminated the enhancing effect of the deformation is only from the solute
segregation formed during deformation. When the materials are annealed at 450C, in
the sample of cold deformation, the recrystallization completes in 10 minutes while in
the sample of A45010/s the recrystallization takes 7000 minutes to start, at which time
the precipitation is complete.
8.4.2. Effect of precipitation on the recrystallization kinetic
As we have observed in this experiment, the basic nucleation mechanisms in
recrystallization after the hot deformation is subgrain growth and strain-induced
boundary migration. The fine precipitates retard both nucleation and growth of the
recrystallized grains. It can be argued that the large recrystallized grain size results from
a reduction in the number of successful nuclei as annealing temperature decreases. In
the following we will develop a model to explain the effect of the precipitation on the
softening kinetics qualitatively. We simplify the analysis by considering the
substructure to be adequately described using two components only. The major
component is considered as an assembly of equiaxed subgrains of mean equivalent
radius, R , mean misorientation, u , and with boundaries of mean energy and mobility,
and M respectively. The minor component we consider as particular subgrains
(effectively sub-critically sized recrystallization nuclei) which have a larger size (
n
R )
and different boundary characteristics ( , ,
n n n
M u ) with respect to that of the average
subgrain assembly [17,18]. According to Cahn and Humphreys the presence of the local
misorientation gradient (covering a number of average subgrain diameters) is necessary
for the formation of the nuclei. During particular subgrain growth its boundary
misorientation
n
u is related to the size of the particular subgrain as [5]
0 0
( )
n n
R R u u = + O (10)
Where
0
u and
0
R are initial values of
n
u and
n
R , and O is the orientation gradient
( /
n n
d dR u O = ).
When a deformed supersaturated solution is annealed, the softening and the
precipitation processes occur simultaneously. The growth rate of the particular subgrain
is given by:
( )
n
n D Z
dR
V M P P
dt
= = (11)
where P
D
and P
Z
are the net driving pressure and Zener pinning pressure respectively.
Chapter 8 158
Fig. 13. The precipitation-temperature-time (PTT) diagram for the samples of
treatment A after different deformation conditions. (a), (b) and (c) the time to 15%
precipitation; (d), (e) and (f), the time to 95% precipitation.
440
460
480
500
520
540
560
0.001 0.01 0.1 1 10 100
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A350Cx1/s
A350Cx0.1/s
CR
No strain
a
440
460
480
500
520
540
560
0.001 0.01 0.1 1 10 100
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A450Cx10/s
A450Cx1/s
A450Cx0.1/s
CR
No strain
b
440
460
480
500
520
540
560
0.001 0.01 0.1 1 10 100
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A550Cx10/s
A550Cx1/s
A550Cx0.1/s
CR
No strain
c
440
460
480
500
520
540
560
0 2 4 6 8 10
Annealing time x10
-4
(min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A350Cx1/s
A350Cx0.1/s
CR
No strain
d
440
460
480
500
520
540
560
0 2 4 6 8 10
Annealing time x10
-4
(min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A450Cx10/s
A450Cx1/s
A450Cx0.1/s
CR
No strain
e
440
460
480
500
520
540
560
0 2 4 6 8 10
Annealing time x10
-4
(min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
A550Cx10/s
A550Cx1/s
A550Cx0.1/s
CR
No strain
f
The net driving pressure acting on the front of a viable nucleus is given by:
( )
n
D
d A
P E
dW
= A (12)
where W and A are the volume and the surface area of the viable nucleus, respectively.
is the specific boundary energy of the viable nucleus. E A is the energy difference
between the two sides of the recrystallized front, which is approximately the stored
energy in the average assembly and for hot deformation it can be estimated by:
/ E R o A = (13)
where o is a geometrical constant and has a value of ~1.5 [5]. The boundary energy
depends on the misorientation u and is given by [5]:
( ) (1 ln( ))
m
m m
u u
u
u u
= (14)
where
m
and
m
u are, respectively, the values of boundary energy and misorientation
for high-angle boundaries, which is commonly taken as 15 and
m
is 0.324 J/m.
The net driving pressure for this particular subgrain to grow is now given by:
2 ( )
ln
n m n
D
n m m
P
R R
u u o
u u
= + O (15)
The first term of the right hand of Eq. (15) is the driving pressure from stored energy
(P
v
), the second term is retarding pressure from boundary curvature (P
c
), and the third
term is a pressure from the deformation gradient.
If the main mechanism of recovery in the subgrain assembly is subgrain coarsening, the
growth rate of a uniform subgrain assembly may be expressed in the form [5]:
( )
4
Z
dR
M P
dt R
= (16)
For a random dispersion of spherical particles (average radius r and volume fraction f
p
)
the drag pressure is given by [19]:
Chapter 8 160
6 /
z p
P f r t = (17)
A more realistic assumption for the precipitate distribution is that the particles lie on
sub-boundaries in the hot worked structure in AA3003. The precipitate-retarding force
corresponding to this case is [19]:
2
3
2
p
z
f
P
r
o
t
= (18)
where o is the average subgrain diameter, which is given by Eq. (5).
The condition for instability, leading to discontinuous growth of the particular subgrain
is written as [17,20]:
0
n
n
dR dR
R R
dt dt
> (19)
Setting the left-hand side of inequality (19) equals to zero and assuming that the
misorientation between a viable nucleus and surrounding is 15, the critical radius of a
viable nucleus can be obtained as:
2
8 ( )
4
2( )
4
n n n n n
n z z
c
n n
z z
M M M M
R P P
M M M M R
R
M
P P
M R

o o
| |
+
|
\ .
=
+
(20)
This equation shows that the critical size for a viable nucleus is a function of
z
P , R ,
as well as the ratio of the mobilities between the high angle grain boundary and low
angle sub-boundary. In general,
z
P will vary during annealing because of the time
dependence of the nucleation and growth of the precipitates. R will increase due to the
concurrent recovery. In addition,
n
M and M will increase as the decrease in the solute
concentration of the matrix. Therefore, R
c
will vary with annealing time and
temperature. An essential condition for the presence of the solution of the Eq. (20) is the
term under the root sign being not less zero. Taking it to be zero, one obtains:
2
( ) 2
8 ( 1)
n
n
zCritical
n n
n
M
M
P
M
R
M
o
(21)
Thus if the value of
z
P given by Eq. (18) is larger than that by Eq. (21) there will be no
nucleation event.
In order to elucidate the evolution of the R
c
with annealing time and annealing
temperature we need to consider the precipitate properties and the boundary properties
in more detail. As derived in the appendix, the volume fraction of the precipitate is
expressed as:
0
0
p
c c
f
c c
|
(22)
where 0.738 = for Mn
3
SiAl
12
, 0.816 = for MnAl
6
, c
|
and
0
c are the Mn
concentrations in the precipitate and in the matrix before precipitation occurs,
respectively. c is the Mn concentration in the matrix at annealing time t for a given
temperature T, which can be obtained on the basis of the decomposition rate as
determined from the resistivity measurement, i.e.
1 exp( )
n d d t
d d
c c
kt
c c
_ _
_ _

