Alain Hazotte - Solid State Transformation and Heat Treatment-Wiley-VCH (2005)

I
Solid State Transformation

and Heat Treatment
Edited by
Alain Hazotte
Deutsche Gesellschaft
für Materialkunde e.V.
Solid State Transformation and Heat Treatment. Edited by Alain Hazotte

Copyright  2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 3-527-31007-X
II
Euromat is the biennal meeting of the Federation of European Materials Societies (FEMS) con-
stituted by its 24 member societies in Europe. The 2003 meeting took place in Lausanne,
Switzerland, and was organised by the French, the German and the Swiss member societies:
Société Française de Métallurgie et de Matériaux (SF2M)

Dezutsche Gesellschaft für Materialkunde (DGM)
Schweizerischer Verband für Materialtechnik (SVMT)
The scientific programme of the EUROMAT 2003 congress was divided into 16 topics that in
turn were substructured into 47 symposia. There will be no publication of a complete set of pro-
ceedings. The present volume of the Euromat Publication series refers to selected papers of the
symposium
Solid State Transformation, Heat Treatment (S3)

III
Solid State Transformation

and Heat Treatment
Edited by
Alain Hazotte
Deutsche Gesellschaft
für Materialkunde e.V.
IV
Editor :
Prof. Alain Hazotte
Laboratoire d’Etude des Textures et Applications aux Matériaux
Université de Metz, Ile du Saulcy
57015 Metz Cedex 01
France
This book was carefully produced. Nevertheless, editor, authors, and publisher do not warrant the information
contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illustrations,
procedural details or otheritems may inadvertently be inaccurate.
Library of Congress Card No.: applied for
British Library Cataloguing-in-Publication Data:

A catalogue record for this book is aailable from the British Library
Bibliografic information published by Die Deutsche Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliografic data is available in the Internet at <http://dnb.ddb.de>.
ISBN 3-527-31007-X
Printed on acid-free paper
Printed in the Federal Republic of Germany
All rights reserved (including those of translation in other languages). No part of this book may be reproduced in
any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into machine language
without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when
not specifically marked as such, are not to be considered unprotected by law.
Composition: W.G.V. Verlagsdienstleistungen GmbH, Weinheim

Printing: betz-druck GmbH, Darmstadt
Bookbinding: J. Schäffer GmbH, Grünstadt
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

V
Preface
The final microstructures of engineering materials are very often obtained through solid state
transformation process(es). A good knowledge of the basic transformation mechanisms, associ-
ated with a tight control of the relevant process parameters, are vital to optimise the material
properties. Thus, one symposium of Euromat 2003, the 8th European Congress and Exhibition
on Advanced Materials and Processes held in Lausanne (CH) on 1-5 september 2003, was logi-
cally devoted to “Solid State Transformations and Heat Treatments”. This symposium was open
to experimental and/or modelling works increasing this knowledge or/and explicitly making use
of it for process/property optimisation. About sixty oral and poster contributions were given.
Both diffusion-controlled and diffusionless transformations were concerned, covering a large
range of metallic or non metallic materials. The sessions gave rise to rich and animated debates
and drew the major directions of future development, specially in the field of transformation
modelling.
This book proposes a selection of original scientific articles relative to communications pre-
sented during this EUROMAT Symposium. Papers are classified in four complementary topics :
• Precipitation and dissolution processes,
• Martensitic and Stress Sensitive Transformations (here ‘stress’ has a large meaning, includ-
ing any mechanical or physical active fields)
• Crystallization (transformation from an amorphous to a crystalline structure)
• Recrystallization and Grain Size Control
I would like to thank all the contributors to this EUROMAT Symposium, even if they finally–
for various reasons- do not appear in this book. I hope that its reader will find as many scientific
interest than I found in its compilation.
Metz, 13 October 2004
A. HAZOTTE
VII
Contents
I Precipitation and Diffusion Processes ................................................................................... 1
Characterization and Modeling of NbC Precipitation in Model Microalloyed Steels................. 3

Perrard, F., Deschamps, A., Donnadieu, P., Bley, F., LTCPM-ENSEEG, Grenoble (F)
Maugis, P., IRSID, Maizieres les Metz (F)
The Effect of Nickel on M23C6 Coarsening Rates in 12CrMoVNb Steels ............................... 12

Vodárek, V., VITKOVICE - Research and Development, Ostrava (Cz)
Strang, A., Department of Engineering, University of Leicester, Leicester (GB)
Effect of Initial Melt Undercooling on Solid-State Recrystallization and Grain Growth

Processes in as-Solidified Superalloy ........................................................................................ 19
Liu, F., Institut für Materialphysik, Universität Göttingen, Göttingen (D) and
State Key Laboratory of Solidification Processing, Northwestern Polytechnical
University, Xi'an, Shaanxi (RC)
Kirchheim, R., Institut für Materialphysik, Universität Göttingen, Göttingen (D)
Yang, G., State Key Laboratory of Solidification Processing, Northwestern
Polytechnical University, Xi'an, Shaanxi (RC)
Modelling the Diffusional Transformation by Application of the Thermodynamic

Extremal Principle ..................................................................................................................... 26
Svoboda, J., Institute of Physics of Materials, Academy of Science of the Czech
Republic, Brno(Cz)
Gamsjäger, E., Fischer, F. D., Institute of Mechanics, Montanuniversität
Leoben, Leoben (A)
Phase Transformations in a Co-Cu-Ni Alloy............................................................................. 34

Guillon, I., LPCES, CNRS, UMR 8648, ICMMO. bât 410, Université de Paris Sud,
Orsay (F) and Laboratoire LURE, CNRS, bât 209d, Orsay (F)
Servant, C., LPCES, CNRS, UMR 8648, ICMMO. bât 410, Université de Paris Sud,
Orsay (F)
Lyon, O., Laboratoire LURE, CNRS, bât 209d, Orsay (F)
The Coupling of Interphase Boundary Migration and Precipitation: Example of a

Microalloyed Nb-Containing Steel. ........................................................................................... 42
Hutchinson, C. R., School of Physics and Materials Engineering, Monash University,
Clayton, Victoria (AUS)
Brechet, Y., Laboratoire de Thermodynamique et Physico-Chimie Métallurgiques,
Domaine Universitaire, St. Martin d'Hères (F)
VIII
Advanced Models for Particle Dissolution in Multi-Component Alloys .................................. 53

Vermolen, F. J., Vuik, K., Department of Applied Mathematical Analysis, Delft
University of Technology, Delft (NL)
van der Zwaag, S., Department of Aerospace Engineering, Delft University of
Technology, Delft (NL)
Modeling of the Precipitation Evolution During Non-isothermal Heat Treatments

in an Al Zn Mg Alloy................................................................................................................. 61
Nicolas, M., Deschamps, A., LTPCM, UMR CNRS 5614, Institut National
Polytechnique de Grenoble, Saint Martin d'Hères (F)
Elastic Strain Energy Study of Directional Coarsening of J' Precipitates in

Single Crystal Superalloys: A 3D Finite Element Analysis ...................................................... 69
Chen, W., Neumann, W., Institute of Physics, Humboldt University of Berlin, Berlin (D)
Liu, Q. K. K., Schumacher, G., Wanderka, N., Department of Structure Research,
Hahn-Meitner Institute Berlin GmbH, Berlin (D)
II Martensitic and Stress Sensitive Transformations ........................................................... 77
Possibilities for the Selective Improvement of the Mechanical Properties of TRIP

Steels Considering as Example Two Steels with Different Carbon Contents .......................... 79
Röthler, B., Traint, S., Pichler, A., Voestalpine Stahl GmbH, Linz (A)
Micromechanical Study of the Martensitic Transformation in TRIP-Assisted

Multiphase Steels ....................................................................................................................... 87
Van Rompaey, T., Blanpain, B., Wollants, P., Department of Metallurgy and
Materials Engineering, Katholieke Universiteit Leuven, Leuven (B)
Lani, F., Jacques, P., Pardoen, T., Division of Physico-Chemistry and Engineering
of Metals, Université Catholique de Louvain-la-Neuve, Louvain-la-Neuve (B)
Effect of Carbon and Nitrogen on the Shape Memory Effect in FeMnSiCrNi SMAs .............. 95
Van Caenegem, N., Duprez, L., De Cooman, B. C., Laboratory for Iron and
Steelmaking, Ghent University, Ghent (B)
Segers, D., Department of Subatomic and Radiation Physics, Ghent
University, Ghent (B)
Rapid Full Annealing under High Magnetic Field .................................................................. 103

Zhang, Y., School of Materials and Metallurgy, Northeastern University,
Shenyang (RC) and LETAM, CNRS-UMR 7078, Université de Metz, Metz (F)
He, C., School of Materials and Metallurgy, Northeastern University,
Shenyang (RC)
Zhao, X., Zuo, L., He, J., Esling, C., LETAM, CNRS-UMR 7078, Université
de Metz, Metz (F)
Nishijima, G., Zhang, T., Watanabe, K., Institute for Materials Research,
Tohoku University, Sendai (J)
IX
III Crystallization................................................................................................................... 111
Crystallization Kinetics and Phase Selection in Amorphous Al-Alloys.................................. 113

Boucharat, N., Rösner, H., Wilde, G., Forschungszentrum Karlsruhe, Institut
für Nanotechnologie, Karlsruhe (D)
Temperature Dependent Critical Size of Embedded Silicon Nano-crystals: A

Molecular Dynamics Study ..................................................................................................... 122
Chiu, Y.-T., Yeh, J.-T., Industrial Technology Research Institute, Taiwan (RC)
Effects of Soft-Impingement and Non-random Nucleation on the Kinetics and

Microstructural Development of Primary Crystallization. ...................................................... 126
Crespo, D., González-Cinca, R., Bruna, P., Departament de Fìsica Aplicada,
EPSC, Universitat Politècnica de Catalunya. Castelldefels (E)
Pineda, E., Departament de Física i Enginyeria Nuclear, ESAB, Universitat
Politècnica de Catalunya. Barcelona (E)
Mechanical Spectroscopy of Quasicrystal Formation from Amorphous Ti- and

Zr-based Alloys........................................................................................................................ 135
Sinning, H.-R., Golovin, I. S., Institut für Werkstoffe, Technische Universität
Braunschweig, Braunschweig (D)
Jianu, A., National Institute of Materials Physics, Bucharest-Magurele (Ro)
Solid-Solid Phase Transformation of Amorphous Titanium Based Alloys............................. 145

Jianu, A., National Institute of Materials Physics, Bucharest-Magurele (Ro)
Sinning, H.-R., Golovin, I. S., Institut für Werkstoffe, Technische Universität
Burkel, E., Mathematisch-Naturwissenschaftliche Fakultät, Universität Rostock,
Rostock (D)
IV Recrystallization and Grain Size Control ...................................................................... 153
Energetics of Three-Dimensional Network Structures .......................................................... 155

Glicksman, M. E., Rensselaer Polytechnic Institute, Troy, N.Y. (USA)
Tempering Peculiarities of Fe-C Alloy by Severe Plastic Deformation.................................. 165

Ivanisenko, Y., Institut für Nanotechnology, Forschungszentrum Karlsruhe,
Karlsruhe (D)
Golovin, I. S., Sinning, H.-R., Institut für Werkstoffe, Technische Universität
Fecht, H.-J., Institut für Nanotechnology, Forschungszentrum Karlsruhe,
Karlsruhe (D) and Division of Materials, Ulm University, Ulm (D)
X
Investigation of Ferrite Softening Processes in a 0.2 wt% C 1.5 wt% Mn Steel

after Hot Deformation.............................................................................................................. 175
Smith, A., Luo, H., Department of Materials Science & Engineering, Delft
University of Technology, Delft (NL)
Hanlon, D. N., IJmuiden Technology Centre, Corus Research, Development
and Technology, IJmuiden (NL)
Sietsma, J., Department of Materials Science & Engineering, Delft University
of Technology, Delft (NL)
van der Zwaag, S., Faculty of Aerospace Engineering, Delft University of
Technology, Delft (NL)
A Preliminary Study on the Grain Refinement in Ultra Low Carbon Steels by

Cyclic Heat Treatment ............................................................................................................. 181
Bayraktar, E., SUPMECA-LISMMA/Paris, School of Mechanical Engineering,
Saint Ouen (F)
Chevalier, J. P., CNAM, Paris (F) and CECM-CNRS, Vitry (F)
Author Index ............................................................................................................................ 189

Subject Index ........................................................................................................................... 191
39
I Precipitation and Diffusion Processes

ISBN: 3-527-31007-X
3
Characterization and Modeling of NbC Precipitation in Model

Microalloyed Steels
F. Perrard*, A. Deschamps*, P. Maugis**, P. Donnadieu*, F. Bley*

* LTCPM –ENSEEG, Grenoble France
** IRSID, Maizières les Metz France
1 Introduction
Precipitation of carbides or carbonitrides containing Nb, Ti or V is extensively used to improve

mechanical properties of microalloyed steels. Carbides are used in austenite to control grain si-
ze, and in ferrite (during coiling) as hardening precipitation [1]. NbC precipitation kinetics in
ferrite has only been studied to a very limited extent [2, 3]. Studying NbC precipitation in ferrite
is actually a difficult task, owing to the complex grain structure (small grains and high dislocati-
on density) and to the low volume fraction of precipitates (10–4–10–3).
In this study, Small-Angle Neutron Scattering (SANS) experiments are used to follow NbC
precipitation kinetics during isothermal heat treatments in two model alloys. Small-Angle Scat-
tering is one of the most appropriate techniques to follow the evolution of the precipitation state
by the determination of statistically relevant values of size and volume fraction of particles.
These results, assorted with literature data, are used to predict precipitation kinetics.
2 Materials and Experimental Technique
2.1 Samples
Two high purity Fe-Nb-C alloys with different niobium content have been studied. Chemical
compositions are given in table 1.
Table 1: Chemical compositions of the steels (weight ppm)
Nb C N S P Al O
Low Nb 400 58 9 10 10 90 53
High Nb 790 110 10 23 10 60 13
Alloys were prepared by vacuum induction melting, cast into ingots, and hot rolled from
50mm to 5mm thickness. After homogenization in austenite during 45 min at 1240 °C and wa-
ter quench, samples (15 × 12 × 3 mm3) were heat treated at three different temperatures
(600, 700 and 800 °C) and various times. All samples were first heat treated in salt bath to keep
a constant heating rate. Longer ageing (i.e. more than one hour) were completed by heat treat-
ments in silica tubes sealed under vacuum.

ISBN: 3-527-31007-X
4
2.2 Small-Angle Neutron Scattering
Experiments were performed on the D22 beamline at Institut Laue Langevin (ILL, Grenoble
FRANCE). Neutrons are monochromated by a velocity selector to a mean wavelength of 6 Å,
the wavelength spread being 'O/ O= 10 %. This wavelength was chosen in order to avoid
double Bragg diffraction problems that can arise form the ferrite microstructure.
Samples were magnetized to saturation (1.3 T) in order to separate the magnetic and nuclear
scattering signals. All measurements were made using the magnetic signal that gives volume
fractions independent from the composition of the precipitates. Three sample – detector distanc-
es (from 2 to 18 m) were used together with appropriate collimation settings to give access to
scattering vectors q (q = 4 ʌ sin T/O) ranging from 0.1 nm–1 to 2.5 nm–1, i.e. mean radius ranging
approximately from 0.7 to 15 nm. Scattered neutrons were recorded with a 96 × 96 cm² active
area 3He multidetector, with a 0.75 × 0.75 cm² pixel size. Data treatment was carried out using
GRASP software developed at ILL [4]: Data files were corrected for electronic noise, detector
response and background noise. Circular averaging around the transmitted beam was carried out
and fitted with a sin²T function to separate the signal into magnetic and nuclear part. Magnetic
scattered intensity calibration was converted into absolute intensity by calibration with a chem-
ical measure of the precipitated fraction (matrix dissolution followed by chemical analysis of
the filtrate).
Precipitate size was evaluated using a pseudo-Guinier analysis: When the Guinier approxi-
mation is fulfilled, scattered intensity I(q) and Guinier radius Rg are linked by this relationship
[5]:
§ q ² Rg2 ·
I ( q ) v exp ¨ ¸ (1)
¨ 3 ¸¹
©
This means also that the Guinier radius can be determined using the scattering vector at the
maximum of Iq² vs q curve (qmax):
3
Rg (2)
qmax
This radius corresponds strictly to the Guinier radius when the Guinier approximation is re-
spected in a wide q-range. Here, this approximation is not fulfilled, which means that this radius
can only be used for comparative purposes, and should be calibrated by an other technique in
order to fully describe the reality. In this paper, radii corresponding to qmax are used to follow
the precipitation reaction. The calibration has to be done by TEM and will be published later.
3 Precipitation Kinetics
3.1 Influence of Temperature
Figure 1 displays the temperature dependence of volume fraction and mean radius versus time
for the high Nb alloy.
5
The precipitation reaction becomes faster when temperature is increasing, which indicates
that the tip of the Temperature-Time-Transformation curves is located above 800 °C.
The temperature dependant coarsening behavior is also evidenced, as well as the decrease in
equilibrium volume fraction when temperature is increased, due to the higher solubility. Finally,
a strong influence of temperature on the nucleation radius is highlighted.
0,1 12
10
0,08
8
0,06
6
0,04 600°C
700°C 4
800°C
0,02
2
0 4 5 6
0 4 5 6
1 10 100 1000 10 10 10 1 10 100 1000 10 10 10
Time (s) Time (s)
Figure 1: Temperature dependence of volume fraction and mean radius vs. time for high Nb alloy
3.2 Influence of Composition
In order to compare the behavior of the two alloys, their volume fractions are normalized to the
same value. Normalized volume fraction for the two alloys at 700 °C versus time, and mean ra-
dius evolution versus time are consigned in figure 2. The high Nb alloy shows precipitation ki-
1,2 10
1
8
0,8
6
Low Nb
0,6 High Nb
4
0,4
2
0,2
0 0
4 5 6 4 5 6
10 100 1000 10 10 10 10 100 1000 10 10 10
Time (s) Time (s)
Figure 2: Influence of composition on normalized volume fraction and mean radius vs. time at 700 °C
6
netics approximately three times faster than the low Nb alloy, whereas the size evolution is
similar.
4 Modeling of Precipitation Kinetics
4.1 Description of the Model
The present model, developed for aluminium alloys by Deschamps and Brechet [6], is partly ba-
sed on the earlier model by Langer and Schwartz, modified by Wagner and Kampmann [7]
(MLS model). In [6], the alloy was considered as a quasi-binary system, which means that an
effective diffusion coefficient was used to describe the diffusion behavior of the two precipitate
components. Subsequently, a ternary approach has been developed by Maugis et al. [8], derived
from a class model description of the precipitation kinetics [3]. In the present model, nucleation,
growth and coarsening steps are considered as coupled regimes, and divided into two stages: a
nucleation and growth regime and a growth and coarsening regime with a simple criterion for
the transition between both.
Classical assumptions are used:
• Spherical and stoechiometric particles, embedded into a homogeneous matrix.

• Growth via a diffusion controlled mechanism.
• Isotropic interfacial energy which varying with temperature.
• Local equilibrium at the precipitate/matrix interface, including Gibbs Thomson effect.
4.1.1 Nucleation and Growth
The nucleation rate is given by equation (3),
dN § 'G · § § t ··
Z E N 0 exp ¨ ¸ ¨1 exp ¨ ¸¸ (3)
dt Nucl © kT ¹ © © W ¹¹
Where N the number of precipitates per unit volume, Z the Zeldovich factor (|1 / 20), E* the
impingement rate (absorption frequency of a niobium atom), N0 the number of nucleation sites
per unit volume, 'G* the activation energy for nucleation, and W the incubation time.
Niobium diffusivity in iron DNb is very slow and will be considered as the rate controlling
process. The evolution of precipitate radius is given by equation (4).
dR DNb X ssNb X iNb

d t Growth R VFe X Nb X Nb (4)
pp i
VNbC
With X ssNb the average atomic fraction of Nb in solid solution, X iNb the atomic fraction of
Nb
Nb at the matrix/precipitate interface, X pp the atomic fraction of niobium in the precipitates
(= 1), VFe the atomic volume of ferrite and VNbC the molecular volume of NbC.
7
For ternary alloys, the interface concentration is determined by the conditions of local equi-
librium and flux compatibility at the interface. It can be shown that the Gibbs-Thomson effect
results in a radius-dependent solubility product K(R):
§R ·
K ( R) K (f)exp ¨ 0 ¸ (5)
© R¹
With K (f) X eNb X eC and R 0 2 J VNbC kT , where K(•) is the solubility product given by
the phase diagram, X eNb and X eC the equilibrium atomic fractions of Nb and C respectively, J
the precipitate/matrix interfacial energy (Jm–2) and R0 the capillarity radius.
Finally, this set of equations leads to the following precipitate radius evolution:
dR DNb X ssNb X iNb 1 dN

R ' R (6)
dt Nucl-Growth R VFe X Nb X Nb N dt
ipp i
VNbC
In this equation, R’ is the radius above which a nucleating precipitate enters the unstable re-
gime of growth. R’ is defined as the radius for which the Gibbs energy of formation
'G(R’) = 'G*–kT [9]. R’ is slightly greater than the critical radius R* and approximates to
1 kB T
R ' R .
2 SJ
4.1.2 Growth and Coarsening
When the mean radius of precipitates is larger than the critical radius, conditions for pure
growth are fulfilled, and equations (3) and (6) reduces to:
d R DNb X ssNb X iNb

°
°° d t Growth R VFe X Nb
i
® VNbC (7)
°d N
° 0
°̄ d t Growth
When the mean radius equals the critical radius, the classical LSW regime is respected
[10, 11]:
d R 4 X eNb R0 DNb § 1 ·
° ¨ ¸
° d t Coars 27 VFe X Nb © R 2 ¹
° VNbC
e
® (8)
°d N 4 DNb R0 X eNb ª R0 X ssNb § 3 · º
° « Nb ¨
N ¸ 3N »
°¯ d t Coars 27 R 3 1 X eNb ¬« R 1 X ss © 4 S R
3
¹ ¼»
8
4.1.3 Transition Criterion
The transition between nucleation-growth and growth-coarsening stages is made continuous

using a coarsening fraction ranging from 0 to 1:
2
ª § f · º
f Coars Sup «1 100 ¨ eqvGT 1¸ , 0 » (9)
«
¬ © fv ¹ »
¼
Where fveqGT is the equilibrium fraction for precipitates of radius R. The coarsening fraction
is used to weight the radius and density evolution rates:
d R dR dR
° dt 1 fCoars f Coars
° d t Growth d t Coars
® (10)
°d N f Coars
dN
°¯ d t dt Coars
4.2 Comparison between Model and Experiment
In order to adjust the model on all experimental results with the same parameters, Nb diffusion
coefficient and interfacial energy are supposed to be adjustable, other parameters being taken
from the literature:
• The chosen Nb diffusion coefficient is: DNb (ms–1) = 5 10–2 exp(–251740 / RT). This value
is in good agreement with literature data [12].
• The solubility product of the literature, proposed by Turkdogan [13] is used:
log10 K(f= –9830 / T + 4.33, with K(f) in weight percent.
• The interfacial energy is supposed to vary with temperature: J(T) = –0.0013 T + 1.74 which
gives reasonable interfacial energies for this range of temperature (between 0.4 and
0.65 Jm–2), and which gives an interfacial energy at 0 K in good agreement with literature
(1.2 Jm–2 has been found at 0 K for the chemical part of the interfacial energy by [14]).
We have found that the only way to have reasonable agreements for the two alloys and all
temperatures is to have a temperature dependant interfacial energy. This decrease with tempera-
ture, due to the entropic term is well known, but it can mask other problems of the model be-
cause interfacial energy is a crucial term in the nucleation and coarsening equations.
Comparison between SANS and modeling results for low Nb alloy is shown in figure 3.
Radius and volume fraction evolutions are in good agreement with simulation. However, it is
important to notice that simulation shows a very sharp increased compared to SANS results (see
for example figure 3 at 600 °C).
Results for high Nb alloy are summarized figure 4.
In this case, good agreement is obtained for low temperatures, but when supersaturation is
diminishing, the model has difficulties to reproduce experimental results.
9
-3
12 1 10
Model - 600°C
10 Model - 700°C -4
Model - 800°C 8 10
SANS - 600°C
8 SANS - 700°C
SANS - 800°C 6 10
-4
6
-4
4 10
4
-4
2 10
2
0
0 0 10 4 5 6
4 5 6 0,01 0,1 1 10 100 1000 10 10 10
0,01 0,1 1 10 100 1000 10 10 10
Time (s) Time (s)
Figure 3: Comparison between SANS and modeling results for low Nb alloy
-3
12 1 10
10 Model - 600°C -4
8 10
Model - 700°C
Model - 800°C
8 SANS - 600°C
-4
SANS - 700°C 6 10
SANS - 800°C
6
-4
4 10
4
-4
2 10
2
0
0 0 10 4 5 6
4 5 6
0,01 0,1 1 10 100 1000 10 10 10 0,01 0,1 1 10 100 1000 10 10 10
Time (s) Time (s)
Figure 4: Comparison between SANS and modeling results for high Nb alloy
We can summarize all these results by Temperature-Time-Transformation (TTT) curves for

both alloys (figure 5). This kind of plot gives a comparison for different steps of the reaction
(10 %, 50 % and 90 %).
It is clearly shown that when supersaturation is high (i.e. low temperatures), model and ex-
periments are in good agreement. As supersaturation is decreasing (High Nb alloy, 800 °C), the
model has difficulties to reproduce correctly the precipitation reaction. This difference at
800 °C can also come from a lack of precision of experimental results due to the very quick re-
action.
10
900 900
a Model - 10% b
850 Model - 50% 850
Model - 90%
800 SANS - 10% 800
SANS - 50%
SANS - 90%
750 750
700 700
650 650
600 600
550 550
500 500
4 5 6 7 4 5 6 7
0,1 1 10 100 1000 10 10 10 10 0,1 1 10 100 1000 10 10 10 10
Time (s) Time (s)
Figure 5: Temperature-Time-Transformation curves for (a) Low Nb alloy, (b) High Nb alloy representing the 10
%, 50 % and 90 % of completion of the precipitation reaction
It is also highlighted that the duration of the reaction, that is to say the time between the nu-
cleation and coarsening stages, is difficult to reproduce: the model is giving very condensed ki-
netics compared to experiments. In contrast, the description of size evolution is comparatively
better.
5 Conclusions
SANS experiments allows us to follow the precipitation reaction for two Fe-Nb-C alloys and
three temperatures, with very small precipitate volume fractions. Results show a nose of the
TTT curve around 800 °C and a small concentration dependence of the kinetics.
Experiments are satisfactorily reproduced by the model, with the same set of parameters for
both alloys. Interfacial energy seems to be the critical parameter for this kind of model.
The limits of this model are notably reached low supersaturations.
Improvement of this model needs to take into account the heterogeneous character of the pre-
cipitation process. The heterogeneous nature of NbC precipitates could explain the small differ-
ence in terms of reaction speed between high and low Nb alloys, and also the slower coarsening
behaviors.
6 Acknowledgements
This work is part of a French scientific program called ”CPR Precipitation”, in collaboration
with Arcelor, Pechiney, CNRS, CEA, INPG, INSA Lyon, Université de Rouen, Université Aix-
Marseille 3. IRSID-Arcelor is acknowledged for providing the material and financial support.
C. Dewhurst, as well as the whole D22-ILL team are gratefully acknowledged for their help
with the SANS experiments and data interpretation.
11
7 References
[1] T. Gladman, The Physical Metallurgy of Microalloyed Steels, The Institute of Materials,
London, 1997
[2] R. Chattopadhyay, Transactions of the Indian Institute of Metals, 1978, 31, 5–6, 322-325
[3] D. Gendt et al., in COM: Conference of Metallurgists, 2000, Ottawa, Canadian Institute
of Mining, Metallurgy and Petroleum, Montreal, Canada
[4] C. Dewhurst, http://whisky.ill.fr/lss/grasp/grasp_main.html
[5] O. Glatter and O. Kratky, Small Angle X-Ray scattering. Academic Press: London, 1982
[6] A. Deschamps and Y. Bréchet, Acta Materialia, 1999, 47, 1, 293–305
[7] R. Wagner and R. Kampmann, Homogeneous second phase precipitation, in Materials
Science and Technology, a comprehensive treatment, VCH, Weinheim, 1991
[8] P. Buessler et al., Iron and Steelmaker, 2003, 30, 9, 33–39
[9] P. Maugis and M. Gouné, in Proceedings of the International Alloy Conference, 2002,
Estoril.
[10] I.M. Lifshitz and V.V. Slyozov, J. Phys. Chem. Solids, 1961, 19, 1/2, 35–50
[11] C. Wagner, Z. Elektrochem., 1961, 65, 581
[12] J. Geise and C. Herzig, Z. Metallkde, 1985, 76, H9, 622
[13] E.T. Turkdogan, Iron and Steelmaker, 196861–75
[14] Z.-G. Yang and M. Enomoto, Materials Science and Engineering A, 2002, A332, 184–
192
12
The Effect of Nickel on M23C6 Coarsening Rates in 12CrMoVNb

Steels
V. Vodárek and A. Stranga

VITKOVICE – Research and Development, Ostrava, Czech Republic
a
Department of Engineering, University of Leicester, Leicester, United Kingdom
1 Abstract
Microstructural studies have been carried out on 12CrMoVNb steels which exhibited sigmoidal
creep rupture behavior when tested at 600 °C. This behavior is associated with marked softe-
ning and microstructural degradation of materials occurring during the creep exposure. In this
paper, the results of studies on precipitation sequences and on M23C6 coarsening rates in a series
of 12CrMoVNb steels containing 0.52, 0.76 and 1.15 wt.%Ni are reported. It has been found
that the M23C6 coarsening rate is approximately three times greater in the highest nickel material
compared with the cast with the lowest nickel content. The results show that a t1/3 law satisfacto-
rily describes the experimental data up to test durations corresponding to the observed inflexi-
ons in the creep rupture data. Beyond the sigmoidal inflexions much reduced M23C6 coarsening
rates were observed in all three casts investigated.
2 Introduction
The modified martensitic 9–12 %Cr steels are regarded as candidate materials for critical com-
ponents in the advanced fossil-fired power plants >1@. The long term creep strength and the oxi-
dation resistance in steam are the most important criteria for such materials.
The 12CrMoVNb steels were originally developed for gas turbine applications >2, 3@. Al-
though these alloys have very high creep strength for durations of 10 000 hours, at temperatures
above about 550 °C the creep strength falls rapidly as durations are extended towards the
100 000 hours which is the typical basis for power plant design. This phenomenon, which is
generally referred to as sigmoidal behavior, is closely associated with microstructural instabili-
ties >3, 4@. Creep resistance of steels in the range of dislocation creep is controlled by the rate of
recovery and recrystallization of the matrix >5@. These processes can be effectively slowed
down by interaction of precipitates with lath/subgrain boundaries. That is why a lot of attention
has to be paid to understanding of the relationships between the chemical composition of steels
and microstructural evolution during long term creep exposure. Marrison and Hogg >6@ reported
the harmful effect of nickel on the creep rupture characteristics of the 12CrMoVNb steels. In-
creased nickel contents cause accelerated microstructural degradation resulting in the sigmoidal
inflexion occurring at progressively shorter creep exposure durations >4, 6@.
In this paper the results of detailed studies on the effect of nickel on the precipitation se-
quences and on the M23C6 coarsening rates in the 12CrMoVNb steels during long term creep/
thermal exposure at 600 °C are reported.

ISBN: 3-527-31007-X
13
3 Experimental
Three casts of 12CrMoVNb steels were investigated which had the highest, mean and lowest
creep rupture strengths, as well as respectively containing the lowest, intermediate and highest
nickel contents. Details of chemical compositions of these materials are shown in Table 1.
Table 1: Chemical compositions of the 12CrMoVNb steels, wt.%

Cast C Si Ni Cr Mo V Nb N
A 0.16 0.28 0.52 11.20 0.61 0.28 0.29 0.074
B 0.14 0.37 0.76 11.10 0.57 0.36 0.32 0.062
C 0.14 0.13 1.15 11.74 0.50 0.29 0.30 0.064
Parameters of commercial quality heat treatment are shown in Table 2. All casts exhibited
sigmoidal behavior in creep rupture tests conducted at 600 °C out to durations of up to 100 000
hours, Fig. 1. As evident the sigmoidal inflexions on the creep rupture curves occur at shorter
times with the increasing nickel contents in the steels.
Table 2: Commercial heat treatments for the 12CrMoVNb steels

Cast Heat Treatment
A 1150 °C/air 650 °C/6 hours/air
B 1160 °C/air 700 °C/6 hours/air
C 1165 °C/air 675 °C/6 hours/air
400
350
300
Stress, MPa
250
200
150 Cast A
100 Cast B
50 Cast C
0
10 100 1000 10000 100000
Creep Rupture Life, Hours
Figure 1: Creep rupture properties of 12CrMoVNb steels at 600 °C
Transmission electron microscopy TEM studies were carried out on thin foils prepared
from both heads and gauge lengths of creep ruptured testpieces using a Philips CM 20 micro-
scope. The thin foils were prepared using a Struers twin jet electropolishing unit operated at
room temperature and 60 V with an electrolyte consisting of 5 % HClO4 in glacial acetic acid.
Electron diffraction and X-ray microanalysis techniques were used for the identification of mi-
14
nor phases. Quantitative evaluation of the size and the number density of minor phase particles
was done using transparencies with redrawn particles from micrographs taken at a constant
magnification. The digitized transparencies were analyzed using the feature analysis software of
a PGT IMIX EDX microanalyser. The total area of micrographs used for image analyses of in-
dividual specimens varied between 0.9 10–4 and 1.8 10–4 mm–2.
4 Results and Discussion
4.1 Precipitation Sequences
In the as-received condition all casts investigated had tempered martensitic microstructures
with no evidence of the presence of any G-ferrite. Prior austenite grain boundaries and martensi-
tic lath boundaries were decorated by M23C6 particles. Large spheroidized primary NbX partic-
les, which had not dissolved during austenitization, were randomly dispersed throughout the
matrix of each cast. Intragranular nitrogen rich M2X and secondary MX particles of the form
V,NbX were present in the matrix depending on the tempering temperature. The fraction of
secondary MX particles increased with increasing temperature of tempering >3, 4, 7@.
Microstructural studies indicated that whilst evolution of minor phases during creep expo-
sure at 600 °C was similar in each cast, the kinetics varied significantly from one cast to anoth-
er. The same minor phases were identified in both heads and gauge lengths of individual
testpieces. Comparisons of the precipitation sequences in each of the 12CrMoVNb steels inves-
tigated are shown in Fig. 2. Particles of M23C6 and NbX phases were present in all casts up to
Laves Laves
Z phase Z phase
MX MX
M2X M2X
M23C6 M23C6
NbX NbX
0 20 40 60 80 100 0 20 40 60 80 100
Exposure Duration x E-03, Hours Exposure Duration x E-03, Hours
a. b.
Laves
Z phase
MX
M2X
M23C6
NbX
0 20 40 60 80 100 Figure 2: Precipitation sequences in the
Exposure Duration x E-03, Hours
12CrMoVNb steels during thermal/creep
exposure at 600 °C, a: Cast A, b: Cast B, c:
c. Cast C
15
the longest durations of creep testing. X-ray diffraction studies indicated that whilst the volume
fraction of the M23C6 phase was approximately the same in all testpieces the fraction of primary
NbX particles was reduced with increasing time of exposure >8@. Furthermore, M2X particles
dissolved due to precipitation of fine platelets of secondary MX particles. This phase also
proved to be thermodynamically unstable, gradually redissolving at longer times of exposure at
600 °C. This was accompanied by precipitation of platelets of modified Z phase, which is a
complex nitride of chromium, niobium and vanadium >9@. The kinetics of these changes were
accelerated by nickel contents in the materials. Furthermore, in Cast A a small amount of Laves
phase of the type Fe2Mo was present after 94 000 hours.
Dissolution of the fine nitrogen rich M2X and MX phases due to precipitation of the thermo-
dynamically more stable modified Z phase and coarsening of M23C6 particles represent the most
important mechanisms of microstructural degradation of the 12CrMoVNb steels during ther-
mal/creep exposure at 600 °C >3, 10@.
4.2 M23C6 Particle Coarsening
Studies on M23C6 coarsening rates in modified 9–12 %Cr steels are complicated by the fact
that several minor phases coexist in the microstructure. Individual minor phases, especially
M23C6 and Laves phase, can not usually be discriminated on conventional bright field TEM
images. That is why most published data represent only apparent coarsening rates. The effecti-
ve discrimination of individual minor phases can be achieved by using the EFTEM technique
>11@. However, in the case of the 12CrMoVNb steels investigated most precipitates are formed
by M23C6 >8@. Due to a very low molybdenum content in the steels the amount of Laves phase is
negligible. Furthermore, the size of most M2X, NbX and Z phase particles is significantly diffe-
rent from that of M23C6 particles. That is why conventional bright field TEM images can be used
for the realistic evaluation of M23C6 coarsening rates in the 12CrMoVNb steels. The results of
measurements on the size and the number density of M23C6 particles in both heads and gauge
lengths of testpieces of Cast A are shown in Fig. 3.
As the particle size increases the particle density reduces. The results prove that the coarsen-
ing of M23C6 particles takes place from the very beginning of creep/thermal exposure at 600 °C.
125
25
100
20 Head
G.L.
-2
,m
Deq., nm
75 15
-12
NA x 10
50 Head 10
G.L.
25 5
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time to Rupture x E-03, Hours Time to Rupture x E-03, Hours
Figure 3: Effects of thermal exposure and creep strain on the size and the number density of M23C6 particles in
Cast A during creep testing at 600 °C
16
Particle coarsening was found to be consistently greater in the creep strained testpiece gauge
lengths compared with their heads, where precipitate growth was primarily due to effects of
thermal exposure.
Most of the microstructural observations on modified 9–12 % Cr steels appear to favor Ost-
wald ripening as the best description of particle coarsening, in spite of many factors which do
not satisfy the assumptions of the theory >12,13@. If particle coarsening is controlled by volume
diffusion of solutes in the matrix then the size of precipitates would be expected to increase in
accordance with an equation of the form:
dt3 d 03 K dt (1)
where dt is the precipitate size at time t, d0 is the precipitate size at t = 0 and Kd is a temperature
dependent rate constant of coarsening.
Studies on M23C6 coarsening rates were carried out on heads of testpieces because the Ost-
wald ripening theory does not take into account the effect of creep strain. Evaluation of the
coarsening rates using equation 1 revealed the presence of inflexions in the data at similar du-
rations to that found for the sigmoidal inflexions in the creep rupture data, Fig. 4.
Furthermore, the results show that Ostwald ripening satisfactorily describes the data up to
test durations corresponding to observed inflexions in the creep rupture data with the Kd value
for Cast C with the highest nickel content being approximately three times greater than that ob-
served for Cast A with the lowest nickel content, Table 3. Beyond the sigmoidal inflexions the
rate of coarsening of M23C6 particles is much reduced in all three casts investigated. At dura-
tions corresponding approximately to the sigmoidal inflexions in the creep rupture data micro-
structural investigations revealed the beginning of two important processes: modified Z phase
precipitation and recrystallization of the matrix.
Table 3: M23C6 coarsening rates in the 12CrMoVNb steels at 600 °C

Cast Kd 1030 >m3 sec–1@
A 3.1
B 5.8
C 8.7
A considerable reduction of the coarsening rate of precipitates in quenched and tempered al-
loys after recrystallization of the matrix has been observed by several authors >12@. These stud-
ies indicate that at least some degree of substructural enhancement of diffusion is responsible
for this effect. Furthermore, coarsening studies in multiphase microstructures should take into
account interactions between individual phases. It can be speculated that precipitation of modi-
fied Z phase makes a contribution to the pronounced reduction of the M23C6 coarsening rate.
Precipitation of Z phase results in a reduction of the solute concentrations in the matrix and also
differences in the interfacial energy between M23C6 and modified Z phase can not be neglected.
Proper calculations of coarsening in the multiphase systems will become possible in near fea-
ture >14@.
Coarsening of M23C6 particles was accompanied by slow changes of their compositions to-
wards the equilibrium composition up to the longest durations of testing. The results of the
EDX analyses on the M23C6 precipitates in all steels are summarized in Fig. 5. The amounts of
17
1,E+6
1,E+6
8,E+5 Kd,C
Kd,B
D3eq., nm3
6,E+5
Kd,A
4,E+5
Cast A
2,E+5 Cast B
Cast C
0,E+0
0 20000 40000 60000 80000 100000
Time to Rupture, Hours
Figure 4: Analysis of M23C6 coarsening rates in heads of creep testpieces after exposure at 600 °C
80
70
60
Wt.% Element
50
Cr
40 Fe
Mo
30
20
10
0
18 19 20 21 22
P = T(20+log t)/1000
Figure 5: Changes in Cr, Fe and Mo contents of M23C6 carbides in all three casts of the 12CrMoVNb steels
nickel in M23C6 particles approximately corresponded to the nickel content in the steels and did
not change during long term exposure. The EDX results indicate that in accordance with Ther-
mocalc >15@ predictions there is an enrichment in chromium and a corresponding depletion in
iron with increasing exposure durations. The enrichment process is driven by the fact that the
temperature of creep/thermal exposure is lower than that associated with the initial tempering of
the steels >16, 17@. It is not yet clear what effects the enrichment process may have on particle
coarsening.
5 Conclusions
The results of studies on the effect of nickel on M23C6 coarsening rates in the 12CrMoVNb
steels during long term creep/thermal exposure at 600 °C can be summarized as follows:
18
1. M23C6 particle coarsening was found to be consistently greater in the creep strained gauge
lengths compared with the heads of testpieces.
2. The t1/3 law satisfactory describes the data up to test durations corresponding to the obser-
ved inflexions in the creep rupture data.
3. Increasing nickel contents cause increased M23C6 coarsening rates. The M23C6 coarsening
rate constant Kd, determined for thermally exposed heads of testpieces, is approximately
three times greater in the high nickel 1.15 wt.% material compared with the low nickel
0.52 wt.% cast.
4. Beyond the sigmoidal inflexions in the creep rupture data much reduced M23C6 coarsening
rates were observed in all three casts investigated.
6 References
[1] M.E. Staubli, K.H. Mayer, T.U. Kern, R.W. Vanstone, in Proc. Parsons 2000, Advanced
Materials for 21st Century Turbines and Power Plants ( Eds.: A. Strang et al.), IOM, Lon-
don, 2000, 98
[2] R. W. Vanstone, in Proc. Quantitative Microscopy of High Temperature Materials (Eds.:
A. Strang et al.), IOM, Sheffield, 2001, 355
[3] A. Strang, V. Foldyna, A. Jakobová, Z. Kuboò, V. Vodárek, J. Lenert, in Proc. Advances
in Turbine Materials, Design and Manufacturing (Eds.: A. Strang et al. ), IOM, Newcastle
upon Tyne, 1997, 603
[4] A. Strang, V. Vodárek, in Proc. Materials for Advanced Power Engineering ( Eds: J.
Lecomte-Beckers et al. ), Forschungszentrum Jülich, Liege, 1998, 601
[5] V. Foldyna, Microstructural Stability of Ferritic 9-12%Cr Steels, COST 501 Report,
Brussels, 1996
[6] T. Marrison, A. Hogg, Proc. Int. Conf. on Creep Strength in Steel and High Temperature
Alloys, ISI Meeting, Sheffield, 1972, 242
[7] A. Strang, V. Vodárek, in Proc. Development and Stability in High Chromium Ferritic
Power Plant Steels (Eds: A. Strang et al.), IOM, Cambridge, 1997, 31
[8] H. K. Chickwanda, Microstructural Stability of 12CrMoVNb Power Plant Steels, PhD
Thesis, Imperial College, London, 1994
[9] A. Strang, V. Vodárek, Materials Science and Technology 1996, 12, 552–556
[10] V. Vodárek, A. Strang, in Proc. Materials for Advanced Power Engineering ( Eds.: J.
Lecomte-Beckers et al.), Forschungszentrum Jülich, 2002, II/1223
[11] F. Hofer, P. Warbichler, Ultramicroscopy 1996, 63, 21–25
[12] J.W. Martin, R.D. Doherty, Stability of Microstructure in Metallic Systems, Cambridge
University Press, Cambridge, 1976
[13] W. Ostwald, Z. Phys. Chem. 1900, 34, 495–498
[14] H.K.D.H. Bhadeshia, in Proc. Parsons 2000, Advanced Materials for 21st Century Turbi-
nes and Power Plants (Eds.: A. Strang et al. ), IOM, London, 2000, 11
[15] B. Sundman, B. Jansson, J.-O. Andersson, Calphad 1985, 9, 153–158
[16] A. Strang, V. Vodárek, H.K.D.H. Bhadeshia, in Proc. Modeling of Microstructural Evolu-
tion in Creep Resistant Materials (Eds.: A. Strang, et al.), IOM, London, 1998, 129
[17] V. Vodárek, A. Strang, Materials Science and Technology 2000, 16, 1207–1211
19
Effect of Initial Melt Undercooling on Solid-State

Recrystallization and Grain Growth Processes in as-Solidified
Superalloy
F. Liu 1,2, G. Yang 2, R. Kirchheim1

1 Institut für Materialphysik, Universität Göttingen, 37073 Göttingen, Germany
2 State Key Laboratory of Solidification Processing, Northwestern Polytechnical University, Xi’an, Shaanxi,
710072, P. R. China
1 Introduction
Experimental studies in solidification process [1], microstructure evolution [2], and grain refi-
nement mechanism [3] have been carried out for undercooled DD3 (Ni alloyed with 9.5 Cr, 5.9
Al, 2.2 Ti, 5.2 W, 5 Co, and 3.8 Mo, wt.%) superalloy melts. If the initial melt undercooling,
'T, is above a critical value, 'T* (= 180 K), then the phase transformation consists of three do-
mains [3]: (1) the first mushy zone (FMZ), i.e. rapid solidification of Ni-based J-solid solution
during recalescence; (2) the second mushy zone (SMZ), i.e. normal solidification by natural
cooling after recalescence; (3) solid-state recrystallization and grain growth, as well as the sub-
sequent gc (Ni3Al(Ti)) precipitation [4].
For 180 K < 'T < 300 K, the as-solidified morphology is composed of the refined granular
crystals with average diameters as 20–50 Pm [1–3]. We suggested that this kind of grain refine-
ment should be caused by solid-state recrystallization as argued by Powell [5], which appears to
occur during, or immediately after, solidification, while the solid metal is very close to the melt-
ing point. As would be expected, 'T not only controls the solidification, but it also influences
the subsequent solid-state recrystallization and grain growth. Here, we concentrated on recrys-
tallization and grain growth associated with rapid solidification, together with an attempt to
highlight the effect of 'T on the recrystallized grain size.
2 Stress Development during High Undercooling Rapid Solidification
In cases where dendrites develop and grow, a coherent dendrite network is usually established
at solid fractions between 10 and 30 percent [6]. The dendrite coherency marks a point in FMZ
[3] where the material starts to develop strength, and where the resistance to material movement
increases drastically. As solid fraction after recalescence further increases with 'T (> 'T*), the
inter-dendritic permeability is reduced, thus resulting in an increasing pressure gradient within
FMZ [7]. The pressure gradient and flow of the liquid means that stresses are created and ex-
changed between the solid dendrites as well. Therefore, the solidification shrinkage, the thermal
contraction, the interactions of the dendrites, and the governing pressure differential for flow [7]
cause the development of large stresses, Vs, in the solid as the flow resistance or solid fraction
increases in FMZ [3].

ISBN: 3-527-31007-X
20
160 P a 2 gl § coh 1 1 1 gs ·
Vs # ¨ gs gs 2ln ¸ Es (1)
R 2 coh
gscoh ¹
f
s
2 g
tf O s ©
2
1 g s 1 g s 1
with gl, gs as the volume fraction of liquid and primary solid (i.e. gscoh < gs < fsR), gscoh and fsR
the solid fraction at the dendrite coherency point and after recalescence, Es the solidification
shrinkage of the primary phase (= (U l–U s)/U l, U l, the density of the liquid, U s, the density of the
primary solid), P the dynamic viscosity of the liquid, a the length of the primary solidification,
O2 the secondary dendrite arm spacing, and tf the primary solidification time, i .e. the recale-
scence time. Since values of tf are similar when 'T > 120 K [8], it is reasonable to assume con-
stant tf in calculations. Values for these parameters are available in [1–3].
The maximum solid fraction after recalescence, fsR, can be written as [9],
CP
fs
R
'H
TR Tm 'T (2)
with Cp as the specific heat, 'H the heat of fusion, Tm, and TR (see Table 2) the melting point
and the maximum recalescence temperature, of the superalloy melt. Substituting Eq.(2) into
Eq.(1) and assuming gscoh to be 0.15 [1] result in,
§ CP 1 ·
CP ¨ 'H TR Tm 'T C ¸
160 P a 2 1 'H TR Tm 'T ¨ 1 P TR Tm 'T ¸
V # 3¨
'H ¸ (3)
tf O22 § CP · ¨ ¸
T T ' T § C ·
¨
© 'H
R m ¸ ¨ 1 ln ¨1
¹ ¨
P
TR Tm 'T ¸ 1.0014 ¸¸
© © 'H ¹ ¹
Applying Eq.(3), the stress development in the dendrite network as a function of 'T is illus-
trated in Fig.1. If 'T > 'T*, the stresses exerted upon the solid network are of higher order of
Figure 1: Stress development in as-solidified DD3 superalloy as a function of 'T

21
magnitudes than its strength and deformation mechanism can probably intervene [10]. This in-
terpretation is well compatible with the microstructure evolution as observed in [3], in which
dendrite network collapses as a result of large stress arising from rapid solidification.
3 Interpretation for Recrystallization
Rapidly solidified DD3 superalloy could be considered as heavily cold-worked materials, which
could transform to its defect-free state by nucleation of fresh defect-free lattice and the growth
of nuclei at the expense of the defect-containing parent material. This heterogeneous transfor-
mation, in which fresh crystals are created, is known as recrystallization. The chemical free en-
ergy, 'G, available for the transformation, is given by
'G = 'H – T'S (4)
The enthalpy increase of materials due to cold work is the stored energy, Es. Hence, 'H is
equal to –Es. The form T'S is normally small compared with Es in present case. Therefore,
'G # Es (5)
In general Es is experimentally determined as a function of the amount of cold work, defor-

mation process, strain rate, and deformation temperature [11]. In the present case, however, Es
can be attributed to the high stress arising from rapid solidification, but not to any other external
agency, that results in the dendrite distortion, fragmentation or plastic deformation, and recrys-
tallization as well. According to a linear relation between the stored energy and stress [12],
O
Es
2Y
V 2
V 02 (6)
where V0 is the stress at which the stored energy is zero, O(= 0–10) a constant of the material
and Y Young’s modulus. Here, we according to [12] choose V0 and O as 0 and 9, respectively.
Recrystallization is correlated with nucleation and growth. The rate of nucleation of spheri-
cal particles may be expressed as,
ª 16 S J 3 V 2 º
« 2
'Gt »
3 Es
[ n Q exp « » (7)
« kT »
« »
«¬ »¼
where –Es / V (V, the average molar volume) replaces the chemical energy, 'Gc*, and the strain
energy of this process has been assumed to be negligible, g the surface energy of liquid metal, k
the Boltzman constant, n the total number of embryos per unit volume (n = 6.02 1023U/M, with
M as the average mole mass, and U as the density of the alloy), Q the frequency of vibration of
atoms and 'Gt* the activation energy required for the transfer of atoms across the interface, ex-
pected to be close to the activation energy for grain boundary (GB) diffusion here [13]. Re-
cognizing that the recrystallization proceeds within a limited temperature range [1–3], it could
22
be assumed as an isothermal process. Since 60 percent solid fraction remains after recalescence
if 'T > 'T*, it is reasonable to assume the maximal recalescence temperature, TR, to be the re-
crystallization temperature [3, 11]. The growth rate of the nuclei may be written as
ª § E ·º § 'S · § 'H t ·
K n G «1 exp ¨ s ¸ » exp ¨ t ¸ exp ¨ (8)
¨ K ¸ ¨ kT ¸¸
¬ © RT ¹ ¼ © ¹ © ¹
where the activation enthalpy 'Ht* is expected to be close to that for diffusion along the GBs
[13], G the jumping distance, and 'St* the entropy change. Since the magnitude of the stored en-
ergy is usually small compared with RT (T = TR), expansion of the term containing Es leads to
§ 'St · § Es · § 'H t ·
K GQ exp ¨¨ ¸¸ ¨ ¸ exp ¨¨ ¸¸ (9)
© k ¹ © RT ¹ © KT ¹
With reference to [14], the final grain size after isothermal recrystallization and grain growth
is proportional to the value of (K /[)1/4, defined as Growth-Nucleation ratio. Substitution of
equation (7) and equation (9) into (K / [)1/4 leads to,
ª 16SJ 3V 2 º
« 2 »
E 3 Es
D v G s exp « » (10)
nRT « KT »
« »
«¬ »¼
with D as the mediate diameter of the recrystallized grain. Combining Eqs.(3) and (6-10) gives
quantitative analysis between the value of (K/ [)1/4 and 'T, as well as that between D and the
value of (K / [)1/4, as shown in Figs.2 (a and b). With increasing 'T, the stored energy is enlar-
ged (see Fig.1), and the change in the nucleation rate becomes much more than that in the
growth rate, then the value of (K/ [)1/4, subsequently, decreases (see Eqs.(7–9) and Fig.2a). Ac-
cordingly, grain size is reduced (see Fig.2b). This indicates that D varies inversely with 'T, evi-
Figure 2: Relation between (a) 'T and (K/ [)1/4, (b) (K/ [)1/4 and the final grain size, obtained in DD3 superalloy
subject to high undercooling rapid solidification and solid-state transformation
23
denced by the corresponding microstructure evolution [1–3]. Nevertheless, the slope of the
curve increases with the reduction of (K/ [)1/4, i.e. the increase of 'T. A clear proportional rela-
tionship between D and (K/ [)1/4 does not hold over the whole undercooling range (see Fig.2b).
4 Effects of Initial Melt Undercooling and Grain Boundary Energy on

the Final Grain Size
The rate of grain growth may be expressed after Eq.(8) as [15],
§ 'G g · § 'Gt ·
K GQ ¨ ¸ exp ¨¨ ¸¸ (11)
© KTR ¹ © KTR ¹
where, 'Gg, the driving force for this process, is the decrease in interfacial free energy, i.e.
change in free energy due to transfer of a mole of the substance from the grain to the outside of
its carved surface,
2J V
'G g (12)
R
with R* as the radius of curvature of the GB, V the volume of the system, and J the interfacial
tension of the GB, i.e. GB energy. Assuming R* to be proportional to D gives dD / dt = k / D,
and another empirical equation can be derived from Eqs.(11–12) as,
2 (13)
D 2 Dt 0 kt
where the rate constant k may be written as
§ Q* ·
k k0 exp ¨ ¸ (14)
© RT ¹
with the activation energy Q* as close to 'Gt*, t the grain growth time, and k0 the temperature-
independent but J-dependent constant.
According to Eqs.(11–14), the final grain size subjected to the isothermal grain growth is de-
termined by Q, t, and J. In the present case, grain growth proceeds by natural cooling after re-
calescence [1–3], so it is reasonable to assume t as constant. Hence the only factor that
determines the final grain size should be g.
According to the Gibbs adsorption equation [16] and the thermodynamic approach of Weiss-
müller [17] and Kirchheim [18], J is reduced by solute segregation, and if it is possible to re-
duce J to zero, there would be no driving force for grain growth in such a binary polycrystalline
closed system. Under the critical assumptions: (1) segregation is of the saturation type, with a
finite number density of segregation sites; (2) the number of solute and solvent atoms is con-
served, and (3) precipitation of an intermetallic compound must be suppressed, the following
equation is deduced [17, 18],
24
J J 0 G 'H seg RT ln X 0 (15)
with J0as the GB energy for pure solvent, G the solute excess at the GBs at saturation, 'Hseg the
segregation enthalpy change per mole solute (here defined as positive for GB enrichment), X0
the bulk solute content, and T (TR) the temperature for grain growth. The second term in brak-
kets in Eq.(15) describes the loss of configurational entropy caused by segregation. Since 'Hseg
is gained by enriching solute atoms at the GBs, J should decrease from its original value J0. Ac-
cording to [18], at constant pressure, temperature and amount of material the change in free en-
ergy can be written as
dG = JdA (16)
Consequently, whenever J is positive, grain growth will decrease the free energy of the sys-
tem. Since systems with J< 0 are not thermodynamically stable, the only case where grain
growth can be suppressed is where J= 0. Since precipitation of solute or stable phase at GBs
and grain growth leads to a more stable equilibrium when compared with an unchanged grain
size and solute segregation, then a metastable thermodynamic equilibrium is reached with satu-
rated GBs for
3 G VM
D
§ J G 'H seg · (17)
X total exp ¨ 0 ¸
© G RT ¹
with D* (D) as grain diameter at the metastable equilibrium.

In the normal solidification after recalescence, the alloying elements are segregated at the
GBs of the as-solidified J-solid solution due to the inevitable equilibrium segregation [1–3].
Suppose that Jc precipitation does not concur with grain growth, J-solid solution could be con-
sidered as an ideal binary polycrystalline closed system with constant 'Hseg. Because of the
large grain size (i.e. in Pm range) [3], it is reasonable to assume that G has reached its saturation
value, independent of 'T. With increasing 'T, TR is reduced, in contrast with an increased X0
due to solute trapping effect [1–3], so more reduction happens to the GB energy according to
Eqs.(15–17). Higher 'T leads to lower GB energy, and subsequently, insufficient grain growth
or further grain size decrease results. This explains why the same change in (K/ [)1/4 gives more
decrease in grain size, as 'T increases (see Fig.2b).
5 Summary
'T not only determines the liquid-solid phase transformation, but its heredity remains to influ-
ence the subsequent solid-state recrystallization and grain growth. Grain refinement occurring
beyond 'T* can be attributed to the solid-state recrystallization. The final grain size of the as-so-
lidified alloy is strongly dependent on its response to recrystallization. With increasing 'T, ra-
pid solidification produces larger strain energy, provides stronger response to recrystallization,
and thus leads to smaller grain size in the as-solidified materials. In connection with the thermo-
dynamical approach developed by Weissmüller and Kirchheim, the much smaller grain size as-
25
sociated with higher 'T can be ascribed to insufficient grain growth in the recrystallized matrix
owing to a significant reduction of GB energy.
6 Acknowledgement
The authors gratefully acknowledge support from the Alexander von Humboldt Foundation and
the Natural Science Foundation of China, the Aeronautical Science Foundation of China, and
State key Fundamental Research of China, under Grant of NO.59871041, NO. 98H53093 and
No. G2000067202, respectively.
7 Reference
[1] F. Liu, and G. C.Yang, Metall. & Mater. Trans. B 2001, 32, 449–456
[2] F. Liu, and G. C.Yang, Mater. Sci. Eng. A 2000, 291, 9–16
[3] F. Liu, and G. C.Yang, J. Crys. Growth. 2001, 231, 295–305
[4] F. Liu, and G. C.Yang, Mater. Trans. 2001, 42(6), 1135–1138
[5] G. L. F. Powell, and L. M. Hogan, Trans. TMS-AIME, 1968, 242, 2133–2142
[6] C. J. Paradis, L. Amberg, and H. J. Thevik, A. Mo, 1995, in the Modeling of casting, Wel-
ding and Advanced Solidification processes VII (Ed: M. Cross, and J. Campell),TMS,
Warrendale, PA, USA, 1986, Pp. 609
[7] G. Chai, G. Backerud, and L Amberg Z. Metallkd, 1995, 86, 54–58
[8] B. B. Wei, Ph.D thesis, Northwestern Polytechnical University, Xi,an, Shaanxi, P. R.
China,1989
[9] T. J. Piccone, Y. Wu, Y. Shiohara, and M. C. Flemings, Mater. Trans. A, 1987, 18,
925–934
[10] J. Pillings, and A. Hellawell, Metall. & Mater. Trans. A. 1996, 27, 229–235
[11] K. Aust, and J. Rutter, in the Recovery and Recrystallization of Metals, (Ed: L. Himmel),
Gordon and Breach Science Publishers, New York, USA, 1963, Chapter 2
[12] M. B Bever, A. L. Titchener, and D. Holt, Prog. Mater. Sci., 1973,17, 1–124
[13] A. K. Jena, and M. C. Chaturvedi, 1992, Phase transformation in Materials, (Englewood
Cliffs, New Jersey) pp. 288
[14] H. Hu, in the Proceedings of the metallurgical reatises, (Ed: T.K. Tien and J.F.Eliot), The
Metallurgical Society AIME, 1981, p. 385
[15] J. W. Cahn, Acta Metall. 1962, 10, 789–809
[16] J. W. Gibbs, Trans Conn. Acad., vol. III (1876), 108-248; (1878) 343–524, Also in: “The
Collected Works of J. W. Gibbs”, Vol. 1, (Longmans, Green and Co., New York 1928), p.
55–354
[17] J. Weissmüller, Nanostructured Materials 1993, 3, 261–271
[18] R. Kirchheim, Acta Mater. 2002, 50, 413–417
26
Modelling the Diffusional Transformation by Application of the

Thermodynamic Extremal Principle
J. Svoboda1, E. Gamsjäger2, F. D. Fischer2

1
Institute of Physics of Materials, Academy of Science of the Czech Republic, CZ-616 62, Brno
2
Institute of Mechanics, Montanuniversität Leoben, A-8700 Leoben

ISBN: 3-527-31007-X
27
28
29
30
31
32
33
34
Phase Transformations in a Co-Cu-Ni Alloy
I. Guillona,b, C. Servanta and O. Lyonb

a
LPCES, CNRS, UMR 8648, ICMMO. bât 410, Université de Paris Sud, 91405 Orsay Cedex, France
E-mail: colette.servcmt@lpces.u-psud.fr
b
Laboratoire LURE, CNRS, bât 209d, BP 34, 91898 Orsay Cedex, France
1 Introduction
Two of the binary phase diagrams of the Co-Cu-Ni ternary one, Co-Cu and Cu-Ni, present a
miscibility gap in the solid state. In 1984 and 1985 Wagner and co-workers, [1, 2], have studied
some alloys belonging to one of these binary Systems, the Co-Cu, by small angle X-ray then
neutron scattering. The unmixing of the alloy reported in the present work was completely un-
explored whereas some ternary isothermal sections and isopleths of the Co-Cu-Ni System were
available in the literature [3]. The 7Co-70Cu-23Ni (at.%) chemical composition studied was
chosen because it is located in the ternary miscibility gap. Moreover, this System can be compa-
red with previously results obtained on the Cu-Fe-Ni system [4]. The Co-Cu-Ni System is very
difficult to study due to the proximity of the atomic number of each element of the alloy. So, se-
veral techniques were used in order to characterize the phase transformations occurring in this
alloy during an ageing treatment: the Anomalous Small-Angle X-ray Scattering, the anomalous
scattering around the Bragg peaks or Anomalous Wide-Angle X-ray Scattering, the Transmissi-
on Electron Microscopy, the dilatometry, and the thermomagnetometry.
2 Sample Preparation
Single crystals with millimetric grain size of composition 7Co-70Cu-23Ni (at.%) were prepared
using the floating-zone technique at LPCES, Orsay, France, from pure metallic powders. The
bar was pressed under an isostatic press until to 2500 bar, and then homogenized for l week at
1338 K in a mixture of Ar and Hz (15 vol.%), then sliced with a diamond saw. A second homo-
genisation (16 h at 1223 K) preceded the water quench. Finally, samples were aged at 823 K in
the two-phase field (miscibility gap) of the ternary phase diagram, in a mixture of Ar and Hz
(15 vol.%). The chemical composition and homogeneity of the samples were determined by
scanning electron microscopy at 15 keV by X-ray microanalysis, applymg the usual correction
(Z, A, F), leading to a precision of 0.5 at.% in the Co, Cu and Ni concentrations.
Samples used, either for A.S.A.X.S., A.W.A.X.S. or T.E.M. experiments, were electro-
chemically thinned with a mixture of 2/3 of methanol and 1/3 of nitric acid and respectively at
room temperature under a voltage of 2 V and at 223 K with a stabilized current of 130 mA for a
section of 7 mm2.

ISBN: 3-527-31007-X
35
3 Experimental Results and Discussion
3.1 A.S.A.X.S. Results
The A.S.A.X.S. pattem, represented m Fig. l. is composed oftwo sets of well-defined maxima
showing a strong scattering anisotropy. The existence of a well-defined maximum is characteri-
stic of a strong correlation of the precipitate positions along a given direction, each set corre-
sponding to one of the three possible faniilies, aligned along one of the soft <100>f.c.c directions
of the Cu enriched matrix.
Figure 1: Two-dimensional isointensity (in logarithmic scale) A.S.AX.S. pattern (recorded at 8322 eV) of a 7Co-
70Cu-23Ni (at %) single crystal aged at 823 K, for 4 hours, oriented parallel to the (010) plane, and showing two
families of precipitates piled-up in the <100>f.c.c. soft directions
Furthermore, the maxima of intensity are embedded in a ring (not due to the weak Laue mo-
notonic scattering) having a practically constant intensity around the direct beam for a given q
value: it means that some precipitates are distributed at random in the matrix. The law of Guini-
er1 applied to a sector I(q) = f(q) by isolating at 45° a part of the ring between two maxima al-
lowed us to calculate a radius of gyration of the precipitates (R||gring) which is slightly lower than
the one (R||ghump) obtained in the hump region for a given ageing duration. This result is obvious
because the detennination of R||ghump may be affected by the interference function due to spatial
arrangement of the precipitates. The growth of R||gring is slightly slower than that of the precipi-
tates with preferential orientation, see Table l. Therefore two spatial arrangements of precipi-
tates seem to be present in the sample, either well aligned in the soft directions <100>f.c.c. of the
Cu enriched matrix or not.
From such sections, some chemical parameters of the precipitates can be calculated in the
parallel and orthogonal direction of the pile-up: the radii of gyration (using the law of Guinier)
in the alignment Rg||, and in the orthogonal RgA direction, the dimensions of the precipitates (de-
termined also by TEM, in the next section), the inter-precipitate distance '|| = 2 S/ qmax, where
1
Developed in "Théorie et Technique de la Radiocristallographie", Dunod, Paris, [5]
36
qmax is the position of the hump in both directions. The slope of the tie-line was calculated equal
to 0.5. All these parameters are summarized in Table l, as a function of the ageing time.
Table 1: Calculated parameters from A.S.A.X.S. patterns for different ageing conditions. The
volume fraction was calculated equal to 13%, by the thermodynamic Calphad method
Aging time Orientation 20min 1h 2h 4h 12h 24h 12 d
R||hump1 Å 23.6 23.6 30.7 32.2 42.9 51.9 86.3
R||ring2 Å 20.7 21.7 25.6 27 33.5 40.9 66
RAhump Å 17.8 19.4 31.4 29.8 50.3 75.4 163.3
/|| 70 70.2 85.3 91.6 119.4 158.2 317.7
/||/ R||hump 3.83 3.45 3.59 3.59 3.6 3.94 4.76
'(Co)/'(Ni)3 0.41 0.54 0.59 0.61 0.5 0.63
1
R||hump 5 3 R||g hump is the Radius of Guinier
2
R||ring 5 3 R||gring is the Radius of Guinier
P
§ CCo CM ·
¨ Co ¸
3 'CCo © P VMc ¹
V c
P
'CNi § CNi CM ·
¨ Ni ¸
© P VMc ¹
V c
From the A.S.A.X.S. spectra, the partial structure functions (P.S.F.) were determined from a
set of over dimensioned equations.
The Variation of the R||hump and /|| parameters evolve in the same way, e.g., the more the pre-
cipitates are bigger, the more the inter-distance parameter increases. The /|| parameter is always
higher than the 2 R||hump one. The /|| / R||hump ratio is constant up to 12 h, and then increases no-
ticeably. The evolution of these two parameters (almost constant up to 1–2 h) let us think to a
two-step decomposition (as we can see by T.E.M. experiments). The first step should occur in
the early stages of ageing up to 1–2 hour where the size of the precipitates increases slowly. Af-
ter this transient period, their size increases faster according to a power kinetics law with a 0.25
exponent.
This hypothesis of two-step decomposition was confirmed by the plot of the scaling behav-
iour of the integrated intensity, I(q,E) / Imax, versus q / qmax as a function of the ageing time
(Fig. 2).
It can be noticed, in Fig. 2, that all curves have the same shape excepted one (the sample aged
for 20 min. at 823 K). So, we suppose that the coarsening is made in two steps, as it produced in
the binary System: Co-Cu, [6].
37
Figure 2: Scaling behaviour of the alloy aged at 823 K showing a two step coarsening
3.2 A.W.A.X.S. Results
From the A.W.A.X.S. patterns recorded around a 200f.c.c. Bragg reflection, Fig. 3 (C1–C4), it
can be seen that at the Co and Ni absorption edges, the hump with the higher intensity is located
in the Gq < 0 region (with Gq = (q–qB) while it is located in the Gq > 0 region at the Cu absorp-
tion edge (8973 eV).
From such sections, we can determined the partial displacements1.
The strong hump with the higher intensity is located in the q[100] > qB region for the Co, Ni
partial displacements, while the Cu one is opposite. From the "Bragg law", we can estimate the
crystalline parameter of the zones, giving rise to the scattered intensity related to the Position
disorder of Co and Ni mainly. It is lower than the mean lattice parameter of the alloy. The Cu
partial displacement corresponds to a dilatation of the Cu region probably centred around the
(Co-Ni) enriched zones. These Co and Ni enriched displaced zones are therefore elastically con-
tracted. It must be noted that the satellites are related to the size of the displaced zones. Their
position tends to the Bragg one, when the ageing time is increasing. So, they can not be directly
related to the crystalline parameter of these zones, because in the coarsening regime the crystal-
line parameters of the unmixed zones are constant.
From these maps, we can plot 'I versus Gq either in the [100] parallel direction of the align-
ment, Fig. 4(C1), or in the [011] orthogonal direction, Fig. 4(C2).
As we did for the A.S.A.X.S. results, we estimated the size of the zones for which the dis-
placements are noticeable in the parallel and perpendicular directions of the precipitate align-
ment: R||(= 2 / L||, where L|| is the width at half intensity of the '|| / versus Gq|| curve,
Fig. 4a.(C1), RA(= 2 / LA, where LAis the width at half intensity of the 'IA versus GqA curve, Fig.
4b.(C2) and their mutual distance: '|| in the parallel direction of the alignment (see Table 2).
2i B 2
1 'I t (q B Gq)( E ) q f q, E FT(Grk ), where FT stands for the Fourier Transform
N
38
Figure 3: Two-dimensional isointensity A.W.A.X.S. patterns of a single-crystal 7Co-70Cu-23N aged for 12

hours at 823 K showing the Variation of Z (in logarithm scale) versus GqA [011] and Gq|| [100] at different photon
energies, (C1) Co absorption edge; (C2) Ni absorption edge; (C3) Cu absorption edge, lower energy; (C4) Cu
absorption edge, higher energy
Figure 4: 'I versus Gq calculated from the partial displaceroents (equation 5) for a 7Co-70Cu-23Ni single crystal
aged for 12 hours at 823 K (C1) in the [100] direction of the precipitate alignment; (C2) in the [011] orthogonal
direction: KCo, • KNi, KCu.
39
Table 2: Calculated parameters for the displaced zones for different ageing conditions at 200f.c.c
Bragg peak from A.W.A.X.S. pattern
T 550 °C D (Å) (100) R|| (Å) RA (Å) D|| / R|| R|| (Å)
T A.W.A.X.S. A.W.A.X.S. A.W.A.X.S. T.E.M.
20 min 77,8 35,1 28 2,22 -
1h 121,7 35,3 29,2 3,45 25
2h 128 44,9 51,4 2,85 33
4h 129,8 47,2 51,9 2,75 48
12 h 179,9 68,9 81,8 2,61 60
24 h 219,5 86,6 109,5 2,53 75
Whatever the ageing duration, the radii R|| and RA of the displaced zones are practically iden-
tical which signifies that these zones have a symmetry of revolution (either spheres, disks or
rods). This symmetry is the same that the one of the precipitates. Their mutual distance /|| tends
to increase faster than R||. Indeed, the /|| / R|| ratio is constant up to 4–12 hours then increases.
The size of the displaced zones is always higher than the 2*radii of gyration of the precipitates.
The mutual distance of the displaced zones (Table 2) is also higher than the interdistance of the
precipitates determined by A.S.A.X.S. (Table l). The Variation of all the Parameters determined
in A.S.AX.S. and A.W.A.X.S. has the same trends.
3.3 Electron Microscopy Results
After a l hour ageing at 823 K, the bright field Images show a particular contrast of matrix de-
formation consisting of rather linear fringe sets, all directed along the same direction <100>f.c.c.
Fig. 5 (C1), two dark linear fringes separated by a triplet ofclear and dark fiinges, resulting of
the presence of the G.P zones in a matrix. In some places, the density of the fringes in the same
direction is so strong that the lines of dark and bright contrast appear to be continuous. The elec-
tron diffraction pattem with the <001>f.c.c. zone axis shows small streaks crossing the Spots of
matrix {220} in the <110>* directions. In addition, each node of the reciprocal lattice consists
of only one spot because of the too weak difference in lattice Parameter between the matrix and
the precipitates.
When the ageing time increases, 24 hours for example, the bright field images shows a ma-
trix deformation contrast as usual (coffee grains for spherical G.P. zones), Fig. 5 (C3). The size
of the coherent G.P. zones can be estimated by measuring the length of the no-contrast line;
their diameter is equal to about 8–10 nm. On the bright field corresponding to a (200) plan, one
finds the directions of alignment net of the coffee grains according to <100>f.c.c. The diffraction
patterns show the same characteristics as previously noted: diffuse scattering around all the
nodes of the reciprocal lattice of the matrix, and very fine lines of scattering in the <110>f.c.c. di-
rections, practically continuous between the nodes as observed in the case of the G.P zones in
the alloy Al-4Cu (weight %), [7], This indicates the presence of planar defects in the {220}f.c.c.
plans of the Co-Cu-Ni matrix, Fig. 5 (C4).
It must be mentioned that elongated superstructure Spots were observed on the electron dif-
fraction pattems of the aged samples, Fig. 5 (C2). Such a result was previously found by neu-
tron diffraction by [8] on Cu-Fe-Ni alloys. We did not succeed to evidence this superstructure
40
by anomalous X-ray diffraction with a four-circle diffractometer. This superstructure may be a

metastable phase formed during the electrothinning of the foils at 223 K. Further experiments
are in progress either on the Co-Cu-Ni or the Cu-Fe-Ni alloys.
Figure 5: T.E.M. results from a sample aged at 823 K a) and b) for 1h, c) and d) for 24h
4 Conclusion
The unmixing, at 823 K, of the 7Co-70Cu-23Ni (at.%) alloy located inside the chemical spin-
odal of the ternary miscibility gap, seems to occur in two successive steps: a transient period un-
til l hour then a coarsening stage up to 12 days. The precipitates adopt two spatial distributions:
the major part consists of three families well aligned along the soft directions <100>f.c.c.of the
Cu enriched matrix and some other precipitates are distributed at random. In the coarsening
step, the radius of the precipitates increases according to a power kinetics law with a 0.25 expo-
nent. The matrix defomiation contrast, as coffee grain, reveals the presence of spherical cohe-
rent G.P. zones. At 823 K, a steady regime is reached from about 2 h of ageing and the slope of
the tie line is calculated equal to 0.5. Although the two-phase model was verified in anomalous
small-angle X-ray scattering, the precipitates seem to have a more complex nanoscopic struc-
41
ture, composed of a Co enriched heart surrounded by a Co less enriched region as revealed by

the determination of two temperatures of Curie then verified by the observation of two acci-
dents on the curve of Variation of the true dilatation coefficient, as well as in differential scan-
ning calorimetry as a function of temperature. The radius of the displaced zones evolves
similarly as the radius of the coherent predpitates. The Co and Ni enriched zones are contracted
while the Cu enriched zones are dilated.
5 References
[1] W. Wagner, R. Poerschke, and H. Wollenberger, Decomposition of Alloys: The Early

States, Oxford, Pergamon, 1984
[2] W. Wagner, R. Poerscke, and H. Wollenberger, Abstracts of the Workshop on Atomic
Transport and Defects in Metais by Neutron Scattering, Jülich, Germany, 1–5, 1985
[3] P. Villars, A Prince and H. Okamoto, Handbook ofternary alloy Phase Diagrams, ASM,
1995, vol. 6, 8178–8192
[4] I. Guillon, O. Lyon and C. Servant, J. Appl. Cryst., 2002, 36,454–459
[5] A. Guinier, Theorie et Technique de la Radiocristallographie, 3"1 ed., Dunod, Paris, 1964
[6] P. Ancrenaz, and C. Servant, J. de Physique I, 1992,2, ?6, 1113–1128
[7] M. Karlik,, These de Doctorat en Science, Ecole Centrale de Paris, 1994
[8] J. Aalders, and C. van Dick, Physical Review B, 1984, 30, 1646–1655
42
The Coupling of Interphase Boundary Migration and

Precipitation: Example of a Microalloyed Nb-Containing Steel
1
C. R. Hutchinson and 2Y. Brechet
1
School of Physics and Materials Engineering, Monash University, Clayton, 3168, Victoria, Australia.
2
Laboratoire de Thermodynamique et Physico-Chimie Métallurgiques, Domaine Universitaire, St. Martin
d’Hères, 38402 Cedex, France.
1 Introduction
The microstructural changes that occur during the processing of many important metals and al-
loys usually result from the competition and interactions between different ‘modes’ of microst-
ructural change. In complex systems the variety of possible behaviours arising from
simultaneous coupled modes of microstructural evolution can indeed be very rich. An intere-
sting example comes from the thermo-mechanical processing of microalloyed steels. In these
materials, strong carbide forming elements such as Nb and/or Ti are added to form precipitates
with C in solution. The goals are twofold: (a) to reduce the C content of the Fe matrix to impro-
ve the deep drawability of the resulting product and (b) to form precipitates that will hinder re-
crystallization of the austenite phase after high temperature deformation, resulting in an
increased ferrite nucleation site density upon cooling and a smaller final ferrite grain size. Du-
ring this thermo-mechanical processing, situations can be imagined where the recrystallization
and recovery of the deformed J, carbide precipitation in the J, the diffusional transformation of
J to D and carbide precipitation in the D all occur simultaneously. It is then the coupled evoluti-
on of all these processes that determines the evolution of the microstructure. The development
of a physically-based description of such a complicated situation, coupling each of these proces-
ses, is a long-term goal of this work and is aimed at developing a description for the design and
on-line monitoring of hot-rolling schedules for these steels. We have recently developed a mo-
del that can quantitatively describe the evolutions in the austenite where simultaneous recrystal-
lization, recovery and precipitations processes occur [1]. The key to a successful quantitative
description was the identification of the important interactions between the different modes of
microstructural change.
In practical cases of thermo-mechanical processing of microalloyed steels, not all of the al-
loying element is removed from the J solid solution in the form of precipitates in the J phase be-
fore the temperature falls below that at which the austenite begins to decompose to ferrite.
Under such cases, the possibility arises for simultaneous J o D transformation and carbide pre-
cipitation. Since both phenomena involve a redistribution of C atoms, it is easy to imagine that
carbide precipitation may influence the J o D transformation and vice versa. In this contributi-
on we present, in the form of a series of order of magnitude calculations, estimates of the effects
of simultaneous carbide precipitation on the J o D transformation in a microalloyed steel. The
objective is to identify which (if any) modes of carbide precipitation influence the kinetics of
the J o D transformation in a quantitatively significant manner before integrating these into a
full numerical model coupling the two processes.

ISBN: 3-527-31007-X
43
2 Modes of Carbide Precipitation
A large variety of types of carbide precipitation in the presence of the J o D transformation

have been studied by many authors over the years, eg. [2]. At the outset, we restrict our attenti-
on to the precipitation of carbides in microalloyed steels (e.g. NbC, TiC, VC etc.) and to cases
where the ferrite formed has an allotriomorphic morphology. At this stage we do not attempt to
deal with the cases of alloyed pearlite (where the ferrite and an alloyed carbide grow co-opera-
tively) or to the case of bainite where the D/J interface is distinctly non-planar even though the
principles to be followed apply equally well to non planar growth morphologies. The cases of
interest have been studied in detail by Honeycombe and co-workers [3–5] over the years and
three distinctly different ‘modes’ of carbide precipitation have been identified.
1. carbide precipitation in the J ahead of the migrating D/J interface

2. carbide precipitation in the D behind the migrating D/J interface
3. carbide precipitation at the migrating D/J interface, leaving behind either rows of carbides
approximately parallel to the D/J interface or fibres approximately perpendicular to the D/J
interface
All three types of precipitation have been observed in Fe-C-Nb steels and for the purpose of
this contribution we will use the alloy studied by Sakuma and Honeycombe [6],
Fe-0.07C-0.031Nb (wt. %), as an example for all calculations. At this point we do not attempt
to identify the conditions under which each mode of precipitation will occur, rather we make a
series of ‘order of magnitude’ calculations of the effects on the D growth rate if the transforma-
tion was accompanied by NbC precipitation in each of its forms.
3 The Effect of NbC Precipitation on the AusteniteoFerrite

Transformation
The diffusional growth of D from J involves both the long range redistribution of C from the
growing D to the J and the rearrangement of the substitutional atoms from the FCC structure of
the J to the BCC structure of the D(Fig. 1) [2]. Both processes consume some of the total ener-
gy driving the transformation and the compositions at the migrating interface must be chosen so
that the two processes give the same interface velocity. Simultaneous carbide precipitation can
influence the kinetics of the J o D transformation in several ways (Fig. 2). Carbide precipitati-
on in the J or at the migrating D/J interface could exert a pinning pressure on the migrating
boundary which would retard its motion. This can be formally described as a decrease in the
driving force, G(t). Any changes in the C composition of the J or D will influence the transfor-
mation rate through the mass balance for C at the interface and changes in the Nb contents will
influence the interfacial compositions of C at the migrating boundary. Furthermore, if Nb in-
teracts strongly with the migrating boundary, then a solute drag effect [7–8] may be expected
and changes to the Nb content of the J will influence the magnitude of this effect.
44
Figure 1: Schematic illustration of the carbon concentration profile across the DJ interphase boundary during
growth. CJDand CDJare the C compositions in the J and D respectively at the interface and Co is far field composi-
tion. M(t) is a kinetic property of the interface where v(t) /M(t) represents the free energy dissipated by the substi-
tutional atom rearrangement at the interface. G(t) is the driving force for this process.
Figure 2: Schematic illustration of the potential effects of NbC precipitation on the JoD transformation (Fig. 1)
in a microalloyed steel
In the following calculations we assume that there is no energetic barrier to the rearrange-
ment of the substitutional atoms at the interface (i.e. infinite intrinsic interface mobility) and
that local equilibrium conditions prevail at the migrating interface. Furthermore, we will as-
sume a concentration independent diffusivity (D) for C so that we can write an analytical ex-
pression for the D / Jinterface velocity (v) as a function of time (t) in a semi-infinite medium:
C J / D C0 D
X|
C J / D CD / J St
45
where CJD is the C concentration in the J at the JD interface, CDJ is the C concentration in the D
at the JD interface, C0 is the far field C concentration, D is the C diffusivity in J and t is time.
All thermodynamic calculations were made assuming that local equilibrium-negligible partition
conditions [9–13] prevail at the migrating interface at all times and we have used the thermody-
namic evaluation of the Fe-C-Nb system due to Huang [14].
3.1 Case 1. NbC Precipitation in the Austenite
NbC precipitation in the Jahead of the migrating boundary can influence the JoD transforma-
tion in at least two ways: (a) the precipitates will exert a pinning (Zener) pressure on the migra-
ting boundary and (b) the removal of C and Nb from solution will change the local chemical
environment encountered by the moving boundary.
3.1.1 Zener Pinning of the Migrating Boundary
NbC precipitation in the J is usually referred to as ‘strain-induced’ precipitation because during

the thermo-mechanical processing of microalloyed steels the precipitation occurs mostly on dis-
locations in the deformed J. To obtain an estimate of the potential pinning pressure of the preci-
pitates we will assume that the distribution is approximately random so that the pinning
pressure can be written (for coherent precipitates) [15–16]:
J /D
3 J int FJ
Pinning Pressure #
R
J /D
where J int is the interfacial energy of the migrating g/D interphase boundary, Fv is the volume
fraction of precipitates and R is precipitate radius. By considering an equilibrium fraction of
precipitates (which for the example alloy treated at 850 °C is ~3.5 · 10–4), a high interfacial en-
J /D
ergy, J int 1 J/m 2 and a small precipitate radius, R = 2 nm, we can obtain an upper estimate
of ~0.5 MPa for the pinning pressure. In comparison, the free energy driving the J o D trans-
formation is typically of the order of 200–1000 J/mol. Using the lower estimate we may appro-
ximate the driving pressure:
'GJ oD
Driving Pressure # # 30 MPa
Vm
where Vm is the molar volume (~6.85 · 10–6 m3/mol for J). It is clear that the potential pinning
pressure of such a low equilibrium precipitate fraction represents less than 2 % of the driving
pressure and we would therefore expect this effect to be negligibly small.
3.1.2 Change in the Local J Chemistry
3.1.2.1 Removal of C from the J Solid Solution

The removal of C from the J solid solution can be considered as an approximate decrease in the
effective far field C0 concentration of the J (Fig. 3a). Again assuming an equilibrium fraction of
46
NbC precipitation in the J precipitated isothermally at 850 °C in an Fe-0.07C-0.031Nb (wt.%)

alloy, the new far field C concentration, Co is 0.0663 (wt.%). Using Eq. 1 to estimate the
growth rate, the effect of carbide precipitation in the J on the rate of a growth can then be writ-
ten:
'X C0 C0
| 0.04
X C J / a C0
The result being that the decrease in the C content of the J due to precipitation gives a small
Figure 3: Schematic illustrations of the effect of NbC precipitation in the J ahead of the migrating boundary on
(a) the far field C concentration and (b) the interfacial C concentrations
acceleration of the transformation of ~4 %.