= =

(23)
where
d
c , c and c
, are the concentrations of Mn in the matrix in the as-deformed

state, after an annealing time t and in equilibrium, respectively.
d
_ ,
t
_ and _
, are the
corresponding resistivities at the start of annealing, after an annealing time t and in
equilibrium, respectively. Values of c
can be calculated from the solvus line as in

Chapter 7. Combining Eqs. (22) and (23), one gets:
0
0
( ) exp( )
n
d
p
c c c c kt
f
c c
|
(24)
where k and n are fitting parameters as listed in Table 5.
Assuming that the precipitates have a simple spherical growth morphology, the growth
rate during the growth stage is:
Mn
D dr c c
dt r c c
|
(25)
By solving Eq. (25) against annealing time we obtain:
Chapter 8 162
2 2
0
0
2 ( )
exp( )
( )
t
n Mn d
D c c
r r kt dt
c c
|
)
(26)
The diffusion coefficient of Mn in aluminum at a given temperature is given by:
0
exp( )
dif
Mn
g
Q
D D
R T
= (27)
where D
Mn
is the diffusion coefficient, c , the instantaneous concentration of Mn in the
matrix, c
|
, the concentration in the particle, c
, the concentration of Mn in matrix in

equilibrium with precipitate assuming a planar interface,
g
R , the gas constant, T,
absolute temperature,
0
r , the average radius of the precipitates in the as-deformed state,
r , the particle radius at annealing time t.
Considering the solute drag effect, the grain boundary mobility in a solute solution is
given by Cahns solution for low driving force [21]:
1
1
'
n
Pure
M c
M
o
| |
= +
|
\ .
(28)
where
2
b gb m
Pure
g
D V
M
b R T
o
= (29)
and
2
( )
' sinh
v B b
B B
N k T E E
ED k T k T
o
o
| |
=
|
\ .
(30)
where M
pure
and c refers to the intrinsic grain boundary mobility and the concentration
of Mn in the matrix,
b
o is the grain boundary width (~1nm),
3
4/
v
N a = the number of
atoms per unit volume, a, the lattice parameter, b, the value of the Burgers vector, D
gb
is the grain boundary self-diffusion coefficient and D is the average value of the
diffusion coefficient in the vicinity of the grain boundary, V
m
is the molar volume of
aluminum,
B
k , Boltzmanns constant, E, the interaction energy between a solute atom
and the grain boundary, which can be estimated by [22]:
3
4
m
E Gr q = (31)
where ( ) /
a m m
r r r q = is the misfit parameter,
a
r , the radius of a solute atom,
m
r , the
atomic radius of a solvent atom, G, the shear modulus and it is given by [23]:
0
1 ( 300)
1
2 933
T
G G
| |
=
|
\ .
(32)
where
4
0
2.54 10 G = MN/m
2
, T is annealing temperature.
The mobility of a sub boundary is related to the average misorientation angle of the
boundary. It has a form of [17]:
4
1 exp( 5( ) )
n
m
M M
u
u
| |
=
|
\ .
(33)
There appears to be no published value of the grain boundary diffusion coefficient. In
this analysis we simply assumed that the activation energy of the grain boundary
diffusion is one half of that for bulk diffusion. The parameters used in the model for the
diffusion coefficients are listed in Table 6.
Table 6. Model input parameters
0
D
ms
-1
dif
Q
kJ/mol
Reference
Bulk diffusion coefficient of Al 10
-5
142 [24]
Grain boundary diffusion of Al 10
-5
71
Mn bulk diffusion in aluminum 410
-5
211 [25]
Mn pipe diffusion in aluminum 410
-5
165
The evolution of the particle fraction and the particle radius is shown in Fig. 14. At
higher annealing temperatures the growth rate of the particle is larger than that at lower
annealing temperature. Therefore at the same amount of the precipitation fraction the
particle size is larger at the high annealing temperature.
Fig. 15 show the driving force evolution during annealing, in which
v
P is the driving
pressure,
c
P is the retarding pressure from the boundary curvature of the viable nucleus,
Chapter 8 164
z
P and
z
P are the pining drag pressures on the subgrain boundary and on the boundary
of the viable nucleus, respectively.
zCritical
P is the pressure calculated from Eq. (21).
z
P
decreases as the annealing time increases because of the rapid growth of the particle
size. There are no possible nuclei present before
z
P decreases to
zCritical
P (the arrow
point in Fig.15).
z
P is larger than
z
P due to the larger size and higher boundary energy
of the viable nucleus.
Fig. 14. The evolution of the particle fraction and particle radius at different
temperatures.
Fig. 15. The driving pressure evolution during annealing at different annealing
temperatures assuming the initial subgrain radius being 1.5 m (for details see text).
0
0.4
0.8
1.2
1.6
2
2.4
1 10 100 1000 10000
Annealing time (s)
P
r
e
s
s
u
r
e
x
1
0
-
5

N
/
m
2
Pv
Pc
Pz
Pz
Pz critical
T=450C
10
0
10
4
10
3
10
2
10
1
0
0.4
0.8
1.2
1.6
2
2.4
0.1 1 10 100 1000
Annealing time (s)
P
r
e
s
s
u
r
e
x
1
0
-
5

N
/
m
2
T=540C
Pz critical
Pz
Pz
Pc
Pv
10
-1
10
3
10
2
10
1
10
0
0
0.04
0.08
0.12
0.16
0.2
0 50 100 150 200 250
Annealing time (s)
A
v
e
r
a
g
e

p
a
r
t
i
c
l
e

r
a
d
i
u
s

(
m
)
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
f
p
T=540C
r
f
p
0
0.02
0.04
0.06
0.08
0.1
0 1000 2000 3000 4000 5000
Annealing time (s)
A
v
e
r
a
g
e

p
a
r
t
i
c
l
e

r
a
d
i
u
s

(
m
)
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
f
p
T=450C
f
p
r
The critical subgrain radius as a function of the annealing time at different temperatures
for 2 initial subgrain sizes is shown in Fig. 16. At a given annealing temperature the
critical nucleus radius first decreases as the pinning pressure decreases and then
increases again due to the increase of the subgrain size. The critical radius decreases as
the annealing temperature increases.