3.1.2.2 Removal of Nb from the J Solid Solution
Thermodynamic Effects
In this contribution we are concerned with the formation of D from J under conditions where
bulk partitioning of Nb between the D and J is not observed. Nevertheless, the decrease in the
Nb content of the J solid solution due to NbC precipitation will exert an influence on the local
equilibrium C composition in the J at the J/ D interface (Fig. 3b). At temperatures around
~850 °C, Nb is a weak ferrite stabiliser and we would therefore expect that from a purely ther-
modynamic viewpoint that the decrease in the Nb (Nb1 o Nb2; Fig. 3b) content of the J will re-
sult in a decrease in CJDand a resulting slowing down of the interface velocity. Again using Eq.
1 for the interface velocity we may write:
'X § C2J / D C0 · § C1J / D C1D / J ·

¨ J/D
D/J ¸ ¨ J/D ¸ 1 | 0.009
X © C2 C2 ¹ © C1 C0 ¹
where C1J / D and C1D / J are the C concentrations in the J and D at the interface, respectively, when
n0 carbide precipitation is observed. C1J / D and C1D / J are the respective interfacial conditions un-
der conditions where an equilibrium fraction of NbC has precipitated in the J, and C0 is the bulk
C content of the alloy. The effect is the small deceleration of the transformation rate by ~1 %.
Kinetic Effects
A second and potentially more important effect of changing the Nb content of the J is the resul-
ting change in the solute drag effect arising from potential interactions of Nb with the moving
47
boundary. Two treatments have been proposed [7–8] for evaluating this effect on migrating
phase boundaries, but in each case the results depend sensitively on unknown parameters such
as the spatial variation in the interaction energy of Nb with the boundary and the diffusivity of
Nb across the boundary. This effect cannot be easily evaluated in the absence of reliable esti-
mates of these parameters.
3.2 Case 2. NbC Precipitation in the Ferrite
If NbC precipitation occurs in the D behind the migrating phase boundary the only potentially
significant effect of the precipitation is to provide an additional flux of C(J2) from the interface
to the NbC precipitates that will effect the interface velocity through the mass balance at the in-
terface (Fig. 4).
Figure 4: Schematic illustration of the C profile across the moving DJ interphase boundary under conditions of
NbC precipitation in the D behind the migrating boundary. J1 is the C flux into the J and J2 is the additional C flux
in the D from the interface to the growing NbC particle.
In this case we are interested in the magnitude of the flux J2 compared to the flux J1. Using a
linearized gradient for the flux in the D we can write the two fluxes as:
wCJ D
J1 DCJ
wx

| C J / D C0 St
int
wCD C D / J C D / Nb C
J2 DCD | DCD
wx int
l*
where x is the direction co-ordinate, CD/NbC is the C concentration in the D at the NbC/D inter-
face and l* is the approximate spacing of NbC precipitates in the D. However, the flux of C to
the NbC precipitates, J2, is limited by the growth of the precipitates and it is the diffusion of Nb
to the precipitates which is the rate limiting step for this process. Therefore, it is the flux of Nb
to the precipitates that will limit the effective C flux from the interface to the precipitates. We
can therefore write:
48
D w Nb D D NbD / J NbD / Nb C
J 2eff J 3Nb DNb | DNb
wx int
l*
D
where DNb is the Nb diffusivity in the D phase [17].
Using the example alloy treated at 850 °C we have evaluated the ratio J 2eff / J1 for various
spacings of NbC precipitates in the D. This ratio is plotted as a function of time in Fig. 5a.
Figure 5: (a) Plot of the ratio of the effective C flux in the D (J2) due to NbC precipitation behind the migrating
interface to the C flux in the J (J1) in the absence of any NbC precipitation. (b) Plot of the ratio of the effective C
flux at the interface (J2) due to interphase boundary NbC precipitation (Case 3) to the C flux in the J (J1) in the
absence of any NbC precipitation.
It is clear that the effect of NbC precipitation in the D depends on the average particle spac-
ing in the D. The transmission electron microscopy study of Sakuma and Honeycombe showed
that at temperatures around 800 °C, when NbC precipitation is observed in the D, the average
particle spacing is of the order of ~0.05–0.1 Pm (Fig. 7 in [6]). Under such conditions the C
flux in the D may be of the order of ~10 or 20 % of the C flux in the J in the absence of precipi-
tation. This is considered to be a relatively small effect but will become much more important if
the average particle spacing decreases substantially.
3.3 Case 3. NbC Precipitation at the Migrating Austenite/Ferrite Interface
3.3.1 Zener Pinning of the Migrating Boundary
In the case of interphase boundary precipitation (IBC) the precipitate distribution encountered
by the migrating boundary is clearly not randomly distributed. Many studies have shown that
the precipitation occurs either in well defined rows approximately parallel to the macroscopic
plane of the DJ interface or as fibres perpendicular to the interface. We will concentrate here
49
on the case of row carbide formation. An example taken from the work of Campbell [18] of
Cr23C6 precipitation in an Fe-0.2C-12Cr (wt.%) alloy shown in Fig. 6a and b clearly illustrates
that precipitation occurs on the relatively immobile terraces of the growth ledges that facilitate
movement of the D/ J interface. The potential pinning pressure arising from such an arrange-
ment will depend on the local crystallography at the interface. We have adopted a simplified ap-
proach to this question and have identified two limiting cases. These are shown in Figs. 6c and
d. If both the macroscopic plane of the D/ J interface and the plane of the riser of the growth
ledges (approximately perpendicular to the macroscopic plane of the D/ J interface) are tightly
constrained by crystallographic considerations (Fig. 3c) then we may expect the potential pin-
ning pressure of IBC to be large because the risers will have difficulty in overcoming the preci-
pitates present on the terrace on which the D is being grown. On the other hand, if the risers are
not constrained by crystallographic considerations then it may be expected that they can bow
between the NbC precipitates (Fig. 6d) and a relatively small pinning pressure may be expected.
Figure 6: (a) Dark field and (b) Bright field transmission electron micrographs of interphase boundary carbide
precipitation of Cr23C6 in an Fe-0.2C-12Cr (wt.%) alloy [18] illustrating that IBC occurs on the immobile sections
of growth ledges at the DJ interface. (c) Schematic diagram of IBC precipitation under conditions where both the
plane of the interface (the terrace of the growth ledges) and the plane of the ledge riser are crystallographically
constrained. (d) Represents the same case shown in (c) except that the plane of the ledge riser in not restricted by
crystallography.
Since D formation usually begins at prior J grain boundaries and growth is favoured into the
J grain to which the D does not have a low order crystallographic relationship, it seems unlikely
that both the macroscopic plane of the D/ J interface and the plane of the risers of the growth
ledges will both be simultaneously tightly constrained by low order crystallographic relation-
ships between the phases. We therefore expect, on average, a situation more like that shown in
Fig. 6d and a relatively small pinning force from NbC precipitation.
50
3.3.2 Change in the Local J Chemistry
3.3.2.1 Removal of C from the J Solid Solution

The precipitation of NbC at the migrating J/ D interface results in an additional C flux that must
be considered in the mass balance and that will effect the J / D interface velocity. This flux can
be evaluated in an analogous way to our treatment of the additional C flux resulting from the
NbC precipitation in the D(Case 2). A schematic illustration of the arrangement is shown in the
Fig. 7. Again, we are interested in a comparison of the usual flux of C into the J as a result of
growth, J1, and the additional C flux that would result if NbC precipitation occurred at the mi-
grating interface, J2.
Figure 7: Schematic illustration of the C profile across a migrating DJ boundary and the simultaneous precipita-
tion of NbC at the boundary (IBC) that give rise to an additional C flux from the interface to the precipitates over
a length scale, l*.
The two fluxes can be written:
wCJ D
J1 DCJ
wx

| C J / D C0 St
int
wCJ C J / D C J / Nb C
J2 DCJ | DCJ
wx int
l*
It is again the diffusion of Nb that will be the rate limiting step in the growth of the NbC pre-
cipitates and it is therefore this Nb flux that will limit the C flux. As was the case for the precip-
itation in the D we can now write:
w Nb J Nb J / D Nb J / 1E&
J 2eff J 3Nb b
DNb b
| DNb
wx int
l*
51
where DNb b is the diffusivity of Nb in the J/ D interphase boundary. We assume here that Nb
diffusion occurs in the interphase boundary and that this can be estimated by the diffusivity of
Nb in a grain boundaries [17]. A recent theoretical treatment of IBC precipitation in steels has
shown that the precipitate row spacing can only be accounted for if the alloying element diffusi-
vity is taken equal to that estimated for the interphase boundary [19]. Using the example alloy
treated at 850 °C we have evaluated the ratio J 2eff / J1 for various spacings of NbC precipitates
in the a. This ratio is plotted as a function of time in Fig. 5b. As expected, the flux depends on
the particle spacing, l*. The observations of Sakuma and Honeycombe suggest that a particle
spacing of 0.1 mm or less is appropriate (Fig. 5, ref. [6]). It is clear from Fig. 5b that the addi-
tional C flux resulting from IBC can be an order of magnitude larger than the usual C flux into
the J. This is a large effect and one that must be considered in our calculations of the growth rate
of a when simultaneous IBC precipitation is observed.
3.3.2.2 Removal of Nb from the J Solid Solution
Thermodynamic Effect
The effect of the removal of Nb from the J solid solution due to IBC can be treated in the same
manner as was considered for the NbC precipitation in the J. It has already been shown that this
results in a negligibly small deceleration of the interface.
Kinetic Effect
The variation in the Nb content at the interface due to IBC precipitation will give rise to a chan-
ge in the magnitude of the solute drag effect of Nb on the migrating DJ interface. However,
whether or not this is a negligible effect depends on the estimates of the unknown interaction
energy of Nb with the boundary and the trans-interface diffusivity of Nb. In the absence of re-
liable measures for these parameters we will not attempt to quantify this effect.
4 Summary
The potential influence of simultaneous NbC precipitation of the J o D transformation in

microalloyed steels was discussed and order of magnitude calculations were carried out to iden-
tify which effects are likely to be important. For the Fe-0.07C-0.031Ni (wt.%) alloy considered
at temperatures ~850 °C, neither NbC precipitation in the J in front of the migrating interface
nor precipitation in the D behind the migrating interface had any significant effects on the D
growth kinetics. However, interphase boundary carbide precipitation can have a large effect on
the D growth rate and should be considered for those conditions where it is known that IBC ac-
companies D growth.
In systems where the precipitate volume fraction can be significantly greater than the rela-
tively low value of 3.5 · 10–4 used in this study, all of the effects of precipitation will be in-
creased and the relative magnitudes of the different contributions should be examined for each
case. With changes in temperature, the equilibrium precipitate fractions and the driving forces
for both the J o D transformation and carbide precipitation will vary as well as the magnitudes
of each of the kinetic terms used in the evaluation of the solute fluxes. It is not obvious that each
of the thermodynamic and kinetic terms that enter into the calculations will vary with tempera-
ture in a manner that conserves the relative effect on D growth of each of the modes of NbC pre-
cipitation discussed here for an Fe-0.07C-0.031Nb (wt.%) alloy at 850 °C. However,
52
calculations of the type presented here can be repeated in a very straightforward manner for
these other conditions.
5 Acknowledgements
CRH greatly acknowledge the financial support of IRSID laboratories (Metz, France) and the
CNRS (France). The authors are also very grateful for experimental assistance and many stimu-
lating discussions with M. Kandel of IRSID.
6 References
[1] H. S. Zurob, C. R. Hutchinson, Y. Brechet and G. Purdy, Acta Materialia, 2002, 50,
3075–3092
[2] R. W. K. Honeycombe and H. K. D. H. Bhadeshia, Steels: Microstructure and Properties,
2nd Edition, Edward Arnold, London 1995
[3] A. T. Davenport and R. W. K. Honeycombe, Proc. Roy. Soc. Lond. A, 1971, 322,
191–205
[4] R. W. K. Honeycombe, Metall. Trans., 1976, 7A, 915–936
[5] R. W. K. Honeycombe, Metal Science, 1980, 6, 201–214
[6] T. Sakuma and R. W. K. Honeycombe, Metal Science, 1984, 18, 449–454
[7] M. Hillert and B. Sundman, Acta Metall., 1976, 24, 731–743
[8] G. R. Purdy and Y. Brechet, Acta metall. mater., 1995, 43, 3763–3774
[9] M. Hillert, Paraequilibrium, Int. Rep., Swedish Inst. Metals Res., (1953)
[10] J. S. Kirkaldy, Can. J. Phys., 1958, 36, 907–916
[11] A. A. Popov and M. S. Mikhalev, Phys. Metals Metallogr., 1959, 7, 36
[12] L. S. Darken, Trans A.I.M.E., 1961, 221, 654–671
[13] M. Hillert, in The Mechanism of Phase Transformations in Crystalline Solids, (1969),
pp. 231–247, Inst. Of Metals, Monograph No. 33
[14] Huang, Zeitschrift fur Metallkunde, 1990, 81, 6, 397–404
[15] M. F. Ashby, J. Harper and J. Lewis, Trans. AIME, 1969, 245, 413
[16] R. D. Doherty, Metal Science, 1982, 16,1
[17] Chr. Herzig, J. Geise and S. V. Divinski, Zeitschrift fur Metallkunde, 2002, 93, 12,
1180–1187
[18] K. Campbell: Ph.D. Dissertation, University of Cambridge, 1971
[19] R. Lagneborg and S. Zajac, Metall. Trans. A, 2001, 32A, 39–50
53
Advanced Models for Particle Dissolution in Multi-Component

Alloys
F. J. Vermolen1, K. Vuik1 and S. van der Zwaag2

1
Department of Applied Mathematical Analysis, Delft University of Technology, The Netherlands
2
Department of Aerospace Engineering, Delft University of Technology, The Netherlands
1 Abstract
The present paper consists of the formulation of a model for particle dissolution in a multi-com-
ponent alloy taking into account cross-diffusion effects. The model consists of a Stefan conditi-
on to compute the velocity of the interface separating the particle and the solvent phase. The
influence of the cross-diffusion terms on the particle dissolution rate is shown and it is conclu-
ded that its impact can be significant.
2 Introduction
In the thermal processing of both ferrous and non-ferrous metals, homogenization of the as-cast
microstructure by annealing at such a high temperature that unwanted precipitates are fully dis-
solved, is required to obtain a microstructure suited to undergo heavy plastic deformation. Alt-
hough precipitate dissolution is not the only process taking place during homogenization, it is
often the most critical of the occurring processes. The minimum temperature at which the an-
nealing should take place can be determined from thermodynamic analysis of the phases pre-
sent. Another important quantity is the minimum annealing time for this temperature. This time,
however, is not a constant but depends on particle size, particle geometry, particle concentrati-
on, overall composition etc.
Due to scientific and industrial relevance of being able to predict the kinetics of particle dis-
solution, many models have been proposed and validated experimentally. The early models on
particle dissolution were based on an analytical solution, for an unbounded medium under the
assumption of local equilibrium at the moving interface, see Whelan [1] for instance. The model
of Nolfi et al [2] incorporates the interfacial reaction between the dissolving particle and the
surrounding phase. Later modelling particle dissolution has been extended to the introduction of
multi-component particles by, among others, Reiso et al [3], Hubert [4] and Atkinson et al [5].
All the mentioned authors consider particle dissolution as a Stefan problem. A recent approach
is the phase-field approach, which is derived from a minimization of an energy functional. This
approach has, among others, been used to simulate dendritic growth. A recent extention to mul-
ti-component phase field computation has been done by Grafe et al [6], where solidification and
solid-state transformations are modelled. Some disadvantages of the phase-field approach are
that no quick estimates of the solution are available and that physically justifiable parameter
values are not easy to obtain. Furthermore, from a numerical point of view they require a mesh
adaptation in the vicinity of the moving interface (see Emmerich [7] for instance) to capture the
interface position correctly. Therefore, we limit ourselves here to view particle dissolution as a

ISBN: 3-527-31007-X
54
Stefan problem, where we solve the multi-component version of Fick’s Law including cross-
diffusion coefficients with a moving interface separating the constituent solvent and solute
phases.
Kale et al [8] consider ternary diffusion in fcc phases and calculates diffusion coefficients in-
cluding cross diffusion coefficients by use of the thermodynamically based Boltzmann-Matano
method. They observe that the cross-diffusion coefficients range up to about a third of the diag-
onal diffusion coefficients. Hence cross-diffusion should be taken into account here. Vergara et
al [9] construct an experimental procedure to obtain cross-diffusion coefficients from experi-
ments. Travis and Gubbins [10] compute cross-diffusion coefficients using Monte-Carlo simu-
lations. The equations for cross-diffusion are motivated for metallic alloys by Kirkaldy and
Young [11] and Glicksmann [12]. They note that the diffusion matrix has real-valued and posi-
tive eigenvalues for an acceptable problem statement. Numerical solutions of cross-diffusion
problems were found by Naumann and Savoca [13] for fixed boundaries. Atkinson et al give a
self-similar solution for a cross-diffusion problem with a moving boundary, describing a phase
transformation in steel. For a more elaborate literature review we refer to Vermolen et al [14].
As far as we know, little has been done, except by Atkinson et al [5], on the extension to
problems that incorporate the movement of the boundary and cross-diffusion. It is our aim in
this work to analyze the moving boundary with cross-diffusion aspects in a finite medium. Ex-
istence of (multiple) solutions, when the diffusion matrix is singular, is stated and proven as
mathematical theorems by Vermolen et al [15]. In the present study we present a model for dis-
solution of stoichiometric multi-component particles combined with cross-diffusion aspects.
Further, we highlight the influence of the cross-terms of the diffusion matrix on the dissolution
kinetics.
3 Model
The as-cast microstructure is simplified into a representative cell containing the atomic struc-
ture of phase D and a single particle of phase E of a specific form, size and location of the cell
boundary. Both a uniform and spatially varying initial composition at t = 0 can be assumed. The
boundary between the phase D and the particle E is referred to as the interface. Particle dissolu-
tion is assumed to proceed via the following subsequent steps: decomposition of the particle,
atoms from the particle crossing the interface and diffusion of these elements in the D phase
away from the interface. Here we assume that the first two steps occur instantaneously and
hence local thermodynamic equilibrium is assumed at the interface. We assume that the dissol-
ving particle is stoichoimetric and that the interfacial concentrations are given by the following
hyperbolic relationship:
sol m1 m2 mn
c c
1
sol
2
csol n K K (T ) . (1)
The factor K is referred to as the solubility product. It has an Arrhenius dependence on tem-
perature. Here, we assume it to be constant and given. The interface concentrations are denoted
by csoli. Further, we note that this equation only holds for concentrations at the interface that are
much lower than the particle concentrations. We denote the moving interface by S(t), where t
denotes time. Since particles dissolve simultaneously in a metal, we introduce a finite cell in
55
which each of the particles dissolve. The concentration profiles of the dissolving particles may
interact and hence soft-impingement occurs. We assume that there is no flux over the cell
boundary. For cases of low overall concentrations in the alloy, the cell size may be large and the
solution resembles the case of an unbounded domain for which a self-similar solution is ob-
tained for the planar (rectangular) case (see Vermolen et al [14] and Atkinson et al [5]). Within
the D phase diffusion takes place and hence there we have
n
wci
wt
¦ Dij 'c j . (2)
j 1
Here the coefficients Dij denote the entries of the diffusion matrix and ci denotes the concen-
tration of species i. The diffusion matrix is notated as follows:
ª D11 ... D1n º

D « ... ... ... » (3)
« »
«¬ Dn1 ... Dnn »¼
The off-diagonal terms are the cross-terms and they are a measure for the importance of
cross-diffusion, which is the interaction between the diffusion of consecutive alloying elements.
When an alloying element is dissolved in the D phase then the resulting stress and elongations in
the crystal structure worsen or facilitate diffusion of the other elements. The off-diagonal entries
of the diffusion matrix are a measure for this mutual influence. When Dij < 0, then alloying ele-
ment j deteriorates the diffusion of element i. Whereas Dij > 0 implies that element j enhances
diffusion of element i. We note that the diffusion matrix depends on time if a non-isothermal
heat treatment is modelled. However, in the present paper we assume that the diffusion matrix
does not depend on the concentration of the alloying elements, time, space. Hence it is constant
and given. Further, we take as initial conditions:
°ci (r ,0) ci 0 for i {1,...., n}

® (4)
°̄ S (0) S0
At the boundary not being an interface we impose a no-flux condition. Furthermore, at the
interface S(t) we have the ‘Dirichlet’ boundary condition ci (S(t),t) = cisol for each alloying ele-
ment. The concentration of the alloying elements in particle is denoted by cparti and this concen-
tration is fixed at all stages of the process. This assumption is based on the constant
stoichiometry during the entire dissolution process. The dissolution rate (interfacial velocity) is
determined from a mass-balance of the atoms. This leads to the following Stefan condition for
the interfacial velocity:
n wc j
ci
part
ci
sol
dSdt ¦ D
j 1
ij
wn
( S (t ), t ) (5)
This implies that the interfacial velocity can be eliminated to get n–1 equations for cisol. Fur-
ther, equation (1) closes the system to get n equations. Note that the set of equations is non-lin-
56
ear. Hence, possibly multiple solutions arise. In Vermolen et al [15] a condition for mass-
conserving solutions is derived to reject possible solutions. Since we consider similtaneous dif-
fusion equations for several chemical elements with a moving boundary, the problem is referred
to as a ‘vector-valued’ Stefan problem. The unknowns are the concentrations, interfacial con-
centrations and the moving boundary position. All concentrations are by necessity non-nega-
tive. Further, a strong coupling exists due the interfacial concentrations in equation (1), the
equations for the moving boundary (5) and the diffusion equation (2). This strong coupling
complicates the solution of the Stefan problem.
4 Solution of the Vector-valued Stefan Problem
First we diagonalize the diffusion matrix D, D = P L P–1, where L = diag(O1, …, On) represents
the diagonal matrix with the eigenvalues on the main diagonal. The columns of the matrix P
consist of the eigenvectors, which correspond to the eigenvalues. Of course if the matrix con-
tains some eigenvalues that are equal and when the diffusion matrix is defective then a Jordan
decomposition is used and L is not completely a diagonal matrix. This situation has been treated
in more detail by Vermolen et al [14]. The matrix P then contains generalized eigenvectors. As
an example we consider a ternary alloy with two diffusing species,
ª D11 D12 º
D «D ,
¬ 21 D22 »¼
then, if the trace of D, tr(D) = D11 + D22 > 0, is positive and the determinant satisfies
0 < det(D) = D11D22–D12D21 < tr(D)2/4, then the eigenvalues are real-valued and positive and
hence the cross-diffusion problem is well-posed. It can be shown that if at least one of the ei-
genvalues is negative then the problem is ill-posed due to instability with respect to small pertu-
rabations, i.e. small perturbations will continue to grow during the entire diffusion process. The
strong coupling in equation (2) is removed by the diagonalization argument to obtain
wui
Oi 'ui (6)
wt
Here the quantities ui denote the entries of the transformed concentrations, u = P–1c. The Ste-
fan condition looses its extra coupling due to cross-diffusion and changes into:
wci
u i
part
ui
sol
dSdt Oi
wn
( S (t ), t ) (7)
Here we also transformed the particle and interfacial concentrations. Furthermore, the condition
in equation (1) changes due to the transformation into
m1 m2 mn
§ n · § n · § n ·
¨¨ ¦ p1 j u j ¸¸ ¨¨ ¦ p2 j u j ¸¸ ¨ ¦ pnj u sol j ¸
sol sol
K K (T ) , (8)
¨j 1 ¸
©j 1 ¹ ©j1 ¹ © ¹
which is slightly more complicated than equation (1).

57
We solve the transformed vector-valued Stefan problem by Finite Differences for the diffu-
sion equation. Further, we use an iterative Trapezoidal Rule for the displacement of the moving
boundary. We compared several methods for the non-linear system: an accuarate Newton
scheme with a discretized approximation for the Jacobian entries, Picard scheme and a relaxed
Picard scheme. It turned out that the differences between the performance of the Picard scheme
and the relaxed Picard scheme are very small. Further, the advantage of a Newton scheme is
that convergence is fast (second order). However, each iteration is expensive since an iteration
takes several (five) concentration profile evaluations. The Picard iteration scheme converges
slowly, order is linear, however, per iteration step only one evaluation of the concentration is
needed. After comparison of the two methods it turned out that the Picard scheme took less
computation time than the Newton scheme. Therefore, the Picard scheme is used. Furthermore,
we compared the scheme based on diagonalization with several numerical schemes where the
full diffusion matrix is taken into account (hence the diagonalization argument is not used).
From this comparison, we omit the details, it turns out that the diagonalization procedure gives
an efficient method in terms of stability and calculation time. For more details on the compari-
son between the various numerical methods, we refer to [17].
In Vermolen et al [14] a self-similar solution is given for a planar and spherical geometry.
The solution is valid for an unbounded domain. From the self-similarity solutions a quasi-binary
solution is derived where cross-diffusion has been taken into account. Maugis [16] derives a
quasi-binary solution for particle growth in a multi-component alloy where the off-diagonal
terms in the diffusion matrix were set equal to zero. At the early stages the solution for the
bounded domain resembles the self-similar solution. This is used to validate the numerical solu-
tion.
We note here that the surface tension in terms of the Gibbs-Thomson effect can be imple-
mented easily for curved geometries. Since in Vermolen et al [14] it is shown that this effect is
negligible for dissolution of a sphere unless the surface tension is unphysically large, we omit
this effect. However, for the case of particle growth, this effect may not be disregarded especial-
ly at the early stages. Further we remark that cases in which the eigenvalues of the diffusion ma-
trix are complex-valued can be dealt with as well. In this paper we omit the treatment of these
cases.
5 Calculated Results
To illustrate the consistency of the numerical method with the self-similar solution, which is de-
rived in [14], we show the interface position as a function of time for the self-similar solution
and the numerical solution in Figure 1. At the early stages of the dissolution process the numeri-
cal solution and the self-similar solution co-incide, which is to be expected since the atoms
hardly reach the outer boundary where no flux of atoms is imposed. However, it can be seen
that due to soft-impingement at the cell boundary the dissolution process slows down after some
time. The self-similarity solution is valid for an unbounded domain so this feature is not incor-
porated here. Hence this solution does not predict the decrease of the dissolution rate due to
soft-impingement and hence the quality of this approximation decreases as time proceeds.
58
1,00E+00
9,50E-01
Position of the interface
9,00E-01
Series1
Series2
8,50E-01
8,00E-01
7,50E-01
0,00E+00 1,00E+01 2,00E+01 3,00E+01 4,00E+01 5,00E+01 6,00E+01 7,00E+01 8,00E+01 9,00E+01 1,00E+02
Time
Figure 1: The interface position as a function of time. Series 1 represents the computed results for the analytical
solution, derived in [14] and series 2 represents the numerical solution.
1,00E+00
9,50E-01
Position of the interface
9,00E-01
Series1
Series2
Series3
8,50E-01
8,00E-01
7,50E-01
0 10 20 30 40 50 60 70 80 90 100
Time
Figure 2: The interface position as a function of time. Series 1, 2 and 3 respectively correspond to D12 = 0,
D12 = –0.5 and D12 = 0.5. All the results were obtained by the use of the numerical method.
59
To illustrate the significance of the influence of the cross-diffusion terms on the dissolution
rate, we show the interface position as a function of time. We take an initial particle size of one
micron, the main diagonal entries of the diffusion matrix are given by D11 = 1, D22 = 2,
cpart1 = 50 = cpart2 to obtain the results in Figure 2. Further, we take the diffusion matrix symmet-
ric, i.e. D12 = D21 for the sake of illustration. It can be seen that once the cross-terms become
‘more negative’ (larger magnitude but negative value), the dissolution rate is delayed. This is to
be expected since if the cross-terms have a negative value then the diffusion process is deterio-
rated. Further, it can be seen that the larger positive values of the cross-terms enhance the disso-
lution process, which is in line with the physical expectations. Furthermore, all curves tend to
the same asymptotic value for large times. This is to be expected since the mass for all of these
situations is similar in the bounded domain and hence the same final interfacial position should
be attained.
6 Conclusions
A model for the dissolution of particles in multi-component alloys taking into account the ef-
fects of cross-diffusion is formulated. A numerical method, based on Finite Differences and a
diagonalization argument of the diffusion matrix, is used to solve the equations. The numerical
results are consistent with the self-similarity solution, which was derived earlier, at the early
stages of the dissolution process. However, as time proceeds, the effects of soft-impingement
become gradually more important and can not be ignored any longer. Hence the accuracy of the
analytical solution deteriorates as time proceeds. Further, it can be seen that if the magnitude of
the cross-diffusion terms is of the same order as the diagonal terms then their impact on the dis-
solution rate is substantial. Finally, it is observed that the diagonalization argument facilitates
the numerical implementation and it also contributes to a decrease of computational time.
7 References
[1] M.J. Whelan, Metals Science Journal, 3 (1969) 95–97

[2] F.V. Nolfi jr, P.G. Shewmon and J.S. Foster, Transactions of the metallurgical society of
AIME, 245 (1969) 1427–1433
[3] O. Reise, N. Ryum and J. Strid, Metallurgical Transactions A, 24A (1993) 2629–2641
[4] R. Hubert, ATB metallurgie, 34-35 (1995) 4–14
[5] C. Atkinson, T. Akbay and R.C. Reed, Acta Metallurgica, 43-5 (1995) 2012–2031
[6] U. Grafe, B. Bottger, J. Tiaden and S.G. Fries, Scripta Materialia, 42, 12 (2000)
1179–1186
[7] H. Emmerich, The diffuse interface approach in materials science: thermodynamic con-
cepts and applications of phase-fields models, Springer Lecture notes in physics, 2003
[8] G.B. Kale, K. Bhamumurty, S.K. Khera and M.K. Asundi, Materials Transactions JIM,
32, 11 (1991) 1034–1041
[9] A. Vergara, L. Paduano, V. Vitagliano and R. Sartorio, Materials Chemistry and Physics,
66 (2000) 126-131
[10] K.P. Travis and K.E. Gubbins, Molecular Simulation, 27, 5-6 (2001) 405–439
60
[11] J.S. Kirkaldy and D.J. Young, Diffusion in the condensed state, The institute of metals,
London, 1987
[12] M.E. Glicksman, Diffusion in solids, John Wiley and Sons, New York, 2000
[13] E.B.Naumann and J. Savoca, AICHE Journal, 47-5 (2001) 1016–1021
[14] F.J. Vermolen, C. Vuik and S. van der Zwaag, Materials Science and Engineering A,
A347 (2003) 265–279
[15] F.J. Vermolen, C. Vuik and S. van der Zwaag, Some mathematical aspects on cross-diffu-
sion and particle dissolution in multi-component alloys, Technical report at the Depart-
ment of Mathematical Analysis: TWA-01-15, Delft University of Technology, The
Netherlands (2001)
[16] P. Maugis, Materials Science Forum, 462-4 (2003) 1313–1318
[17] F.J. Vermolen and C. Vuik, Solution of vector-valued Stefan problems with cross-diffu-
sion, Technical report at the Department of Mathematical Analysis: TWA-03-14, Delft
University of Technology, The Netherlands (2003)
61
Modeling of the Precipitation Evolution During Non–isothermal

Heat Treatments in an Al Zn Mg Alloy
M. Nicolas, A. Deschamps
LTPCM, UMR CNRS 5614, Institut National Polytechnique de Grenoble, Saint Martin d’Hères, France
1 Introduction
During the process of most age hardening aluminum alloys, non-isothermal precipitation plays
a key role. A particularly important practical situation is met during welding, when the heat af-
fected zone is subjected to a rapid temperature peak, during which precipitate microstructures
can evolve. The purpose of the present study is to develop a model capable of predicting the be-
havior of an initial precipitate microstructure subjected to a wide range of non-isothermal situa-
tions, including complicated cases like the temperature profiles encountered in the heat-affected
zones (HAZ) of welds. This will be achieved by combining a quantitative study of precipitation
kinetics during non-isothermal situations, and a precipitation model capable of dealing with
continuous temperature changes.
This study is carried out on a medium-strength 7000 series alloy (AA7108.50) based on the
Al-Zn-Mg system, used in the automotive industry for making high strength, welded structural
parts. The strength of this alloy is mainly controlled by precipitation, which occurs through a
complex sequence of precipitation involving both metastable phases like GP zones and the K'-
phase, the stable phase being K of equilibrium composition MgZn2 [1].
The quantitative study of precipitation is carried out using in-situ small-angle X-ray scatter-
ing. The modeling approach is based on the approach originally developed by Wagner and
Kampmann [2], based on the discretization of a precipitate size distribution into size classes,
and the description of the behavior of each class by classical growth/dissolution laws has prov-
en successful [3,4]. We will adapt such a model to the present ternary alloy.
2 Materials and Experimental Techniques
The alloy of study is the commercial Al-Zn-Mg alloy 7108.50 (Al-5wt%Zn-0.8wt%Mg-

0.16wt%Zr). The initial temper used is the T7 state, which is obtained by two-step ageing per-
formed after solution treatment at 480 °C and water quench. The first step consists in a 6-hours
stage at 100 °C, and the second step consists in 6 h at 170 °C.
Small-Angle X-ray Scattering has been used to determine quantitatively the features of the
precipitate microstructure: mean radius and volume fraction. All SAXS experiments presented
here have been carried out on the D2AM beam-line (BM02-CRG) of the European Synchrotron
Radiation Facility (ESRF). Details of the experiment set-up can be found in [5]. The precipitate
size was evaluated using the Guinier approximation, which provides the Gyration radius Rg of
the particles (q is the scattering vector (Å–1))

I v exp q 2 Rg2 3 (1)

ISBN: 3-527-31007-X
62
The mean radius of the particle size distribution was subsequently determined by comparison
with TEM results, and the following ratio was found to give a good description of all types of
precipitates:
Rmean (TEM) = 0.95 Rg (SAXS) (2)
In the following, all precipitate sizes determined from SAXS experiments are converted in ”re-
al-space” mean radius using this relationship. The precipitate volume fraction was evaluated
using the measured integrated intensity Qo:
f
Q0
³ I (q)q²dq
0
2S ²('U)² f v (1 f v ) (3)
where 'U is the electron density contrast between the precipitate and matrix. The detail of the
evaluation of this contrast term is found in [5].
In-situ SAXS experiments using a rapid furnace have been used to follow the precipitation
evolution during non-isothermal heat treatments. Two types of heat treatments have been inves-
tigated: reversion treatments characterized by an isothermal holding at the desired temperature
subsequent to a fast heating ramp (300 K/min) and continuous heating at constant heating rate.
3 Results
Figure 1 shows the evolution of volume fraction and precipitate size during reversion experi-
ments at temperatures ranging from 200 to 300 °C. The general behavior is a fast dissolution,
which extent increases with increasing reversion temperature, followed by a slow increase of
volume fraction, up to the equilibrium volume fraction at the reversion temperature, which de-
creases with increasing temperature. During the first stage, the average precipitate radius stays
roughly constant, whereas it increases quickly during the second stage (coarsening stage).
3 150
Volume fraction (%)
Particle radius (Å)
280°C
2.5
240°C
2 200°C 100
200°C
220°C
1.5 200°C
240°C
260°C
1 50
280°C
0.5
300°C 300°C
0 4
0 4
1 10 100 1000 10 1 10 100 1000 10
Time (s) Time (s)
Figure 1: Evolution of precipitate volume fraction and size during an isothermal reversion treatment as a function
of temperature (dots: experimental results, lines: model predictions)
63
Figure 2 shows the evolution of volume fraction during continuous heating as a function of
heating rate. A classical sigmoïdal decrease of the volume fraction is observed. In the initial part
of the dissolution, an increase in heating rate results in a positive shift in dissolution tempera-
ture, as expected from kinetic reasons. Less classical is the subsequent behavior for the slow
heating rates: a temporary interruption of dissolution is observed around 250 °C, before com-
plete dissolution occurs at higher temperature.
3 3
Volume fraction (%)
Volume fraction (%)

10°/min 10°/min
2.5 20°/min 2.5 20°/min
50°/min 50°/min
2 100°/min 2 100°/min
300°/min 300°/min
1.5 1.5
1 1
0.5 0.5
0 0
150 200 250 300 350 400 150 200 250 300 350 400
Temperature (°C) Temperature (°C)
Figure 2: Evolution of precipitate volume fraction during continuous heating as a function of heating rate (left
experimental results, right model predictions)
4 Modeling of Non-isothermal Precipitation Kinetics
We have developed a model based on the methodology initially proposed by Wagner and
Kampmann [2]. Starting from the particle size distribution of the initial state (determined expe-
rimentally), the stability of each size class is estimated at each time step and thus, precipitates
will either grow or shrink depending on their size respective to the critical radius. The particle
size distribution and remaining matrix concentration are recalculated and used for the next time
step. Outputs of the models are the particle size distribution, the volume fraction, the mean par-
ticle size, the concentration in the matrix, etc.
Some simplifying assumptions have been used to limit the complexity of the model:
• The composition of the precipitates is reduced to the stœchiometric equilibrium composi-

tion of the K-phase: MgZn2, and the solvus boundary of the K-phase in the ternary Al-Zn-
Mg system is described by a solubility product.
• The growth/dissolution of the MgZn2 precipitates is controlled by diffusion to the particle/
matrix interface.
• Nucleation is not considered since we are interested in the evolution of a pre-precipitated
microstructure.
The model is based on the classical law governing both dissolution and growth and applied
to each size class of the particle distribution:
64
dr X X i (r ) D
Q (4)
dt X p X i (r ) r
where r is the radius at time t, X the mean solute fraction in the matrix (at%), Xi(r) the solute
concentration at the particle/matrix interface (at%), Xp the solute concentration of the precipitate
(at%) and D the volume diffusion coefficient in the matrix.
All features are relative to the limiting species regarding diffusion, in our case magnesium.
X is obtained by a mass balance between precipitates and matrix. The equilibrium composition
MgZn2 of the K-phase has been considered for the value of Xp. The diffusion coefficient is de-
termined by an Arrhenius law:
Qdiff
D D0 exp( ) (5)
RT
Xi(r) is the solute fraction at the particle/matrix interface. In the case of a diffusion-controlled
process, Xi(r) is approximated by the equilibrium solute fraction of the matrix in presence of a
particle of radius r at the considered temperature. In a ternary system, the Gibbs-Thomson mod-
ified equilibrium diagram is not sufficient to determine univocally Xi(r). An additional equation
is provided by the flux equality at the interface, which is a necessary condition to satisfy the st-
oechiometry of the precipitate :
DZn X i,Zn (r ) X Zn 2 DMg X i,Mg (r ) X Mg (6)
DZn and DMg are the respective diffusion coefficients for Zn and Mg, Xi,Zn(r) and Xi,Mg(r) the
interfacial solute fraction of Zn and Mg of a particle of radius r and X Zn and X Mg the mean
solute fractions in Zn and Mg in the matrix far from the particle.
The equilibrium solvus boundary is simply described by a solubility product:
ǻ Sq ǻ H q
ln K f o
ln (X Zn o
)² X Mg R

RT
(7)
o o
where X Zn and X Mg are the equilibrium atomic fractions of zinc and magnesium respectively,
R the universal gas constant, T the absolute temperature, 'S° and 'H° are the entropy and ent-
halpy, respectively, associated with the precipitation/dissolution of the K-phase. When dealing
with ternary alloys, the Gibbs-Thompson effect has been generalized to the solubility product,
defining a solvus boundary for a precipitate of size r:
§ 2J Vm ·
K (r ) K f exp ¨ ¸ (8)
© r RT ¹
J is the particle/matrix interfacial energy, Vm is the molar volume of the precipitate MgZn2.
Xi(r) (referring to the limiting specie, here Mg) is then determined by the intersection be-
tween the solubility limit K(r) and the flux equality equation. In the phase diagram representa-
tion, the flux equality equation is a straight line going through the point corresponding to the
65
CZn
Flux equality line
t t0
stoechiometric line
C0
Ci(t∞) Ci(ro,to)
Ci(r,t)
CMg
Figure 3: Representation of the equilibrium and Gibbs-modified solubility product in the ternary phase diagram.
The intersection of the flux equality line with the solubility product line defines the solute concentration at the pre-
cipitate interface.
mean concentration in the matrix, which slope is DZn / 2DMg. The determination of Xi(r) is illus-
trated in figure 3 for a given size class of size r in the growth regime.
5 Application of the Model
The above precipitation model has been calibrated on the reversion experiments presented abo-
ve. Subsequently, we will study its applicability to other types of non-isothermal heat treat-
ments.
The parameters which need to be calibrated in the model are thermodynamic constants (the
solubility product through 'S° and 'H°), the interfacial energy J of the precipitates and the dif-
fusion coefficients for both zinc and magnesium. The parameter set obtained is as follows:
'H° = 75 kJ mol–1;

'S° = 21.4 J mol–1 K–1;
Do,Mg = 1.3 10–8 m2 s–1;
Do,Zn = 1.3 10–7 m2 s–1;
Qdiff,Mg = Qdiff,Zn = 88.3 kJ mol–1;
J= 0.65 J m–2.
These parameters are reasonable when compared to literature values. Only the interface en-
ergy is somewhat higher than the range of values generally accepted for incoherent precipitates
(~0.3–0.4 J m–2). This discrepancy may arise from the composition assumption of MgZn2-pre-
cipitates which was chosen, whereas the reality is probably much closer to a Zn/Mg ratio of 1.
The capacity of the model to describe precipitation evolution during reversion treatments is
shown in figure 1. The agreement is very satisfying both in terms of volume fraction and parti-
cle size evolution.
66
With the help of the model, the precipitate microstructure evolution during reversion can
now be better understood. For the sake of simplicity, we can recall the expression of the critical
radius in a regular solution, for a binary alloy A-B, where precipitates are pure B:
2 J vat
R* (9)
k T ln( X X eq )
It can be readily observed that the critical radius decreases with increasing solute content,
and increases with increasing temperature (through the temperature dependance of Xeq).
Thus, when an initially present microstructure is subjected to a sudden temperature increase,
the critical radius becomes instantaneously much larger than the size of any precipitate present.
Consequently all precipitates are destabilized and begin to dissolve. This dissolution results in
an increase of the solid solution content, which results in a decrease of the critical radius. When
R* becomes smaller than some surviving precipitates, the dissolution rate decreases. When R*
becomes approximately equal to the average precipitate size, dissolution stops and precipitates
start to coarsen. This coarsening behavior increases again both the average and critical radius,
and thus the solid solution is progressively desaturated until the equilibrium volume fraction at
the reversion temperature is reached. This dynamic behavior is illustrated in figure 4.
3 150
Volume fraction (%)
A r*
2.5
D
2 100
B D
1.5
C
1 C 50
A B
rmean
0.5
Dissolution Growth
0 4
0 4
0.1 1 10 100 1000 10 1 10 100 1000 10
Time (s) Time (s)
Figure 4: Evolution of the average and critical radii during the reversion experiment at 240 °C, as compared to the
evolution of the volume fraction (model predictions)
This model can then be directly applied to the continuous heating experiments shown above.
The result of the modeling is shown in figure 2. It can be seen that the transition from the con-
tinuous dissolution behavior to the interrupted dissolution behavior is correctly predicted.
Again, the comparison between the average and critical radii for the two extreme heating rates
can be used to better understand the cause of this transition. This is illustrated in figure 5. When
the heating rate is low, dissolution starts at a relatively low temperature, and thus for a given
temperature the solute content is higher, and the critical radius is smaller. This dissolution oc-
curs to an extent such as, in a given temperature range, the average and critical radii become
equal, which promotes a coarsening regime, and thus an interruption of the dissolution curve.
On the contrary, in the case of a high heating rate, dissolution starts at a higher temperature, and
67
150 150

10°C/min 100°C/min
100 100
50 50
r r
mean mean
r* r*
0 0
50 100 150 200 250 300 50 100 150 200 250 300
Temperature (°C) Temperature (°C)
Figure 5: Evolution of the precipitate size and critical radius during continous heating, for the 10 and 100 °C/min
heating rates (model predictions)
the critical radius stays at all times much larger than the average radius, resulting in continuous
dissolution.
Finally, this model has been applied to the thermal profiles met in the heat-affected zones of
MIG welds. The temperature history has been measured by thermocouples placed at different
distances from the weld line, and calculated continuously by FEM in all other points. The tem-
perature profile for each point of the HAZ is then used to predict the microstructure present at
the end of the welding cycle. These predictions are compared with the experimental measure-
ments of microstructure profiles, achieved by scanning a HAZ with an X-ray beam and evaluat-
ing the precipitate size and volume fraction by small angle scattering.
The comparison is shown in Figure 6. The HAZ can be separated in two zones: first, close to
the weld line, all precipitates are dissolved by the welding process, and some GP zones re-pre-
cipitate during the last stages of cooling to room temperature. Further from the weld line, only
partial dissolution occurs, accompanied by an increase in average precipitate size. Except for
2
Fv (%)
500
Temperature (°C)
Weldsim - DFW = 10mm

400 Weldsim - DFW = 17mm 1
Weldsim - DFW = 30mm SAXS
Model
300 0
40
R (Å)
200 30 R (η)
20
100
10
0
0
0 10 20 30 40 50 60 70 0 5 10 15 20 25 30 35 40
Time (s) Distance from weld line (mm)
Figure 6: Thermal profiles in the HAZ of a MIG weld for three distances from the weld line, and microstructural
profile in the HAZ, both measured by SAXS and predicted by the model.
68
the re-increase in volume fraction due to GP zone precipitation, the microstructural profile is
correctly predicted by the precipitation model.
6 Conclusions
This paper presents a quantitative study of the stability of precipitates in an AlZnMg alloy sub-
jected to three types of non-isothermal heat treatments: isothermal reversion, continuous ramp
heating, and temperature peaks met in the heat-affected zones of MIG welds. Precipitation mo-
dels considering the complete size distribution have been adapted to the present precipitation
phenomena. The model, adjusted to the reversion experiments, has proven successful in predic-
ting in detail the behavior in the two other types of temperature histories.
The dissolution behavior can be well understood in terms of the relative values of the aver-
age and critical radius, the latter being dependent both on temperature and the instantaneous sol-
ute content.
7 Acknowledgements
The authors would like to thank F. Bley, F. Livet, and the team of ESRF / D2AM, for help with
the SAXS experiments. Hydro Automotive Structures is thanked for providing the material and
for financial support. Pr. Y. Bréchet is thanked for fruitful discussions.
8 References
[1] S. P. Ringer, K. Hono, Materials Characterization 2000, 44, 101–131

[2] R. Wagner, R. Kampmann, Materials Science and Technology: A comprehensive treat-
ment, vol. 5, VCH, Weinheim, 1991, 213
[3] O. R. Myhr , Ø. Grong, Acta Materialia 2000, 48 (7), 1605–1615
[4] J. D. Robson, P. B. Prangnell, Acta Materialia 2001, 49, 599–613
[5] A. Deschamps, M. Nicolas, Acta Materialia (in press)
69
Elastic Strain Energy Study of Directional Coarsening of

J’ Precipitates in Single Crystal Superalloys: A 3D Finite
Element Analysis
W. Chen1, Q. K. K. Liu2, G. Schumacher2, N. Wanderka2, and W. Neumann1

1
Institute of Physics, Humboldt University of Berlin, Germany
2
Department of Structure Research, Hahn-Meitner Institute Berlin GmbH, Germany
1 Introduction
Single crystal nickel base superalloys are widely used as blade materials in aero engines and in
gas turbines for power generation. Their superior strengths at elevated temperatures are attribu-
ted to high volume fraction of initially cuboidal intermetallic J’ precipitates with L12 superlat-
tice structure. The J’ precipitates generally have a unimodal size distribution with an average
size of about 450 nm in edge length and are embedded coherently in the fcc J matrix with a re-
gular distribution. Under service conditions at the material temperature up to 1273 K the J’ pre-
cipitates in single crystal superalloys are, however, not stable. Single crystal blades after long
term exposure under service conditions showed a pronounced morphological change of the J’
precipitates (formation of J’ rafts) [1]. Under laboratory conditions two kinds of J’ rafts, plate-
like geometry perpendicular to the external load direction and rod-like geometry with the rod
axis parallel to the external load, have been observed frequently in creep-deformed single cry-
stal superalloys with load axis parallel to one of the <100> crystallographic orientations [e.g.
2–6]. The appearance of J’ rafts in this connection depends on the nature of the external loading,
the sign of the lattice mismatch between the J’ and J phases, and the respective elastic constants
[2, 3]. Consequently, the high temperature mechanical behaviour of single crystal superalloys is
influenced by the microstructural evolution occurring during creep deformation [4, 5].
The formation of J’ rafts in single crystal superalloys during creep deformation has been in-
vestigated extensively. From the thermodynamic viewpoint all three energy terms, i.e. the
chemical potential, the interface energy and the elastic strain energy, in principle would influ-
ence the rafting processes. A recent study of Nabarro [7] has shown that the chemical potential
term of the driving force for J’ rafting induced by external loading is much higher than the elas-
tic strain energy term caused by the lattice mismatch, the difference in elastic constants of the J’
and J phases, and the external loading. The value of the chemical potential term, however, van-
ishes rapidly as soon as the J’ rafting process starts and has, therefore, essentially no influence
on the J’ rafting processes. The elastic strain energy term is thought to play an important role in
the rafting processes of the J’ phase under creep loading conditions. Various analytical methods
have been employed to estimate the elastic strain energy terms for J’ rafting [e.g. 2, 3, 8–10].
Generally the J’ rafting was considered to occur simultaneously with creep deformation under
external mechanical loading. Recent experimental studies have revealed that a small amount of
creep deformation of the order of 0.1 % is enough to cause J’ rafts formation during the subse-
quent annealing without external loading [11–14]. These observations lead us to re-examine
carefully the existing models that assume the J’ rafts formation taking place only under external
mechanical loading. In the present study the elastic strain energies for all three morphologies of

ISBN: 3-527-31007-X
70
the initially cuboidal J’ precipitates as well as of the plate-like and rod-like J’ rafts at 1223 K,
which is a relevant temperature for gas turbine blades, were calculated using the elasticity theo-
ry and the three dimensional finite element method. The results are discussed in the light of the
minimum energy criterion.
2 Unit Cell and Material Parameters
The typical initial microstructure of single crystal superalloys, as shown e.g. in Figure 1a, con-
sists of cuboidal J’ precipitates which are embedded in the J matrix quite regularly. Figures 1b
and 1c show plate-like and rod-like J’ rafts in a model single crystal superalloy SC16 after ten-
sile and compressive creep deformation, respectively. In order to simplify the finite element cal-
culation it is assumed in the present study that both the initial J’ precipitates and the J’ rafts built
during creep deformation have a strictly periodical distribution in the fcc J matrix. The calcula-
tions therefore can be reduced to a volume which contains only one J’ precipitate or one J’ raft
together with the corresponding J matrix. Due to the four-fold symmetry of the microstructures
with respect to the load axis parallel to [100] crystal orientation only 1/8 of the above volume
(referred as unit cell in the present paper) is necessary for the finite element calculations.
3µm 3 µm
a) b) c)
Figure 1: Micrographs of a) initial microstructure (TEM, dark field image using the super lattice reflection 001),
b) plate-like J’ rafts (SEM) after tensile creep deformation (T = 1223 K, V = 150 MPa) and c) rod-like J’ rafts
(SEM) induced by compressive creep deformation (T = 1223 K, V = –150 MPa) in SC16
For the initial microstructure the geometrical parameters of the J’ precipitates in SC16 (Ni-
3.5Al-16Cr-3Mo-3.5Ta-3.5Ti in wt.%) after full heat treatments [6] were used in the present
study. The alloy contains cuboidal J’ precipitates with a volume fraction of about 40 vol.%. The
cuboidal J’ precipitates have an average edge length of about 450 nm and a radius of rounded
corner of about 1/4 of the edge length. In the present study the unit cell of the initial microstruc-
ture is generated employing the above morphological parameters and is plotted in Figure 2a.
Microscopic investigations [6] have shown that the volume of the J’ rafts is generally much
larger than the initial cuboidal J’ precipitates, see also Figures 1a–1c. Without making the finite
element calculations too unwieldy but still realistic, we assumed that the J’ rafts are formed by
coalescence of four cuboidal J’ precipitates either perpendicular to the load axis (plate-like J’
rafts) or along the load axis (rod-like J’ rafts). We expect that this rough microstructural ap-
proach is sufficient to bring out the trend and other qualitative behaviours of the J’ rafting. The
71
a b c
Figure 2: Unit cells used in the present study for microstructures with a) cuboidal J’ precipitates, b) plate-like J’
rafts and c) rod-like J’ rafts
thickness of the plate-like J’ rafts and the cross section of the rod-like J’ rafts were chosen to
have the same dimension as the edge length of the initial cuboidal J’ precipitates of 450 nm. The
radius of the rounded edges of the two types of the unit cells of J’ rafts was kept the same as that
of the initial cuboidal J’ precipitates. The volume of the unit cells of both kinds of J’ rafts was
chosen to be four times of that of an initial cuboidal J’ precipitate in order to keep the J’ volume
fraction of 40 % unchanged. Figures 2b and 2c show respectively the unit cells employed in the
present study.
Due to the requirements on symmetry and on periodicity in distribution of the initial J’ pre-
cipitates as well as the J’ rafts in the J matrix, the surfaces of all three unit cells were kept to be
movable during the whole finite element calculations, but remained always planar.
A series of lattice mismatch parameters in a range of –0.005 to +0.005 was used to study the
influence of lattice mismatch on elastic strain energies. The thermal expansion method was used
to introduce lattice mismatch into the unit cell: It was assumed that all three linear thermal ex-
pansion coefficients of the J’ phase have a fictive value of 10–5 K–1 and those of the J matrix a
value of 0 K–1. Different lattice mismatch levels can be reached by varying the temperature of
the unit cell. A detailed analysis was carried out, however, for the calculation using a lattice
mismatch parameter of –0.001 which is considered to be a representative value for most rele-
vant single crystal superalloys used in gas turbines. The elastic constants of the J’ precipitates
and of the J matrix at 1223 K are not available for the SC16 alloy studied in the present work.
As approximations we have used data taken from literature [15] and are summarised in Table 1.
To study the influence of the external load on the distribution of lattice distortion and on the
elastic strain energies the creep loads corresponding to the elastic strains of ±0.1 % along [100]
crystal orientation (parallel to x axis in the coordinate system used in the present study) were in-
cluded additionally in the calculations of elastic strain energies under tensile and compressive
loading, respectively.
The present investigation was carried out using a commercial finite element software pack-
age MARC 2001 and MENTAT 2001 [16].
72
Table 1: Elastic constants of J’ and J phases at 1223 K (in N/nm2)

c11 c12 c44
J’ 2.04 10–7 1.41 10–7 9.90 10–8
J 1.68 10–7 1.14 10–7 9.00 10–8
3.1 Elastic Strain Energies Induced by Lattice Mismatch
Figures 3a–3c show the elastic strain energy density calculated for the cuboidal J’ precipitates
and for the J’ rafts with plate-like and rod-like morphologies, resp., using a realistic lattice mis-
match of –0.001 for SC16 at 1223 K [17]. Further analyses show that the distribution of elastic
strain energy density is essentially independent of the data of the lattice mismatch. The absolute
values of elastic strain energy density vary, however, with the lattice mismatch parameters used
in the calculations.
Figure 3: Elastic strain energy density of unit cells with a) cuboidal J’ precipitate, b) plate-like and c) rod-like J’
raft induced by a lattice mismatch of –0.001 at 1223 K. Note that the volumes of b) and c) are four times larger
than a). (in x10–13 N nm–2)
By integrating the elastic strain energy density over the volume of unit cells we obtain the to-
tal elastic strain energies Etotal for different lattice mismatch parameters in the range from
–0.005 to +0.005. Figure 4 shows an example of the results of the elastic strain energy calcula-
tions carried out for the cuboidal J’ precipitates at 1223 K. The total elastic strain energies for
the negative lattice mismatches are slightly smaller than those obtained using the positive lattice
mismatches. The same tendency could also be observed for the two kinds of J’ rafts. This differ-
ence is caused by the application of the thermal expansion technique in the present study to in-
troduce the lattice mismatch into the unit cells, described briefly in the previous section. The
final volume of the unit cell at the end of finite element calculation is slightly larger for positive
lattice mismatch than for negative lattice mismatch with the same absolute value. Since the total
elastic strain energy is obtained by integrating the elastic strain energy density over the final
volume, hence we see the above discrepancy. Apart from this negligibly small difference the to-
tal elastic strain energies in the positive and negative range of lattice mismatch are essentially
symmetrical. The following discussion is concentrated on the total elastic strain energies in su-
peralloys with negative lattice mismatches.
73
2.5 -7
1.0x10 Cuboidal
Plate-like
2.0
Eaverage in x10 N*nm

-8
Etotal in x10 N*nm
8.0x10 Rod-like
1.5 -8
6.0x10
-5
-5
-8
1.0 4.0x10
-8
0.5 2.0x10
0.0 0.0
-0.006 -0.003 0.000 0.003 0.006 -0.0050 -0.0025 0.0000

Lattice Mismatch Lattice Mismatch
Figure 4: Total elastic strain energy Etotal induced Figure 5: Average elastic strain energies of unit cells
by lattice mismatch for cuboidal J’ precipitates at with cuboidal, plate-like and rod-like J’ precipitates
1223 K at 1223 K
Since by construction of the unit cells the volume of the plate-like and rod-like J’ rafts are
larger than that of the initial cuboidal J’ precipitates, the total strain energy can not be used di-
rectly as a criterion of comparison. Instead, the volume-averaged elastic strain energies Eaverage
were calculated for all three morphologies and are plotted in Figure 5. We obtain from the
present study the following relationship
Eaverage(cuboidal) > Eaverage(rod-like) > Eaverage(plate-like).
In [18] similar results were also reported for the elastic strain energy estimates using strongly
simplified morphologies of the J’ precipitates as well as the J’ rafts in case that no elastic misfit
is applied in the calculations. The difference in the volume-averaged elastic strain energies
among the cuboidal J’ precipitates, the plate-like and the rod-like J’ rafts increases with increas-
ing absolute values of lattice mismatch between the J’ and J phases [Figure 5].
The fact that the cuboidal morphology has the higher value of the volume-averaged elastic
strain energy than the two other morphologies can be understood from the point of view of the
difference in the volumes of the respective unit cells. The region of high elastic strain energy
density is essentially found at the interphase interface between the J’ and J phases, see Figures
3a–3c. The ratio of interphase interface to volume in the unit cells generally becomes smaller
with increasing J’ size. As we used larger volumes to represent the plate-like and rod-like mor-
phologies, the volume of the relatively unstrained material in the unit cells increases with the
volume of the J’ phase. Hence, the volume averaged elastic strain energy decreases with respect
to the volume of J’ phase contained in the unit cells.
Based on the present results, the coalescence of initially cuboidal J’ precipitates, from the
point of view of the elastic strain energy, can take place even without external mechanical load-
ing. Such phenomena were observed in superalloys after a long term ageing treatment at high
temperature [e.g. in 19]. This kind of coalescence of the J’ precipitates, however, did not show
an orientation preference as observed in creep-deformed superalloys. Recent experimental in-
vestigations have revealed the J’ rafts formation during high temperature annealing in pre-
strained single crystal superalloys under appropriate creep conditions [11–14]. Our finite ele-
ment results indicate that J’ rafting would occur without external mechanical loading since both
the J’ rafts morphologies are energetically favoured compared to the initial cuboidal J’ mor-
74
phology (see Figure 5), provided that a starting barrier would be overcome during the pre-de-
formation at high temperature. A possible mechanism has been suggested by Reed and co-wor-
kers [14].
3.2 Elastic Strain Energies Under Creep Loading
Figures 6a–6c show the results of the elastic strain energy density calculated for the cuboidal J’
precipitate and for the J’ rafts with plate-like as well as rod-like morphologies, resp., using the
same material parameters as employed in Figures 3 and an external tensile strain load of 0.1 %
additionally. Similar calculations were also carried out for all three J’ morphologies under com-
pressive creep loading.
Figure 6: Elastic strain energy density of unit cells with a) cuboidal J’ precipitate, b) plate-like and c) rod-like J’
raft under tensile strain loading of 0.1 % at 1223 K (in x10–13 N nm–2)
The volume-averaged elastic strain energies Eaverage,loaded for cuboidal J’ precipitates and both
kinds of J’ rafts under tensile and compressive loading are shown in Figure 7. The application
of the external mechanical loading leads to an increase in the average elastic strain energies.
However, the inequalities among the volume-averaged elastic strain energies of three J’ mor-
phologies remain unchanged compared to the case of no mechanical loading:
Eaverage,loaded (cuboidal) > Eaverage,loaded (rod-like) > Eaverage,loaded (plate-like).
This behaviour can be understood in the sense of the Colonnetti’s theorem that the internal
energy of a solid strained both by internal and external load does not contain the energy term
caused by interaction of the internal and the external load [20]. The elastic strain energy is
therefore the sum of the elastic strain energy induced by the lattice mismatch and the elastic
work done by the external load. Since the latter per unit volume is approximately the same for
all three morphologies, the inequality relationship survives under loading condition.
A different inequality relationship has been given by several studies [e.g. 2, 3] using the elas-
tic inclusion theory developed by Eshelby [21]. The discrepancy between the results presented
in this paper and those obtained using the elastic inclusion methods could be caused by some as-
75
9
tensile
8
compressive
N*nm
7
6
-14
5
Eaverage in x 10
4
3
2
1
0
Cuboidal Plate-like Rod-like
Morphology of γ' Precipitates
Figure 7: Volume-averaged elastic strain energies as function of J’ morphologies under mechanical loading
sumptions which are indispensable for efforts to obtain an analytic solution. Typically, a spher-
ical morphology is assumed in most of the analytic calculations as an alternative of the cuboidal
geometry of the J’ precipitates. This assumption generally leads to an underestimation of the
elastic strain energy for a realistic microstructure in single crystal superalloys.
For the same absolute value, the sign of the mechanical loading does not change significantly
the volume-averaged elastic strain energies (see Figure 7), although compressive loading does
lead to a slightly larger volume-averaged elastic strain energy than tensile loading. The forma-
tion of the plate-like J’ rafts under tensile loading and of the rod-like J’ rafts under compressive
loading in a single crystal superalloy with negative lattice mismatch could not be explained us-
ing arguments of minimum elastic strain energy. The difference in the local microstructural ev-
olution under tensile and compressive loading might play a key role in the development of the
respective J’ rafts morphologies and should be investigated in detail for the understanding the J’
rafts formation behaviour during high temperature annealing in a pre-deformed single crystal
superalloy under creep loading conditions.
4 Conclusion
Using three dimensional finite element simulation we have carried out the elasticity analyses for
single crystal superalloys with an initial microstructure of cuboidal J’ precipitates, and those
with plate-like J’ rafts formed by coalescence of four cuboidal J’ precipitates in the direction
perpendicular to the load axis as well as those with rod-like J’ rafts formed by alignment of four
cuboidal J’ precipitates along load axis during creep deformation.
The results of the present study show that the volume-averaged elastic strain energy (elastic
strain energy per volume) induced by the lattice mismatch in superalloys with cuboidal J’ pre-
cipitates is larger than those with plate-like or rod-like J’ rafts. From the point of view of the
elastic strain energy the coalescence of cuboidal J’ precipitates into a plate-like or a rod-like
morphology is energetically favoured. This explains the formation of the J’ blocks observed fre-
quently in the superalloys after a long term ageing treatment at high temperature.
76
With or without mechanical loading, our finite element results show the same trend for the
average elastic strain energies: Eaverage (cuboidal) > Eaverage (rod-like) > Eaverage (plate-like). The
inequality relationship is independent of the sign of the applied creep load. This indicates that
the J’ rafting processes observed experimentally could not be explained by the minimum elastic
strain energy criterion alone. Microstructural changes induced by creep deformation must be in-
volved in the modelling.
5 Acknowledgement
The financial support of the DFG (Deutsche Forschungsgemeinschaft) under the grant number
NE 646/5-2 and WA 1378/1-2 is gratefully acknowledged.
6 References
[1] H. Biermann, B. Grossmann, T. Schneider, H. Feng and H. Mughrabi, Superalloys 1996,

Minerals, Metals and Materials Society, 1996, p. 201
[2] A. Pineau, Acta metall., 1976, 24, 559
[3] F.R.N. Nabarro, Metall. Mater. Trans., 1996, 27A, p. 513
[4] D.D. Pearson, F.D. Lemkey and B.H. Kear, Superalloys 1980, ASM, 1980, p. 513
[5] M. Otto, U. Tetzlaff, and H. Mughrabi, Microstructure and Mechanical Properties of
Metallic High Temperature Materials, Wiley-VCH, 1999, p. 425
[6] D. Mukherji, H. Gabrisch, W. Chen, H.J. Fecht and R.P. Wahi, Acta mater, 1997, 45,
p. 3143
[7] F.R.N. Nabarro, Scr. mater., 1997, 37, p. 497
[8] J. C. Chang and S. M. Allen, J. Mater. Res., 1991, 6, p. 1843
[9] A. G. Khachaturyan, S. Semenovskaya and T. Tsakalakovs, Phys. Rev., 1995, B52, p. 1
[10] T. Ohashi, K. Hidaka and S. Imano, Acta mater., 1997, 45, p. 1801
[11] J.Y. Buffiere and M. Ignat, Acta metal. mater., 1995, 43, p. 1791
[12] M. Veron, Y. Brechet and F. Louchet, Scr. mater., 1996, 34, p. 1883
[13] N. Matan, D.C. Cox, C.M.F. Rae and R.C. Reed, Acta mater., 1999, 47, p. 2031
[14] P. Henderson, L. Berglin and C. Jansson, Scr. mater., 1999, 40, p. 229
[15] M. Fahrmann, W. Hermann, E. Fahrmann, A. Boegli, T.M. Pollock and H.G. Sockel,
Mat. Sci. Eng., 1999, A260, p. 212
[16] MARC User’s Guide, MSC Software Corp., Santa Ana, CA., USA, 2001
[17] W. Chen, N. Darowski, I. Zizak, G. Schumacher, H. Klingelhöffer, N. Wanderka and W.
Neumann, Mat. Sci. Forum, 2003, 426–432, p. 4555
[18] T. Ichitsubo, D. Koumoto, M. Hirao, K. Tanaka, M. Osawa, T. Yokokawa and H.
Harada, Acta mater., 2003, 51, p. 4033
[19] D. Bettge, Doctoral thesis, Technical University Berlin, 1996, p. 37
[20] T. Mura, Micromechanics of Defects in Solids, Martinus Hijhoff Pub., 1987, p. 211
[21] J.D. Eshelby, Proc. Roc. Soc., 1957, A241, p. 376
39
II Martensitic and Stress Sensitive Transformations

ISBN: 3-527-31007-X
79
Possibilities for the Selective Improvement of the Mechanical

Properties of TRIP Steels Considering as Example Two Steels
with Different Carbon Contents
B. Röthler, S. Traint, A. Pichler

Voestalpine Stahl GmbH, Linz, Austria
1 Introduction
The TRIP effect (TRIP stands for transformation induced plasticity) was first described by
Zackay et. al. almost 35 years ago [1]. The high content of expensive alloying elements such as
Ni prohibited industrial use of these steel grades for automotive parts. By applying a special
two-step heat treatment in combination with alloying elements that hinder carbide precipitation
(Si, Al), it is possible to sufficiently stabilize the retained austenite at room temperature even in
low alloyed steels. Increasingly high safety standards in the automotive industry and the trend
towards saving costs and energy pushed this development of low alloyed TRIP steel sheets with
high strengths and excellent formability. Another advantage of these steels is the possibility to
save weight without compromising the high safety standards.
In the case of low alloyed TRIP steels the retained austenite is mainly stabilized by carbon
and also to some extent by manganese. The conventional composition of these steel grades is in
the range of 0.15–0.4 C, 1–2 Si, 1–2 Mn (mass contents in %). The stability of the retained
austenite against strain induced martensitic transformation is crucial for an optimum TRIP ef-
fect. If all austenite already transforms at early stages of straining, the initial hardening is high,
but the achievable elongations are low. To delay necking and to reach high elongations, the
martensite should transform gradually in a wide range of strain.
In order to keep production and processing costs low, good weldability and easy coatability
are required, too. Since the high silicon mass content in conventional low alloyed TRIP steels is
detrimental to these processing features, the silicon content has to be reduced and other alloying
elements having a similar effect on the phase transformation must be considered. Possible can-
didates to substitute or reduce silicon reported in the literature are aluminium, phosphorus and/
or copper, which are supposed to have a suppressing effect on carbide formation during the bai-
nitic reaction, too [2–11].
At the moment aluminium seems to be the most promising candidate for substituting silicon.
Several investigations dealing with alternative alloying concepts for low alloyed TRIP steels
study the influence of aluminium [2–9]. Although aluminium was found to retard carbide pre-
cipitation during isothermal holding in the bainitic range, a deterioration of the attainable tensile
strengths is reported when completely substituting silicon by aluminium [5–7]. This may be at-
tributed to the relatively weak solid solution strengthening effect of aluminium compared to sil-
icon. One possibility to compensate this deficiency is to increase the fraction of retained
austenite by raising the carbon content. There are only very few publications dealing with the
effects of carbon in Al-TRIP. In this work the effects of different carbon contents on the me-
chanical properties of a TRIP steel are taken as an example for the several strategies which have
been developed to improve the mechanical properties of TRIP steels: The key point is the frac-

ISBN: 3-527-31007-X
80
tion and properties (morphology, grain size, hardness, etc.) of the different phases (austenite,
ferrite, martensite, eventually also carbides), which can be affected by varying the alloying con-
cept (stabilization of D and/or J; solid solution hardening, grain refining,…) or by changing the
annealing conditions.
2 Experimental
2.1 Production of the Material
All experiments were performed with commercially hot and cold rolled material. The chemical
composition of the two steels investigated is shown in Table 1
Table 1: Chemical composition of the investigated steel grades, mass contents in %
Alloy C Si Mn Al
“LC” 0.2 0.3 1.5 1.2
“HC” 0.3 0.3 1.5 1.2
2.2 Annealing Simulations
The annealing simulations with the multi-purpose annealing simulator (MULTIPAS) in the la-
boratory at voestalpine Stahl GmbH [10] were conducted on industrially cold-rolled material. In
order to characterize production via continuous annealing, the influence of the annealing tempe-
rature, the starting temperature for rapid cooling (i.e. quenching temperature) and the overaging
temperature was investigated for a line speed of 80 m/min (table 2). While investigating the in-
fluence of the annealing temperature the quenching temperature and the overaging temperature
were kept constant at 700 °C and 400 °C, respectively. For the variation of the quenching tem-
perature the annealing and overaging were conducted at 800 °C and 400 °C, respectively, and
for the overaging simulations the annealing and the quenching temperatures were fixed at
800 °C and 700 °C. The sheets were characterized in the as-annealed condition without temper-
rolling.
Table 2: Annealing parameters (v: line speed, m/min; TAN: annealing temperature, °C; tAN:
annealing time, s; TQ: quenching temperature, °C; CR: cooling rate, K/s; TOA: overaging tempe-
rature, °C; tOA: overaging time, s)
V TAN tAN TQ CR TOA tOA
80 840–760 130 720–660 23–37 450–300 460
2.3 Characterization of the Microstructure and Mechanical Testing
Standard light microscopy was used to reveal the microstructure of the heat treated samples.
The specimens for the metallographic investigations of the microstructure were conventionally
prepared and etched with Nital and Klemm’s [11, 12] agents.
81
The content of retained austenite was measured using a magnetic volumetric method [13].
Mechanical properties were measured on a Roell-Korthaus RKM 250 tensile testing machine
according to European Standard EN 10 002. The strain was applied to the tensile specimens par-
allel to the rolling direction. The n-values were calculated for different strain ranges.
2.4 Thermodynamical Calculations
Thermodynamical calculations were performed to estimate the influence of the carbon content
on the equilibrium fraction of austenite formed in the two alloys during intercritical annealing
using the program ThermoCalc and the database TCFE 2000.
2.5 Dilatometric Experiments
The dilatometric investigations were conducted on a Bähr dilatometer DIL 805 A/D with sam-
ples prepared from cold rolled material (1.4 × 3.5 × 10 mm3). In all experiments a heating rate
of 25 K/s was applied. The specimens were held at an annealing temperature of 800 °C for 60 s.
The cooling rate for the subsequent cooling to the overaging temperature was 70 K/s for all di-
latometric experiments. 500 °C and 600 s, 475 °C and 600 s, 450 °C and 600 s, 400 °C and
900 s, 350 °C and 1200 s and 300 °C and 1200 s were chosen as overaging temperature and
time. Austenite and ferrite exhibit different temperature dependencies of thermal expansion.
Therefore the dilatation measured during the isothermal holding resulting from the austenite
transformation depends on the temperature and on the fraction of transformed austenite. Taking
this into account, the measured dilatation was related to the bainitic temperature according to
the normalization procedure given in [6].
3 Results
3.1 Thermodynamical Calculations
The thermodynamical calculations performed with ThermoCalc show that for the high-carbon
steel the equilibrium fraction of austenite is higher than for the low-carbon steel, which had to
be expected in accordance with the Fe-C-phase diagram. For an intercritical annealing tempera-
ture of 800 °C the alloy with a mass content of C of 0.3 % contains about 50 % more austenite
(60 %) than the one with only 0.2 % carbon (40 %).
3.2 Dilatometric Experiments
The enrichment of the austenite with carbon during the bainitic transformation in the overaging
zone is the critical step in the processing of low alloyed TRIP steels. Therefore, the amount and
kinetics of the transformation in this zone were investigated in detail with a dilatometer. The di-
latation was measured as a function of the overaging temperature and time, while the intercriti-
cal annealing parameters were kept constant for all experiments. Figures 1a, b show the
82
16 16 16 16
TB=300 °C TB=450 °C
14 14 14 TB=475 °C 14
TB=350 °C
12 12 12 TB=500 °C 12
norm. ∆l [%]
norm. ∆l [%]
TB=400 °C
[%]
[%]
10 10 10 10
8 8 8 8
6 6 6 6
4 4 4 4
2 2 2 2
0 0
0 0
0 120 240 360 480 600 720 840 0 120 240 360 480 600
a) time tOA [s] b) time tOA [s]
Figure 1: Normalized length change as a function of isothermal holding time for steel “LC” (Tan = 800 °C,
tan = 60 s; dT/dt = 70 K/s)
normalized length change as a function of the overaging time for overaging temperatures bet-
ween 300°C and 500 °C for alloy “LC“. The holding temperature affects the transformation be-
haviour (one- or two-step transformation), the amount of austenite transformed during the
bainitic holding and the kinetics of the transformation. In the case of isothermal holding bet-
ween 300 and 400 °C the transformation shows a typical one-step behaviour. Increasing the
temperature within this range results in a decrease of the transformed fraction of austenite and a
considerable acceleration of the transformation kinetics. At 400 °C the transformation is finis-
hed after only 240 s.
At holding temperatures above 400 °C the transformation behaviour is altered: At 450 °C the
bainite formation proceeds in two steps, which can be characterized as a fast first transformation
at short holding times followed by a delayed second transformation. At 475 °C the first transfor-
mation step is shifted to even shorter holding times (tB = 60 s). The dilatometric experiments
conducted with the high-carbon steel show a very similar behaviour. Only the amount of austen-
ite transformed during the holding in the overaging zone is somewhat higher than in the case of
the low-carbon variant.
3.3 Mechanical Properties
3.3.1 Variation of the Annealing Temperature
The impact of the different temperatures applied during the heat treatment of the cold rolled ma-
terial was investigated in different annealing simulations. Figure 2 reveals that the intercritical
annealing temperature seems to have nearly no influence on the mechanical properties. Neither
the yield and tensile strength nor the elongation values are considerably affected by the variati-
on of the annealing temperature between 760 and 840 °C. The tensile strength of the low-carbon
steel is about 150 MPa lower, and the total elongation is a bit higher than for the high-carbon
steel within the whole temperature range. The yield strength and the elongation-features (uni-
form elongation and total elongation) are quite the same for both alloys.
83
35
700
Rp0.2 ; Rm [MPa]
High Carbon
Low Carbon
AU ; A80 [%]
30
600
25
500 High Carbon

20 Low Carbon
a) 760 800 840 b) 760 800 840

Tan [°C] Tan [°C]
Figure 2: Influence of the annealing temperatures on the mechanical properties (TO = 700 °C; TOA = 400 °C;
v = 80 m/min); a) Rp0.2, Rm, b) Au, A80
3.3.2 Variation of the Quenching Temperature
For the variation of the quenching temperature the diagrams Rp0.2 and Rm vs. TQ and AU and A80
vs. TQ show a similar pattern: The temperature seems to have nearly no influence on the mecha-
nical properties of both steels when varied between 660 °C and 720 °C. Strikingly the yield
strength and the elongation features are again in the same range for “HC” and “LC”, while the
tensile strength is again about 150 MPa higher for steel “HC” than for steel “LC”.
3.3.3 Variation of the Overaging Temperature
In contrast to the variation of the annealing and the quenching temperature the overaging tem-
perature remarkably influences the mechanical properties of the investigated steels (see also Fi-
gure 3).
Again the run of the curves is quite similar for the high- and the low-carbon steel, and only
the tensile strength level differs by approximately 150 MPa. For temperatures between 375 °C
and 425 °C the yield strength of both alloys shows a maximum, whereas the tensile strengths are
at a minimum level in this temperature range. Below 375 °C both steels show a dual phase steel-
900 High Carbon High Carbon

35
Low Carbon Low Carbon
800
Rp0.2 ; Rm [MPa]
30
AU ; A80 [%]
700
600 25
500
20
400
300 15
280 320 360 400 440 480 280 320 360 400 440 480
a) TOA [°C] b) TOA [°C]
Figure 3: Influence of the overaging temperature on the mechanical properties (Tan = 800 °C; TO = 700 °C; v = 80
m/min); a) Rp0.2, Rm, b) Au, A80
84
like behaviour: with decreasing temperature the yield strengths decrease and the tensile
strengths increase (Fig. 3a). Figure 3b shows increasing elongations with increasing overaging
temperature and a maximum for steel “HC” and “LC” at 425 °C and 400 °C, respectively. At
temperatures above these maxima the elongation remarkably drops.
The content of the retained austenite measured for the high-carbon steel is somewhat higher
than the one for the low-carbon steel. For both alloys the austenite fraction decreases with in-
creasing overaging temperatures.
Below 375 °C “LC” and “HC” exhibit quite high n4-6 values and lower n values for higher
strain. Between 375 °C and 425 °C the behaviour is exactly reverse (see also Figure 4a and b).
18 0.40
Low Carbon
0.36 n4-6
16 n18-20
0.32
0.28
fv RA [%]
14
0.24
n [-]
12 0.20
High Carbon
Low Carbon 0.16
10
0.12
8 0.08
280 320 360 400 440 480 280 320 360 400 440 480
a) TOA [°C] b) TOA [°C]
Figure 4: Influence of the overaging temperature a) on the content of retained austenite and b) on the n-values
(Tan = 800 °C; TQ = 700 °C; v = 80 m/min)
4 Discussion
Results of the thermodynamical calculations and the fact, that for the steel with the higher car-
bon content the fraction of austenite transformed into bainite during the isothermal holding in
the overaging zone is higher than for the low carbon alloy, can thoroughly be explained by the
Fe-C phase diagram. The two step behaviour observed in the dilatometric measurements at hig-
her overaging temperatures can be explained by the incomplete bainite reaction phenomenon
and the T0 concept [14, 15]. In order to allow ferrite formation, carbon has to diffuse into the re-
tained austenite, surrounding the growing ferrite lath. Therefore formation of bainitic ferrite can
only proceed as long as the carbon concentration of the retained austenite is below the maxi-
mum carbon content according to the T0 concept (first step). Eventual precipitation of carbides
deprives the austenite from carbon, and consequently the bainite reaction can go on further (se-
cond step).
4.1 Mechanical Properties
There is a striking discrepancy between the effects of the carbon content on the yield and the
tensile strengths. Consistent for all experimental series (variation of annealing, quenching and
85
overaging temperature), the yield strength and the elongation features are in the same range for
“HC” and “LC”, while the tensile strength is about 150 MPa higher for steel “HC” than for steel
“LC”. This could be explained as follows: The higher carbon content of steel “HC” results in
more austenite formed during the intercritical annealing, which means that subsequently more
bainite can be achieved, and more retained austenite can be stabilized. Since bainite and auste-
nite are harder than ferrite, this in turn leads to an increased tensile strength. Normally this
should also raise the yield strength, but in the case of this TRIP steel it seems that the fraction of
harder particles in the ferrite matrix is too low to considerably affect the yield strength of the
composite material.
The variation of the intercritical annealing temperature reveals that this parameter seems to
have nearly no influence on the mechanical properties. This can be attributed to the effect, that
the “excess” austenite, which is formed at higher intercritical annealing temperatures, is com-
pensated by an increased formation of ferrite during cooling. This might occur, because the
austenite formed at higher temperatures is poorer in carbon, and therefore more easily trans-
forms into ferrite while cooling to the overaging zone.
In contrast to the variation of the annealing and the quenching temperatures the overaging
temperature remarkably influences the mechanical properties of the investigated steels. For in-
termediate overaging temperatures the yield strengths and the elongations of both alloys show a
maximum, whereas the tensile strengths are at a minimum level – which agrees well with the
relatively low content of retained austenite and its stability in this temperature range (see also
Fig. 4). This behaviour can be attributed to an optimum stabilization of the retained austenite by
carbon enrichment. Since kinetics of this carbon enrichment are too slow below 375 °C, the
austenite is not as stable as at higher temperatures, and easily transforms into martensite. There-
fore both steels show a dual phase steel-like behaviour: the yield strength decreases, the tensile
strength increases and the elongations decrease with lower temperatures (Fig. 3a, b). The re-
markable drop of elongations at temperatures above 425 °C and 400 °C for steel “HC” and
“LC”, respectively, can be explained by carbides, which may form at this relatively high tem-
peratures. By the precipitation of carbides the austenite is deprived of carbon, which again
means a destabilization of the austenite.
5 Conclusion
In this work the influence of the carbon content on the phase transformations during the an-
nealing treatment and consequently on the mechanical properties of low alloyed aluminium
TRIP steels are studied. These investigations are performed by means of dilatometric experi-
ments, thermodynamical calculations, tensile tests, and microstructural investigations.
The main results are stated in brief: Increasing the carbon content in low alloyed aluminium
TRIP steels increases the amount of intercritically formed austenite. Consequently also more
bainite can be achieved. For the effects on the mechanical properties this means: when applying
the same (specially designed) heat treatment, raising the carbon content by 0.1 % (mass content)
adds approximately 150 MPa to the tensile strength, while the yield stress is nearly unaffected.
Since elongation values are as good as with lower carbon contents one can conclude, that a
higher carbon content – together with the appropriate heat treatment – is one possibility to attain
an improved strength – ductility combination.
86
6 References
[1] V. F. Zackay, E. R. Parker, D. Fahr, R. Busch; Transactions of the ASM 60 (1967)

252–259
[2] S. Traint, A. Pichler, K. Hauzenberger, P. Stiaszny, E. Werner; Proc. 42nd Mechanical
Working and Steel Processing Conf., Oct. 23–25, 2000, Toronto, XXXVIII, 549–561
[3] M. De Meyer, D. Vanderschueren, B.C. De Cooman, ISIJ Int. 39, 1999, 8, 813–822
[4] M. De Meyer, D. Vanderschueren, B.C. De Cooman; Proc. 42nd Mechanical Working and
Steel Processing Conf., Oct. 23–25, 2000, Toronto, XXXVIII, 265–276
[5] E. Girault, A. Mertens, P. Jacques, Y. Houbaert, B. Verlinden, J. Van Humbeeck, Scr.
Mater. 44, 2001, 885–892
[6] S. Traint, A. Pichler, P. Stiaszny, K. Spiradek-Hahn, E. Werner; Proc. 43rd Mechanical
Working and Steel Processing Conf., Oct. 29–31, 2001, Charlotte, XXXIX, 449–467
[7] P. Jacques, E. Girault, A. Mertens, B. Verlinden, J. Van Humbeeck, F. Delannay, ISIJ Int.
41, 2001, 9, 1068–1074
[8] F. C. Chen, C. P. Chou, P. Li, S. L. Chu, Mat. Sci. & Eng. A160, 1993, 261–70
[9] A. Pichler, S. Traint, H. Pauli, H. Mildner, J. Szinyur, M. Blaimschein, P. Stiaszny, E.
Werner; Proc. 43rd Mechanical Working and Steel Processing Conf., Oct. 29–31, 2001,
Charlotte, XXXIX, 411–434
[10] P. Stiaszny, E. Tragl, A. Pichler, R. Angerer, M. Hackl, W. Hackl; Proc. TERMEC 2000,
Las Vegas
[11] F. S. LePera, Journ. Met., 1980, 38–39
[12] G. Petzow, Metallographisches, keramographisches, plastographisches Ätzen, 6th ed.,
Gebrüder Bornträger, Berlin, Stuttgart, 1994
[13] E. Wirthl, R. Angerer, K. Hauzenberger; AISE Conf. Sept. 2000, Chicago, 61–64
[14] H. K. D. H. Bhadeshia, Bainite in steels, 1st ed., The Institute of Materials, London, 1992
[15] S. Traint, A. Pichler, K. Hauzenberger, P. Stiaszny, E. Werner; Int. Conf. On TRIP-Aided
High Strength Ferrous Alloys, June 19–21, Ghent, 2202, 121–128.
87
Micromechanical Study of the Martensitic Transformation in

TRIP-Assisted Multiphase Steels
T. Van Rompaeya, F. Lanib, P. Jacquesb, B. Blanpaina, P. Wollantsa, and T. Pardoenb

a
Department of Metallurgy and Materials Engineering, Katholieke Universiteit Leuven, Leuven, Belgium
b
Division of Physico-Chemistry and Engineering of Metals, Université Catholique de Louvain-la-Neuve, Lou-
vain-la-Neuve, Belgium
1 Abstract
A finite element embedded cell model of an austenite inclusion in a ferrite grain, surrounded by
an homogenous matrix representing the behavior of a TRIP-assisted multiphase steel matrix, is
developed in order to obtain a better understanding of the mechanics of the TRIP-effect. The
transformation of a lenticular martensite plate is imposed inside the austenite grain under loa-
ding and after some amount of plastic deformation. The values for the mechanical driving force
and the elastic and plastic accommodation energies associated to the transformation strains are
calculated as a function of the orientation of the martensite lens with respect to the tensile loa-
ding direction.
The calculations show that the elastic and plastic accommodation terms are of the same order
of magnitude as the mechanical driving force. The present model also predicts which variant
orientation is energetically most favorable with respect to the externally applied stress state.
2 Introduction
TRIP-assisted (Transformation-Induced Plasticity) multiphase steels show remarkable mecha-

nical properties of formability and crashworthiness resulting from the occurrence of concurrent
modes of plastic deformation, viz. dislocation strengthening and mechanically-induced marten-
sitic transformation. The aim of the present study is to contribute to a better understanding of
the mechanics of the transformation of martensite at the scale of one isolated transforming au-
stenite grain. This is an important step towards the development of multi-scale models leading
to the prediction of the macroscopic properties of TRIP-steels. It complements earlier studies by
Fischer and coworkers [1–3].
Since martensitic transformation is associated with considerable shape and volume changes,
the local stress state varies drastically during the transformation, both inside the newly-formed
martensitic inclusion, and in the surrounding austenite and ferrite phases. This change and the
resulting change in strain energy strongly affect the transformation thermodynamics. Local
stresses in the retained austenite grain working through the transformation strain act as mechan-
ical driving forces, whereas the accommodation work in the surrounding material opposes the
transformation. It is therefore necessary to include mechanical considerations in the thermody-
namic description of the martensitic transformation. Analytic solutions (e.g. Eshelby or Mori-
Tanaka approximations) exist only for specific simplified situations. In order to perform the