If we assume that the probability for a nucleation event in association with a possible
site is proportional to the number of subgrains able to grow then the smaller the R
c
, the
more possible nucleation sites and the smaller the grain size. This explains why the
recrystallized grain size decreases as annealing temperature increases.
Fig. 16. The critical subgrain radius as a function of the annealing time at different
temperatures for 2 initial subgrain sizes.
Another point to be noted here is that, for the precipitate volume fraction encountered,
heterogeneous precipitation on the scale of substructure seems to be essential for the
retarding the recrystallization growth, and so plastic deformation plays an important role
in the geometry as well as the kinetics of precipitation. The precipitates existing prior to
the plastic deformation have no appreciable retarding effect on recrystallization even
though the particles may be small enough. This is why the recrystallization kinetics in
samples with treatment B is faster at high annealing temperature while it is slower at
lower annealing temperature.
2
3
4
5
6
7
8
9
10
11
0.1 1 10 100 1000 10000 100000
Annealing time (s)
C
r
i
t
i
c
a
l

n
u
c
l
e
u
s

r
a
d
i
u
s

(
m
)
2.5m
500C
2.5m
540C
1.5m
540C
1.5m
450C 1.5m
500C
R
=2.5m
T=450C
10
-1
10
5
10
4
10
3
10
2
10
1
10
0
Chapter 8 166
8.4.3. Interaction between precipitation and recrystallization
In essence, there is a strong two-way interaction between recrystallization and
precipitation after deformation: the precipitation reaction is accelerated by deformation
and weakened by recovery and recrystallization. The recrystallization is retarded by the
precipitation of MnAl
6
particles on subgrain boundaries. The inter play of precipitation
and recrystallization phenomena can be studied by referring to Recrystallization-
precipitation-temperature-time diagrams (RPTT). The RPTT diagram is constructed
from experimental data for samples of treatment A, as shown in Fig. 17 for three
different deformation conditions. Here the start-up of recrystallization (taken as 5 %
recrystallized) and complete recrystallization (95 % recrystallized) are plotted, together
with the conditions for 15 % precipitation. The stored energy in the samples of
A350C0.1/s is slightly higher than that in the samples of A450C10/s. However, the
recrystallization kinetics in the sample A450C10/s is much faster than that in the
sample of A350C0.1/s. The precipitation kinetics in the sample of A350C0.1/s is
faster than that in the sample of A450C10/s at early stage of annealing, and therefore,
more precipitates form in the sample of A350C0.1/s, which provide a higher
retarding pressure to the recrystallizing front.
Fig. 17. Recrystallization-precipitation-temperature-time (RPTT) diagram for samples
of treatment A after three different deformation conditions. P
0.15
, R
s
and R
f
are the time
to 15% precipitation, start-up of recrystallization and completion of recrystallization.
Labels 1, 2, 3 indicate the deformation conditions A450C10/s, A350C0.1/s and
A350C1/s, respectively.
440
460
480
500
520
540
560
1.E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06 1.E+08
T
e
m
p
e
r
a
t
u
r
e

(
C
)
R
s
3 R
s
1 R
s
2
R
f
3 R
f
1 R
f
2
P
0.15
2 P
0.15
1
P
0.15
3
10
-2
10
8
10
6
10
4
10
2
10
0
10
-4
A complete recrystallization-precipitation-temperature-time diagram in a wider
temperature range is established for material with treatment A after cold deformation,
as shown in Fig. 18. As we have analyzed in Chapter 7, at higher temperature the
precipitate to form is MnAl
6
while at lower temperature the precipitate is MnAl
12.
If the
recrystallization starts before conspicuous precipitates have formed, for example, when
cold rolled material is annealed above 450C, a fast recrystallization is present and a
fine recrystallized grain size is obtained. However, if conspicuous precipitates have
formed before recrystallization start, such as at annealing temperatures below 450C, a
retarding effect on recrystallization is due to happen and a larger grain size is obtained.
Fig. 18. Recrystallization-precipitation-temperature-time (RPTT) diagram for samples
of treatment A after cold rolling. CR-50% cold rolling, NS-no strain.
300
350
400
450
500
550
600
1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Time (min)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
t
r0.05
t
p0.15
,CR t
p0.15
,NS
t
r0.95
MnAl
6
MnAl
12
10
-2
10
4
10
2
10
0
10
6
t
p0.95
,CR t
p0.95
,NS
Chapter 8 168
8.5. Conclusions
(1). For AA3003 alloy, preheat treatments, which changes Mn content in supersaturated
solution and the distribution of the second phase particles, have a large effect on the
mechanical behavior of the alloy, softening kinetics and the fully recrystallized grain
size.
(2). The softening kinetics is either fast or sluggish depending on the preheat treatments,
the deformation conditions and the annealing temperature. When the softening kinetics
is fast, the recrystallization kinetics in the material with a lower supersaturated Mn in
matrix is faster than that in the material with a higher supersaturated Mn in matrix.
When the softening kinetics is slow, an opposite trend is found.
(3). When recrystallization is slow the contribution of the recovery to the softening can
be as high as 70%. When recrystallization is fast recovery is responsible for only ~25%
of the softening.
(4). The recrystallized grain size in AA3003 after cold or hot deformation increases as
the annealing temperature decreases. The size increase at lower temperatures is
primarily in the RD direction.
(5). Recrystallization kinetics is not simply proportional to the value of the Zener-
Holloman parameter as a result of precipitation effect.
(6). The dynamic decomposition, which includes the solute atom redistribution and the
formation of the precipitates, during hot deformation is related strongly to the strain
rate. The static decomposition kinetics of supersaturated alloy during annealing at a
given temperature is affected by the deformation conditions, recovery and
recrystallization processes.
(7). Particle stimulated nucleation (PSN) is only observed in the cold deformed samples
annealed at higher temperatures (>500C). Nucleation by subgrain growth and strain
induced boundary migration (SIBM) are the dominant nucleation mechanisms in the hot
deformed samples annealed in the temperature range 450-540C.

(8). A theoretical framework has been established in which precipitation reaction and
recrystallization kinetics has been linked in a semi quantitative manner. The fine
precipitates present prior to deformation have less appreciable retarding effect on
recrystallization.
References
1. L.F. Mondolfo, Aluminium Alloys: Structure and Properties, Butterworth and Co.,
London, 1976.
2. E. Nes, J.D. Embury, Z. Metallkde. 66 (1975) 589-593.
3. G. Hausch, P. Furrer, H. Warlimont, Z. Metallkde. 69 (1978) 174-180.
4. P. Furrer, N. Rheinfall, H. Warlimout, Hanau, Aluminium 54 (1978) 135-142.
5. F.J. Humphreys, M. Haltherly, Recrystallization and Related Annealing Phenomena,
Pergamon, London, 1996.
6. E. Nes, Aluminium 52 (1976) 560-563.
7. F. Gatto, G. Camona, M. Conserva, P. Fiorini, Mater. Sci. Eng. A3 (1968) 56-61.
8. P.L. Morris, B.J. Duggan, Met. Sci. January (1978) 1-7.
9. V. Hansen, B. Andersson, J.E. Tibballs, J. Gjonnes, Metall. Mater. Trans. B 26
(1995) 839-849.
10. R. Sandstrm, Z. Metallkde. 71 (1980) 741-751.
11. C.M. Sellars, W.J. McGregor Tegart, Int. Met. Rev. 17 (1972) 1-24.
12. H. Shi, A.J. McLaren, C.M. Sellars, R. Shahani, R. Bolingbroke, Mater. Sci. Techn.
13 (1997) 210-216.
13. R. Akeret, Z. Metallkde. 61 (1970) 3-10.
14. H. Zhang, E.V. Konopleva, H.J. McQueen, Mater. Sci. Eng. A319-321 (2001) 711-
715.
15. O.D. Sherby, P.M. Burke, Pro. Mat. Sci. 13 (1967) 325-332.
16. S.P. Chen, N.C.W. Kuijpers, S.van der Zwaag, Mater. Sci. Eng. A341 (2003) 296-
306.
17. F.J. Humphreys, Acta Mater. 45 (1997) 4231-4240.
18. F.J. Humphreys, Acta Mater. 45 (1997) 5031-5039.
19. S.S. Hansen, J.B. Vander Sande, M. Cohen, Metall. Trans. A 11 (1980) 387-402.
20. C.V. Thompson, H.J. Frost, F. Spaepen, Acta Metall. 35 (1987) 887-890.
21. J.W. Cahn, Acta Metall. 10 (1962) 789-798.
22. N.F. Fiore, C.L. Bauer, Pro. Mat. Sci. 13 (1967) 85-88.
23. H.J. Frost, M.F. Ashby, Deformation Mechanism Maps, Pergamon Press, 1982.
24. R.D. Doherty, J.A. Szpunar, Acta Mater. 32 (1984) 1789-1798.
25. E.A. Brandes, G.B. Brook, Smithells Metals Reference Book, 1988.
Chapter 8 170
Appendix: The precipitation volume fraction
Assuming a system containing N
0
mole atoms, N
p
in the precipitate, N
m
in the matrix,
the molar (atom) volume of the precipitate and the matrix is
p
V ,
m
V , respectively,
p
M and
m
M are the molar atom weight of precipitate and matrix, respectively.
The volume fraction of the precipitate is expressed as:
(1 )
p p p p p p
p
m m p p m m p m
N V N V x V
f
N V N V N V x V
= ~ =
+
(1)
0
/
p p
x N N = is the mole fraction of the precipitate.
We use c
|
and c to represent the Mn concentration in the precipitate and the matrix at
annealing time t for a given temperature T,
0
c is the average Mn concentration of the
alloy before precipitation occurs. The weight fraction of precipitate is
p
W , then
0
(1 )
p p
c W c W c
|
+ = (2)
From (2) we can derive
0
p
c c
W
c c
|
(3)
The relationship between the mole fraction and the weight fraction is:
1
/
p p p
p
p p m
W W W
x
M M M
| |
= + |
|
\ .
(4)
Substitute (3), (4) into (1) we got:
0
0
p
m
p
m p
V
M c c
f
V M c c
|
(5)
/
p m
V V relates to the lattice parameters. The lattice parameter of the precipitate
Al
12
Mn
3
Si is a
p
=12.65 , the number of atoms in the unit cell is 138. The lattice
parameter of the precipitate Al
6
Mn is a=6.5 , b=7.55 , c=8.87 , the number of
atoms in the unit cell is 28. The lattice parameter of pure aluminum is a
al
=4.0496 .
The volume fraction of the precipitate is expressed as:
0
0
p
c c
f
c c
|
(6)
where 0.738 = for Mn
3
SiAl
12
, 0.816 = for Mn Al
6
.
Chapter 8 172
S.P. Chen
Summary
Microstructural modeling of thermomechanical processing is well established as a
valuable procedure for optimizing processing conditions in the steel industry, while the
similar work for aluminum industry is still an opening problem. Recovery,
recrystallization and grain growth are core elements of this processing. This thesis
concentrates on the recovery and recrystallization behavior of aluminum alloys with a
focal point on the static softening kinetics after deformation at elevated temperatures.
The aim of the project is to investigate recovery and recrystallization processes in
aluminum alloys as a function of deformation condition, thermal history, composition
and precipitate content, and therefore to develop a validated physical model to predict
the recrystallization kinetics as a result of such processes. We performed experiment
and built models to achieve this objective.
After a general introduction, in Chapter 2 the quantification of the recrystallization
behavior in AA1050 using different techniques, optical microscopy, Electron Back
Scattering Diffraction (EBSD) and microhardness is described. Based on this study, a
new methodology for the determination of the recrystallized volume fraction from
anodically etched aluminum alloys using optical microscopy is proposed. The method
involves the creation of a composite image from multiple micrographs taken at a series
of orientations. The multiple orientation image method is shown to consistently yield a
recrystallized volume fraction which is significantly higher than that determined from a
single image while multiple orientation imaging and Orientation Imaging Microscopy
(OIM) results are found to be in good agreement. Furthermore it is shown that, after the
subtraction of the effect of concurrent recovery using the rule of mixtures,
microhardness indentation can also be used to determine the recrystallized volume
fraction. The multiple orientation image method in combination with microhardness test
is employed throughout this work to determine the recrystallization kinetics.
In Chapter 3, a physical model to predict the recrystallization kinetics of single phase
polycrystalline metals, based on a single grain representation of deformed
microstructure (characterized by a mean subgrain size and mean misorientation of
subgrain boundaries), is presented. The model takes into account the grain geometry,
the position and the density of the nucleation sites. The selected geometry is a regular
Summary 174
tetrakaidecahedron, combining topological features of a random Voronoi distribution
characteristic for polycrystalline material with the advantages of a single-grain
calculation. The model employs empirically determined relationships from existing
literature to describe the deformed microstructure and in so doing, enables the
prediction of the recrystallization behavior when only the deformation strain and the
recrystallization temperature are known. The boundary mobility and the driving force as
well as the nucleation density are related to the true plastic strain of deformation
through the microstructure. The model also describes the effects of concurrent recovery
on the overall recrystallization kinetics. The model predictions show that the grain