ISBN: 3-527-31007-X
88
stress analysis in a more realistic geometry, it is necessary to rely on numerical approaches,

such as the finite element method.
An adequate thermodynamic criterion for the growth of a martensite region, including the ef-
fect of the evolving stress state, was developed by Fischer and Reisner [1]. According to this
criterion the martensitic transformation is governed by an energy balance between driving terms
and dragging terms:
F m F c l W b W e W p , with (1)
Fm ³V
o
:H T dV , (2)
M
:
1
We W :C:W dV ,
2 :³
(3)
tf
Wp ³ ³W :H
p
dVd t. (4)
ts :
The integral in expression \* MERGEFORMAT (3) is the mechanical driving force F m, it

represents the work available due to the action of the transformation strain HT in the local stress
field at the start of the transformation V o. F c is the chemical driving force, it corresponds to the
chemical free energy difference between the austenite and martensite. The right hand side of ex-
pression \* MERGEFORMAT (2) includes the various dragging forces. The transformation bar-
rier W O is associated to the extra energy required to replace the austenitic fcc lattice by the
trigonal lattice of the martensite. The two other dragging energy terms are the elastic and plastic
accommodation energies W e and W p, respectively. The tensor W represents the stress variation
in the material due to the transformation, C is the fourth-order compliance tensor, and H p corre-
sponds to the plastic strain rate tensor. W e represents the energy that is stored elastically in the
material due to the stress variation t, cf. expression \* MERGEFORMAT (4). W p equals the en-
ergy dissipated through plastic flow during the transformation (expression \* MERGEFOR-
MAT). The volume integration in the term F m is limited to the volume of transformating region
: M, whereas for the accommodation terms the volume integration involves the total volume W.
In order to get a more quantitative appraisal of Fischer's criterion, the mechanical driving
force F m, and both accommodation energy terms W e and W p are investigated numerically
through finite element (FE) simulations of a „approximately self-consistent“ model of a unit
cell of a TRIP-assisted steel subjected to a uniaxial tensile load. In this work the influence of the
orientation of the martensite lens with respect to the tensile direction is studied. From the nu-
merical simulations both qualitative and quantitative information can be obtained.
89
3 Unit Cell Model
3.1 Unit Cell
As a result of the specific thermomechanical processing the microstructure of TRIP-assisted

steels contains three phases: a ferrite matrix in which bainite and retained austenite are disper-
sed. The austenite is metastable, and it transforms into martensite under mechanical loading du-
ring plastic deformation. In these types of steel lenticular plates of martensite are formed in
successive bursts. In this work only the formation of the first lens of martensite is considered.
Reisner et al. [2] present an analysis of the interaction energies when a second variant trans-
forms in the neighborhood of a first variant.
The unit cell is modelled in 2D using generalized plane strain conditions. This assumes that
the in-plane displacements do not depend on the out-of-plane displacement, which is uniform,
and is not necessarily equal to zero. The unit cell is shown in Figure 1. An austenite grain, : A,
is surrounded by a matrix receiving the behavior of the ferrite-bainite aggregate,: FB. The vol-
ume fraction of austenite is 16 % which is a typical value for industrial TRIP-assisted mul-
tiphase steels and which corresponds to the alloy studied by Furnémont [4] (see further). In
order to minimize boundary effects the unit cell is embedded in a large matrix having on aver-
age the properties of TRIP-assisted steel, : T (Figure 1a).
Upon deformation the metastable austenite phase transforms partially into martensite. The
product phase is lenticular shaped, and its orientation with respect to the tensile direction is giv-
en by T. The volume fraction of martensite with respect to the initial austenite volume amounts
to 25 % (Figure 1b). Again, this value agrees with observations by Jacques et al. [5] showing
that one austenite grain (with a grain size equal to about one micron) typically involves 3 to 8
martensite lenses with different crystallographic orientations (different variants).
The mechanical properties of ferrite (with bainite) and austenite were obtained from the
work of Furnémont et al. [4,6], in which the mechanical properties of the individual phases in a
multiphase TRIP-steel were determined by combining neutron diffraction observations and dig-
ital image correlation methods. The flow curve can be represented as a Swift law:
N
V y Vˆ y 1 h H p (5)
with Vy and H p the yield stress and the equivalent plastic strain respectively. The coefficients
are Vˆ yA = 720 MPa, hA = 50, NA = 0.25, and Vˆ yFB = 531 MPa, hFB=110, NFB= 0.22 for the austeni-
te and the ferrite-bainite matrix, respectively. Both the austenite and the ferrite-bainite matrix
are modelled using J2 flow theory. Martensite is supposed to show a linear elastic behavior with
the elastic modulus equal to 200 GPa. The homogenized material : T surrounding the unit cell
is also modelled using J2 flow theory with the uniaxial stress strain response obtained from stan-
dard tensile tests. Considering that the present analysis deals only with uniaxial tensile loading
conditions, the error introduced by assuming J2 flow theory will be minor and therefore the mo-
del can be considered as "approximately self-consistent".
For the transformation strain tensor HT, typical values for TRIP-assisted steels are taken. The
volume dilatation G associated with the transformation equals 3 %, and the transformation shear
J is 20 %.
90
3.2 Finite Element Model
The model is subjected to a uniaxial tensile test, i.e. the boundary planes normal to the tensile
axis are displaced the same amount in opposite directions while leaving the transverse bounda-
ries free (Figure 1a).
Figure 1: Schematic representation of the 2D model. a) The unit cell is embedded in a large matrix of TRIP-assi-
sted steel. b) The unit cell consists of a ferrite grain, with an austenite inclusion. Upon deformation a lenticular
martensite plate is formed inside the austenite phase. c) Finite element discretization of the unit cell.
The mesh was generated and the calculations were performed using the general purpose fi-
nite element program ABAQUS. The mesh is shown in Figure 1c. Initially the regions : M and
: A are both retained austenite. The volume : M is "artificially" transformed into martensite af-
ter a certain amount of uniaxial deformation using the following procedure. For the calculations
shown in this work, the equivalent plastic deformation prior to the transformation is always
91
equal to 5 %. When this level of plastic strain is attained on average in the homogenized materi-
al the martensitic transformation is simulated using an artificial thermal expansion in the len-
ticular region : M. Anisotropic thermal expansions are autorized by ABAQUS allowing to
represent both the shear and the dilatation components of the transformation strain. Similarly,
the mechanical properties of the transforming elements : M are defined as temperature depend-
ent, thus allowing to simulate the change in properties between austenite and martensite. During
the transformation, no external deformation is applied. Note that in the present simulation no
crystallographic constraint is provided for selecting the most favorable variant. In reality, with
the hypothesis introduced in expression \* MERGEFORMAT (2), the selected crystallographic
variant will be the one closest to the orientation corresponding to the maximum mechanical
driving force.
Figure 2: Distribution of the equivalent von Mises stress in the retained austenite phase at the end of the transfor-
mation. The stress level at the start of the transformation is indicated. Lighter (darker) shades of gray indicate
compression (tension) as compared to the initial situation. For clarity, the stress distribution in the martensite lens
was excluded.
4 Results
At the end of each simulation increment, the finite element software generates output data for
every element in the mesh. This data includes the local stress tensor V o, the total strain tensor H,
and the plastic strain tensor H p at the integration point, and the volume V of the element.
Figure 2 shows the distribution of the equivalent von Mises stress in the retained austenite : A
at the end of the transformation. The stress level at the start of the transformation is indicated in
92
the graph. Lighter (darker) shades of gray show regions in compression (tension) with respect to
the initial situation. The initial stress in the phases surrounding the austenite is lower than that in
the austenite phase itself. For clarity, the stress distribution in the martensite lens was excluded.
The tensile stress in this phase varies from 2 to 4 times the initial stress in the retained austenite.
It can also be observed from Figure 2 that the shape and the volume of the martensitic region
: M has changed.
Postprocessing routines were developed in order to calculate the values of the mechanical
driving force F m, and the elastic and plastic accommodation energies W e and W p. The mechan-
ical driving force F m is a local term, and is integrated over every element in region : M. Its val-
ue is calculated at the start of the transformation. The accommodation energy terms requires to
sum over every element in the mesh :. At the end of the transformation the elastically stored
energy W e is calculated. The increment of plastic dissipated energy is calculated at the end of
each time increment during the transformation step. The total dissipated energy W p is obtained
by summing all these contributions over the entire transformation time interval. The energy
terms are given as energy densities per volume of formed martensite.
In Figure 3 the values of the mechanical driving force, and the elastic and plastic accommo-
dation energy terms are plotted as a function of the orientation T of the martensite lens, see Fig-
ure 1. The energy difference W e + W p – F m is given as a function of T in Figure 4.
120
100
-3
Energy Density, MJ.m
80
60
40
m
F
20 e
W
p
W
0
0 10 20 30 40 50 60 70 80 90
Orientation of Lens θ, °
Figure 3: Mechanical driving force F m, elastic and plastic accommodation energy terms W e and W p as a function
of the martensite orientation T
93
180
160
140
-3
Energy Density, MJ.m
120
100
80
60
40
20 e p m
W +W -F
0
0 10 20 30 40 50 60 70 80 90
Orientation of Lens θ, °
Figure 4: W e + W p– F m as a function of the martensite orientation T
5 Discussion
The stress distribution in Figure 2 shows that in the austenite phase two regions of compres-
sive stress occur, more or less parallel to the martensitic lens. This is mainly due to the volume
dilatation associated with the martensitic transformation. The compressive stress in the sur-
rounding ferrite phase can be attributed to the overall volume increase of the austenite grain
: A + : M. The region with tensile stresses in the vicinity of the martensite lens is due to the
large amount of transformation shear. In the martensite phase itself large tensile stresses are
present, varying from 2 to 4 times the initial stress in the parent austenite phase. This is because
martensite is a very hard phase modelled with linear elastic behavior.
The values of the different energy terms are of the same order of magnitude (Figure 3). This
is different than the results by Fischer and coworkers [1] showing the mechanical driving force
at least ten times larger than W e and W p. In the latter study, full austenitic steels are considered
with a much smaller amount of retained austenite transforming simultaneously in a unit cell.
Their results were obtained with numerical simulations, similar to those presented by Reisner et
al. [2]. Our results presented here agree with the general assumption that the accommodation
terms play an important role in the strain-induced transformation in TRIP-assisted steels.
While the term F m can be seen as the driving force required to initiate the transformation, the
energy difference W e + W p – F m is more related to the growth of the lens: the smaller this dif-
ference, the larger will be the lens. In other words, W e + W p – F m represents a measure for the
amount of energy needed for the growth of a martensite lens with a specified orientation. From
Figure 4 it is then clear that a lens at 30°, if nucleated, will lead to the largest possible volume of
transformation for that transformation step.
94
6 Conclusion
The embedded cell model presented in this report provides an attractive way to study the me-
chanics of strain-induced martensitic transformation and the TRIP-effect at the scale of one
transforming grain. It allows to study, qualitatively and quantitatively, the energy terms invol-
ved in the transformation, as a function of the external applied load, and of microstructural fea-
tures, such as the orientation of the martensite lens. The model also allows to get a clear view on
the stress distribution in the material as a result of the martensitic transformation. Moreover,
this model is flexible. For example, other externally applied stress states can easily be imple-
mented.
From the calculations it is observed that the mechanical driving force, the elastic accommo-
dation energy, and the plastic accommodation energy are of the same order of magnitude. It is
also shown that the amount of energy needed to grow a lenticular martensitic plate, strongly de-
pends on the orientation of the lens with respect to the main loading direction.
7 Acknowledgements
F. Lani acknowledges the support of the Fond National de la Recherche Scientifique, FNRS,
Belgium. This research was carried out under the Interuniversity Attraction Poles (IAP) Pro-
gramme, financed by the Belgian State, Federal Office for Scientific, Technical and Cultural
Affaires, under contract P8/05.
8 References
[1] F.D. Fischer, G. Reisner, Acta Mater. 1998, 46, 2095–2102

[2] G. Reisner, F.D. Fischer, Y.H. Wen, E.A. Werner, Met. Trans. A 1999, 30A, 2583–2590
[3] F. Marketz, F.D. Fischer, Comp. Mater. Sci. 1994, 3, 307–325
[4] Q. Furnémont, PhD Thesis, Université Catholique de Louvain-la-Neuve, 2003
[5] P. Jacques, Q. Furnémont, T. Pardoen, F. Delannay, Acta Mater. 2001, 49, 139–152
[6] Q. Furnémont, F. Lani, T. Pardoen, F. Delannay, P. Jacques, submitted to Acta Mater
95
Effect of Carbon and Nitrogen on the Shape Memory Effect in

FeMnSiCrNi SMAs
N. Van Caenegem1, L. Duprez1, D. Segers2, B. C. De Cooman1

1
Laboratory for Iron and Steelmaking, Ghent University, Technologiepark 903, 9052 Ghent, Belgium
2
Department of Subatomic and Radiation Physics, Ghent University, Proeftuinstraat 86, 9000 Ghent, Belgium
e-mail: Bruno.Decooman@UGent.be
1 Abstract
Fe-Mn-Si-Cr-Ni alloys are one-way shape memory alloys (SMA), which make use of the
J!Hstress-induced martensitic transformation. In the present study, the effects of alloying
addition of carbon and nitrogen on the shape memory effect (SME) is reported. C and N
strengthen the austenite. C increases the stacking fault energy (SFE) of the SMA. The influence
of N on the SFE depends on the content of N. Whereas the strengthening of the parent phase is
beneficial, the lowering of the SFE is not; it is shown that C additions of 0.2 % and low N addi-
tions of about 0.1 % to ferrous SMAs do not have a favorable effect on the SME. The characte-
risation of the martensitic transformation and the phase components was carried out using light
optical microscopy, X-ray diffraction (XRD) and transmission electron microscopy (TEM). The
SME was evaluated by means of a bending test.
2 Introduction
The shape memory alloys are materials characterized by a shape memory effect (SME) and su-
perelasticity, which ordinary metals and alloys do not have. This unique behavior is found in
Au-Cd, Ni-Ti, Cu-Zn-Al and Cu-Al-Ni which are non-ferrous shape memory alloys. It is also
known that ferrous alloys can exhibit SME such as Fe-Pd, Fe-Pt, Fe-Ni-C, Fe-Ni-Ti-Co, and Fe-
Ni-Cr, although the shape recovery is generally less complete. Ferrous SMA based on Fe-Mn
alloy system have received much attention since they may become an important type for one-
way SMAs due to their cost-effectiveness [1]. The shape memory effect in FeMnSi alloys was
first reported by Sato et al. [2]. They found that a single crystal of an Fe-30%Mn-1%Si alloy ex-
hibited a complete shape memory effect when deformed in the <414> direction. Sato et al. [3]
and Murakami et al. [4] also succeeded in developing polycrystalline FeMnSi shape memory al-
loys. Moriya et al. [5] and Otsuka et al. [1] developed FeCrNiMnSi and FeCrNiMnSiCo shape
memory alloys, which are stainless SMAs.
Fe-Mn-Si-Cr-Ni alloys are one-way shape memory alloys (SMA) with a fcc crystal structure
(Figure 1). By deforming the material, the fcc Jo hcp H stress-induced martensitic transforma-
tion takes place. During heating, the reverse H o J transformation occurs and the initial J-fcc
crystal is obtained and the initial shape is recovered.
The Jo H transformation proceeds by the motion of one a/6(112) Shockley partial disloca-
tion on every second {111} austenite plane. There are 3 possible shear directions on every
{111} plane. When we decrease the temperature under the Ms temperature, the austenite will

ISBN: 3-527-31007-X
96
Force, N
σ, stress, MPa
Ms
As
it
ten
aus e
+ si
te
nsi
n
rte
ma te
rte
ma
it
ten 3. Straining: reorientation
aus e
of ε variants
Length, mm
ε, strain, %
2. Cooling: formation of ε
4. Inverse strain: return to

1. sample fully γ fully γ sample
Temperature, °C
Figure 1: Schematic principle of the one-way shape memory effect. By cooling under the Ms thermally-induced
H martensite starts to form. By deforming, further transformation to stress-induced martensite occurs and existing
H variants reorient. With unloading, there is an elastic spring back. By heating the transformed specimen, the mar-
tensite transforms to austenite and the original shape is recovered.
transform to the hcp H martensite. This thermally-induced or spontaneous transformation to H

martensite is known to be self-accommodating. The H band usually consists of 3 laminated vari-
ants with different <112> shear directions, which minimizes the total shape strain of the trans-
formed region (Figure 2 left).
1/6 [211]
1/6 [112] 1/6 [121]
1/6 [121]
fcc
T σ
Figure 2: Schematic presentation of the fcc to hcp transformation: (left) formation of the self-accomodating
phase by the movement of three different shockley partial dislocations during the spontaneous thermally-induced
martensitic transformation, (right) formation of stress-induced martensite by the movement of only one type of
Shockley partial dislocation.
When we apply stress on a SMA, its Ms temperature increases due to the thermodynamic ef-
fect of the applied stress, and it is possible to obtain a martensite structure at higher temperature.
The stress-induced H transformation is generally accomplished by the selective motion of a sin-
gle type of Shockley partial dislocation being most favorable to the direction of the applied
stress. The stress-induced H band will therefore generally consist of a single variant (Figure 2
right). In the case of thermally-induced H band, the total shape strain is almost zero by self-ac-
commodation of 3 kinds of variants. The stress-induced H band produces a large transformation
shear which effectively contributes to the shape deformation for the SME.
97
In order to reach a complete shape memory effect, several conditions must be fulfilled. The
yield stress of the J matrix should be as high as possible to avoid deformation by slip. The stack-
ing fault energy has to be low. The deformation must result in the stress-induced H martensite
formation only, i.e. the temperature of deformation has to be between Ms and Md. The shape
strain of stress-induced martensite should be completely reversible, which means that the mar-
tensite interface should remain mobile. The required martensite morphology for ferrous SMA is
of the thin plate type. Intersections between H-variants can obstruct the forward and the reverse
transformation, as do slip, interactions, precipitations and forest dislocations.
The shape memory effect in Fe-Mn-Si alloys depends strongly on the alloy composition. Al-
loying elements can control the Ms and Md temperature, the stacking fault energy, the flow
stress and the precipitation state.
Mn and Si lowers the SFE. In addition Si increases the yield stress of the austenite parent
phase. Mn, Cr and Ni lower the Ms temperature.
In the present study, the effect of C and N on the SME is investigated. C is known to lower
the Ms temperature. C and N both strengthen the austenite, which is favorable for the SME. C
increases the SFE however, and Yakubtsov et al. have reported [6] that the effect of N on the
SFE depends on the content of nitrogen. There exists an optimum in N content, at lower N con-
tent the SFE increases, at higher N content the SFE decreases.
3 Experimental
Five different ferrous SMAs compositions were prepared for the present study. The alloys were
prepared in an induction furnace operated in air and the material was ingot-cast in a Cu-mold.
Carbon was added by means of graphite. Nitrogen was added by the means of ferromanganese
containing 7.7 % N. The compositions of the SMA are listed in table 1.
Table 1: Alloy compositions in wt%

Alloy Mn Si Cr Ni C N Fe
FeMnSiCrNi 12 6 9 5 0 - Balance
FeMnSiCrNi+0.2%C 12 6 9 5 0.2 - Balance
FeMnSiCrNi 14 6 10 7 - 0 Balance
FeMnSiCrNi+0.12%N 12 6 9 7 - 0.12 Balance
FeMnSiCrNi+0.13%N 11 6 9 7 - 0.13 Balance
Directly after solidification, the ingots were hot rolled and air cooled. After cooling they
were reheated to 1100 °C during 15 min and water quenched to room temperature, to suppress
the formation of ferrite.
The characterisation of martensitic transformation and the phase components of the alloys
were carried out by light optical microscopy (LOM), X-ray diffraction (XRD) and transmission
electron microscopy (TEM). The shape memory effect was evaluated by means of a bending
test.
98
The specimens, examined by light optical microscopy, were first mechanically polished,
then electrolytically in a solution of 20 % perchloric acid and 80 % butylcellosolve and finally
etched in a solution of 1.2% K2S2O5 and 0.5% NH4HF2 in distilled water.
Room temperature X-ray diffraction was used to determine the phase content and the volume
fraction of each phase. The measurements were done on a Siemens D5000 diffractometer using
either Mo (OMo = 0.071 nm) or Cu-radiation (OCu = 0.154 nm).
Thin foils of the SMAs were examined in a Philips EM420, operated at 120 kV.
The mechanical properties were determined on room temperature by tensile testing on an In-
stron 5569 with an initial crosshead speed of 2,7 mm/min, which was increased to 27 mm/min
at a strain of 3.4 %.
Figure 3: XRD patterns before and after a tensile deformation to fracture for (a) FeMnSiCrNi alloy, (b) FeMn-
SiCrNi+0.2 %C alloy, (c) FeMnSiCrNi alloy, (d) FeMnSiCrNi+0.12 %N alloy and (e) FeMnSiCrNi+0.13 %N
alloy. Clear hcp Hmartensite peaks are observed for deformed alloys only. In (a) and (b), the strained samples
were measured with a Cu tube, the angles are calculated to present the result in the same scale. Their intensities
are multiplied by 100.
99
4.1 X-ray Diffraction
We determined the phases present in a undeformed and a deformed SMA by means of XRD.
The 2TB range of the diffractogram containing the {100}H, {002}H and {101}H peaks is shown in
Figure 3. The undeformed alloys did not contain the H phase; some bcc D phase was however
present. The FeMnSiCrNi+0.12%N contained a relatively large amount of bcc D phase.
The strained samples all had lower diffraction peak intensities. H martensite is clearly formed
as a result of deformation. The alloys without C and N shows the highest intensity of the {101}H
peak.
4.2 Tensile Test
The solid solution strengthening of the parent J phase as a result of the C and N-alloying is clear
from the stress-strain curves shown in Figure 4. The yield strength of the alloy without C was
V0.2% = 258.9 MPa and of the C-alloyed SMA was V0.2% = 351.5MPa. The yield strength of the
alloys investigated for the influence of nitrogen are 279 MPa for the alloy without N, 343 MPa
for the one with 0.12 %N and 329 MPa for the 0.13 %N alloyed SMA.
1200
1400
+N +0.2% C
1000
1200
σ, true stress MPa
800
σ, true stress MPa
1000
FeMnSiCrNi
FeMnSiCrNi
800 600
600
400
400
FeMnSiCrNi
200
200 FeMnSiCrNi+ 0.12% N FeMnSiCrNi
FeMnSiCrNi+ 0.13% N FeMnSiCrNi+0.2% C
0 0
0 10 20 30 40 50 0 5 10 15 20 25 30
ε, true strain % ε, true strain %
Figure 4: True stress-true strain curves for (a) the FeMnSiCrNi +C SMA (b) the FeMnSiCrNi+N alloys.
In the case of the 0.12 %N-containing alloy the high yield strength is probably due to the sec-
ond phase strengthening by the bcc D phase.
4.3 Transmission Electron Microscopy
Figure 5 is a TEM bright field (BF) with the corresponding selected area diffraction pattern
(SADP). In the BF intersecting thin plate H martensite can be observed. At the intersection of
two variants of the H martensite, D’ martensite is formed. The formation of D’ martensite is very
likely one of the reasons for incomplete recovery in ferrous SMAs. It impedes the reverse moti-
on of the Shockley partial dislocations. This is the reason why the formation of several H mar-
tensite variants must be avoided. The deformed material should therefore not contain different H
100
{111}γ
{011}α’ ε/γ γ
{0001}ε
α
ε/γ
{101}α’
{200}γ
0.2 µm
{220}γ
Figure 5: (left) TEM bright field of FeMnCrNi+6%Si, hot rolled, air cooled, annealed at 1100 °C-15min and
water quenched at room temperature, bending deformation: 3.4 %, [011]J zone axis. (right) SADP shows diffrac-
tion spots of the [-111]D zone axis, the [011]J and two variants of the [2-1-10]H
martensite variants within one single grain. A single variant behavior must be promoted by trai-
ning the stress-induced martensite. Training is a processing method for SMAs which involves
successive cycles of deformation and heating.
The presence of bcc D, fcc J and hcp H phases could be ascertained by means of SADP. In the
diffraction pattern of D’ martensite, J and H martensite, the following orientation relationships
were present: {0001}H // {011}D’ (Bogers-Burgers) and {011}D // {111}J(Kurjumov-Sachs).
4.4 Bending Test
The bending test, as illustrated in Figure 6: (a) Bending test: a straight sample with thickness t is
bend 180° to a radius of 25 mm, unloaded and annealed at 420 °C. Note the considerable elastic
spring back. (b) result of the bending test, influence of the annealing temperature and the
amount of prestrain on the shape recovery, (c) influence of carbon on the shape recovery, (d) in-
fluence of nitrogen on the shape recovery.(a), was used to evaluate the shape recovery. Samples
of 150 × 15 × 1.7 mm and 100 × 15 × 1.7 mm were bent to a 50 mm or 10 mm diameter to obtain
a strain of 3.4 % or 17 %, respectively. After deformation the samples were heated in an oil bath
(T < 200 °C) or a salt bath (200 °C < T < 450 °C). The samples recovered their initial shape
partly and the recovery strain Hwas calculated by means of the following equation:
Tm
H Hp (1)
180 T e
where Hp is the prestrain, Te is the elastic spring-back angle after deformation and Tm is the ben-
ding angle after the strain recovery anneal.
To see the effect of the prestrain on the shape recovery H/ Hp was plotted as a function of the
annealing temperature. This is shown in Figure 6b. The recovery strain increased with increas-
ing annealing temperature. Annealing at 50 °C did not result in shape recovery. At 80 °C, the re-
covery strain was limited. At 270 °C the recovery strain was very large and reached its
101
maximum. It is clear from the data that a higher prestrain reduces the SME. This is very likely
due to the presence of dislocations in the parent phase which prevent the easy motion of the
transformation interface. The addition of C (Figure 6c) and N (Figure 6d) to the FeMnSiCrNi
alloy clearly lowers the recovery strain.
Bending 100% recovery

1.0
FeMnSiCrNi+0.2% C 3.4% strain
0.8 FeMnSiCrNi+0.2% C 17% strain
Heating 0.6
θm/(180-θe)
θe θm
0.4
Unloading
0.2
d 0.0
Original position 0 50 100 150 200 250 300 350 400 450 500
t Temperature,°C
(a) (b)
1.0
100% recovery 100% recovery
1.0
FeMnSiCrNi+0.2% C 3.4% strain FeMnSiCrNi
0.8 FeMnSiCrNi 3.4% strain 0.8 FeMnSiCrNi+0.12% N
FeMnSiCrNi+0.13% N
0.6 0.6
θm/(180-θe)
θm/(180-θe)
50% recovery
0.4 0.4
0.2 0.2
0.0 0.0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Temperature,°C Temperature,°C
(c) (d)
Figure 6: (a) Bending test: a straight sample with thickness t is bend 180° to a radius of 25 mm, unloaded and
annealed at 420 °C. Note the considerable elastic spring back. (b) result of the bending test, influence of the
annealing temperature and the amount of prestrain on the shape recovery, (c) influence of carbon on the shape
recovery, (d) influence of nitrogen on the shape recovery.
5 Conclusions
The addition of low C or low N content, i.e. 0.2 % C and 0.1 % N, does not improve the SME of
FeMnSiCrNi SMAs. Although C and N strengthen the parent austenite phase, C lowers the
stacking fault energy which is detrimental to the shape memory effect.
The intersection of two different H variants was observed in TEM to lead to the formation of
D ferrite in ferrous SMAs. This fact is very likely responsible for the incomplete recovery of the
FeMnSiCrNi shape memory alloys.
102
6 References
[1] H. Otsuka, H. Yamada, T. Maruyama, H. Tanahashi, S. Matsuda and M. Murakami, Trans

ISIJ, 1990, 30, 674
[2] A. Sato, E. Chishima, K. Soma and T. Mori, Acta Metallurgica, 1982, 30, 1177
[3] A. Sato, E. Chishima, K. Soma and T. Mori, Acta Metallurgica, 1986, 34, 287
[4] M. Murakami, H. Otsuka, H.G. Suzuki and S. Matsuda, Proceeding of the International
Conference on Martensitic transformations (ICOMAT-86), Japan Inst. Met., 1986, 985
[5] Y. Moriya, T. Sampei and I. Kozasu, Annual Meeting at Yokohama, Conf. Abstr, JIM,
1989, 222
[6] I. Yakubtsov, A. Ariapour, D.D. Perovic, Materials Science forum, 1999, Vols 318–320,
p. 121–130
103
Rapid Full Annealing under High Magnetic Field
Y. Zhang1, 2, C. He1, X. Zhao1, L. Zuo1, C. Esling2, J. He1, G. Nishijima3, T. Zhang3 and

K. Watanabe3
1
School of Materials and Metallurgy, Northeastern University, Shenyang, China
2
LETAM, CNRS-UMR 7078, Université de Metz, Metz, France
3
Institute for Materials Research, Tohoku University, Sendai, Japan
1 Abstract
The kinetic effect of a magnetic field on high temperature phase transformation during cooling
in 42CrMo steel is investigated. Results show that the magnetic field can considerably increase
the driving force for the transformation from austenite to ferrite by enhancing the Gibbs free en-
ergy difference between the two phases. Even at the high cooling rate of 46 qC/min, the decom-
position of austenite in a 14 Tesla high magnetic field is still ferritic and pearlitic, instead of
being bainitic as is usually observed. A microstructure of fine and randomly distributed ferrite
grains and pearlite colonies is obtained. Therefore, this rapid magnetic annealing process can
also effectively prevent the formation of the banded structure that occurs commonly during the
conventional full annealing due to previous hot working. Image analyses and hardness tests
show that the amount of ferrite obtained in this way is almost the same as that obtained by the
conventional annealing and the hardness is still within the optimum range for the subsequent
machining. Magnetic annealing has thus the merits of improving the microstructure by avoiding
banded structure and also optimizing the process by greatly shortening the cooling time. There-
fore, it is a promising approach for the innovation of conventional processes. Probing into this
issue is of both theoretical significance and technical interest.
2 Introduction
The study of the influence of a magnetic field on heat treatment and the phase transformations
of ferro-alloys was initiated in the middle of the past century. [1, 2] However, most of the litera-
tures [1–6] have focused on martensitic phase transformation in a limited range of materials
with low martensitic start temperatures from the physical, metallographical and crystallographi-
cal point of view. Quite recently, with the development of superconducting materials and the
progress of cryocooling technique, thermal treatment at higher temperatures in a magnetic field
higher than 10 Teslas has been possible and research [7, 8] on the effect of high magnetic field
on high-temperature phase transformation has been developed. As the research is still in its in-
itial stage, it is still basically of the fundamental sort. However, just like most scientific resear-
ches, their ultimate goal is to develop practical applications in production and achieve progress
in technology. Attempts in applying high magnetic field to practical heat treatment processes
are necessary and have technical importance.
In the present work, medium carbon-content alloy steel was annealed conventionally without
a magnetic field, and annealed rapidly with a 14 T magnetic field. The aim was to investigate
the kinetic influence of the magnetic field on the austenite-ferrite transformation from a techno-

ISBN: 3-527-31007-X
104
logical point of view. Moreover, the possible improvement to the conventional annealing proc-
ess is also analyzed.
3 Experimental
The material used in this study was 42CrMo, a Chinese structural steel with chemical composi-
tion (wt%) of 0.38–0.45 %C, 0.20–0.40 %Si, 0.50–0.80 %Mn, 0.90–1.20 %Cr,
0.15–0.25 %Mo, d0.04 %P, d0.04 %S and d0.30 %Cu. Specimens of dimensions
20 mm u10 mm u2 mm were cut from a hot-rolled rod with their longitudinal direction paral-
lel to the rolling direction. Magnetic field heat treatment was carried out in the furnace set in a
15-T cryocooled superconducting magnet of 52 mm in bore size at the High Magnetic Field La-
boratory for Superconducting Materials, Institute for Materials Research, Tohoku University
[9]. Specimens were kept in the central – zero magnetic force – region, with their hot-rolling di-
rection parallel to the axis of the magnet. Magnetic field annealing was performed at 880 qC for
austenization for 33 min. Its cooling was conducted at the rate of 46 qC/min within a 14-T ma-
gnetic field. Conventional full annealing was carried out in an ordinary chamber furnace. Speci-
mens were heated up to 860 qC and isothermal held for 30 min and cooled naturally inside the
furnace at a cooling rate of about 1 qC/min. The transformed microstructures were etched and
analyzed with an image analyzer. Their original austenite grain structure was also obtained and
observed. Vicker’s hardness was tested with a 5 kg load.
4 Results
The microstructure obtained by conventional annealing is shown in Fig. 1. It can be seen that it
consists of typical ferrite (bright areas) and pearlite (dark areas). Ferrite grains and pearlite colo-
nies aligned alternately along the direction of the previous rolling (horizontal in the picture).
The original austenite grain structure was also obtained and shown in Fig. 2. It is found that alo-
ng the same rolling direction (horizontal in the picture) fine- and coarse-grained zones are alter-
nately distributed; this can be attributed to the inhomogeneous deformation occurring in the hot-
rolling stage. The microstructure of the specimens rapidly annealed in the 14 T high magnetic
field is shown in Fig 3. It still consists of a typical mixture of ferrite and pearlite, but they distri-
bute randomly with obviously smaller average sizes. Image analyses show that the area percen-
tage of ferrite is about 24.4 % in conventionally fully annealed specimens and about 23.1 % in
magnetically rapidly annealed specimens. The two values are very close to each other. Hardness
values – expressed as Brinell Hardness – of the specimens rapidly annealed are in the range of
HB192~210, which is just within the optimum hardness range of HB160~230 for machining.
5 Discussion
The microstructure in Fig. 1 obtained through conventional full annealing displays the typical
banded structure. Its formation can obviously be associated with the previous hot-working sta-
ge. As seen in Fig. 2, the original austenite grain sizes are not homogeneous. Fine and coarse
105
50µm
Figure 1: Banded microstructure of 42CrMo obtained by conventional annealing. Specimens were heated up to
860 qC and held for 30 minutes and cooled inside the furnace at a cooling rate of about 1 qC/min (the rolling
direction is horizontal in the picture)
grains are distributed alternately along the same rolling direction, which indicates that the pre-
vious deformation was not homogeneous. The fine-grained zones used to be the heavily defor-
med areas, while the coarse-grained zones the weak deformation areas. Evidently, this
inhomogeneous austenite grain structure will exert considerable influence on the subsequent
phase transformation during annealing. With fully austenitized 42CrMo, austenite transforms
first to ferrite between Ar3 and Ar1 temperatures and then to pearlite below Ar1 during the follo-
wing slow cooling. As furnace cooling in the conventional full annealing is very slow - about 1
qC/min, the nucleation of proeutectoid ferrite occurs at higher temperatures with relatively lo-
wer undercooling degrees, and thus with a smaller driving force. In such a case, nucleation of
ferrite mainly happens on austenite grain boundaries. Considering that grain boundaries are dis-
tributed densely within the fine-grained zones, the amount of nucleation in those areas is much
higher than that in the coarse-grained ones. With the formation of ferrite, the excessive carbon
atoms diffuse out. Then the coarse-grained zones receive more carbon atoms due to the high nu-
50µm
Figure 2: Original austenite grain structure of 42CrMo after rolling (the rolling direction is horizontal in the
picture)
106
cleation rate in the fine-grained zones and become rich in carbon. Consequently, the probability
for ferrite nucleation in these areas tends to decrease. Therefore the final banded microstructure
of alternately distributed ferrite and pearlite was obtained.
Structural steels with medium carbon content are commonly used to manufacture key struc-
ture components under heavy loads such as gearboxes, crankshafts, connecting rods, fastening
pieces and the like. Most of them are forged for blanks. Therefore, banded structures occur eas-
ily in subsequent cooling or full annealing whose aim is to modify the microstructure and im-
prove machinability. This kind of microstructure is detrimental to materials as it creates
anisotropy in performance; it should therefore be avoided. To eliminate it, normalizing is usual-
ly applied. But it always yields a bainitic or martensitic microstructure that is high in hardness
and unfavorable to machining, high temperature tempering is indispensable. As the method is
not economical in furnace batches and is complicate in operation, it is not satisfactory.
However, rapid annealing in high magnetic field could offer a potential alternative. As seen
in Fig. 3, when rapidly cooled at 46 qC/min in 14 T after having been fully austenitized at
880 qC, the microstructure is still composed of ferrite and pearlite, but they distribute randomly
and the grain sizes are smaller than those obtained by conventional annealing. Usually, at such a
high cooling rate, the specimens of 42CrMo are composed mostly of bainite [10], but when a
14 T magnetic field is applied, as was done in this work, the microstructure still remains as fer-
ritic and pearlitic. Obviously, the magnetic field has a strong effect to accelerate the transforma-
tion from austenite to ferrite.
50µm
Figure 3: Microstructure of 42CrMo obtained by rapid annealing in a 14-Tesla magnetic field. Specimens were
heated up to 880 °C and held for 33 minutes and cooled at a rate of 46 °C/min (the rolling direction is horizontal
in the picture)
According to the Johnson-Mehl equation, the amount of a new phase produced during solid
state phase transformation f is related to its nucleation rate N , growth rate v and transformation
time t:
A 3 4
f 1 exp( Nv t ) (1)
4
107
By substituting the nucleation and the growth rate equations [11] on the basis of the features of
proeutectoid ferritic transformation, the kinetic equation of proeutectoid ferritic transformation
from austenite can be obtained as follows:
1 Q V3 xJ x
ln t A ln(ln )B C 2
E ln J (2)
1 f RT 'GV x xD
where A, B, C, R and E are constants; Q and T are the activation energy for diffusion and the ab-
solute temperature; V and 'GV are the interfacial energy and the driving force or Gibbs volume
free energy difference between ferrite and austenite; xJ and xD are the solubility values of auste-
nite and ferrite at temperature T, and x the carbon content of the material. As magnetic proper-
ties, such as magnetizations or susceptibilities, of the product ferrite and the parent austenite are
different, the applied magnetic field inevitably has an influence on the decomposition of auste-
nite by modifying the stability of both phases. As both austenite and ferrite can be magnetized
to some extents in a high magnetic field, their Gibbs free energies will then be lowered. The ex-
tra energy difference induced by the magnetic field between these two phases can be expressed
as follows:
'G M WD WJ (3)
where WD and WJ are the energies lowered by magnetic field for ferrite and austenite, respec-
tively. In the case of a transformation in a magnetic field, the third item in Eq. (2) should be
complemented by a magnetic term as:
V3
C (4)
('GV 'G M )2
Since the degree of magnetization of ferrite is higher than that of austenite, the decrease in
the amount of energy is larger for ferrite than for austenite. Thus, 'GM is negative i.e. has the
same sign as 'GV. As a result, the transformation time from austenite to ferrite in the magnetic
field is sufficiently reduced or, in other words, the transformation is greatly accelerated. There-
fore, even at a high cooling rate of 46 qC/min, the staying time at ferritic transformation temper-
ature range is still sufficient for austenite to transform into ferrite and thus the final
microstructure of ferrite and pearlite is obtained. Due to the introduction of 'GM, the overall
driving force for the transformation into ferrite under the magnetic field is greatly increased and
thus the nucleation barrier considerably reduced. As a consequence, ferrite is nucleated at a high
nucleation rate. More sites inside grains other than those along the austenite grain boundaries
are available for nucleation. Therefore, the final microstructure of randomly distributed ferrite
grains and pearlite colonies with smaller sizes is obtained, as shown in Fig. 3.
In addition, image analysis shows that the area percentage of ferrite obtained by rapid an-
nealing in the 14 T magnetic field is 23.1 % and very close to the figure 24.4 % obtained by
conventional annealing. Besides, the hardness of the specimens rapidly annealed in the 14 T
field ranges from HB192 to 210, which lies just within the optimum hardness range of HB160-
230 for machining. So the microstructure and hardness of specimens rapidly annealed in the
magnetic field fully meet the technical requirements.
108
It can thus be seen that rapid annealing in a high magnetic field can effectively avoid the for-
mation of the banded microstructure occurring after hot working. Moreover, it improves the
microstructure by refining grains and providing a more uniform distribution. Finally, it also
greatly enhances the productivity by greatly reducing the cooling time required by conventional
annealing. For the specimen size used in the present work, the cooling time from 860 qC down
to the furnace discharging temperature of 550 qC is 300 minutes (5 hours) for conventional an-
nealing, whereas it is only 7.2 minutes for rapid magnetic field annealing, i.e. about 45 times
shorter. The above results lead to a promising approach for a reforming of the conventional
processing methods.
6 Conclusions
1. The banded structure in 42CrMo in this work obtained during conventional full annealing is
related to the inhomogeneous deformation occurring in the previous hot-rolling process.
During the slow cooling of the conventional annealing, the nucleation of ferrite starts at
higher temperature with lower undercooling degree. Consequently, the austenite grain
boundaries are the main sites for ferrite nucleation. The fine-grained zones resulted from
the previous heavy deformation offer more nucleation sites than the coarse-grained zones
and finally lead to the formation of the banded microstructure.
2. The magnetic field can obviously increase the driving force for the transformation from
austenite to ferrite by increasing the Gibbs free energy difference between the two phases
and therefore accelerate the transformation. Even at the high cooling rate of 46 qC/min, the
decomposition of austenite in the high magnetic field of 14 T is still ferritic and pearlitic,
instead of bainitic as usual, with a fine and randomly distributed grain structure. The
amounts of ferrite obtained without and with the magnetic field are almost the same and the
hardness of the specimens rapidly annealed in the 14 T field is also favorable to the subse-
quent machining.
3. With the combined merits of an improved microstructure and enhanced process by shor-
tening the cooling time for annealing, thus leaving out additional heat treatments to elimi-
nate the banded structure in one, the rapid annealing in a high magnetic field is a promising
approach for an updating of conventional processes. Gaining new insights into the kinetic
effect of a high magnetic field on phase transformation is of both theoretical and technical
significances.
7 Acknowledgments
This study was financially supported by the key project of National Natural Science Foundation
of China (Grant No. 50234020), the National High Technology Research and Development
Program of China (Grant No. 2002AA336010), the TRAPOYT in Higher Education Institutions
of MOE, P.R.C., the Chinese-French PRA Projects (PRA MX00-03). One of the authors (Y.Z.)
expresses her thanks to Mr. M. Peiffer, Préfecture de Metz, for having made possible her parti-
cipation to EUROMAT 2003 at Lausanne.
109
8 References
[1] M. A. Krivoglaz, V. D. Sadovskiy, Fiz. Metal. Metall. 1964, 18, 23–27

[2] P. A. Malinen, V. D. Sadovskiy, Fiz. Metal. Metall. 1966, 21, 787–788
[3] T. Kakeshita, K. Shimizu, S. Funada, M. Date, Acta Metall. 1985, 33, 1381–1389
[4] T. Kakeshita, K. Kuroiwa, K. Shimizu, T. Ikeda, A. Yamagishi, M. Date, Mat. Trans. JIM
1993; 34; 415–422
[5] T. Kakeshita, T. Yamamoto, K. Shimizu, K. Sugiyama, S. Endo, Mat. Trans. JIM 995, 36,
1018–1022
[6] T. Kakeshita, T. Saburi, K. Kindo, S. Endo. Jpn. J. Appl. Phys. 1997, 36, 7083–7049
[7] K. I. Maruta, M. Shimotomai, Mat. Trans. JIM 2000, 41, 902–906
[8] H. Ohtsuka, Y. Xu, H. Wada, Mat. Trans. JIM 2000, 41, 907–910
[9] K. Watanabe, S. Awaji, M. Motokawa, Y. Mikami, Jpn. J. Appl. Phys. 1998, 37,
L1148–1150
[10] Y. D. Zhang, Ph. D dissertation, 2003, p. 45
[11] H. B. Chang, Z. G. Li, T. Y. Hsu (Z. Y. Xu), X. Y. Ruan, Acta Metal. Sinic (English Let-
ters) 1998, 11, 207–214
39
III Crystallization

ISBN: 3-527-31007-X
113
Crystallization Kinetics and Phase Selection in Amorphous

Al-Alloys
N. Boucharat, H. Rösner, G. Wilde

Forschungszentrum Karlsruhe, Institut für Nanotechnologie, Karlsruhe, Germany
1 Introduction
The recently increased interest in the development of new metastable materials have brought
metallic glasses into a focus of attention due to their attractive role as effective precursors for
nano-scale microstructures. Using rapid solidification or other non-equilibrium processes, glas-
ses can be produced that partially devitrify, yielding typical nanocrystal dispersions with num-
ber densities of the order of 1017–1018 m–3. However, the recent discovery of new amorphous
alloy compositions such as Al-TM (transition metal)-RE (rare earth element) alloys open a new
opportunity in terms of both fundamental understanding and advanced applications, e.g. due to
the exceptional mechanical properties [1]. These glass forming alloys present an unusually high
nanocrystal number density (1021–1023 m–3) [2] that offers improved performance in structural
applications and exceptional properties with the combination of low specific weight and high
tensile strength [1]. The high solute content of the nanostructured Al alloys compared to crystal-
line Al alloys appears to be one important factor, but the interaction between the Al nanocrystal
dispersion and the shear bands that develop upon deformation of the amorphous matrix is also
likely to be important [3]. The partial devitrification upon controlled thermal treatments of the
glasses results in a microstructure consisting of a fine dispersion of Al crystallites in a residual
Al-based amorphous matrix. While the origin of the Al-nanocrystals is still uncertain, experi-
mental studies together with kinetics analyses indicated that the nanocrystals seem to originate
from quenched-in nuclei that can grow during subsequent heat treatment processes at tempera-
tures near the glass transition [4]. However, the large disparities in the component diffusivities
that are mainly due to the large atomic size difference is the key parameter that accounts for the
rather high stability of the nanoscaled structure to significant changes in size scale [5]. Indeed,
the overlap of the solute diffusion fields of neighboring crystallites leads to a considerable slo-
wing-down of the nanocrystal growth and moreover involves the formation of pronounced con-
centration gradients near the nanocrystal/amorphous interfaces that can markly affect the
nucleation conditions for different phases [6, 7]. The scope of this investigation is to survey the
impact of different thermal treatments at temperatures below the glass transition on the nano-
crystallization and the crystallization sequence of the Al88Y7Fe5 ternary alloy in the presence of
concentration gradients. Moreover, one key of the strategy for a wider understanding of the ori-
gin of the nanocrystals that develop during primary crystallization is to promote the nucleation
density by incorporating immiscible elements e.g. Pb or In into the melt prior to quenching. The
results show that the processing pathways act strongly on the microstructure and the phase sel-
ection and that the control of the concentration gradient represents a new opportunity to achieve
specific structures.