geometry change during deformation has a large effect on the recrystallization kinetics.
Chapter 3 lays the foundation for the kinetic model of the single grain approach. It is
further modified to incorporate the textural components and then applied to predict the
recrystallization kinetics in the AA1050 after cold deformation in Chapter 4. After
introducing the deformation inhomogenity from grain to grain due to difference in
Taylor factor, the single grain approach could predict the JMAK exponent of the
recrystallization kinetics correctly. The simulation shows that the JMAK exponent
depends on the grain geometry, nucleation site density, the initial grain size prior to
deformation and the main textural components after deformation. The effect of grain
size on recrystallization kinetics is predicted to depend on the amount of the prior strain
applied. This prediction is well supported by literature data.
In order to explain the difference between the laboratory simulation results and real
industrial processing, in Chapter 5 FEM is applied to calculate and compare the strain,
strain rate and stress distributions during plane strain compression in Gleeble 3500
stimulator and in the industrial hot rolling operation. It is shown that the evolution of
the equivalent strain rate and temperature during hot rolling operation and in constant
nominal strain rate PSC testing is quite different. In the hot rolling operation, the
highest strain rate distribution is concentrated just beneath the surface of the slab around
the entry point. The strain rate in the center of the rolling slab experiences a peak
variation, i.e. first increases and then decreases as the slab passes the roll gap. The
equivalent stress, which is a combined effect of the strain rate, strain and the
temperature, reaches a maximum values at about one third to two third of the arc contact
length from the entry, the position of the peak corresponding to that of the equivalent
strain rate. While in the PSC experiment the distribution of the equivalent strain, strain
rate and stress depend on the ratio of the tool width to the thickness of the sample in the
compression direction and the friction at the surfaces between the tool and the
specimen. In Chapter 6, the actual distribution of the strain, strain rate and stress in the
PSC samples, as determined by using FEM simulation were used as input parameters
for the recrystallization kinetic model.
Summary 175
The physical model based on the single grain approach is refined and extended to apply
to the prediction of the recrystallization kinetics of a single-phase metal following hot
deformation in Chapter 6. The model accounts properly for the effect of the concurrent
recovery and textural components in the deformed microstructure on the
recrystallization kinetics. As a physical model, all the input parameters are related to the
microstructure and can be determined from experiment only leaving the pre-exponent
factor of the mobility as an adjustable parameter. Experimental work is conducted on
the recovery and recrystallization kinetics in AA1050 following plastic deformation at
elevated temperatures as encountered during break down rolling to extract the physical
input parameters and to validate the model. The Plane strain compression (PSC) test is
used to simulate the hot rolling deformation. The predictions are in good agreement
with the experimental results.
As for most of industrial alloys such as AA3003, decomposition, recovery and
recrystallization take place in the same temperature range. Therefore, the interaction
between precipitation and recrystallization has profound implications for the control of
the microstructure. When recrystallization starts earlier than precipitation, the
recrystallization kinetics is fast and the finial recrystallized grain size is usually small.
However when precipitation starts earlier than recrystallization, recrystallization process
is sluggish and coarse-grained finial structure is due to obtain. In Chapter 7, nucleation
kinetics of the precipitation of MnAl
6
and MnAl
12
in the aluminum alloy AA3003 alloy
have been investigated experimentally and theoretically. The results show that cold
rolling enhances the rate of precipitation and this effect increases as the annealing
temperature decreases. Micro-segregation of the Mn solute atom at the dislocation
network during cold deformation is found to have a significant effect on the nucleation
kinetics of the precipitation of MnAl
6
and MnAl
12
in AA3003 in addition to the effect of
dislocations, which increase the nucleation site density and reduce the nucleation
barrier. A model to predict the start times of the precipitation during isothermal holding
is constructed by considering the effects of dislocations and recovery as well as micro-
segregation of Mn on the nucleation kinetics of precipitates.
In Chapter 8, experiments on AA3003 are described which were performed to study the
effect of the hot deformation parameters on the precipitation and recrystallization
behavior. It is shown that the solute content of Mn in the matrix has a large effect on the
deformation, decomposition and softening behavior of the alloy. The decomposition
process retards the softening processes and the later, in turn, reduce the rate for the
formation of the precipitates. Dynamic precipitation during concurrent deformation
depends on the strain rate and deformation temperature. The static precipitation during
annealing after deformation is enhanced by deformation but weakened by recovery and
Summary 176
recrystallization. The softening kinetics is either fast or sluggish depending on the pre-
heat treatments, deformation conditions and the annealing temperature. When
recrystallization is slow the contribution of the recovery to the softening can be as high
as 70%. When recrystallization is fast recovery is responsible for only ~25% of the
softening. Recrystallization kinetics is not simply proportional to the value of the Zener-
Hollomon parameter. Particle stimulated nucleation (PSN) is believed to be an
important nucleation mechanism in the cold deformed samples annealed at higher
temperatures (>500C). However, nucleation by subgrain growth and strain induced
boundary migration (SIBM) are the dominant nucleation mechanisms in the hot
deformed samples annealed in the temperature range 450-540C.
In this thesis it has been shown that the use of the single grain approach provides a new
insight into the recrystallization kinetics in a single phase metals after hot or cold
deformation. When combined with a precipitation model it has the potential to predict
the recrystallization kinetics in supersaturated alloys. The models form a strong
foundation for larger scale modeling of the thermomechanical behavior of industrial
aluminum alloys. However, in order to use the model for quantitative predictions, more