ISBN: 3-527-31007-X
114
2 Experimental Procedure
Ingots of Al88Y7Fe5, Al87Y7Fe5Pb1, Al87Y7Fe5In1 were synthesized by arc-melting from high pu-
rity Al, Pb, In (99.999 %), Y (99.99 %) and Fe (99.995 %). Glassy ribbons with an approximate
thickness of 40 µm were produced using a single-roller melt-spinning technique. For glasses
containing Pb and In, the melt-spinning experiments were performed under inert gas atmosphe-
re to avoid excessive loss of Pb and In during quenching. Energy dispersive X-ray analysis
(EDX) in a scanning electron microscope (SEM, Leo 1530) confirmed that the compositions of
the samples after melt-spinning correspond to the nominal starting composition. X-ray diffrac-
tion (Philips X’Pert) using Cu-KD radiation as well as careful microstructural analyses perfor-
med by transmission electron microscopy (TEM, Philips CM 30 ST and Tecnai, F20 STEM)
coupled with selected area electron diffraction (SAED) have confirmed the homogeneous and
amorphous structure of the as-spun Al88Y7Fe5 ribbons. Thin foils of ribbons for TEM were pre-
pared by twin-jet electro-polishing (1/3 nitric acid and 2/3 methanol) to avoid any sample-hea-
ting effects. Continuous heating and isothermal annealing experiments were monitored by
differential scanning calorimetry (DSC, Perkin Elmer Pyris 1 and TA-instrument Q100). With
this method, signals of about 1 µW can reproducibly be analyzed.
3 Al-Nanocrystal Development
The thermal response of an as -spun Al88Y7Fe5 sample displays a two-step crystallization pro-
cess during continuous heating in DSC and no endothermic glass transition signal (Fig. 1a). As
verified by XRD and TEM analyses, the primary crystallization reaction that starts at about
Tx = 276 °C yields a high number density of almost pure Al-nanocrystals. Moreover, upon hea-
ting the glass to temperatures well below the primary crystallization onset temperature, Al-na-
nocrystals develop slowly, which allows a more accurate experimental monitoring, offering a
new advantage in the understanding of the kinetic factors involved in the crystallization process.
The underlying mechanism that accounts for the extremely high density is closely related to the
kinetic competition between crystallization and vitrification in the melt that occurs during rapid
quenching. However, the nature of nucleation catalysts that are active during rapid quenching is
still a matter of discussion. Several proposals involving solute effects, phase separation or quen-
ched-in nuclei [8] and heterogeneous nucleation have been advanced to account for the high na-
nocrystal number density, but recent crystallization measurements and kinetics analyses [4]
provide new evidence for the role of the as-quenched structure on nanocrystal production. The
analyses strongly indicate that the initial quenching process is not sufficiently rapid to comple-
tely prevent the formation of nuclei. Under growth-control conditions, upon heating a sample
containing quenched-in nuclei, rapid crystallization ensues at the glass transition which then
roughly coincides with the crystallization temperature. As illustrated in Fig. 1b by TEM images,
samples held at 245 °C for 10 min show at early stages of the reaction a low number density of
small spherical nanocrystals. The increase of the holding time from 10 to 30 min involves a
rather fast nanocrystal growth (Fig. 1c) with a considerable increase of the nanocrystal number
density and the crystal size that is coupled with the development of a dendritic shape. At a later
stage of the crystallization there is, however, a rather high stability of the nanoscaled structure
[9]. These results are confirmed by DSC measurements that exhibit a decrease of the exother-
mic signals with increasing annealing time due to the drastic slowing-down of the nanocrystal
115
growth at rather early stages of the crystallization (Fig 1a). The combination of two effects, the
rejection of the solute during nanocrystal growth [10] and the large difference of the component
diffusivities [5], cause the formation of pronounced concentration gradients at the interface of
the growing nanocrystal that account for the dendritic growth morphologies. This result has also
been shown independently by direct experimental analyses using atom-probe techniques [7].
Moreover, at rather early stages of the growth, the overlap of the diffusion fields of neighboring
particles due to the sluggish solute diffusivities limits the crystal growth [5], which accounts for
the stabilization of the nanoscale microstructure.
a b c
Figure 1: a) Continuous heating DSC traces at 20 °C min–1 of as-spun Al88Y7Fe5 and samples that have been
annealed at 245 °C for 30 min, 60 min and more than 150 min, respectively. The thermal responses show a pri-
mary crystallisation reaction corresponding to the formation of a high number density of Al-nanocrystals. The
DSC signals of annealed samples present an additional exothermic peak related to the formation of a transient
phase. Bright-field TEM analyses of Al88Y7Fe5 samples after annealing at 245 °C for b) 10 min and c) 30 min,
indicating that Al-nanocrystals develops during isothermal treatments. Corresponding dark-field TEM images
show the shape evolution of the nanocrystals from a spherical to a dendritic morphology.
4 Nanocrystallization Catalysis
Based upon the experimental results discussed above, it is apparent that the formation of the ex-
treme nanocrystal number densities in Al-based glasses is inherently coupled to the presence of
a high initial nucleation site density [4]. The key strategy in enhancing the nanocrystal number
density, and thus to improve both property performance and microstructure stability, is to pro-
mote the nucleation density of nanocrystals while minimizing the change of the amorphous ma-
trix phase. As reported [11], an increase of the number density of Al-nanocrystals in Al88Y4Ni8
alloys has been observed by substitution of 1 at% Cu for Al. The heterogeneous catalysis of Cu
concerning the development of Fe-nanocrystals has also been confirmed in iron-based amorp-
hous alloys [12]. According to these experimental observations, one opportunity to modify the
density of nuclei in Al88Y7Fe5 undercooled melts that can lead to large variations of the resul-
ting microstructure is given by small variations of the processing pathway such as the precipita-
tion of an immiscible liquid component. For this reason, 1 at% Pb and 1 at% In were substituted
for Al in an Al88Y7Fe5 alloy during melt spinning. One of the main aspect in the selection of
both elements resides in the fact that Pb and In are not reported to form stoichiometric com-
pounds with the other constituent elements and are both immiscible in the majority component,
116
i.e. in Al. Thus the particles can then serve as internal nucleation sites to increase the nucleation
site density of Al-nanocrystals. On the other hand, the phase-separated regions introduce a large
amount of additional internal interface area that could favor the formation of quenched-in nu-
clei. However, the large difference in the melting temperatures of both elements could have an
impact on the nanocrystallization kinetics since Pb melts at temperature above the onset tempe-
rature of the primary crystallization in Pb-free samples and In melts at lower temperature. As
shown on the TEM images in Fig. 2a, the immiscibility of Pb in liquid Al results in a liquid pha-
se separation and consequently leads to a microstructure consisting of an amorphous matrix and
spherical regions of nanocrystalline Pb after quenching. This has been confirmed by XRD ana-
lyses that indicate typical Pb reflection peaks superimposed on the scattering maximum corre-
sponding to the amorphous phase. The situation is rather different in the In-containing sample
since XRD and TEM/SAED analyses on an as-spun sample do not show any evidence of cry-
stalline phases. However, the energy dispersive X-ray analysis (EDX) in a scanning electron
microscope confirmed that the composition of both samples agrees well with the nominal star-
ting composition Al87Y7Fe5Pb1 and Al87Y7Fe5In1. Moreover, XRD measurements of partially
crystallized Al87Y7Fe5In1 samples show small broad peaks in addition to the Al peaks corre-
sponding to the [111] diffraction of pure In. Finally, small endothermic peaks corresponding to
the melting of Pb and In, respectively, are detected by thermal measurements (Fig. 3a, b). It is
interesting to note that the small sizes of Pb and In particles and the presence of the crystallized
AlYFe matrix involves for both elements a reduction of the melting points compared to those of
the bulk elements. Identical melting behavior of Pb and In particles has been also reported in as-
spun Al-Pb and Al-In alloys as a consequence of the inoculant particle sizes and morphologies
[13, 14].
a b
Figure 2: a) Bright-field TEM image of an as-spun Al87Y7Fe5Pb1 sample showing a microstructure consisting of
an amorphous matrix and spherical regions of nanocrystalline Pb. In insert, the corresponding SAED pattern. b)
Bright-field TEM image of Al87Y7Fe5Pb1 sample that has been held at 230 °C for 10 min showing a microstructure
consisting of an amorphous matrix, nanocrystal-line Pb particles and spherical nanocrystal particles. c) HRTEM
image indicating fringes of an Al-nanocrystal that developed into the amorphous matrix independently of the Pb
particles.
Upon devitrification (Fig. 3a, b), the crystallization reactions of as-spun Al87Y7Fe5Pb1 and
Al87Y7Fe5In1 samples follow the same sequence as observed in the as-spun Al88Y7Fe5 sample
(Fig. 1a) that has also been confirmed by XRD measurements. The primary crystallization peak
has an onset temperature that is strongly shifted to lower temperature and does not display a rel-
atively sharp onset but a gradual heat flow at the beginning of the reaction as well as a wider
117
Al87Y7Fe5In1
Al87Y7Fe5Pb1
a b
Figure 3: Continuous heating DSC traces at 20 °C min–1 of a) as-spun Al87Y7Fe5Pb1 and b) as-spun Al87Y7Fe5In1.
The thermal responses show in both cases a marked shift of the primary crystallisation peak to lower tempera-
tures. The endothermic peaks corresponding to the melting of Pb and In are depicted and show also a shift to
lower temperatures.
temperature range compared to the thermal response of the as-spun Al88Y7Fe5 sample (Fig 1a).
These significant changes indicate that the incorporation of Pb and In particles into the ternary
alloy alters the nucleation and growth conditions of the Al-nanocrystals. These results are con-
sistent with the TEM analyses on Al87Y7Fe5Pb1 samples that have been annealed at 230 °C for
10 min. The micrographs in Fig. 2b, c show nanocrystalline Pb particles dispersed in the amor-
phous matrix that are surrounded by a high number of small spherical regions that have not been
observed in the as-spun samples (Fig. 2a ) and that correspond to Al-nanocrystals. It is clearly
evident that a higher number density of smaller spherical Al-nanocrystals developed into the
amorphous matrix compared to those that formed in the Pb-free sample after similar thermal
treatment (Fig. 1b). Nevertheless, the strong increase of the number density seems to limit the
further nanocrystal growth since sizes of few nanometers are obtained. In addition, in recent in-
vestigations by careful HRTEM analyses [15, 2], both Al-nanocrystals formed on Pb particles
and Al-nanocrystals growing independently into the matrix have been observed demonstrating
that Pb nanodispersions act as heterogeneous nucleation sites increasing the nucleation rate by
lowering the energetic barrier. The observations of Al-nanocrystals that are not attached to Pb
particles suggest however that a single mechanism may not apply to the formation of Al-nanoc-
rystals. In In-containing samples the situation is rather different since liquid In particles can not
serve as heterogeneous nucleants. However, the considerable shift of the DSC signal to a lower
temperature confirms that in addition to conventional heterogeneous catalysis, other mecha-
nisms have to be considered for a full description of the nanocrystallization process. In particu-
lar, the inoculants introduce a large amount of additional internal interface area, which acts to
decrease the heat transfer during the initial quenching. This effect leads to a slowing-down of
the local cooling rate so that the viscosity around the inoculant particles is maintained suffi-
ciently low to promote the retaining of additional pre-existing nuclei. By subsequent thermal
treatments, additional Al-nanocrystals can then develop yielding a considerable increase of the
nanocrystal number density. However, the more pronounced decrease of the onset temperature
of the primary crystallization reaction in In-containing samples compared to the Pb-containing
samples remains surprising, suggesting that additional kinetic factors such as the size of the par-
ticles could impact the nanocrystallization process. Thus, the present results demonstrate that
the incorporation of Pb particles in the amorphous matrix is effective in promoting the crystalli-
zation of nano-sized Al, yielding a significant increase in the number density. The significant
118
changes in the DSC signals (Fig. 3a, b) coupled to the different melting behavior of Pb and In
indicate that the inoculant particles have a large impact on the kinetic balance between crystalli-
zation and vitrification during rapid quenching that controls the generation of the nanocrystal
dispersion.
5 Phase Sequence Evolution During Devitrification
Besides results obtained by atom-probe measurements [7] and by kinetics modeling [5], the nu-
cleation conditions for different phases can be markedly affected by the presence of a concen-
tration gradient as well as by its steepness and curvature. Since the concentration gradient is
related to the low diffusivity of the components, thermal treatments have been performed in or-
der to investigate the impact of a changing concentration gradient on the nucleation selection. In
agreement with the onset temperature of the primary crystallization peaks (Fig. 1a, Fig. 3a,b),
Al88Y7Fe5 samples were annealed at 245 °C, Al87Y7Fe5Pb1 at 230 °C and Al87Y7Fe5In1 at 190 °C
for various periods of time. The calorimetric signals of Al88Y7Fe5 samples given in Fig. 1a show
an additional maximum in the range of 320–360 °C between the primary and the intermetallic
crystallization reaction for all samples that had been annealed at 245 °C. Moreover, annealing
the samples for periods exceeding 150 min did show a marked decrease of the crystallization si-
gnal of this additional phase during subsequent heating experiments. This additional peak was
not observed on the thermal response of the as-spun samples and careful XRD measurements of
samples that had been heated up to 348 °C did not show any crystalline phase except the Al-na-
nocrystals [9]. In contrast, XRD measurement on samples that have been annealed at 245 °C for
60 min and that were subsequently reheated up to 355 °C exhibited strong peaks- in addition to
the Al-reflections- that correspond to the formation of a so-called transient phase (Fig. 4a).
Al
b 10 nm
Transient phase
Al-nanocrystal
a
Transient
phase
Figure 4: Al88Y7Fe5 samples that have been annealed at 245 °C for 60 min and subsequently heated at
20 °C min–1 up to 355 °C a) XRD pattern shows additional peaks (·) to the Al-reflection peaks (|) corresponding to
the transient phase. b) HRTEM image shows an Al-nanocrystal surrounded by fringes corresponding to the tran-
sient phase.
119
It is worth noting that the precipitation of the transient phase affects considerably the crystal-
lization signal of the intermetallic phase (Fig. 1a) since it corresponds to the full crystallization
of the residual amorphous phase. The situation seems to be very different in the presence of in-
oculant Pb and In particles. In Fig. 5, representative examples of the typical thermal responses
of annealed Al87Y7Fe5Pb1 samples at 230 °C are given that do not show any additional signal
compared to the Pb-free samples for similar annealing times (Fig. 1a). Moreover, XRD meas-
urements of Al87Y7Fe5Pb1 that was annealed at 230 °C for 60 min and that has been reheated
subsequently up to 325 °C, do not show additional peaks to the Al and Pb reflections. Similar
results have been obtained on Al87Y7Fe5In1 after annealing at 190 °C. These results show that
the presence of inoculant particles considerably modifies the phase evolution sequence of
Al88Y7Fe5 during devitrification in inhibiting the formation of the transient phase.
Figure 5: Continuous heating DSC traces at 20 °C min–1 of annealed Al87Y7Fe5Pb1 at 230 °C for 10, 60 and
120 min. The thermal responses do not show any exothermic signal in addition to the nanocrystallization of Al
and the crystallization of the equilibrium intermetallic phases.
In order to analyze the reason for the nucleation of the transient phase, HRTEM analyses
have been performed. Fig. 4b shows a HRTEM image of an Al-nanocrystal that has been
formed by holding the Al88Y7Fe5 sample at 245 °C for 60 min and by re-heating it up to 355 °C.
The sample regions surrounding the nanocrystals exhibit the presence of fringes that can not be
attributed to the fcc-Al and that correspond to the transient phase. Detailed analyses [9] indicate
that the interface between the Al-nanocrystal and the transient phase seems to be semi-coherent
indicating that the transient phase nucleated at the Al-nanocrystal/glass interface. This situation
can be analyzed within the framework of nucleation in a concentration gradient as proposed by
Desré [6]. After short time annealing (e.g. during heating the as-spun material at a constant
rate), the resulting concentration gradient is rather steep [5] and prevents the formation of any
additional phase near the interface. Moreover, the width of spatial regions that have a composi-
tion suitable for the nucleation of an intermetallic phase can be smaller than the critical radius
for nucleation. However, the diffusion fields are broadened if limited interdiffusion has oc-
curred during a pre-annealing treatment. The resulting decrease of the concentration gradient in-
volves two effects: the activation barrier for nucleation is decreased and the spatial regions of
compositions that allow the nucleation of an additional phase with limited stability range are
120
broadened. Thus, nucleation and growth of metastable phases as indicated in Fig. 4a,b are ex-
pected to be favored during the development from initial sharp gradients towards full diffusion-
al equilibrium. At the late stages of the interdiffusion process, the solute concentration along the
diffusion path is not sufficient for the nucleation of metastable phases. Thus, the nucleation of
the stable phase is again favored and – as observed experimentally - the calorimetric signal due
to the formation of the transient phase disappears after long annealing periods (Fig. 1a). Howev-
er, since the steepness and the curvature of the concentration gradient is strongly dependent on
the effectiveness of the solute to diffuse into the matrix, the distance between neighboring na-
nocrystals appears to be a predominant parameter for the nucleation of the transient phase. The
strong increase of the nanocrystal number density that has been observed in Pb- containing sam-
ples limits the interdiffusion in promoting early impingement of the diffusion fields that strong-
ly affects the nucleation kinetics of the transient phase. In this case, the specific steepness and
curvature of the concentration gradient that are suitable for the stability of the transient phase
can hardly be reached. This limits or even inhibits the nucleation of the transient phase as
shown in Fig. 5. Thus, the present results are in complete qualitative agreement with the conclu-
sions drawn from the concept of nucleation in a concentration gradient [6], indicating the exist-
ence of a critical value of the concentration gradient that favors the nucleation of the transient
phase.
6 Summary
Along with the effort in extending the understanding of the crystallization behavior of marginal
glass forming alloys, new investigations have been performed specially in order to increase the
number density of Al-nanocrystals dispersed within an amorphous matrix . In the efforts in pro-
moting the nucleation sites, immiscible Pb and In particles have been incorporated into the ter-
nary alloy. The results give evidence to the nanocrystal-promoting effect of the inoculants that
offers new opportunities in producing nanostructured materials with a high number density of
nanocrystals. In addition to the inherent nanostructure synthesis option, the combined investiga-
tions with incorporated Pb and In particles indicated that different mechanisms apply in the ca-
talysis process. Moreover, it is demonstrated that the formation of a pronounced concentration
gradient at the particle/amorphous matrix interface is one of the key parameters that provides
for the kinetic stability of the nanostructure and the modification of the phase evolution se-
quence of the crystallization reactions. These investigations indicate the attractive opportuni-
ties, e.g. to produce specific microstructures and to select nucleating phase by controlling the
concentration gradient.
7 Acknowledgements
The authors gratefully acknowledge the support by the DFG (Emmy Noether program and cen-
ter for functional nanostructures, CFN).
121
8 References
[1] A.L. Greer, Science 1995, 267, 1947–1953

[2] G. Wilde, N. Boucharat, R.J. Hebert, H. Roesner, W.S. Tong, J.H. Perepezko, Adv. Engr.
Mat. 2003, 5 (3), 125–130
[3] H.Chen, Y. He, G.J. Shiflet, S.J. Poon, Nature 1994, 367 (10), 541–543
[4] G. Wilde, H. Sieber, J. H. Perepezko, Scripta Mater. 1999, 40, 779–783
[5] D.R. Allen, J.C. Foley, J.H. Perepezko, Acta Mat. 1998, 46, 431–440
[6] F. Hodaj, P.J. Desré, Acta Mater. 1996, 44, 4485–4490
[7] K. Hono, Y. Zhang, A. Inoue, T. Sakurai, Mater. Sci. Eng. 1997, A226–228, 498–502
[8] A. L. Greer, in Proceedings 22nd RisØ International Symp. on Mat. Sci., Roskilde, Dan-
mark, 2001, pp.461
[9] N. Boucharat, H. Rösner, J.H. Perepezko, G. Wilde, Mater. Sci. Eng. A, in press
[10] A.A. Csontos, G.J. Shiflet, Nanostruct. Mater. 1997, 9, 281–289
[11] S.J. Hong, P.J. Warren, B.S. Chun, Mater. Sci. Eng. 2001, A304–306, 362–366
[12] T. Ohkubo, H. Kai, D.H. Ping, K. Hono, Y. Hirotsu, Scripta mater. 2001, 44, 971–976
[13] D.L. Zhang, B. Cantor, Acta metall. Mater. 1991, 39 (7), 1595–1602
[14] H. Roesner, P. Scheer, J. Weissmüller, G. Wilde, Phil. Mag. Lett. 2003, 83 (8), 511–523
[15] G. Wilde, R.I. Wu, J.H. Perepezko, Adv. Solid-state Physics 2000, 40, 391
122
Temperature Dependent Critical Size of Embedded Silicon

Nanocrystals: A Molecular Dynamics Study
Y.-T. Chiu, J.-T. Yeh

Industrial Technology Research Institute, Taiwan
1 Abstract
In this research, we adopted a molecular dynamics simulation to study the temperature depen-
dence of the critical size of silicon nucleation. For a nanocrystal embedded in a corresponding
amorphous matrix, the critical size has strong temperature dependence. To study the above phe-
nomenon at the atomic level, we used the Stillinger-Weber potential [1] to modify a parallel
molecular dynamics program, DL_POLY [2], to handle silicon systems. A spherical silicon cry-
stal embedded in a corresponding amorphous matrix was employed as a computational model,
and it was found that the critical size decreases when the temperature is reduced, which is con-
sistent with existent thermodynamics models.
2 Introduction
Melting and crystallization occur in pulsed laser annealing of low temperature poly-silicon
(LTPS) processing of thin film transistors (TFT) in flat panel displays (FPD). The qualities of
the silicon grains influence the performance of the transistors. Therefore, controlling the condi-
tions of processing is critical important. Understanding the mechanism of annealing is essential.
However, in a nanoscale, the melting temperatures of crystals are size dependent [3]. In other
words, for a specific temperature, a critical size exists. A crystal with a size larger than the criti-
cal size will grow, while one sized smaller will melt.
The size dependence of crystal melting is a general phenomenon. To explain the size de-
pendence of crystal melting, Gibbs-Thompson’s equation is a well-known thermodynamics
model [4]:
Tm a
1 , (1)
TB rc
in which TB is the bulk melting temperature, Tm the melting temperature of the nanocrystal, rc
the critical size at temperature Tm, and a the parameter to be determined. However, for crystals
with only several nano-meters, feasibility of the above equation is still worth studying. Recent-
ly, the Celestini model [4], which takes into account the quasi-liquid-layer between the core of
crystal and the matrix, predicted melting temperature as
Tm b [
1 (1 ), (2)
TB rc rc 2[

ISBN: 3-527-31007-X
123
2J sl
b . (3)
U Lrc
In the above equations, Jsl represent the solid-liquid interfacial energy, U the density, L the la-
tent heat for melting, [ the correlation length of interaction. In Celestini’s research, Jsl and [ are
obtained by fitting their numerical results.
On the other hand, simulations of the nanocrystal melting and crystallization by molecular
dynamics at the atomic level can enhance the comprehension and application of the nanosized
grain growth mechanism. The molecular dynamics [5] has been used to study melting and crys-
tallization speed of bulk silicon with free surfaces [6,7], and satisfactory results were obtained.
In this research, we adopted a molecular dynamics simulation to study the temperature depend-
ence of the critical size of melting and crystallization of silicon nanocrystals embedded in the
corresponding amorphous matrices.
3 Method
In order to understand the temperature dependence of the critical size, we employed the Stillin-
ger-Weber potential to modify DL_POLY so that it is capable of handling systems containing
silicon atoms.
The Stinninger-Weber potential has been used widely to simulate silicon system since it was
published. It consists of two terms. The first term varies with the distance between two parti-
cles, while the other varies with the angle among three particles. On the other hand, DL_POLY
is a parallel molecular dynamics code developed by the Daresbury Laboratory in the United
Kingdom. It has been ported to various computers including low cost PC-clusters. However, the
Stillinger-Weber potential does not exist in the official version of DL_POLY.
In this research, the DL_POLY subroutines FORGEN and THBFRC were separately modi-
fied to calculate forces generated by the above two terms of the Stillinger-Weber potential. In
addition, since the Stillinger-Weber potential is a short-ranged force field, the long range cor-
rection of pressure and energy in the subroutine LRCORREC were forced to be zero. All simu-
lations were executed in a parallel mode on a Beowulf-structured PC-cluster with thirty-two
processors.
The accuracy of the modified program is verified by calculating material characteristics from
the numerical output and comparing these values with the ones from related literatures. These
material properties are the bulk modulus, coefficient of thermal expansion, melting speeds [6],
and crystallization speeds [7].
To find the temperature dependent critical size of embedded silicon nanocrystals, a model
consisted of a spherical silicon crystal embedded in a cubic matrix made up of amorphous sili-
con atoms was used. The periodic boundary conditions and the NPT ensemble were used while
the crystal radius was set at several different values for each distinct temperature level. Then,
the critical radius of the nanocrystal for a specific temperature level was determined by averag-
ing the radii of two runs. In the above runs, the small crystal melted and the large one grew. In
the simulations, these two radii values differed from each other by only 0.1 nm or 0.05 nm.
124
4 Result and Discussion
Figure 1 shows the temperature dependent critical radius of the embedded silicon nanocrystal.
In this figure, square hollow marks represent the present result. The dashed line represents the
Gibbs-Thompson equation (1) and the solid line, the Celestini model (2). In addition, the bulk
melting temperature, 1685 K, is plotted for comparison. Therefore, it is obvious that the Gibbs-
Thompson equation, which is in the order of the inverse of the radius, is not adequate to descri-
be the relationship between melting temperatures and critical radii. However, as the same figure
shows, the present molecular dynamics result is consistent with the Celestini model.
Figure 1: The temperature dependent critical radius of an embedded silicon nanocrystal
Furthermore, by utilizing our results, the parameter D in equation (1) was found to be
0.404nm, b and [ in equation (2) to be 0.242 nm and 0.376 nm, respectively. It is well esta-
blished that the density of silicon is 2533 kg/m3 and latent heat is 31022 J/mole [8]. Thus, the
interfacial energy was calculated to be 0.338 J/m2, which is very close to the value of 0.34 J/m2
reported in the literature [9].
Figure 2 represents the development of an embedded silicon crystal with a radius of 1.7 nm
at 1300 K. On the contrary, figure 3 indicates the melting of one with a radius of 1.3 nm at
1200 K. Smaller crystals melt at lower temperatures, while larger crystals grow at higher tem-
peratures. The above phenomenon is only apparent in a nanoscale.
125
Figure 2: The embedded silicon crystal with radius 1.7 nm grows at 1300 K
Figure 3: The embedded silicon crystal with radius 1.7 nm melts at 1200 K
5 Acknowledgement
The authors would like to thank professor Maruyama of the University of Tokyo in Japan for
providing the post processing software, PVwin.
6 References
[1] F. H. Stillinger, T. A.Weber, Phys. Rev. B, 1985, 31, 5262–5271

[2] W. Smith, T.R. Forster, The DL_POLY_2 User Manual, ver 2.13, CLRC, Daresbery
Laboratory, Warrington, UK, 2001
[3] A. N. Goldstein, C. M. Echer and A. P. Alivisatos, Science, 1992, 256, 1425–1427
[4] F. Celestini, R. J.-M. Pellenq, P. Bordarier and B. Rousseau, Z. Phys. D. 1996, 37,
49–53
[5] J. M. Haile, Molecular Dynamics Simulation, John Wiley & Sons, New York, 1992
[6] S. R. Phillpot, S. Yip and D. Wolf, Computers in Physics, 1989, Nov/Dec 20–31
[7] M. H. Garbow, G. H. Gilmer, A. F. Bakker, MRS Symp. Proc. 1989, 141, 349–354
[8] Markku Ylilammi, http://www.geocities.com/Silicon Valley/Bay/4104/silicon.html, 1998
[9] S. R. Stiffler, M. O. Thompson, and P. S. Peersy, Phys. Rev. Lett. 1988, 60, 2519–2522
126
Effects of Soft-Impingement and Non-random Nucleation on the

Kinetics and Microstructural Development of Primary
Crystallization
P. Bruna*, E. Pineda**, D. Crespo* and R. González-Cinca*

*Departament de Física Aplicada, EPSC, Universitat Politècnica de Catalunya. Castelldefels, Spain
**Departament de Física i Enginyeria Nuclear, ESAB, Universitat Politècnica de Catalunya. Barcelona, Spain
1 Introduction
A variety of nano-structured materials are obtained by primary crystallization of rapidly-quen-

ched amorphous precursors, the new phase developing by nucleation and diffusion controlled
growth [1,2]. In many cases slowly diffusing elements are added in order to control the crystal-
lite growth and obtain refined grain size distributions [3]. Surface instability is not observed in
such crystallizations due to the reduced grain size, but the concentration profiles around the gro-
wing crystallites affect both the nucleation and the growth of neighbor particles [4]. On the one
hand, the stabilization in the crystalline grain surroundings inhibits the nucleation process, thus
reducing the probability of nucleation of new particles. In this work this non-homogeneously
distributed nucleation probability will be referred as Non-Random Nucleation (NRN). On the
other hand, the overlapping of the slowest solute concentration profiles reduces the gradients
and consequently the growth rate of the grain boundary region facing a neighbor crystallite. The
grain growth of the grains is then stopped before direct impingement and, besides, it becomes
non-isotropic because of its dependence on the local distribution of neighbor particles. This be-
havior is commonly referred as diffusion controlled Growth with Soft-Impingement (GSI).
Both NRN and GSI mechanisms are difficult to deal with and, though being classical prob-
lems in metals’ and alloys’ phase transformations, there is not common agreement on what
overall kinetics and microstructure characteristics can be attributed to each one. In particular
there exists discussion concerning to the applicability in this cases of the Avrami equation,
which is commonly used to analyze calorimetric data of primary crystallization [5]. The Kol-
mogorov-Johnson & Mehl-Avrami (KJMA) equation
x t 1 exp ª¬ x t º¼ (1)
describes the evolution of the transformed volume fraction, x(t), in terms of the so called exten-
ded volume fraction, x t , which gives the volume the new phase would occupy if there was
not interference between neighbor particles, that is, neglecting overlapping [6–9]. The extended
transformed volume fraction can be computed from the nucleation and growth rates, namely I(t)
and G(W, t), as
3
t 4S ª t
x t ³0 I W dW G W , t c dt cº (2)
3 ¬« ³tc ¼»
where W defines the nucleation time of the particles.

ISBN: 3-527-31007-X
127
However, as both parameters are difficult to determine independently it is common to define

a reaction constant k and to assume a potential dependence on the transformation time through
the so-called Avrami exponent n, thus having

x t 1 exp kt n (3)
The KJMA equation is strictly valid only for transformations driven by random nucleation,
isotropic growth and direct impingement between neighbor particles [10]. In such cases the val-
ue of n is related to the nucleation and growth mechanisms governing the transformation [5].
Even if the KJMA conditions are utterly fulfilled, time dependent nucleation or growth pro-
tocols lead to complex experimental n behaviors which can be difficult to interpret [11]. In addi-
tion, if NRN and GSI are expected the KJMA conditions are not satisfied and then the KJMA
equation is not well-grounded. In such case there is a lack of foundations in the interpretation of
experimental results. Several theoretical models dealing with such transformations have been
presented during the recent years [4,12,13]. However, some of their results are contradictory
and the complexity of the experimental systems make difficult to discern their validity. As an
example of an experimental system, reference [14] shows the varying Avrami exponents calcu-
lated from calorimetric data for the nano-crystallization of a FINEMET alloy.
As already stated GSI and NRN are related to the evolution of the slowest solute concentra-
tion field in the meta-stable amorphous phase; this fact makes the phase-field simulations a be-
coming method for their study. Indeed, phase-field simulations allow the integration of the
diffusing field without explicitly tracking the moving grain interfaces. In this work, we show
phase-field simulations of primary crystallization which clarify the contributions of each mech-
anism to the transformation kinetics and the final microstructure of such phase transformations.
Previous phase-field simulations of transformations with GSI and NRN qualitatively confirmed
the validity of this approach for the simulation of phase transformations with such characteris-
tics [15,16]. In this work a quantitative analysis of the results of the phase-fields simulations al-
lows us to check the validity of the theoretical GSI and NRN models proposed up to date.
2 Phase-field Model Simulations
The phase-field model used in this work is a solidification model previously developed for den-
dritic formation studies [17]. Here, we have suppressed the anisotropic terms of the model ob-
taining a simplified model with equations
wI § 1 ·
H 2m I 1 I ¨ I 30HE'uI 1 I ¸ (4a)
wt © 2 ¹
wu 1 wI
wt '

30I 2 60I 3 30I 4
wt
2u (4b)
&
where I r , t is the phase-field which distinguishes between transformed and untransformed si-
&
tes (that is between crystalline and super-cooled liquid sites) and u r , t is a normalized diffusi-
on field which in our simulations describes the slowest solute concentration value at each site.
The parameter ' means respectively the under-cooling or the super-saturation values in the ca-
128
&
ses of u r , t being a temperature field or a solute concentration field. In the present work ' (the
super-saturation value) coincides with the final crystallized volume fraction, that is
c* c m t 0 (5)
'
c* c xt
where cxt and c* are the crystalline and the super-cooled liquid equilibrium concentrations, and
cm (t = 0) is the initial solute concentration of the amorphous matrix. The remaining parameters
needed in equations (4) are described in more detail elsewhere [17].
When a transformed crystalline site is fixed in the under-cooled matrix (that would corre-
spond to a nucleation event), the integration of the model induces a progressive growth of the
grain, the value of the field u being constant at the crystalline phase while diffusing along the
untransformed matrix. During the first stage of the grain growth the interface velocity is con-
trolled by equation (4a), and once the equilibrium value of the field u is reached at the interface
the growth is controlled by the diffusion of the u-field, given by equation (4b).
In order to analyze the kinetics given by equations (4), simulations of an isolated grain grow-
ing without interference in a grid of 400 × 400 points were performed. Figure 1 shows the grain
radius and the interface velocity evolution in such simulations. The behavior of the grain growth
can be well described by a first linear growth stage, plus a second diffusion controlled growth
stage. The linear growth velocity is dependent of the phase-field model parameters and can be
fitted to a constant value U ~ ' / m. Afterwards, the velocity during the second stage is well de-
scribed by the classical diffusion controlled growth rate of a spherical particle firstly deduced
by Zener [18]. Such velocity is dependent on ' and the diffusion coefficient of the u field,
namely D. The transition between both growth modes takes place when the diffusion controlled
velocity becomes lower than the linear one, as observed in figure 1.
diffusion controlled
growth velocity (arbitrary units)
linear growth growth (Zener)

grain radius (arbitrary units)
growth (Zener) linear growth
t (arbitrary units) t (arbitrary units)
Figure 1: Grain radius versus time for a phase-field simulation of an isolated grain. Dots: results from phase-field
simulations. Continuous Line: theoretical prediction using the parameters of equations (4).
Considering the above discussion, the kinetics given by the phase-field model are qualita-
tively similar to the one expected in primary crystallizations, where, in many cases, diffusion
controlled growth becomes the dominant growth mechanism after a first interface controlled
growth stage [4]. This fact will allows us to check the theoretical models used in primary crys-
tallization studies in a controlled system where all the parameters are explicitly known.
129
3 Kinetic Models for NRN and GSI
As previously said, several theoretical models have been proposed to describe the NRN and
GSI effects in primary crystallization. In the case of NRN, different approaches lead to different
equations for the crystallized fraction evolution. These equations are basically generalizations
of the KJMA equation which are able to describe some of the deviations from the KJMA theory
observed in experimental data [12,13,19]. The main difference between the models is the degree
of deviation from the KJMA equation attributed to NRN. A previous work from the authors
also dealt with NRN transformations by means of stochastic simulations [20]. In that work the
main parameters were the length of the particles’ corona where the nucleation is inhibited, and
the degree of nucleation inhibition in such zone . After examining a wide range of the parame-
ters values one could conclude that NRN does not affect significantly the overall kinetics and,
moreover, the micro-structure developed is only slightly different from the one obtained in a
system with an average reduction of the nucleation rate in all the untransformed matrix. Hence,
the lost of spatial randomness in the nucleation probability does not seem to be responsible for
the severe decrease in the transformation kinetics observed in experimental data, which must be
then attributed to GSI. The previous discussion is in agreement with other previous studies on
primary crystallization [21].
There are also some theoretical models dealing with the GSI effects. Here we will detail a
mean-field model developed by Clavaguera-Mora et al. [4] which gave very good results in the
study of FINEMET alloys de-vitrification. Roughly speaking, the model is based in two differ-
ent effects occurring in primary crystallization. Firstly a geometrical effect due to the exhaust-
ing of the transformable space not only due to the increase of the crystallized volume but also
due to the associate increase of the stabilized zone. Secondly the model states a chemical effect
due to the reduction of the kinetic parameters because of the progressive stabilization of the ma-
trix. In order to take into account the geometrical effect, the model proposes that the extended
fraction term in the KJMA equation must be normalized by the final crystallized volume, that
means
x t ' 1 1 exp ª¬ x t ' 1 º¼ (6)
This equation is in fact the most common application of the KJMA equation to the calorimet-
ric data obtained from primary crystallizations. In order to take into account the chemical effect
the model states that the kinetic parameters must be reduced by a soft-impingement factor
1 x t
M x t (7)
1 'x t
which coincides with the average variation of the solute concentration in the untransformed
phase
c* c m t c* c m t 0 (8)
xt
M t
*
c c c* c xt
Hence, implying a first order approximation, the kinetic parameters become

130
U t U 0M t , D t D0M t , I t I 0M t (9)
Summarizing, the model proposes a modified KJMA equation where kinetic parameters
have a time decreasing dependence given by the soft impingement factor M(t). This model al-
lowed the authors to reproduce the overall kinetics in a FINEMET alloy isothermal nano-crys-
tallization for different annealing temperatures [4,14]. Particularly interesting for the present
work is the fact that the factor M(t) was able to explain the final low Avrami exponents, which
are obtained very commonly in primary crystallization.
4 Results and Discussion: Simulations Compared with Kinetic Models
Multiple grain phase-field simulations were performed in a grid of 1000 × 1000 with reflecting
boundary conditions. Several simulations were performed for values of ' ranging between 0.3
and 0.7. As the analysis of the results suggested no relevant differences related to the ' value,
here we will just show the results obtained for the '= 0.5 simulations for the sake of simplicity.
Firstly we will analyze the results obtained for preexisting nuclei transformations. In such simu-
lations 125 nuclei were settled randomly before the start of the transformation and no further
nucleation was allowed. Consequently NRN does not appear in these simulations and they are
particularly appropriate to study the GSI effects separately.
1 3
constant growth
2
n(t)
x(t)
0.5
1
growth
0 0
0 1 2 3 4 0 0.5 1
t (arbitrary units) x(t)
Figure 2: Transformed fraction evolution (left) and Avrami exponent versus transformed fraction (right). Phase-
field simulation (symbols), classical KJMA equation (dashed line), modified KJMA equation - equations (8) to
(11) - including the soft impingement effect (continuous line).
Figure 2 shows both the transformed fraction evolution and the varying Avrami exponents
obtained in such simulations, compared to the models described above. It can be seen that the
kinetics of the presented simulations is more close to the classical KJMA equation without no
modifications than to the model proposed to describe the GSI effect through the soft impinge-
ment factor, described by equations (8) to (11). The kinetics of our simulations can be well un-
derstood considering a transformation of preexisting nuclei shifting from an interface controlled
growth (n = 2) to a diffusion controlled growth (n = 1) as the transformation advances.
131
Another result of our simulations is the grain size distribution obtained in simulations of
transformations with preexisting nuclei. Figure 3 shows that the grain size distribution, normal-
ized by the final transformed fraction ', coincides with the one expected for a Voronoi tessella-
tion [22]. This result means that the GSI effect on the grain size distribution is essentially a
reduction of the mean grain size without affecting the distribution shape.
grain density
0 1 2 3
grain area (A/<A>)
Figure 3: Comparison of grain size distributions obtained for a phase-field simulation of a transformation with
preexisting nuclei (Bars) and a Voronoi tessellation (line).
Further simulations were designed to analyze the NRN effect. In these simulations nuclei ap-
peared at each time step with a nucleation probability depending of the local value of the con-
centration field u(r, t). This nucleation probability was chosen to follow a law similar to the one
expected in real systems [23], that means
ª & ª § ·º
& § cm r , t ·º 1 1
I r ,t I 0 exp « Q / ln 2 ¨ I 0 exp « A ¨ 2 2 ¸» (10)
¨ * ¸¸ »» &
«¬ ¨© ln 1 u r , t / 2 ln 1/ 2 ¸¹ »¼
«¬ © c ¹¼
With this prescription, the nucleation rate is fixed I0 at untransformed matrix points having
the initial solute concentration, cm (t = 0), and becomes null at the points with the equilibrium
concentration c*. The parameter A, together with the concentration profiles shape, determines
the thickness of the corona around the crystalline grains where the nucleation is almost inhibit-
ed. In this work the phase-field simulations of NRN transformations were performed with
A = 0.1. Values of A = 1 induce very large reduction of the nucleation in the overall amorphous
matrix, thus inducing transformations with kinetics very similar to the transformations with pre-
existing nuclei shown above. On the opposite limit, simulations performed with values of
A = 0.01 show very small NRN effects, only noticeable at the very end of the transformation.
Analysis of the NRN simulations lead to conclusions similar as in the preexisting nuclei
case. Figure 4 shows the transformed fraction evolution and the varying Avrami exponent cal-
culated for the NRN simulations compared to the ones expected by the classical KJMA equa-
tion and the theoretical model including the soft impingement factor. As in the previous case,
the phase-field simulations including GSI and NRN exhibit slight deviations from the classical
KJMA equation. Now, the combined effects of GSI and NRN induce larger differences between
simulations and the KJMA model than in the preexisting nuclei transformations; such differenc-
es are more noticeable in the final values of the Avrami exponent. However, the simulations do
132
not show the important transformed fraction evolution delays and low final Avrami exponent
values found in experimental data on primary transformations and expected by the soft im-
pingement model of equations (8) to (11). The Avrami exponent behavior corresponds to a con-
stant nucleation transformation shifting from an interface controlled growth (n = 3) to a
diffusion controlled growth (n = 2) as the time since the crystallite nucleation elapses.
1 3
constant growth
2
x(t)
n(t)
0.5 diffusion
controlled growth
1
0 0
0 1 2 3 0 0.5 1
t (arbitrary units) x(t)
Figure 4: Transformed fraction evolution (left) and Avrami exponent versus transformed fraction (right). Phase-
field simulation (symbols), classical KJMA equation (dashed line), modified KJMA equation including the soft
impingement effect (continuous line).
The analysis of the micro-structure obtained in NRN simulations leads to conclusions similar
to these found in the above mentioned previous work using stochastic simulations [19]. The
grain size distribution can be well described with an average reduction of the nucleation rate,
that means assuming a randomly distributed time decreasing nucleation of the type
ª § cm t ·º &
I t I 0 exp « Q / ln 2 ¨ * ¸ » , cm t cm r , t (11)
¨ c ¸»
¬« © ¹¼
As a consequence, the main effect of NRN on the transformation is to narrow the grain size
distribution due to the overall inhibition of new nuclei at the latest times of the transformation,
but almost independent of the spatial non-random distribution of such nuclei in the untrans-
formed phase.
5 Conclusions
The theoretical models used in analyzing the primary crystallization kinetics have been checked
by means of phase-field simulations. The effect of the overlapping concentration profiles (diffu-
sion controlled growth with soft impingement) is found to produce slight delays to the grain
growth kinetics. Moreover, a nucleation affected by the amorphous matrix stabilization around
133
the growing crystallites (non-random nucleation), it is found to induce only non-significant de-
viations of the overall kinetics. In the case of preexisting nuclei transformations, the grain size
distribution is well described by a Voronoi tessellation, showing that the soft impingement has
no effect on the final space partition between the grains. In transformations with non-random
nucleation, the grain density evolution can be well described by an average reduction of the nu-
cleation rate. Summarizing, the effects of soft impingement and non-random nucleation obser-
ved in the kinetics and micro-structure are in all cases smaller than those expected, the Avrami
equation being able to describe the transformed fraction evolution within a moderate error and
the Avrami exponent maintaining its meaning. The important delays and complex Avrami ex-
ponent behaviors commonly observed in real transformations kinetics, which are generally des-
cribed by kinetic models assuming soft impingement effects, must be attributed to other
phenomena (as for example, changes in the properties of the super-cooled liquid matrix due to
its varying composition). Further work will extend the model to 3D and will be devoted to the
study of the effect of the untransformed matrix varying composition in the overall kinetics.
6 Acknowledgements
E. Pineda, P. Bruna and D. Crespo are supported by CICYT (grant MAT2001-0957) and Gene-
ralitat de Catalunya (grant 2001SGR00190). R. González-Cinca is supported by European
Commision (Research Training Network HPRN-CT-2002-00312), Direccion General de Inve-
stigación Científica y Técnica (project BFM2000-0624-C03-02), Comissionat per a Universi-
tats i Recerca (project 2001SGR0021) and Departament d’Universitats Recerca i Societat de la
Informació (project 2002XT00010).
7 References
[1] A. Inoue, Prog. Mater. Sci., 1998, 43, 365–520