experimental data on the grain boundary mobility are required.
S.P. Chen
Samenvatting
Het modelleren van de microstructuurontwikkeling tijdens thermomechanische
processen wordt erkend als een waardevolle procedure voor het optimaliseren van de
procesparameters in de staalindustrie, terwijl dergelijke modelvorming voor aluminium
momenteel een belangrijk onderwerp van onderzoek is. Herstel, rekristallisatie en
korrelgroei zijn de voornaamste elementen tijdens het proces van aluminiumproductie.
In dit proefschrift wordt onderzoek beschreven dat is gericht op het herstel- en
rekristallisatiegedrag van aluminiumlegeringen, met bijzondere aandacht voor de
kinetiek van statisch herstel en rekristallisatie na deformatie op hoge temperaturen.
Het doel van het onderzoek is het bestuderen van herstel- en rekristallisatieprocessen in
aluminiumlegeringen als functie van de deformatiecondities, de thermische
geschiedenis, de chemische samenstelling en de aanwezigheid van precipitaten, hetgeen
beoogd wordt te leiden tot een fysisch model dat de rekristallisatiekinetiek voorspelt
tijdens thermomechanische behandeling. In het onderzoek zijn zowel experimenten
uitgevoerd als modellen ontwikkeld om dit doel te bereiken.
Na een algemene inleiding wordt in hoofdstuk 2 een quantitatieve studie beschreven van
het rekristallisatiegedrag in AA1050, met behulp van verschillende experimentele
technieken: optische microscopie, Electron Back Scattering Diffraction (EBSD) en
microhardheidsmetingen. Gebaseerd op dit onderzoek wordt een nieuwe methodologie
voorgesteld voor de bepaling van de volumefractie gerekristalliseerd materiaal met
behulp van optische microscopie aan anodisch getste aluminiumlegeringen. De
methode maakt gebruik van een samengesteld beeld, verkregen uit verschillende
microstructuurfotos genomen bij een reeks van orintaties. Deze multiple orientation
imaging methode resulteert consistent in een volumefractie gerekristalliseerd materiaal
die significant hoger is dan de volumefractie bepaald aan de hand van een enkele
opname. De resultaten van de multiple orientation imaging methode en de Orientation
Imaging Microscopy (OIM) vertonen een goede overeenkomst. Bovendien wordt
aangetoond dat ook microhardheidsmetingen, gebruik makend van een mengregel en na
aftrek van de effecten van gelijktijdig herstel, ook gebruikt kunnen worden om de
fractie gerekristalliseerd materiaal te bepalen. De multiple orientation imaging methode
wordt, in combinatie met microhardheidsmetingen, in de hele dissertatie gebruikt om de
rekristallisatiekinetiek te bepalen.
Samenvatting 178
In hoofdstuk 3 wordt een fysisch model gepresenteerd, dat de rekristallisatiekinetiek
voor nfasig polykristallijn materiaal voorspelt. Het model is gebaseerd op een
weergave van de gedeformeerde microstructuur door een enkele korrel, gekarakteriseerd
door de gemiddelde afmeting van de subkorrels en de gemiddelde misorintatie over
een subkorrelgrens. Het model neemt de korrelgeometrie en de kiemposities en -
dichtheid in rekening. De gekozen geometrie is een regelmatig tetrakaidecahedron,
hetgeen de topologische eigenschappen van een willekeurige Voronoi distributie voor
polykristallijn materiaal combineert met de voordelen van berekeningen aan een enkele
korrel. Het model gebruikt empirisch bepaalde verbanden uit de literatuur om de
gedeformeerde microstructuur te beschrijven, en is zo in staat om het
rekristallisatiegedrag te voorspellen op basis van slechts de plastische rek en de
rekristallisatietemperatuur. De grensvlakmobiliteit, de drijvende kracht en de
kiemdichtheid zijn gerelateerd aan de ware plastische rek via de microstructuur. Het
model beschrijft ook de effecten van gelijktijdig herstel op de rekristallisatiekinetiek.
Het model laat zien dat de verandering van de korrelgeometrie tijdens de deformatie een
grote invloed heeft op de rekristallisatiekinetiek.
Hoofdstuk 3 legt de basis voor het kinetiekmodel gebaseerd op de single-grain
benadering, dat toegepast wordt in de volgende hoofdstukken. Het model wordt verder
ontwikkeld door textuurcomponenten in rekening te brengen, en vervolgens toegepast in
hoofdstuk 4 voor de voorspelling van de rekristallisatiekinetiek in AA1050 na koude
deformatie. Variaties in de deformatie per korrel worden in rekening gebracht via
verschillen in de Taylor-factor, waarna de single-grain benadering een goede
voorspelling levert van de exponent in de Johnson-Mehl-Avrami-Kolmogorov (JMAK)
beschrijving van de rekristallisatiekinetiek. De simulaties laten zien dat de JMAK-
exponent afhangt van de korrelgeometrie, de dichtheid van kiemplaatsen, de
oorspronkelijke korrelgrootte (vr deformatie), en de belangrijkste textuur
componenten na deformatie. Er wordt door het model voorspeld dat de invloed van de
korrelgrootte op de rekristallisatiekinetiek afhangt van de hoeveelheid toegepaste
deformatie. Deze voorspelling is in overeenstemming met gegevens in de literatuur.
Om de verschillen tussen laboratoriumexperimenten en industrile procescondities te
bestuderen wordt in hoofdstuk 5 Eindige-Elementen Modellering (Finite-Element
Modelling, FEM) toegepast om de rek, de reksnelheid en de spanningsverdelingen
gedurende plane-strain compressie (PSC) uit te rekenen en te vergelijken in een Gleeble
3500 simulator en onder de condities van het industrile warmwalsproces. Er wordt
aangetoond dat de ontwikkeling van de equivalente reksnelheid en de temperatuur heel
verschillend is gedurende het warmwalsproces en gedurende PSC tests bij nominaal
constante reksnelheid. Gedurende het warmwalsproces treedt de maximale reksnelheid
Samenvatting 179
op vlak onder het oppervlak van de plaat, dicht bij het punt waar de plaat de wals ingaat.
De maximale variatie in reksnelheid wordt gevonden in het midden van de plaat: eerst
neemt de reksnelheid toe; later vindt een afname plaats als de plaat de wals passeert. De
equivalente spanning, die afhangt van een combinatie van rek, reksnelheid en
temperatuur, bereikt een maximum op ongeveer een derde tot twee-derde van het
walscontact vanaf het beginpunt, waarbij het maximum overeenkomt met het maximum
in equivalente reksnelheid. In een PSC experiment hangt de verdeling van de
equivalente rek, reksnelheid en spanning af van de verhouding van de walshoogte tot de
dikte van de plaat, en van de wrijving tussen het instrument en het preparaat. In
hoofdstuk 6 wordt de eigenlijke verdeling van rek, reksnelheid en spanning in PSC
preparaten, bepaald met FEM simulaties, gebruikt als invoergegevens voor het model
voor de rekristallisatiekinetiek.
Het fysische model gebaseerd op de single-grain benadering is verder verfijnd en
ontwikkeld in de toepassing voor de rekristallisatiekinetiek voor een nfasig metaal na
warmdeformatie in hoofdstuk 6. Het model geeft een goede representatie van de
effecten van gelijktijdig herstel en de textuurcomponenten in de gedeformeerde
microstructuur op de rekristallisatiekinetiek. Omdat het model een fysische basis heeft
kunnen alle invoergegevens gerelateerd worden aan de microstructuur, en verkregen
worden uit experimenten, waarbij alleen de pre-exponentile factor van de mobiliteit
een aan te passen parameter is. Om de fysische invoergegevens te bepalen en het model
te valideren, zijn experimenten uitgevoerd aan de rekristallisatiekinetiek in AA1050 na
deformatie op hoge temperatuur, zoals toegepast wordt gedurende break-down walsen.
De plane-strain compressie test wordt gebruikt om warmwalsen te simuleren. De
modelvoorspellingen zijn goed in overeenstemming met de experimentele resultaten.
Voor de meeste industrile legeringen, zoals AA3003, vinden decompositie, herstel en
rekristallisatie plaats in hetzelfde temperatuurstraject. De interactie tussen precipitatie
en rekristallisatie heeft daarom vergaande consequenties voor de controle van de
microstructuur. Als rekristallisatie eerder start dan precipitatie is de
rekristallisatiekinetiek snel en vormt zich een fijne gerekristalliseerde microstructuur.
Als daarentegen de precipitatie eerder start dan de rekristallisatie is het
rekristallisatieproces relatief traag en wordt een vrij grove microstructuur verkregen. In
hoofdstuk 7 wordt de nucleatiekinetiek van de precipitatie van MnAl
6
en MnAl
12
in de
aluminiumlegering AA3003 experimenteel en theoretisch bestudeerd. De resultaten
laten zien dat koudwalsen de precipitatiesnelheid verhoogt, welk effect toeneemt bij
afnemende gloeitemperatuur. Microsegregatie van Mn in oplossing op het
dislokatienetwerk gedurende koude deformatie heeft een significante invloed op de
nucleatiekinetiek voor de precipitatie van MnAl
6
en MnAl
12
in AA3003. Daarbij komt
nog het effect van de dislokaties, die de dichtheid van kiemplaatsen doen toenemen en
Samenvatting 180
de barrire voor kiemvorming doen afnemen. Een model is opgesteld dat de starttijden
voor precipitatie tijdens isotherme warmtebehandeling voorspelt, gebaseerd op de
invloeden van zowel dislokaties en herstel als microsegregatie van Mn op de
nucleatiekinetiek.
In hoofdstuk 8 worden experimenten aan AA3003 beschreven, die uitgevoerd werden
om de invloed van de warmvervormingsparameters op het precipitatie- en
rekristallisatiegedrag te bestuderen. Er wordt aangetoond dat het gehalte van Mn in
oplossing in de matrix een grote invloed heeft op het gedrag van de legering in
deformatie, decompositie, en herstel en rekristallisatie. Het decompositieproces
vertraagt het herstel en rekristallisatie, die op hun beurt de vormingssnelheid van
precipitaten doen afnemen. Het verloop van dynamische precipitatie gedurende
gelijktijdige deformatie hangt af van de reksnelheid en de deformatietemperatuur.
Statische precipitatie, gedurende een gloeibehandeling na de deformatie, neemt toe door
de deformatie, maar neemt af als gevolg van herstel en rekristallisatie. De kinetiek van
herstel en rekristallisatie kan zowel snel als traag zijn, afhankelijk van de voorafgaande
warmtebehandelingen, de vervormingscondities en de gloeitemperatuur. In het geval
van trage rekristallisatie kan het aandeel van herstel wel 70% bedragen; in het geval van
trage rekristallisatie slechts 25%. De rekristallisatiekinetiek is niet simpelweg evenredig
met de Zener-Hollomon parameter. Door deeltjes gestimuleerde kiemvorming (particle-
stimulated nucleation, PSN) is waarschijnlijk een belangrijk kiemvormingsmechanisme
in koudvervormde preparaten die op een hoge temperatuur (hoger dan 500C) gegloeid
worden. In het temperatuurgebied van 450 tot 540C zijn kiemvorming door groei van
subkorrels en kiemvorming door rek genduceerde grensvlakverplaatsing (strain-
induced boundary migration, SIBM) de belangrijkste mechanismen.
In deze dissertatie wordt aangetoond dat het gebruik van een fysisch model op basis van
de single-grain benadering nieuw inzicht levert in de rekristallisatiekinetiek in nfasig
metaal na koude of warme deformatie. Als dit model gecombineerd wordt met een
precipitatiemodel ontstaat de mogelijkheid om de rekristallisatiekinetiek in
oververzadigde legeringen te voorspellen. De modellen vormen een sterke basis voor de
modellering van het thermomechanische gedrag van industrile aluminiumlegeringen op
een grotere schaal. Voor het gebruik van de modellen voor quantitatieve voorspellingen
zijn echter meer experimentele gegevens met betrekking tot de grensvlakmobiliteit
nodig.
181
Publications
S.P. Chen, Z.J. Lok and S. van der Zwaag
,
On the precipitation and recrystallization
behavior in an AA3003 following hot deformation, in Proceedings of int. conf. on
Thermomechanical processing: Mechanics, Microstructure and Control, 23-26 June
2002, The university of Sheffield, England, 171-180.
S.P. Chen, N.C.W. Kuijpers and S. van der Zwaag, Effect of microsegregation and
dislocations on the nucleation kinetics of precipitation in aluminum alloy AA3003,
Mater. Sci. Eng. A341, 2003, 296-306.
S.P. Chen, D.N. Hanlon, S. van der Zwaag, Y.T. Pei, and J.TH.M. de Hosson,
Quantification of the recrystallization behavior in Al-alloy AA1050, J. Mat. Sci., 37,
2002, 989-995.
S.P. Chen, I. Todd, and S. van der Zwaag, Modeling the kinetics of grain boundary
nucleated recrystallization processes, Metall. Mater. Trans. 33A, March, 2002, 529-539.
S.P. Chen and S. van der Zwaag, EBSD study on the recrystallization behavior in
AA1050, in 'Nederlandse Vereniging Voor Microscopie, Jaarboek 2001', 2001, 40-41.
S.P. Chen, T. Zuidwijk, and S. van der Zwaag, Recovery and recrystallization kinetics
in AA 1050 after simulated break down rolling, in 'Recrystallization and Grain Growth,
Proceedings of the First Joint International Conference', G. Gottstein and D.A. Molodov
(Eds), Springer, Volume 2, 2001, 821-830.
S.P. Chen, Y. Van Leeuwen, S.B. Davenport, D.N. Hanlon, and S. van der Zwaag, in
'Microstructures, Mechanical Properties and Processes', The effect of grain geometry on
the kinetics of recrystallization processes, Y. Brechet (Eds), Euromat1999, 3, 1999, 96-
101.
S.P. Chen, M.S. Vossenberg, F.J. Vermolen, J. van de Langkruis, and S. van der Zwaag,
Dissolution of particles in an Al-Mg-Si alloy during DSC runs, Mater. Sci. Eng. A, 272,
1999, 250-256.
S.P. Chen, K.M. Mussert, and S. van der Zwaag, Precipitation Kinetics in Al6061 and
an Al6061-alumina Particle Composite, J. Mat. Sci., 33, 1998, 4477-4483.
182
S.P. Chen and S. van der Zwaag, A comparative FE study of hot rolling and PSC testing
of AA1050, accepted for publication in J. Test. and Eval., 2003.
S.P. Chen and S. van der Zwaag, A single grain approach applied to modeling
recrystallization kinetics in single-phase metal, submitted to Metall. Mater. Trans. A,
2002.
S.P. Chen, A. Miroux and S. van der Zwaag, Modeling recrystallization kinetics in
AA1050 following simulated break down rolling, to be submitted to Acta Mater. 2003.
S.P. Chen, Z.J. Lok, A. Miroux and S. van der Zwaag,