[2] M. E. McHenry, M. A. Willard, D. E. Laughlin, Prog. Mater. Sci., 1999, 44, 291–433
[3] Y. Yoshizawa, S. Oguma, K. Yamauchi, J. Appl. Phys., 1988, 64, 6044–6046
[4] M. T. Clavaguera-Mora, N. Clavaguera, D. Crespo, T. Pradell, Prog. Mater. Sci., 2002,
47, 559–619
[5] W. Christian, The Theory of Transformations in Metals and Alloys, Pergamon Press,
Oxford, 1975, Chapter 12
[6] A. N. Kolmogorov, Bull. Acad. Sci. USSR, Phys. Ser., 1937, 1, 355–359
[7] W. A. Johnson, P. A. Mehl, Trans. Am. Inst. Mining and Metallurgical Engineers, 1939,
135, 416–58
[8] M. Avrami, J. Chem. Phys., 1939, 7, 1103–1112
[9] M. Avrami, J. Chem. Phys., 1940, 8, 212–224
[10] V. Sessa, M. Fanfoni and M. Tomellini, Phys. Rev. B, 1996, 54, 836–841
[11] T. Pradell, D. Crespo, N. Clavaguera, M. T. Clavaguera-Mora, Proceedings of the Fifth
International Workshop on Non-Crystalline Solids. Non-Crystalline and Nanoscale Mate-
rials (Ed: R. Rivas, M. A. Lopez-Quintela,), World Scientific, Singapore, 1998,
p. 317–322
[12] H. Hermann, Europhys. Lett., 1998, 44, 245–250
134
[13] M. Tomellini, M. Fanfoni, M. Volpe, Phys. Rev. B, 2002, 65, 14031–(1–4)

[14] T. Pradell, D. Crespo, N. Clavaguera, M. T. Clavaguera-Mora, J. Phys.: Condens. Matter,
1998, 10, 3833–3844
[15] J. P. Simmons, C. Shen, Y. Wang, Scripta mater., 2000, 43, 935–942
[16] L. Granasy, T. Borzsonyi, T. Pusztai, Phys. Rev. Lett., 2002, 88, 206105–(1–4)
[17] R. Gonzalez-Cinca, L. Ramirez-Piscina, Phys. Rev. E, 2001, 63, 051602–(1–9)
[18] C. Zener, J. Appl. Phys., 1949, 20, 950–953
[19] E. Pineda, D. Crespo, T. Pradell, Philos. Mag. A, 2001, 82, 107–121
[20] E. Pineda, D. Crespo, J. Non-Cryst. Solids, 2003, 317, 85–90
[21] D. Crespo, T. Pradell, M. T. Clavaguera-Mora and N. Clavaguera, Phys. Rev B, 1997, 55,
3435–3444
[22] D. Wearie, J. P. Kermode, J Wejchert, Philos. Mag. B, 1986, 53, L101–L105
[23] D. Turnbull, J. Appl. Phys., 1950, 21, 1022–1028
135
Mechanical Spectroscopy of Quasicrystal Formation from

Amorphous Ti- and Zr-based Alloys
H.-R. Sinning1, I. S. Golovin1, A. Jianu2

1
Institut für Werkstoffe, Technische Universität Braunschweig, Germany
2
National Institute of Materials Physics, Bucharest-Magurele, Romania
1 Abstract
Rapidly quenched Ti-Zr-Ni and Ti-Zr-Ni-Ag alloys have been produced in an almost conti-
nuous sequence from apparently amorphous over nano-quasicrystalline to poly-quasicrystalline
states. Influences of such structural changes on Young’s modulus, damping, and on the Snoek-
type relaxation peak of absorbed hydrogen (with hydrogen as a probe) are studied and briefly
discussed with respect to ordering processes, relations between amorphous and icosahedral qua-
sicrystalline structures, and the nature of the amorphous-quasicrystalline transition.
2 Introduction
Amorphous metals and icosahedral quasicrystals are interconnected in many aspects, both from
fundamental and applied viewpoints. Most of these common aspects are consequences of the
strong structural similarity, at least in the range of short-range order, between these two classes
of alloy phases.
From the viewpoint of application, quasicrystal formation from the amorphous state may on
one side restrict the production and use of bulk metallic glasses: if the nucleation of (periodic)
crystals is effectively suppressed – which is a precondition for the production of amorphous al-
loys in bulk form – it may still be possible to destabilize the glass (and hence to limit the attain-
able dimensions of bulk production) by the nucleation of quasicrystals. Accordingly, most of
the more recent observations of quasicrystal formation from the amorphous state were made in
bulk glass-forming systems [1–6]. On the other hand, the precipitation of fine quasicrystalline
particles from the amorphous matrix can effectively improve the mechanical properties of bulk
metallic glasses [7–9].
On the fundamental side, there is the question how far (in a double sense) the structures of
amorphous and icosahedral quasicrystalline phases resemble each other, i.e. how far they can be
described by the same concept or principle, and up to which length scale they are comparable in
real space. On the atomic level, polytetrahedral packing is obvious in both cases, which in three-
dimensional space necessarily means a non-periodic arrangement of non-regular tetrahedra.
Starting from an arbitrary ”central” atom, the first nearest-neighbor shell preferably forms an
icosahedron, around which the number of possible atomic configurations increases rapidly to-
wards higher coordination shells. This viewpoint leads to various types of cluster models, which
exist both for quasicrystals [10–13] and for the amorphous state [14]. Without going into de-
tails, models with overlapping clusters (each atom belongs to at least one cluster) must be dis-
tinguished from those with separated clusters (each atom belongs to at most one cluster); in the

ISBN: 3-527-31007-X
136
latter case, so-called ”glue atoms” between the clusters may form more disordered boundary re-
gions. Looking at the relation between atomic order and electronic states, local spherical perio-
dicity induced by electronic resonances (Hume-Rothery stabilization) has been discovered as a
main principle for all those structures that do not exhibit long-range translational periodicity
[15–17].
Hence, closer studies of amorphous-quasicrystalline (a/qc) transitions are highly desirable
from quite different, fundamental as well as application-oriented viewpoints. Two main ques-
tions about the nature of the a/qc transition and about the related understanding of the atomic
structures, already addressed in the early days of quasicrystal research [18–20], are still open:
1. Is the a/qc transition a sharp one, or a continuous process with a gradual structural change?
Some support for the latter possibility was indeed found in early electron diffraction experi-
ments [18] as well as in recent high resolution TEM images of a/qc interfaces [9].
2. Is there any structural difference, at the limiting ”X-ray amorphous” case of fine grain size
where the diffraction patterns become indistinguishable, between an amorphous and a
nano-quasicrystalline state [19, 20] – or is the amorphous phase nothing else than a highly
defective quasicrystal?
Rapidly quenched Ti-Zr-Ni(-Ag) alloys may be considered as a model system to study these
questions experimentally, since it has been shown that these alloys can be produced, depending
on composition and quenching rate, in X-ray amorphous / nano-quasicrystalline states as well as
with well-developed poly-quasicrystalline microstructures [21, 22].
In the present paper the method of Mechanical Spectroscopy, combined with TEM and elec-
tron diffraction, is applied to different aspects of the amorphous-to-quasicrystalline transition
and subsequent ordering processes. Besides in-situ measurements of the changes of Young’s
modulus and damping during the respective annealing treatments – reflecting order-induced
changes of atomic bonding forces and of defect motion, respectively – the hydrogen absorption
capability of Ti- and Zr-based alloys is used to study also the related properties of the Snoek-
type relaxation peak of interstitial hydrogen. In the sense of hydrogen as a probe, this relaxation
mechanism is known to be sensitive to the local atomic order around the jumping hydrogen at-
oms [23–25], which led to several comparative studies between amorphous and crystalline
structures [24, 26] as well as between different amorphous alloys [25, 27, 28]. Recently this me-
thodic approach was extended also to icosahedral quasicrystals, first in Zr69.5Cu12Ni11Al7.5 [29–
31] and then at different compositions in the Ti-Zr-Ni system [32, 33], revealing significant dif-
ferences in short-range order between the different icosahedral phases, but not between amor-
phous and icosahedral Zr69.5Cu12Ni11Al7.5. It is not yet clear, however, how amorphous or nano-
quasicrystalline Ti-Zr-Ni(-Ag) alloys fit into this picture, especially with respect to the above
open questions. First related results on the annealing behavior of these rapidly quenched model
alloys are therefore presented in this paper.
3 Experimental Procedure
Rapidly quenched ribbons of Ti-Zr-Ni and Ti-Zr-Ni-Ag alloys, at compositions specified be-
low, were melt-spun at the National Institute of Materials Physics (Bucharest, Romania) as des-
cribed in detail elsewhere [21, 22, 34]. Transmission electron microscopy (TEM) and electron
137
diffraction were used to characterise the initial microstructures and their subsequent transforma-
tions during annealing. For mechanical spectroscopy, the 0.02–0.06 mm thin samples were pla-
ced in a cantilever beam configuration by careful clamping at one end, and electrostatically
excited to flexural vibrations (0.1–20 kHz) in a standard vibrating-reed measurement chamber
with a high-frequency (40 MHz) detection circuit [33, 35]. The resonance frequency f and me-
chanical damping (internal friction, loss factor) Q–1 of the sample vibrations were measured un-
der vacuum either as a function of temperature (between 90 and 900 K) during linear heating
with 1 or 2 K/min, or isothermally as a function of time. Between these mechanical tests, hydro-
gen could be introduced by in-situ gas-phase charging (usually at 473 K / 1 bar H2) without tou-
ching the geometry of the vibrating-reed samples. In that way, the hydrogen concentration cH in
a sample could be varied successively from about 0.001 (initial impurity content) to 0.3 H/M, as
estimated empirically from the height of the respective damping peak [24, 31] which was roug-
hly calibrated by heat extraction analysis.
4 Experimental Results
4.1 Microstructure
With respect to their initial structural states after melt-spinning, the Ti-based alloys used in this
study form two groups: one with a well-developed poly-quasicrystalline microstructure with a
grain size of some 100 nm (called hereafter ”group A”: Ti53Zr27Ni20, Ti41Zr42Ni17; Fig. 1a), and
another one which is at the resolution limit between nano-quasicrystalline and amorphous struc-
tures (”group B”: Ti45Zr38Ni17, Ti51Zr27Ni20Ag2, Ti48Zr27Ni20Ag5; Fig. 1b). All the alloys of the
second group are ”X-ray amorphous”, i.e. their X-ray diffraction patterns are characteristic of
an amorphous structure [34]. However, the TEM micrographs show finer differences: whereas
some nanograin structure is still visible in Ti51Zr27Ni20Ag2 (Fig. 1b), the alloy Ti48Zr27Ni20Ag5
can hardly be distinguished from an amorphous phase even in TEM. Some isolated, ”larger” (~
10 nm) nanograins, embedded in an apparently amorphous matrix (within the resolution of con-
ventional TEM), are found in Ti45Zr38Ni17.
a b
500 nm 100 nm
Figure 1: Two examples of TEM micrographs of the as-quenched ribbons: (a) Ti41Zr42Ni17; (b) Ti51Zr27Ni20Ag2
138
Annealing of the ”group B” alloys results in a continuous coarsening of the nano-qc mor-
phology (Fig. 2, Table 1), in good agreement with X-ray diffraction [21]. In the case of
Ti51Zr27Ni20Ag2, this finally leads to a well-defined poly-quasicrystalline microstructure with
grain sizes around 100 nm (Fig. 2b), similar to that obtained earlier from amorphous
Zr69.5Cu12Ni11Al7.5 [29]. The difference is, however, that in the ”group B” Ti-based alloys the
formation of clearly recognizable quasicrystals is entirely a solid-state process, whereas in bulk
glasses like Zr69.5Cu12Ni11Al7.5 (”group C”) the quasicrystals form in the highly undercooled
melt above the glass transition temperature.
a b
Figure 2: Coarsening of the nano-quasicrystalline structure in Ti51Zr27Ni20Ag2 after heating with a constant rate
of 2 K/min up to: (a) 753 K, (b) 886 K. The length of the bar is 100 nm
Table 1: Rough estimate of average grain sizes in Ti51Zr27Ni20Ag2 after different annealing tre-
atments (heating with 2 K/min up to the temperatures indicated, except last column)
Annealing temperature as quenched 573 K 753 K 873 K 886 K isothermal 24 h 753 K
Grain size [nm] <5 5–10 ~ 15 50–100 ~ 100 20–30
4.2 Annealing Effects on Young’s Modulus and Intrinsic Damping
A first overview over the transformation behavior of the as-quenched ribbons during annealing
is given in Fig. 3 in form of the changes of Young’s modulus E – determined from the square of
the vibrating-reed resonance frequency f – during linear heating with a constant rate of 2 K/min.
Because absolute values of E were not available, the curves were arbitrarily normalized with re-
spect to the starting value at room temperature. However, as this normalization has no physical
meaning (the initial state after rapid quenching is not well defined), only the relative changes of
E can be compared to each other.
If a solid is structurally stable and does not undergo a transformation, then only the reversi-
ble temperature dependence should be seen which is normally a smooth, monotonically de-
creasing function E(T). Superimposed on this reversible behavior are irreversible ordering
processes, which strengthen the atomic bonds and hence increase the elastic stiffness of the sol-
139
id. Fig. 3 now clearly shows that these ordering processes are very different between the three
groups of alloys mentioned above. The typical behavior of metallic glasses is represented by
curve C, where the irreversible structural relaxation (slow increase of E between 400 and 600
K) is well separated from the effects occurring in the vicinity of crystallization, of which in Fig.
3 only the first, sharp increase due to the transformation into a metastable quasicrystalline state
around 680 K is shown [29, 31]. However, the initially X-ray amorphous alloys of group B do
not show this separation but are characterized by an extremely long (400–800 K), continuous,
and unusually strong increase (about 50 % when extrapolating the reversible slope back to room
temperature!), which roughly looks like an amplified combination of a structural relaxation ef-
fect (like curve C below 600 K) with a further modulus increase like in poly-qc material (group
A) at higher temperature. Moreover, the magnitude of this huge irreversible modulus increase
seems to be quite reproducible, when compared e.g. to the scatter between the two curves meas-
ured for different ribbons of alloy A1. Not to be discussed in the present context is the final
modulus increase for A1 (lower curve) and B3 above 840 K, indicating further transformations
into crystalline phases as confirmed by TEM. Up to the onset of this effect in some of the Ti-
based alloys, only the icosahedral phase was observed.
A1: Ti53Zr27Ni20
A2: Ti41Zr42Ni17
B1: Ti45Zr38Ni17
B2: Ti51Zr27Ni20Ag2
B3: Ti48Zr27Ni20Ag5
C: Zr69.5Cu12Ni11Al7.5
Figure 3: Relative changes of Young’s modulus E (v f2) at first heating with 2 K/min (oscillation frequency f bet-
ween 250 and 750 Hz)
The damping Q–1 shows different characteristics because it is dominated by a much stronger
temperature dependence of its reversible component, behind which the irreversible changes are
hidden and only seen relative to a second temperature scan (Fig. 4). This comparison also shows
a remarkable difference between the ”B” and ”C” alloys: whereas in the former, Ti-based alloys
the transition from X-ray amorphous to quasicrystalline always means a decrease of damping
(Fig. 4a), quasicrystalline Zr69.5Cu12Ni11Al7.5 shows a higher damping than in the previous
amorphous state (Fig. 4b), which is still far from being understood [33].
Because of the strong reversible temperature dependence, the irreversible changes of damp-
ing during the transformation are better studied in isothermal experiments, which was done in
140
0.1
-1 a b
Q
0.01
1 0.01
"X-ray amorphous"
2
0.001 2 quasicrystalline
quasicrystalline 0.001 1
amorphous
300 400 500 600 700 800 900 300 400 500 600 700
T (K) T (K)
Figure 4: Temperature dependence of damping at f = 440 r 80 Hz, comparison between the first and the second
run of heating with 2 K/min: (a) Ti45Zr38Ni17, (b) Zr69.5Cu12Ni11Al7.5.
Fig. 5a for Ti51Zr27Ni20Ag2 at 733 K. The resulting changes of both E and Q–1 are widely linear
in log t and do not show any discontinuity at the apparent transition from ”amorphous” to ”qua-
sicrystalline” X-ray patterns in related synchrotron radiation experiments [34]. The isothermal
modulus increase, however, covers only a small fraction of the total increase, which in its major
part already occurred during heating-up (Fig. 5b).
Figure 5: (a) Isothermal changes of damping and resonance frequency (Young’s modulus) of Ti51Zr27Ni20Ag2 at
T = 733 K after (b) heating up with 4 K/min; the arrows and circles denote the following isothermal changes
4.3 Shape of the Snoek-Type Hydrogen Relaxation Peak
The general characteristics of the well-known hydrogen damping peak in amorphous metals
(see reviews [24, 27]) – a broadened, asymmetric peak shifting to lower temperatures with
increasing height and H concentration – were found recently also for icosahedral quasicrystals
[29–33], and are now confirmed for the case of nano-quasicrystalline structures (see the examp-
le in Fig. 6). Since the effect is so similar in amorphous and icosahedral structures, in particular
for Zr69.5Cu12Ni11Al7.5 [29, 31], the detection of finer differences e.g. during the gradual de-
velopment of quasicrystalline order requires a careful peak-shape analysis. Therefore, the peaks
measured under different conditions were re-plotted in Fig. 7 as a function of reciprocal tempe-
rature (which is more reasonable from relaxation physics [36]), normalized in height after sub-
141
traction of a linear background QB–1, and corrected for the shift in peak temperature in the case
of Fig. 7b. We can see that a significant narrowing of the peak occurs only in the ”later” coarse-
ning stages (and also there only at low H concentration) where the E increase in Fig. 3 is already
completed, whereas in the early stages at grain sizes up to 15 nm the peak shape practically does
not change during annealing. It should be noted that these annealing-induced changes of peak
shape are generally smaller than those observed earlier on the ”group A” alloy Ti53Zr27Ni20
[32, 33].
Figure 6: Growth of the hydrogen damping peak in slightly annealed (2 K/mino573 K) Ti51Zr27Ni20Ag2, after
stepwise charging at 1 bar H2 / 473 K for 0, 6, and 18 h; frequency ~ 350 Hz
Figure 7: Normalized plots (see text) of the H damping peak in Ti51Zr27Ni20Ag2 at different grain sizes, as sym-
bolized by the line thickness (see Table 1 for the respective heat treatments): (a) at low cH ~ 0.005 H/M, grain size
(1) 5–10 nm, (2) ~15 nm, (3) 20–30 nm, (4) 50–100 nm; (b) at varying cH: roughly 0.005 H/M (solid lines), 0.03
H/M (dashed), 0.2 H/M (dotted); grain sizes 5–10 nm and 20–30 nm.
142
5 Discussion and Conclusions
Combining mainly four parameters (grain size, Young’s modulus, intrinsic damping, width of the H
peak) with three groups of alloys (cf. Section 3.1), we can give here only a preliminary and simplified
interpretation of a few main points.
As seen most clearly for the X-ray amorphous ”group B” alloys, Young’s modulus E and the
H peak are obviously sensitive to quite different parts of a broad spectrum of ordering process-
es. The surprisingly strong, relative increase of E in Fig. 3 points to a highly perturbed and dis-
ordered initial state (as compared e.g. to bulk glasses, curve C); this disorder might be of a more
extended, delocalized type analogous to ”free volume” in metallic glasses, and possibly related
to the fraction of ”grain boundaries” between the nano-quasicrystals. On the other hand, the lo-
cal diffusion jumps underlying the hydrogen peak directly react to changes in local atomic con-
figurations. Taking the present and earlier results [29–31] together, it turns out that the a/qc
transition has only very little influence on this local order as detected by the H peak, but that in
this respect the quasicrystalline order sharpens much later. In other words, quasicrystals with
grain sizes below 100 nm, even when showing clear icosahedral diffraction patterns, are appar-
ently less perfect than those at larger grain size.
As concerns the two open questions mentioned in the Introduction, the present results tenta-
tively support the idea that amorphous structures can form a continuous transition to the quasic-
rystalline state, in the sense that there is no substantial difference between amorphous and nano-
quasicrystalline structures at the X-ray amorphous limit of fine grain size. But this is not neces-
sarily true for all amorphous structures. One can exclude neither the existence of different
amorphous structure types (e.g. homogeneous – heterogeneous/polycluster [14]), nor different
types of a/qc transitions (continuous – discontinuous; solid state transformation – qc formation
in the undercooled melt).
6 Acknowledgement
We wish to thank Mrs. Chr. Grusewski for the TEM investigations. Financial support by the
DFG priority programme ”Quasicrystals” (grant number Si 408/8-1) is gratefully acknowled-
ged.
7 References
[1] U. Köster, J. Meinhardt, S. Roos, H. Liebertz, Appl. Phys. Lett. 1996, 69, 179–181
[2] J. Eckert, N. Mattern, M. Zinkevitch, M. Seidel, Mater. Trans. JIM 1998, 39, 623–632
[3] L.Q. Xing, J. Eckert, W. Löser, L. Schultz, Appl. Phys. Lett. 1998, 73, 2110–2112
[4] A. Inoue, T. Zhang, J. Saida, M. Matsushita, M.W. Chen, T. Sakurai, Mater. Trans. JIM
1999, 40, 1181–1184
[5] J.K. Lee, G. Choi, D.H. Kim, W.T. Kim, Appl. Phys. Lett. 2000, 77, 978–980
[6] N. Wanderka, M.-P. Macht, M. Seidel, S. Mechler, K. Ståhl, J.Z. Jiang, Appl. Phys. Lett.
2000, 77, 3935–3937
[7] L.Q. Xing, J. Eckert, W. Löser, L. Schultz, Appl. Phys. Lett. 1999, 74, 664–666
143
[8] A. Inoue, T. Zhang, J. Saida, M. Matsushita, Mater. Trans. JIM 2000, 41, 1511–1520
[9] Y.C. Kim, J.H. Na, J.M. Park, D.H. Kim, J.K. Lee, W.T. Kim, Appl. Phys. Lett. 2003, 83,
3093–3095
[10] C. Janot, M. de Boissieu, Phys. Rev. Lett. 1994, 72, 1674–1677
[11] K. Hiraga, K. Sugiyama, T. Ohsuna, Phil. Mag. A 1998, 78, 1051–1064
[12] D. Gratias, F. Puyraimond, M. Quiquandon, A. Katz, Phys. Rev. B 2000, 63, 024202
[13] C. Janot, J.-M. Dubois in Quasicrystals, An Introduction to Structure, Physical Properties,
and Applications (Eds.: J.-B. Suck, M. Schreiber, P. Häussler), Springer Series in Materi-
als Science, Vol. 55, Springer, Berlin Heidelberg, 2002, Chapter 8, p. 183
[14] A.S. Bakai in Glassy Metals III (Eds.: H. Beck, H.-J. Güntherodt), Topics in Applied Phy-
sics, Vol. 72, Springer, Berlin Heidelberg, 1994, Chapter 6, p. 209
[15] P. Häussler, H. Nowak, R. Haberkern, Mater. Sci. Eng. A 2000, 294–296, 283–286
[16] P. Häussler, J. Barzola-Quiquia, J Non-Cryst. Solids 2002, 312–314, 498–501
[17] P. Häussler, J. Barzola-Quiquia, R. Haberkern, C. Madel, M. Lang, K. Khedhri, D. Dek-
ker in Quasicrystals, Structure and Physical Properties (Ed.: H.-R. Trebin), WILEY-
VCH, Weinheim, Germany, 2003, Chapter 3.5, p. 289
[18] K. Urban, N. Moser, H. Kronmüller, phys. stat. sol. (a) 1985, 91, 411–422
[19] J.L. Robertson, S.C. Moss, K.G. Kreider, Phys. Rev. Lett. 1988, 60, 2062-2065
[20] L.C. Chen, F. Spaepen, Nature 1988, 336, 366–368
[21] R. Nicula, U. Ponkratz, A. Jianu, C. Schick, E. Burkel, Mater. Sci. Eng. A 2000, 294–296,
90–92
[22] F. Vasiliu, A. Jianu, R. Nicula, E. Burkel, J. Metastable Nanocryst. Mater. 2000, 8 (also
Mater. Sci. Forum 2000, 343-346) 33–36
[23] P.M. Richards, Phys. Rev. B 1983, 27, 2059–2072
[24] H.-R. Sinning, Def. Diff. Forum 1995, 123–124, 1–30
[25] F. Winter, S. Adam, H.-R. Sinning, J. de Physique IV, Colloque C8, 1996, 55–58
[26] H.-R. Sinning, J. Alloys Comp. 1994, 211/212, 216–221
[27] H.-R. Sinning, G. Steckler, R. Scarfone, Def. Diff. Forum 1999, 167–168, 1–15
[28] H.-R. Sinning, J. Alloys Comp. 2000, 310, 224–228
[29] R. Scarfone, H.-R. Sinning, J. Alloys Comp. 2000, 310, 229–232
[30] H.-R. Sinning, R. Scarfone, J. Metastable Nanocryst. Mater. 2000, 8 (also Mater. Sci.
Forum 2000, 343-346) 738–743
[31] R. Scarfone, Dissertation (Ph.D. thesis), Techn. Univ. Braunschweig 2002, available at
http://www.biblio.tu-bs.de/ediss/data/20020410a/20020410a.html
[32] H.-R. Sinning, R. Scarfone, I.S. Golovin, Mater. Sci. Eng. A 2003, in press
[33] H.-R. Sinning, R. Scarfone, I.S. Golovin in Quasicrystals, Structure and Physical Proper-
ties (Ed.: H.-R. Trebin), WILEY-VCH, Weinheim, Germany, 2003, Chapter 5.6, p. 536
[34] A. Jianu, H.-R. Sinning, I.S. Golovin, R. Nicula, E. Burkel, this volume
[35] H.-R. Sinning, J. Phys.: Condensed Matter 1991, 3, 2005–2020
[36] A.S. Nowick, B.S. Berry, Anelastic Relaxation in Crystalline Solids, Academic Press,
New York, 1972
144
Solid-Solid Phase Transformation of Amorphous Titanium Based

Alloys
A. Jianu1, H.-R. Sinning2, I. S. Golovin2, E. Burkel3

1
National Institute of Materials Physics, Bucharest-Magurele, Romania
2
Institut für Werkstoffe, Technische Universität Braunschweig, Germany
3
Mathematisch-Naturwissenschaftliche Fakultät, Universität Rostock, Germany
1 Introduction
Ti-based alloys are an important class of engineering materials for which high strength and ther-
mal stability are essential. New amorphous multicomponent Ti-based alloys have been produ-
ced, during recent years, by melt casting or rapid solidification of the melt [1–5]. It is also
known that icosahedral quasicrystal formation occurred frequently in titanium-based alloys
[6–8] and an alloy system with stable icosahedral structure was also reported [9]. Amorphous
and icosahedral alloys are structurally related due to the presence of tetrahedral local configura-
tion of the atoms. Many amorphous alloys have been found to transform during primary crystal-
lization process into icosahedral phase [10–15]. A high degree of crystallization leads to
embrittlement, but if it can be arrested when the crystallites are nanometer size, the resulting
nano-quasicrystalline composites (materials with nanograins dispersed in an amorphous matrix)
actually have greater tensile strength that the original amorphous materials [16–18]. In order to
increase the nucleation site density and to reduce the growth of the quasicrystallites, some addi-
tional elements have been added. The growth of quasicrystals may be controlled by factors as
phason strain [6], for example, and in this respect it is known that Ag addition to Al-Cu-Fe ico-
sahedral phase enhances the phason disorder in the first coordination sphere [19–20].
The aim of this study is to analyse the effect of the gradual substitution of Ti by Ag in the Ti-
Zr-Ni rapidly solidified alloy. The ternary phase diagram of the Ti-Zr-Ni alloy system shows
narrow compositional ranges for the equilibrium icosahedral phase [21] centred around the
Ti53Zr27Ni20 composition. The structure of as-quenched Ti-Zr-Ni-Ag alloys and the phase evolu-
tion during thermal annealing were investigated using high-resolution powder diffraction, dif-
ferential scanning calorimetry (DSC), transmission electron microscopy (TEM), and in situ
high-pressure/high-temperature X-ray powder diffraction.
2 Experimental Details
Multicomponent alloys with (Ti1-xAgx)53Zr27Ni20 (x = 0, 2, 5, 7, 10, 15, 20 at.%) composition

were studied in the present research. Primary alloy ingots were obtained from high-purity ele-
ments by arc melting in water-cooled copper mould. The ingots were further induction-melted
in quartz crucibles and rapidly quenched onto a copper wheel (single-roller melt spinning tech-
nique), in protective gas Ar atmosphere, as described elsewhere in detail [22].
The structure of the as-quenched ribbons was characterized by high-resolution powder dif-
fraction using synchrotron radiation (beamline B2 - HASYLAB - HAMBURG). The morpholo-

ISBN: 3-527-31007-X
145
gy of the as-quenched and heat-treated ribbons was examined by transmission electron

microscopy.
Thermal analysis was performed in a differential scanning calorimeter. The measurements
were performed at a heating rate of 20 °C/min., under a purified argon flow.
In situ high-temperature non-isothermal and isothermal energy dispersive X-ray powder dif-
fraction experiments and in situ high-pressure/high-temperature non-isothermal X-ray powder
diffraction were performed in transmission mode at F2.1 beamline (HASYLAB - HAMBURG)
[23]. A cubic sample assembly (boron made, edge length 8 mm) was fixed or compressed by six
tungsten carbide anvils in a 250-ton hydraulic press. The heating was provided by the AC pow-
er source through a cylindrical graphite heater (1.5 mm inner diameter) via two opposite anvils.
Small pieces of alloy samples were loaded inside the heater into the BN container. The temper-
ature was measured by means of a thermocouple in contact with the sample. The scattered syn-
chrotron radiation was detected using a Ge solid state detector with an energy range of
2–80 keV, positioned at small diffraction angles (3.5° dTd4°).
The samples were heated in steps of 20 qC. After the temperature was stabilized, each dif-
fraction pattern was collected using an average exposure time of about 10 minutes.
A single-phase icosahedral sample was obtained for the Ti53Zr27Ni20 composition using a linear
wheel speed of 20 m/s. Cahn (N, M) indices [24] were used for indexing the icosahedral phase
diffraction lines (Fig. 1a). All icosahedral diffraction lines could be attributed to a P-type icosa-
hedral phase with the quasilattice parameter a = 5.059 r0.005 Å (hypercubic lattice parameter
A = 7.1545 Å). Fig. 1a shows also the simulated spectra of the b.c.c. crystalline approximant
(W-phase) [25] in order to demonstrate the single-phase icosahedral nature of the as-quenched
Ti53Zr27Ni20 alloy structure.
a b
Figure 1: X-Ray powder diffraction pattern (a) and microstructure morphology (b) of single-phase icsoahedral
structure of as-quenched Ti53Zr27Ni20 alloy
The microstructure of the as-quenched Ti53Zr27Ni20 sample consists of a compact aggregation

of large icosahedral grains with grain sizes between 2 and 6 Pm. The TEM micrographs reveal a
faceted morphology and particulate contrast in the form of radial striations, typical for icosahe-
dral phase microstructures (Fig.1b). The selected area electron diffraction (SAED) pattern, ob-
tained along the fivefold icosahedral direction, exhibits sharp diffraction spots.
146
While the Ti53Zr27Ni20 alloy ribbons with single-phase icosahedral structure exhibit high brit-
tleness, fracturing when bent by less than 40°, the substitution of a part of Ti with Ag changes
drastically the behaviour of the ribbons during the bending test. The rapid solidification of al-
loys with composition of (Ti1-xAgx)53Zr27Ni20 (x d 10 at.%) leads to the formation of high duc-
tile ribbons (which could be bent by 180° without fracture for many times), even if the linear
wheel speed is reduced down to only 9 m/s. The thickness of the ribbons showing high ductility
is in the 30–120 Pm range.
The X-ray analysis of the samples confirms the formation of an amorphous single-phase
(Fig. 2a). However, the bright-field micrograph and selected area electron diffraction patterns of
Ti51Ag2Zr27Ni20 shown in Fig. 2b, suggest that some nanograins (2–7 nm) with diffuse grain
boundaries still exist. For samples with 5, 7 and 10 at % silver content, the amorphous structure
was also confirmed by electron microscopy.
b
a
Figure 2: (a)X-ray powder diffraction patterns of (Ti1-xAgx)53Zr27Ni20 alloys (x = 2, 5, 7, 10) in the as-quenched
state (b) and bright-field micrograph and selected area electron diffraction pattern of Ti51Ag2Zr27Ni20 rapidly solidi-
fied alloy
For as-quenched samples with x = 15 and 20 at.% Ag, additional diffraction lines belonging
to a face-centered-cubic Ag single-phase (lattice constant a = 4,086 Å) could also be identified
(Fig. 3a). The corresponding TEM micrograph of the Ti33Zr27Ni20Ag20 alloy is shown in Fig. 3b.
For both Ag-rich samples, rapid solidification leads to the formation of Ag/TiZrNiAg in-situ na-
nocomposites consisting of isolated quasi-spherical Ag nanoparticles with an average diameter
of about 30 nm embedded in an amorphous matrix. The volume and size distribution of Ag na-
nograins (30 nm) appears to be more uniform in the 20 at.% Ag sample (Fig. 3b). The presence
of Ag nanograins indicates the existence of a solubility limit of Ag in the amorphous matrix in
the range of 10-15 at.%.
The alloys with 15 and 20 at.% have reasonably good bending ductility (> 90° before frac-
ture).
Fig. 4a shows the DSC traces of the as-quenched (Ti1-xAgx)53Zr27Ni20 alloys. The alloys with
2 and 5 at.% of silver exhibit a first very broad exothermic peak in the temperature range of
510–610 °C and 500–582 °C, respectively. In the interval where the DSC analysis has been per-
147
a b
Figure 3: (a) X-ray powder diffraction pattern and (b) TEM micrograph of Ti33Ag20Zr27Ni20 rapidly solidified
alloy
a
b c
Figure 4: (a) DSC curves of melt-spun Ti-Zr-Ni-Ag alloys and (b, c) in situ energy dispersive x-ray powder dif-
fraction patterns recorded for Ti51Ag2Zr27Ni20 and Ti48Ag5Zr27Ni20 amorphous alloys under various temperatures
formed, the alloy with a content of 5 at.% Ag shows a second exothermic peak. The onset and
peak temperature of this second peak are 608 and 645 °C, respectively.
With the increase of silver content, two well-defined exothermic peaks are seen for each
Ti46Ag7Zr27Ni20 and Ti43Ag10Zr27Ni20 alloys, in the temperature range of 516–630 °C. The onset
and the peak temperatures decrease when the silver content increases from 7 at.% to 10 at.%.
The alloy with 15 at.% Ag shows a major exothermic peak, which is formed by the three super-
imposed peaks, while for 20 at.%, two exothermic events take place in the range of 500–580 °C.
The phase transformations of Ti-Zr-Ni-Ag alloys during heating were further analysed by in
situ energy dispersive X-ray powder diffraction experiments using synchrotron radiation.
Fig. 4 b, c exemplifies the diffraction patterns obtained for alloys with 2 and 5 at.% Ag content,
during the heating process. A broad amorphous peak, located at Q |2.5 Å–1, together with a few
148
Bragg peaks from BN and NaCl, are observed in the diffraction patterns recorded at room tem-
perature.
The phase transformation process during the heating procedure is quite similar for the alloys
with 2 and 5 at.% Ag. However, the change in the patterns shape of the Ti48Ag5Zr27Ni20 alloy
first occurs at higher temperature than for the alloy with 2 at.% Ag, namely at 520 °C instead of
440 °C.
No intermetallic crystalline compounds are detected, indicating that only an amorphous to
quasicrystal transition occurs in the temperature range of 440–580 °C. The icosahedral phase,
which forms in the two alloys, is metastable and further transforms into a b.c.c crystalline ap-
proximant, with the lattice constant a0 = 14,317 Å [25].
In situ high-pressure/high-temperature non-isothermal X-ray powder diffraction measure-
ments of Ti51Ag2Zr27Ni20 alloy were performed in transmission mode at a pressure of 2.5 GPa
(Fig. 5a).
Figure 5: (a)In situ energy dispersive x-ray powder diffraction patterns recorded for Ti51Ag2Zr27Ni20 rapidly soli-
dified alloy under 2,5 GPa (b) and in situ energy dispersive X-ray powder diffraction patterns recorded for the
same alloy at 460 °C under isothermal conditions
For this alloy with some nanograin structure still visible (Fig. 2a), the change in shape is de-
tected in the pattern recorded at 500 °C. Two new Bragg peaks appear and can be indexed to
(18, 29) and (20, 32) peaks for a primitive icosahedral structure. No intermetallic crystalline
compounds are detected in the temperature range of 440–650 °C. The temperature enhancement
of the amorphous-to-quasicrystalline phase transformation with pressure, observed in this alloy,
indicates that atomic mobility is involved in the formation and growth of the icosahedral phase
from amorphous structure.
In order to monitor the amorphous-to-quasicrystalline phase transformation of the
Ti51Ag2Zr27Ni20 amorphous alloy, in situ isothermal energy-dispersive X-ray powder diffraction
experiments were carried out at ambient pressure and 460 °C. The quasicrystalline phase was
detected after 320 min. by the change of the pattern shape indicating a slow process of transfor-
149
mation for this temperature (Fig. 5b). With increasing time, the icosahedral peaks increase in in-
tensity, proving that the volume fraction of the quasicrystalline phase increases. During the
isothermal annealing, a discontinuity of the growing process was observed after 450 min. It was
not confirmed by the isothermal changes of the damping and resonance frequency (Young’s
modulus), performed at 460 °C in related mechanical spectroscopy experiments [26]. No inter-
metallic compounds were detected from diffraction patterns recorded at this temperature and
annealing time, showing that only an amorphous-to-quasicristalline phase transformation oc-
curs.
Heat treatment in vacuum, followed by TEM investigation, was applied to alloys with 2 and
5 at.% Ag. The microstructure of the samples, annealed at 480 °C, is homogeneous and consists
of nanoquasicrystals with 20–30 nm grain sizes [26].
When the annealing temperature was increased up to 610 °C, the grains size of the icosahe-
dral structure of the Ti51Ag2Zr27Ni20 alloy increases up to about 100 nm [26], while the micro-
structure of the alloy with 5 at.% Ag consists mainly of grains of b.c.c.-approximant phase and
minor icosahedral phase (Fig. 6).
Figure 6: The microstructure morphology of thermal annealed Ti48Ag5 Zr27Ni20at 610 °C
The bright field electron micrograph shows traces of icosahedral grains and the lamellar
structure of the crystalline approximant (with a radial orientation similar to the radial striations
of the icosahedral phase) grown on former icosahedral sites. The selected area diffraction in
Fig. 6 shows a twofold I-phase diffraction pattern overlapped with a [001] b.c.c. pattern.
Fig. 7 shows diffraction patterns of Ti46Zr27Ni20Ag7 and Ti33Zr27Ni20Ag20, recorded during in
situ high-temperature non-isothermal energy dispersive X-ray powder diffraction experiments.
The structure changes with temperature are quite similar for alloys containing 7, 10, 15 and 20
at.% Ag. The first structural transition is from amorphous to a mixture of C14 Laves hexagonal
and b.c.c. crystalline approximant phase. More investigations by electron microscopy and x-ray
diffraction, on annealed samples of this group, are under way in order to clarify the structure of
the crystalline phases formed during the second stage of the structural transformation.
150
a b
Figure 7: In situ energy dispersive x-ray powder diffraction patterns recorded for Ti46Ag7Zr27Ni20 and
Ti33Ag20Zr27Ni20 amorphous alloys under various temperatures
4 Conclusions
The quaternary (Ti1-xAgx)53Zr27Ni20 amorphous alloys were produced by the melt-spinning tech-
nique. The composition of these alloys was obtained by the substitution of Ti from the parent
Ti53Zr27Ni20 alloy (which forms during rapid solidification a single-phase icosahedral structure)
with Ag, element with positive heat of mixing against Ni. A slight substitution of Ti by Ag
(x=2) causes a significant structural disorder of the icosahedral phase, leading to an amorphous
phase.
On heating, the (Ti1-xAgx)53Zr27Ni20 amorphous alloys suffer a two-stage transformation.
Metastable icosahedral phase with nanoscale grain dimensions has been obtained during pri-
mary crystallization of amorphous alloys with 2 and 5 at.% Ag. The formation of a large frac-
tion of nanoscale icosahedral phase through a continuous transition with a reduced exothermal
effect suggests a high nucleation rate and a relationship between the icosahedral structure and
the local atomic configuration of the Ti-based amorphous alloy.
For higher Ag contents, the amorphous alloys transform into a mixture of C14 Laves hexag-
onal and b.c.c crystalline approximant. Both these structures are related to the icosahedral and
amorphous phase through the local atomic configuration built with tetrahedral units.
151
5 References
[1] A. Inoue, N. Nishiyama, K. Amiya, T. Zhang, T. Masumoto, Mater. Lett. 1994, 19,
131–135
[2] K. Amiya, N. Nishiyama, A. Inoue, T. Masumoto, Mater. Sci. Eng. 1994, A179/180,
692–696
[3] T. Zhang, A. Inoue, T. Masumoto, Mater. Sci. Eng. 1994 A181/182, 1423–1426
[4] T. Zhang, A. Inoue, Mater. Trans. 1998, JIM 39, 1001–1005
[5] D. Louzguine, A. Inoue, J. Mater. Res. 1999, 14, 11, 4426–4430
[6] K.F. Kelton, Int. Mater. Rev. 1993, 38, 105–137
[7] J.L. Libbert, K.F. Kelton, A.I. Goldman, W.B. Yelon, Phys. Rev. 1994, B49,
11675–11681
[8] W.J. Kim, K.F. Kelton, Philos. Mag. Lett. 1995, A72, 1397–1408
[9] R.M. Stroud, K.F. Kelton, S.T. Misture, J. Mater. Res. 1997, 12, 434–438
[10] U. Köster, J. Meihardt, S. Roos, A. Rüdiger, Mater. Sci. Forum 1996, 225, 311–316
[11] J. Eckert, N. Mattern, M. Zinkevitch, M Seidel, Mater. Trans. 1998, JIM 39, 623–632
[12] B.S. Murty, D.H. Ping, K. Hono, A. Inoue, Appl. Phys. Lett. 2000, 77, 1102–1104
[13] A. Inoue, T. Zhang, M.W. Chen, T. Sakurai, J. Saida, M. Matsushita 2000, Apply. Phys.
Lett, 76, 967–969
[14] L.Q. Xing, J. Eckert, W. Löser, L Schultz 1999, Apply. Phys. Lett. 74, 664–666
[15] B.S. Murty, D.H. Ping, K. Hono, A. Inoue, Scr. Mater. 2000, 43, 103–107
[16] H. Chen, Y. He, G.J. Shiflet, S.J. Poon, Nature, 1994, 367, 541–543
[17] C. Fan, A. Takeuki, A. Inoe, Mater. Trans. 1999, JIM40, 42–51
[18] C. Fan, D.V. Louzguine, C. Li, A. Inoue, Appl. Phys. Lett. 1999, 75, 340–342
[19] R. Manaila, D. Macovei, A. Jianu, R. Popescu, R. Nicula, F. Zavaliche, A. Deveny,
Y. Xie, Phys. Stat. Sol. 1996, B 193, 283–294
[20] R. Popescu, D. Macovei, M. Manciu, F. Zavaliche, D. Fratiloiu, A. Jianu, A. Devenyi, R.
Manaila, Y. Xie, T. Hu, B.R. Orton, R.J. Cernik, C.C. Tang, J. of Phys.: Condens. Matter.
1997, 9, 7523–7528
[21] J. P. Davis, E. H. Majzoub, J. M. Simmons and K. F. Kelton, Mat. Sci. Eng. 2000,
A294–296, 104–107
[22] F. Vasiliu, A. Jianu, R. Nicula, E. Burkel, Mater. Sci. Forum 2000, 343-3, 33–36
[23] P. Zinn, J. Lauterjung, R. Wirth, Z. für Kristallographie 1997, 212, 691–698
[24] J.W. Cahn, D. Shechtman, D.J. Gratias, J. Mater. Res. 1986, 1, 13–26
[25] W.J. Kim, P.C. Gibbons, K.F. Kelton, W.B. Yelon, Phys. Rev. 1998, B 58, 2578–2585
[26] H.-R. Sinning, I. S. Golovin A. Jianu, this volume
39
IV Recrystallization and Grain Size Control