The

effect of dispersoids and
concurrent precipitation on recrystallization kinetics in hot deformed AA3003, in
preparation.
183
Acknowledgements
The research described in this thesis was carried out at the Laboratory for Materials
Science, Delft University of Technology. The contributions of many people to the
development of this thesis are greatly acknowledged.
I am extremely grateful to my promoter Prof. Dr. ir. S. van der Zwaag, it was his kind
acceptance that gave me an opportunity to pursue my Ph. D degree at the Delft
University of Technology. His inspirational guidance, continuing support and
enthusiastic encouragement were tremendous for me. I also appreciated very much for
kind care that he gave to my family.
I wish to express my gratitude to Dr. Yvonne van Leeuwen for her phase transformation
model. The recrystallization model in this thesis is developed from the phase
transformation model. I am grateful to Dr. J. Wang and N.C.W. Kuijpers for beneficial
discussions on some interesting topics, to Dr. J. Sietsma, R. Lok and S. Kruijver for
translation of the summary and propositions from English into Dutch, to Dr. A. Miroux
for critical reading of the manuscript.
I would like to acknowledge all technicians in the Group of Microstructure Control of
Metals (MCM), particularly, T. de Haan, N. Geerlofs, T. Zuidwijk and E. Peekstok.
Their good cooperation was indispensable for the success of the project.
I would like to thank all the present and former colleagues in MCM Group for their help
and their discussions about my research and various other subjects. Also thanks go to
Drs. T. Hurd, M.R. van der Winden, J. van de Langkruis and C. Liu from the Aluminum
group, Corus R&D for their invaluable discussions and suggestions in the project.
Further I would like to thank all the members in the Aluminium Cluster of NIMR for
the beneficial questions and discussions during and after cluster meetings.
I would like to take this opportunity to acknowledge The Netherlands Institute for Metal
Research for offering me this rewarding position.
Special thanks go to my parents for their enlightening: effort, endeavor and endurance.
Last, but surely not least, I am greatly indebted to my wife, Ying, to my daughters, Sijia
and Silu, for their love, patience, understanding, and support in every aspect.
184
185
About the author
Shangping Chen was born on August 25, 1962 in Shaanxi Province, P. R. China. He
obtained a BSc degree in 1983 and a MSc degree in 1987 in Materials Science and
Engineering in Xian Jiaotong university, Shaanxi, China. He was employed as a
research assistant in 1988, and a lecturer in 1990, and an associate professor in 1996, at
Department of Mechanical Engineering, Shaanxi Institute of Technology. In 1997, he
was awarded to work as a Visiting Scientist for one year in Laboratory for Materials
Science, Delft University of Technology, the Netherlands, within a scientific
cooperation project between the Royal Dutch academy of sciences and the Chinese
academy of sciences. Since 1
st
December 1998, he has been studying for a PhD degree
in the Netherlands Institute for Metal Research (NIMR) under the guidance of Prof.
dr.ir. S.Van der Zwaag. The main research work during these four years is documented
in the thesis entitled Recovery and Recrystallization kinetics in AA1050 and AA3003
Aluminium Alloys.

Recovery and Recrystallization Kinetics in AA1050 and AA3003 Aluminium Alloys

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Recovery and Recrystallization Kinetics in AA1050 and AA3003 Aluminium Alloys

Enviado por

Direitos autorais:

Formatos disponíveis

Recovery and Recrystallization Kinetics in

AA1050 and AA3003 Aluminium Alloys

(which is calculated from the current experiment and this value

when calculating the nucleation site

. The corresponding nucleation site density and

, resulting in an apparent activation energy of

, the equilibrium concentrations of Mn at the annealing temperature. C

, the saturation level of the conductivity of the fully

can be calculated from equation (19) by inputting the solubility of Mn

against temperature is shown in Fig. 3, too. The precipitation fraction as a function

is the saturation level of the resistivity of

are the same as above. k and

, are the concentrations of Mn in the matrix in the as-deformed

can be calculated from the solvus line as in

, the concentration of Mn in matrix in

Você também pode gostar