ISBN: 3-527-31007-X
155
Energetics of Three-Dimensional Network Structures
M. E. Glicksman
Rensselaer Polytechnic Institute, Troy, N.Y., USA
1 Abstract
Integral geometry and topology is applied to the problem of space-filling in isotropic network
structures, such as polycrystalline solids and foams. The theory developed is based on represen-
ting network elements, such as grains and bubbles, as average N-hedra, where N equals the
number of contacting neighbors in the network. Average N-hedra satisfy Young-Laplace ther-
modynamic equilibrium at the triple lines (edges) and quadra-junctions (vertices) and act as
”proxiesîfor irregular cells with equivalent topology. This analysis yields the metric, energetic,
and kinetics behavior for average N-hedra as a function of their topological class. This paper
concentrates on the metric and energetic properties of average N-hedra; the kinetic properties
will be discussed elsewhere. The new approach should prove useful for constructing more accu-
rate descriptions of three-dimensional microstructures. As shown here, the availability of analy-
tic relations for the curvatures, areas, volumes, and free energies, provides precise benchmarks
to test numerical simulations of the properties of average N-hedra, and to guide further quantita-
tive experiments on network energetics and kinetics in three-dimensional microstructures.
2 Introduction
The energetics and growth kinetics of space-filling (contiguous) grains remain important topics
within the context of microstructure evolution. The foundation for grain growth in two dimensi-
ons was established a half-century ago by C.S. Smith [1], J. von Neumann [2], and W.W. Mul-
lins [3]. Von Neumann proved in two dimensions (R2) that an isotropic array of contiguous cells
(grains or bubbles) with a uniform mobility, M, obeys the kinetic law da / dt = (SJM / 3) (n–6),
n t2, where n denotes the number of sides or vertices of a two-dimensional grain or bubble, and
a is its area. The von Neumann-Mullins relationship may be viewed as the kinetic growth law
for isotropic two dimensional systems, and provides an excellent approximation for the beha-
vior of grains forming polycrystalline networks in thin films, or, in general, where one spatial
dimension is suppressed relative to the other two [4–7]. The von Neumann-Mullins relati-
onship, called the ”n–6î rule, was known empirically by experimentalists [8, 9] even before its
precise mathematical form was derived by von Neumann.
The Helmholz free energy of an isotropic two-dimensional assembly of grains of unit thickness,
F2, is given by
J gb
E2 ¦ Pi (1)
2 i

ISBN: 3-527-31007-X
156
where Jgb is the specific grain boundary free energy, and Pi is the total perimeter of the im grain.
The factor 1/2 adjusts for the fact that the summation in Eq.(1) counts every grain boundary
twice. Graner et al. [10] showed that the total perimeter of a two-dimensional polycrystal may
be written conveniently in terms of the total grain area, namely that
J gb J gb
E2
2
¦ e(ni ) Ai |
2
e(6)¦ Ai (2)
i i
where the coefficients e( ni ) { Pi Ai form a discrete function that varies smoothly and weakly
with the number of sides, n. Specifically, for regular two-dimensional grains with curved sides
meeting at 120° e(2) | 3.78 decreasing steadily to e(f) | 3.71. Recently, Vaz et al. [11] con-fir-
med that Eq.(2) provides an accurate energy estimate for large isotropic grain or bubble arrays
in R2.
The situation regarding the energetics and kinetics of three dimensional polycrystals remains
under analysis [12–17], numerical simulation [18–23] and experiment [24, 25] that continue
vigorously until the present. This paper reports on an analysis of idealized ”average N-hedraî
that may be used to estimate the excess free energy of an isotropic polycrystal in R3, in direct
analogy to Eq.(2) for the free energy of two-dimensional grains as used by Graner et al.
3 Geometry of Interfaces
Quantitative discussion of the evolution or shortening of curves in R2 [2, 3, 26], and the more
general problem of surface evolution in R3 [27, 28], requires a mathematical foundation based
on differential geometry. Consider the small area on a differentiable interface, or surface, for-
mally termed a coordinate patch. The geometry of a smooth surface – one that is bounded and
exhibits well-defined derivatives – may be specified through the local shape tensor [29, 30]
N
ª
« 1
0 º»
N ij « »
(3)
«¬ 0 N 2 »¼
where N1 and N2 are the principal curvatures on the coordinate patch. Two independent scalar in-
variants may be derived using eq.(3) that define the mean curvature, H, as one-half the trace of
the shape matrix, [30],
1 ªN 0 º» 1
H { Tr «« 1 » (N 1 N 2 ) (4)
2 «¬ 0 N 2 »¼ 2
and the Gaussian curvature, K, as the determinant,
ªN1 0 º»
K { Det «« » N 1N 2 (5)
¬« 0 N 2 ¼»
These curvatures characterize the local geometry of an interface over any small, suffi-ciently
uniform patch. Note that the curvatures H and K have different units and, more impor-tantly,
157
represent independent attributes of the local geometry. Mean curvature, H, relates changes of
area, A, to the volume, V, swept by the surface or boundary moving through its embedding
space, R3. Specifically, H equals the coefficient connecting area and volume, and is defined as
1GA (6)
H
2 GV
The Gaussian curvature relates to the topological, i.e., space-filling, properties of a sur-face.
Specifically, K relates area change to changes in the spherical image, or solid angle, :. The to-
tal Gaussian curvature of any simply-connected object in R3 equals 4S, which corre-sponds to
mapping the the surface normals enclosing a body onto the unit ball of orientation. Thus, at any
point on a surface,
d:
K (7)
dA
It may be shown from the definitions of H and K that for equicurved spherical interfaces
K=H. (8)
4 Average N-hedra
4.1 General
In this brief paper we present the concept of „average N-hedraî“ as mathematical objects in R3
with N identical faces, capable of contacting N neighbors in a network. In short, average
N-hedra represent a geometrical extension to all integers 2 d N d f of the geometrical and topo-
logical properties of constructible uniformly curved regular polyhedra. Average N-hedra have N
identical faces and 3 (N – 2) identical edges that meet three at a time at 2 (N – 2) identical trihe-
dral vertices. Only five constructible examples of average N-hedra exist [22]: viz., the cases
N = 2, 3, 4, 6 and 12. It is for these special N-values only that the number of sides per face,
<p> = 6 – 12 / N equals an integer. All other lN-hedra have non-integer <p>-values, making
them non-constructible abstractions. Nevertheless, as will be shown, the average N-hedra can
serve as ”proxiesî for real polyhedral grains within the same topological class.Grains in real po-
lycrystals are generally irregular polyhedra with curved faces that exhibit variable shapes (trian-
gles, quadrilaterals, pentagons, hexagons, etc.) with sides of variable length.The key
correspondence between irregular network polyhedral cells and the average N-hedra is that the
two have identical topological (space filling) properties, but the latter are easily analyzed regar-
ding their geometric, energetic, and kinetic properties.
158
Figure 1: Isolated N-hedra. Sketches of the five constructible average N-hedra. Top row, left, center and right:
N = 2, 3 and 4. Middle row, left and center: N = 6, and 12. Excepting N = 2, which lacks vertices and may not be
incorporated in a network as a grain shape, the cases N = 4, 6 and 12 are each based on ”primitiveî (flat faced) Pla-
tonic solids with same number of faces (tetrahedron, cube, and dodecahedron, respectively). The trihedron, N = 3,
with two vertices, does not have a corresponding Platonic solid. The sixth object, middle row, right, is the Kelvin
tetrakaidecahedron, with N = 14. This polyhedron is not an average N-hedron as the faces are not identical, consi-
sting of squares and hexagons. The three objects shown on the bottom row are constructible hedra (none are aver-
age N-hedra) and were simulated by Cox and Fortes. Note that the faces for N d 13 bulge outward, so H t 0,
whereas the faces for N t 14 are concave, so H d . Provided by [32].
5 Space Filling in R3
5.1 Gauss-Bonnet Theorem
All bounded differentiable surfaces in R3, obey a fundamental topological law of filling space
known as the Gauss-Bonnet theorem [29,31]. The Gauss-Bonnet theorem provides our starting
point for the description of space-filling network cells bounded by uniformly curved surfaces
meeting at piecewise continuous edges in R3. Unit normals completely covering any isolated
159
N-hedron, if gathered and placed at the center of the unit ball, will cover its surface. The trans-
&
formation is described as n(R 3 ) o S (R 2 ) , which implies that normals covering the surfaces of
a simply connected closed body in R3 are mapped to the two-dimensional spherical manifold, S.
This mapping of the normals provides the topological space-filling requirement that the total
spherical image in R3 is : tot = 4S.
5.2 Contributions to : tot
:tot ³³ Ki dA 3( N 2):1 2( N 2):0 4S (9)
Note that the three contributions appearing in the middle of Eqs.(1) consist of 1) the total (inte-
gral) Gaussian curvature, 2) the images of the curved edges that each turn through a fixed planar
angle Z = : 1, and 3) the fixed image contribution, : 0 from each of the equilibrium vertices.
The contributions to the spherical image from the curved symmetrical edges on aver-age
N-hedra is proportional to the planar turning angle between adjacent vertices, which may be de-
termined by elementary methods from the symmetry properties of the average N-hedra as
§ D S · § 1·
Z S 2arctan ¨ sin tan ¸ 2arccos ¨ ¸ (10)
© 2 p !¹ © 3¹
where D is the exterior angle between the face normals located at the geometric center of adja-
cent faces. This angle may be shown to be the following smooth discrete function of N,
§ S · S (2 N 3)
D 4arctan 1 2sec ¨ ¸ cos (11)
© 2( N 2) ¹ 6( N 2)
Equilibria at triple lines is satisfied if adjacent faces intersect with an internal dihedral angle
of 120°, and correspondingly, equilibria at vertices occur if the normals on the three faces meet-
ing at the vertices differ in their orientationby exactly 60°. The equilibria at triple lines and ver-
tices, specified through the ”Plateau rules,î also guarantees that the quadra-junction angle where
three triple lines lying on a polyhedral grain meet a fourth to form a network node is the so-
called tetrahedral angle, arcos(–1/3) |109.47...° [33]. Finally, it may be shown that each equi-
librium vertex on any isotropic network polyhedron contributes a fixed amount, : 0, to the
spherical image given by
1
: 0 2S 3arccos( ) 0551287… (12)
3
160
Figure 2: The spherical image, : 0, produced by the triad of unit normals at an equilibrium vertex obeying the
Plateau rules for a 3-D network. The Plateau rules specify that the three normal vectors at an equilibrium vertex
make an angle of 60° with respect to each other. The area subtended on the unit ball of orientation is the con-stant,
: 0 = 0.551287… . Thus, the contribution to the total spherical image, : tot = 4S, for all the vertices on an average
N-hedron is equal to 2 (N - 2) 0.551287…
6 Metric Properties
6.1 Curvature
The metrical properties of average N-hedra may be specified by a single length scale, the vertex
separation, O. For comparative purposes, we choose O = 1. The radius of curvature of the faces,
Rc = H–1, may be shown by elementary methods to be
§D S ·ª S Z D Z Dº
Rc 2sin ¨ ¸ «cot csc sin cos sin » (13)
© 2 6 ¹¬ p ! 2 2 2 2¼
6.2 Areas
The total face areas of the average N-hedra (O = 1) represent an important geometric property,
as the area of a network cell is proportional to its work of formation, or Helmholz free energy.
The total face area, A(N), maybe found as the fraction, G(N), of the area of a full sphere of radi-
us Rc, thus
4S 3
A( N ) G ( N ) Rc (14)
3
It is easy to show using integral geometry that the fraction, G, of the area of a sphere, Asph, of
radius Rc = H–1, which equals the total face area of an average N-hedron is given by
1
G( N ) ³³ KdA (15)
4S
161
because for spheres, Eq.(8) holds and K / H2 = 1, so
Asph Rc2 ³³ KdA (16)
Equation(9) may be solved exactly for the total integral of the Gaussian curvature, which
when substituted into Eq.(15) and then Eq.(14) yields with help of Eq.(13) the exact face areas
for the average N-hedra.
6.3 Volumes
The volume, V, of any average N-hedron may be found by formulating an integral using spheri-
cal coordinates, the origin for which is located at the centroid of the polyhedron. This procedure
leads to the formula
N 2 3
V Rc F ( N ) (17)
12
where the function F (N) is given by
3 8S N 1 § 2 S · § 2 S ·
F (N ) 22 57 arccos 33arcsin ¨ cos ¸ tanarcsin ¨ cos ¸ (18)
N 2 3 © 3 p ! ¹ © 3 p !¹
7 Energetics
The Helmholz free energy of an isotropic polycrystal in three dimensions is given by
J gb
E3
2
¦ Ai ( N ) (19)
i
Recently, Cox and Fortes [22] suggested and showed by computer simulations that an analo-
gous situation for estimating the free energy of an isotropic polycrystal exists in R3. Specifical-
ly, Cox and Fortes showed that the analog of Eq.(2) in R3 is
J gb 2
E3
2
¦ ei ( N )Vi 3
(20)
i
where the function e(N) is defined as
A( N )
e( N ) { 2 (21)
V3
162
Equation (21) is easily evaluated employing the analytical expressions derived here for the
area, A(N), Eq.(14), and the volume, V, Eq.(17). Indeed, as shown in Figure (3), this quantity
does vary slowly with N, as suggested in [22]. Figure (3) also provides a comparison of the
present analytic results obtained from Eq.(13) with values reported by Cox and Fortes for sever-
al N-hedra using Brakke’s surface evolver [18,19] to evaluate the areas and volume. The values
for e(N) agree well with the simulated values, especially for those cases where the constructible
N-hedra correspond to average N-hedra (N = 3, 4, 6 and 12). For the three cases reported in [22]
where the constructible N-hedra are not average N-hedra, the simulations yield higher values.
Figure 3: Comparison of computer simulations [22] with the present analysis of the dimensionless quantity,
e(N) = A(N) / V–2/3. This scale-independent quantity varies slowly with N within a rather narrow range of values.
The square symbols are the simulations of four constructible average N-hedra that can participate in a polycrystal-
line network. The crosses represent constructible polyhedra and exhibit values less than 2
DeHoff [33] among others [34] has shown that the average number of faces per grain in an
isotropic polycrystal is <N> = 13.397, corresponding to the ”idealî flat-faced grain that satisfies
the Plateaurules. The value of e(N) | 5.254, so the total free energy of an isotropic polycrystal
may be expressed as
2
E3 | 263J gb ¦ Vi 3 (22)
i
163
7 Conclusions
1. Average N-hedra are used to represent the topological space-filling properties and geometry
of irregular polyhedral grains in a three-dimensional polycrystals.
2. The curvatures (both mean and Gaussian), total face areas, and volumes of were derived
exactly for all the average N-hedra using their symmetry properties.
3. As suggested recently by Cox and Fortes, the total energy of an isotropic polycrystal can be
expressed through the dimensionless ratio of the total face area to the two-thirds power of
the volume. This ratio was calculated from the exact expressionsderived here for the area
and volume, and compared to recently published computer simulations. Good agreement
was found.
4. Using the average number of faces per grain derived by deHoff, <N> = 13.397, the total
free energy of an isotropic polycrystal is found to be
2
E3 | 263J gb ¦ i Vi 3
8 Acknowledgement
The author is pleased to acknowledge the support provided to him from the Alexander von
Humboldt Stiftung as a Senior Research Preissträger at the Institut für Metallkunde und Metall-
physik, RWTH-Aachen, Germany.
9 References
[1] C. S. Smith, „Grain Shapes and other Metallurgical Applications of Topology“, Chapter
in Metal Interfaces, American Society for Metals, Cleveland, OH, 1952, p. 65
[2] J. von Neumann, in Metal Interfaces, written discussion, American Society for Metals,
Cleveland, OH, 1952, p. 108
[3] W. W. Mullins, J. Appl. Phys. 27, 1956, p. 900
[4] V. Fradkov, M. Palmer, J. Nordberg, M. E. Glicksman, and K. Rajan, Physica D 66, 1993,
p. 50
[5] V. E. Fradkov, M. Palmer, M. E. Glicksman, and K. Rajan, Acta Metall. Mater., 42, No.
8, 1994, p. 2719
[6] M. A. Palmer, V. E. Fradkov, M. E. Glicksman, and K. Rajan, Scripta Met. et Mat., 30,
1994, p. 633
[7] M. A. Palmer, M. E. Glicksman, K. Rajan, V. Fradkov, and J. Nordberg, Metall. and Mat.
Trans. A, 26A, 1995, p. 1061
[8] E. B. Matzke, Am. J. Botany, 33, 1946, p. 58
[9] E. B. Matzke, in Metal Interfaces, written discussion, American Society for Metals, Cle-
veland, OH, 1952, p.110
[10] F. Graner, Y. Jiang, E. Janiaud, and C. Flament, Phys. Rev. E., 63, 2001, p. 402
[11] M. F. Vaz, M.A. Fortes, F. Graner, Phil. Mag. Lett., 82, 2002, p. 575
[12] W. W. Mullins, J. Appl. Phys., 59, 1986, p. 1341
164
[13] W. W. Mullins,Acta Metall., 37, 1989, p. 2979

[14] D. Weaire and J. A. Glazier, Phil. Mag. Lett., 68, 1998, p. 363
[15] J. A. Glazier,Phys. Rev. Lett., 70, 1993, p. 2170
[16] C. Monnereau and M. Vignes-Adler, Phys. Rev. Lett., 80, 1998, p. 5228
[17] D. Wu, Private Communication, 2003
[18] K. Brakke, Exper. Math. 1, 1992, p. 141
[19] K. Brakke, (http://geom.umn.edu/software/evolver), (2002)
[20] C. Monnereau, N. Pittet, and D. Weaire, Europhys. Lett., 52, 2000, p. 361
[21] S. Hilgenfeldt, A. M. Kraynik, S. A. Koehler, and H. A. Stone, Phys. Rev. Lettr., 86,
2001, p. 2685
[22] S. J. Cox and M. A. Fortes, Phil. Mag. Lettr., 83, 2003, p. 28
[23] A. Rollett, Private Communication, 2003
[24] C. P. Gonatus, J. S. Leigh, A. G. Yodh, J. A. Glazier, and B. Prause, Phys. Rev. Lettr., 75,
1995, p. 573
[25] J. A. Glazier and B. Prause, in Foams, Emulsions and Their Applications, P. Zitha et al.
eds., MIT-Verlag, Bremen, 2000, p. 120
[26] G. Huisken, J. Differential Geometry,20, 1984, p. 237
[27] M. Goldberg, Tohoku Math. J., 40, 1934, p. 226
[28] Donald A. Drew, SIAM J. Appl. Math., 50, No.3, 1990, p.649
[29] D. Struik, Lectures on Classical Differential Geometry, Addison-Wesley, Reading, MA,
1950
[30] A. Gray, 2nd ed., pp. 373-380, CRC Press, Boca Raton, FL, 1997
[31] Martin M. Lipschutz, Differential Geometry, Schaum’s Outlines, McGraw-Hill, New
York, 1981
[32] S. J. Cox Private Communication, 2003
[33] R. T. DeHoff, Acta Metall. Mater., 42, no. 8, 1994, p. 2633
[34] H. S. M. Coxeter, M.S. Longuet-Higgins, and J.C.P. Miller, Phil. Trans. Roy. Soc. Lon-
don, Series A, 236, 1954, p.401
165
Tempering Peculiarities of Fe-C Alloy by Severe Plastic

Deformation
Y. Ivanisenko1, I. S. Golovin2, H.-R. Sinning2 and H.-J. Fecht1,3

1
Institut für Nanotechnology, Forschungszentrum Karlsruhe, 76021 Karlsruhe, Germany.
2
Institut für Werkstoffe, Technische Universität Braunschweig, 38106 Braunschweig, Germany.
3
Division of Materials, Ulm University, 89081 Ulm, Germany.
1 Introduction
Severe plastic deformation at the room temperature of carbon steels leads to formation of nano-
crystalline structure and total cementite dissolution [1–3]. The resulting Fe-C alloy displays et-
ching properties characteristic for martensite as well as very high hardness. However
estimations of the lattice parameter by the X-ray diffraction methods [3] have not confirmed the
conventional martensite formation: neither a change of the lattice parameter nor a splitting of
XRD peaks was observed. It was suggested that carbon atoms segregate to the dislocations and
grain boundaries [3]. Upon annealing, a strong interaction between C and dislocation cores re-
sults in retardation of H-carbide precipitation and stabilizes the nanocrystalline structure in fer-
rite [4]. However the question about the distribution of carbon atoms in ferrite matrix after
cementite decomposition is not quiet clear till now.
The internal friction (IF) method is known to be a useful tool to study distribution of C in
iron and steels (for general review see [5, 6]): one of the most well known IF effect, the carbon
Snoek peak, is due to carbon atom jumps in D-Fe [7]. The aim of this paper is to study redistri-
bution of carbon during tempering of Fe-C alloy obtained by severe plastic deformation.
2 Experimental
The samples of eutectoid carbon steel UIC 860 (0.6–0.8 wt.% C, 0.8–1.3 wt.% Mn, 0.1–0.5
wt.% Si, 0.04 wt.% P(max), 0.04 wt.% S(max), Fe-balance) 10 mm in diameter and 0.3 mm in
thickness were subjected to High Pressure Torsion (HPT) deformation as described in [3]. Sam-
ples were rotated to N = 5 turns. The microstructure of the HPT-processed and annealed at
623 K samples was investigated by means of Leo 1550 scanning electron microscope (FEG-
SEM) operating at 2 kV. The surface of the sample was polished and etched with a 5 % solution
of HNO3 in ethanol (Nital). The lattice parameter of ferrite was determined using the Debye-
Sherrer method with Nelson-Riley extrapolation [8]. TEM investigations were carried out using
the Philips CM20 electron microscope at an accelerating voltage of 200 kV. The microhardness
of the samples was measured with a load of 200 g using a Buehler MMT 7 device. At least 10
indentations for each sample were made.
A standard vibrating-reed technique [1] with electrostatic excitation and high frequency (40
MHz) detection circuit was used to measure the internal friction (IF) Q–1 and resonance frequen-
cy f of flexural vibrations during heating with 1 to 3 K/min, at temperatures between 90 and
900 K with vacuum 10–5 mbar and better. Several heating runs with subsequent increasing of

ISBN: 3-527-31007-X
166
upper temperature limit were made for each specimen. The damping measurements were done
by counting the number of freely decaying vibrations between two pre-selected amplitude lev-
els. The specimens for IF investigations with dimensions about 8 u2 u0.2 mm were cut out of
the HPT-processed steel washer as shown on Fig. 1. and placed in cantilever beam configura-
tion by clamping at one end.
Figure 1: A sketch showing how the specimen for IF experiments was cut out of the HPT-processed washer.
Approximate arrangement of zones with nanocrystalline and cellular structure are shown. D – degree of cementite
dissolution. See text for details.
3 Results
3.1 Characterisation of the as-Processed Specimen
In the present experiments samples have been obtained by HPT straining. The shear strain J in
the certain point of the sample is a function of the distance of this point from the sample
center R:
2S NR
J (1)
h
where N – is a number of turns and h – is a thickness of the sample. For that reason the structure
of the specimen for IF experiments is different on its edges and in the middle. At the distance
more than 3 mm from the center it consists of fibres parallel to shear planes (Fig. 2). During the
etching procedure these fibres are resistant to etching and remain shiny, whereas the structure of
deformed pearlite in the inner areas of the specimen become dark and display the cementite la-
mellas by etching (Fig. 2). The etching properties of the shiny regions seem to be analogous to
the martensite obtained by quenching of carbon steels [2]. At high magnification shown on the
SEM image (Fig. 3) one can see that the fibrous regions have a very smooth homogenous sur-
face. No cementite lamellae are seen. TEM investigation has shown that the microstructure of
the fibres is homogeneous nanocrystalline with a mean grain size of 10 nm (Fig. 4a). The SAD
167
Figure 2: Macrostructure of the HPT-processed spe- Figure 3: SEM observation of the microstructure of
cimen: the boundary between non-etching fibrous fibrous regions of HPT-processed specimen
structure and pearlite is shown. Optical microscopy.1
patterns from areas of 0.5 Pm2 (Fig. 4a, insert) show Debye-Sherrer circles characteristic of po-
lycrystalline structures with very small crystallite sizes. The diffraction patterns do not display
the cementite spots (112), (120), which are normally found within the (110)D circle. The
microstructure of the inner part of the specimen is cellular type as observed by TEM (Fig. 4b).
These observations are consistent with our previous investigation of the structure of the studied
steel subjected to the high pressure torsion deformation in a range of shear strains [3].
The relative amount of the remaining cementite phase was determined along a chord of the
specimen, i.e. along length of the IF specimen by the saturation magnetization method using the
amplitude of the MS drop at the cementite Curie temperature as described in [3]. It was found
Figure 4: Bright field TEM images of the microstructure of the HPT-processed specimen: (a) nanocrystalline
structure on the periphery1 and (b) cellular structure in the inner areas. Note that the rests of cementite lamellae
still can be observed
1
Reprinted from [3] with permission of Elsevier
168
that degree of cementite dissolution changes from 100 % at the ends of the specimen to 50 % in
its middle part. Coercive force is decreased due to HPT deformation from 11.4 Ö in pearlitic
specimen to 5.9 Ö in HPT specimen.
XRD pattern of the as-processed specimen displays broadened ferrite peaks, no splitting of
XRD reflexes due to tetragonal martensite formation is observed. The estimated lattice parame-
ter of ferrite is 2.870 r0.007 Å. This value corresponds within the experimental error limits to
the ASTM data a = 2.8664 Å for coarse grained non-deformed ferrite.
The hardness of the HPT-processed specimen as measured on the periphery of the specimen
(HV = 9.4 GPa) is higher than that is in the inner areas (6.7 GPa).
3.2 Tempering Behaviour
3.2.1 Thermomagnetic Experiment
Fig. 5 shows the thermomagnetic curves of the as-processed specimen recorded separately for
its periphery (Fig. 5a) and inner part (Fig. 5b). In the temperature range of 293–623 K monoto-
nous decrease of magnetisation saturation (MS) typical of ferrite is observed for the periphery of
the specimen, indicating that no notable carbides precipitation occurs in this temperature range.
A sharp irreversible drop of MS in the temperature range of 623–673 K (Fig. 4a, AB) corre-
sponds to the precipitation of paramagnetic at these temperature interstitial carbides and cemen-
tite [11]. The analysis of the curve for the middle part (Fig. 4b) is more difficult. Due to
remaining cementite the weak fluctuation of the saturation magnetization at the cementite Curie
temperature may overlap with possible reactions caused by the carbides precipitations. This
overlapping possibly leads to the smooth decrease of MS curve without distinct steps.
Figure 5: Saturation magnetization as a Figure 6: Correlation between hardness and tempering

function of temperature of the HPT-processed temperature for HPT processed specimen: (a) periphery
specimen for: (a) periphery of the specimen of the specimen and (b) for the inner parts
and (b) for the inner parts
169
3.2.2 Etching Behaviour
The fibrous macrostructure at the edges of the specimen which exhibits a high resistance to et-
ching in as-processed state still persisting up to annealing at 473 K for 1 hour.
3.2.3 Microhardness
The high value of hardness of Fe-C alloy produced by HPT is preserved up to a temperature
of 623K and then decreases monotonously (Fig. 6a,b).
3.2.4 Microstructure Evolution
The structure of the HPT specimen was studied in its inner part after annealing at 623 K, 1 h.
(Fig. 7a). It is seen that the structure is characterised by a more homogeneous diffraction con-
trast as compared to the as-processed state, indicating a decrease of the internal stresses.
Some relaxed grains with narrow boundaries having a mean size of about 100–150 nm can be
seen (Fig. 7a). Arrays of very fine carbides appear inside the ferrite grains; at high magnificati-
ons one can see that the crystallite lattice of precipitates is coherent to that of iron. Dark contrast
of the images of carbides is due to coherency strains. The temperature of 623 K corresponds to
the interstitial iron carbides T' precipitation in tempering of martensite [12, 13].
Figure 7: Microstructure of the inner area of the HP specimen after annealing at 623 K, 1 h. (a) Bright field image
and diffraction pattern (onsert); (b) HRTEM image of the precipitated carbide particle.
3.2.5 Internal Friction
We have measured the Fe-0.8 wt.%C specimen in two different structural states: first state is in-
itial pearlitic specimen (before HPT) after hot rolling at 1120 Ʉ followed by air cooling, and se-
cond state – after HPT deformation by 5 turns (see eq.(1)). The temperature dependencies of IF
(TDIF) for these specimens are presented in Fig. 8. For comparison reason we also have measu-
170
red water quenched Fe-0.01 wt.%C specimen, i.e. practically ferritic specimen. Temperature
dependencies of resonance frequency are shown in Fig. 8 for the HPT specimen only.
A few IF peaks can be distinguished in our measurements. Well defined carbon Snoek peak
is doubtless seen around 380 Ʉ at TDIF curve of quenched Fe-0.01 wt.%C specimen (at fre-
quency 544 Hz). This peak is denoted as the S-peak. The peak is accompanied by typical for a
relaxation peak decrease in modulus (E ~ f 2) sometimes called as Kronig-Kramers relation be-
tween Q–1 and f 2 (shown in the insert of Fig. 7). Two IF peaks denoted as peak P1 and P2 are
observed in the ”as received”, i.e. thermo-mechanical treated pearlitic specimen: P1 at about
293K (f |760 Hz) and P2 at about 600 K (f |720 Hz). The P1 peak is accompanied by Kronig-
Kramers predicted decrease in the resonance frequency (not shown in the Figure 8).
This effect is either much smaller or it is absent in the case of the P2 peak. After 30 min an-
nealing at 993 K (followed by water cooling in quartz ampoule), which is close but still below
to the temperature of transformation to austenite, no pronounced IF peaks were recorded at
TDIF curve, while the damping background decreases.
The temperature dependent Q–1 and f curves both of the HPT-processed specimen are shown
in Fig. 8. We increased step by step the upper limit of the measurements temperature to produce
step by step annealing of the specimen, only selected TDIF curves are shown in the Figure. The
P2-type peak is clearly seen in the ”fresh” HPT specimen, very small P1-type peak can be also
distinguished. Both P2 and P1 peaks in the HTP specimen decrease due to annealing similarly
to pearlitic specimen. It is notable that IF background in the HPT specimens is much higher
than that is in the pearlitic specimen in a wide range of temperatures. Relatively similar but
stronger effect of increase in the IF background is known for severe deformed Cu [14] at elevat-
ed temperatures.
4 Discussion
The present results are in a good agreement with and extend the previous research [4] which has
shown, that behaviour of Fe-C alloy obtained by severe plastic deformation by high pressure
torsion follows the path typical for conventional martensite. A sequence of carbides precipitati-
on at the temperatures corresponding to the low and medium tempering (373–523 K and
523–673 K respectively) was observed [4]. However unlike a conventional martensite, in the
case of HPT-processed steel precipitation of the H-carbide due to low temperature tempering
was retarded indicating the distribution of capacious part of carbon atoms at the dislocations
and non-equilibrium grain boundaries [15]. The high values of hardness (Fig. 6a) and etching
resistance persisting up to annealing at 623 K for the periphery areas of the specimen as well as
overall appearance of the thermomagnetic curve (Fig. 5a) observed in the present experiment
clearly show retardation of H-carbide precipitation in that part of the specimen. It is difficult to
make a similar conclusion for the inner part of the specimen. Though the degree of cementite
dissolution in that area was roughly estimated to be 50 %, specimen was darkened by etching in
Nital and displayed cementite lamellae (Fig. 2, Fig. 4b). The monotonous character of decrea-
sing of the MS curve in temperature range of 373–723 K for the inner part (Fig. 5b) caused by
the small amount of remaining cementite and precipitating carbides and possible overlapping of
several small drops situated close to each other makes impossible to conclude about the tempe-
ratures of precipitation. However the stability of microhardness in the temperature range of
373–623 K shows that no notable precipitation process occurs in that temperature range. There-
171
fore we can suppose that there are no principal difference in the annealing behaviour of the ma-
terial of the inner and periphery parts of the specimen. As such there is no difference in
distribution of carbon atoms in ferrite after cementite dissolution.
Our first IF results can be discussed at this stage mainly on a qualitative level because of
some ”technical” problems still not solved in this research. First, this is a problem of structural
inhomogeneity along the HTP specimen. Second, some scattering from one HPT specimen to
another one. Another disadvantage of this research – the absence of IF tests at different frequen-
cies.
With increase in temperature the damping background in the HPT specimen becomes higher
as compared with the pearlitic specimen both in as-received state and even after preheating to
828K. Higher damping capacity of the HPT state at room and elevated temperatures is a prom-
ising for different application if it is considered in combination with higher yield stress of the
HPT material. Engineering materials for damping applications are often characterised by so-
called Sugimoto parameter [16]:
D= (SQ01–1)V0.2 (2)
where Q01–1 is internal friction measured at a stress level 0.1 V0.2. Obviously, a simultaneous
increase in Q01–1 and V0.2 both increases the Sugimoto parameter D.
In most cases the damping increases with temperature as: (see A.Rivière in [6]):
Qb–1 = Aexp(–U / kT) (3)
where U is an activation energy. However, it is not easy to associate the value U with a specific
dislocation mechanism. In case of severe deformed Cu and Ni a specific dislocation behaviour
in the temperature range below recrystallisation temperature was discussed in [18] and a rather
general dependence for A in eq.(3) is proposed: Qb–1 ~ U for a fixed temperature, where U is
the dislocation density. In the range of recrystallisation Qb–1 ~ Ul n, where l is the dislocation
loop length (R. Schaller in [6]). If the dislocation density and the dislocation loop length are not
changed with temperature, the damping increases with temperature, and a higher damping back-
ground in the HPT specimen is a result of a higher dislocation density. If recovery or recrystal-
lisation takes place, the dislocation substructure is changed, and the damping decreases. It is
notable, that even after annealing to 830 K the background damping in the HPT specimen is at
least twice higher than that is in the pearlitic specimen.
Apart of the Snoek peak, the IF peaks presented in Fig. 8 are nor yet explained. From the
temperature position of the relaxation peak and corresponding frequency one can roughly esti-
mate the activation energy of related process by so-called Marx-Wert equation [17]:
H = RTm ln(kBTm / hfm) + Tm'S (4)
where R and h are universal gas and Planck constants, respectively, and the entropy of activati-
on 'S = 1.1 10–4 eV/K. If the peaks presented in Fig. 8 are relaxation peaks, then the activation
energy of the peak P1 in pearlitic specimen (this peak is accompanied by typical for relaxation
peak decrease in the resonance frequency) is about HP1 |0.62 eV. Activation energy of the
small peak P1 in HPT-processed specimen is HP1 |0.58 eV, from which one can suppose a si-
milar origin of these peaks. In general a HPT process decreases the P1 peak height, but some
172
Figure 8: Temperature dependent internal friction and resonance frequency (G ~ f2) in Fe-0.8 wt.%C specimen
scattering from this tendency has been recorded from one specimen to another. The activation
energy of the P1 peak is too small to be the carbon Snoek relaxation.
The carbon Snoek peak in D-Fe with the activation energy about 0.85 eV is observed at
314 K using 1 Hz. The activation energy of the S peak (T = 387 K, f = 541 Hz) in the Fe-
0.01 wt.%C specimen used in this research and estimated by eq.(3) is 0.83 eV. For resonance
frequency f | 900 Hz used for the HPT specimen measurements, the carbon Snoek peak should
be observed at about 670 K. Indeed some asymmetry of the P1 peak (from the high temperature
side) in pearlitic specimen and weak deviation of IF from exponential increase with temperature
in HPT specimen can be a result of very low Snoek peak contribution to the TDIF curve.
Since the height of this effect is very low, it can be concluded that no pronounced sign of the
Snoek peak is observed in the HPT state. This is an indirect evidence that carbon in the HPT
specimen is situated not in the solid solution but at dislocations or grain boundaries.
Contrary to the P1 peak, which decreases with deformation, an internal friction at elevated
temperatures (including the P2-like peak) increases due to HPT both in absolute value of IF and
in relative height of the P2 peak.
We suppose that the P2 peak is not a relaxation peak, but is a result of a superposition of two
opposite tendencies: increase in damping with temperature in such a non-equilibrium structure
overlaps with the effect of annealing, i.e. recovery, polygonisation and may be grains growth,
which decrease dislocation density and correspondingly damping above a certain temperature
(which should be below recrystallisation temperature). These two opposite tendencies produce
IF ”peak” at a temperature corresponding to the beginning of recovery processes. In case of the
HPT carbon steel the P2 peak is observed at a temperature of |700 K, which corresponds to the
beginning of hardness decrease (Fig. 6). Important argument in favour of this second approach
is a non-equilibrium microstructure of HPT specimens with a grain size as small as 10 nm (at
the edges of the IF specimen) and high dislocation density of about 1011–1012 cm–2 arranged in a
173
cell walls and non-equilibrium grain boundaries [15]. Similar, but smaller in absolute values, ef-
fect can be also observed in thermo-mechanically treated pearlitic specimen.
Similar viewpoint is considered for severely deformed Cu and Ni [18]: this interpretation is
based on behaviour of non-equilibrium grain boundaries and grain boundary dislocations. It is
notable that the peaks discussed in [18] in severe deformed Cu and Ni are observed at the tem-
peratures corresponding (0.3–0.34) Tm (Tm is the melting point), which is exactly the case for
our HPT specimen: 0.32 Tm.
For better understanding inelastic properties of Fe-C HPT specimens and in particular to find
out the origin of the IF peak around 670 K we plan to carry out additional experiments (TDIF
tests with different resonance frequency, isothermal experiments at different temperatures,
study of amplitude dependent IF).
5 Conclusion
The tempering behaviour of the Fe-C alloy produced by severe plastic deformation is studied.
The starting material is obtained in two structural states – nanocrystalline with total dissolution
of Fe3C at the periphery of the specimen and with a cell structure in the inner area, where the de-
gree of cementite dissolution was 50 %. It has been shown that that there are no principal diffe-
rence in the annealing behaviour of the material of the inner and periphery parts of the
specimen. Precipitation of epsilon carbide at the temperatures, corresponding to the low tempe-
ring of martensite is retarded. Such behaviour is related with the arrangement of carbon atoms
mainly on the nonequilibrium grain boundaries and in segregations on the dislocations forming
cell walls. The onset of carbides precipitation at the temperature of 623 K coincides with the on-
set of microstructure relaxation, namely recovery of non-equilibrium grain boundaries that me-
ans rearrangement and annihilation of dislocations.
Internal friction experiments have not shown the Snoek peak in the HPT processed speci-
men. This is consistent with the estimations of the lattice parameter in the HPT specimen, which
have not revealed any changes comparing with non-deformed state. The IF peak at |580 K (the
P2 peak) is tentatively explained by superposition of two opposite tendencies: increase in IF
background with temperature and decrease in dislocation density, and correspondingly in damp-
ing, above |0.3 Tm due to various recovery processes in the HPT specimen.
6 Acknowledgement
High Pressure Torsion processing of the steel specimen was carried out in the Institute of Phy-
sics of Perspective Materials (IPPM) USATU, Ufa, Russia. Authors are grateful to Professor
R.Z. Valiev from IPPM for very useful discussions, Dr. F. Banhart (Ulm University) for perfor-
ming the TEM investigations and Dr. Weissmueller (Forschungszentrum Karlsruhe) for the
help with the estimation of the lattice parameter.
174
7 References
[1] A.V. Korznikov, Yu.V. Ivanisenko, D.V. Laptionok, I. M. Safarov, V.P Pilyugin, R. Z.
Valiev, NanoSrtuctured Materials, 1994, 4, 159–167
[2] K. Hono, M. Omuma, M. Murayama, S. Nishida, A. Yoshie, T. Takahashi, Scripta Mater,
2001, 44, 977–983
[3] Yu. Ivanisenko, W. Lojkowski, R. Z. Valiev, H.-J. Fecht, Acta Mater, 2003, 51,
5555–5570
[4] Yu. Ivanisenko, R. Wunderlich, R. Z. Valiev, H.-J. Fecht, Scripta Mater, 2003, 49,
947–952
[5] A. S. Nowick, B. S. Berry, Inelastic Relaxation in Crystalline Solids, Academic Press:
New York, 1972
[6] R. Schaller, G. Fantozzi, G. Gremaud (Eds), Mechanical Spectroscopy Q–1 2001 with
Applications to Materials Science. Trans. Tech. Publication LTD, 2001
[7] J. L. Snoek, Physica VIII, 1941, n. 7, p. 711–733
[8] Klug & Alexander 'X-ray Diffraction Procedures', pp. 661–665
[9] H.-R. Sinning, J. Phys.: Condensed Matter 1991, 3, 2005
[10] G. Petzow, Aetzen. Stuttgart: Gebrueder Borntraeger Berlin, 1994
[11] Steel. A Handbook for Materials Research and Engineering vol. 1. (Ed. by Verein Deut-
cher Eisenhüttenleute), Berlin, Springer-Verlag, 1992
[12] Grandle J, Sucksmith W. J. Iron Steel Inst. 1941;168:142
[13] Nakagura S, Suzuki T, Kusinoki M. Trans. of JIM 1981;10:699
[14] R. R. Mulyukov, A. I. Pshenichnyuk, J. of Alloys and Comp. 2003, 355, 26–30
[15] R. Z. Valiev, V. Yu. Gertsman, O. A. Kaibyshev, Phys Stat Sol (a) 1986, 61, 95–102
[16] K. Sugimito, N. Kinzoku Gakkai Kaiho, 1975, 14 (7), 491
[17] C. Wert, I. Marx, Acta Metall. 1953, 1, 113–122
[18] M. U. Griaznov, A. N. Sisoev, V. N. Chuvildeev. Material Science (in Russian), 1999, 6,
31-39; 1999, 7, 8–15; 1999, 8, 7–12
175
Investigation of Ferrite Softening Processes in a 0.2 wt% C

1.5 wt% Mn Steel after Hot Deformation
A. Smith1,3, D. N. Hanlon2, H. Luo1,3, J. Sietsma3 and S. van der Zwaag4

1
Netherlands Institute for Metals Research, 2628 AL Delft, the Netherlands,
2
IJmuiden Technology Centre, Corus Research, Development and Technology, P.O. Box 10.000, 1970 CA
IJmuiden, The Netherlands.
3
Department of Materials Science & Engineering, Delft University of Technology, 2628 AL Delft, the Nether-
lands
4
Faculty of Aerospace Engineering, Delft University of Technology, 2629 HS Delft, the Netherlands
1 Introduction
During industrial rolling of steels control of recovery and recrystallisation is important to opti-
mise the properties of the final product. They are therefore important phenomena to study.
Whilst much research has been carried out on recrystallisation kinetics in austenite e.g.
[1–3], relatively little attention has been given to recovery kinetics e.g. [4].
The present paper investigates the static recovery kinetics in ferrite in a low alloy C-Mn
steel. The approach has been to experimentally determine the kinetics of recovery using the
stress relaxation technique for a variety of temperatures, strains and strain rates. The kinetic
data has been analysed using a model already existing in the literature [5]. From this model the
activation energy for recovery has been determined for a range of temperatures, strains and
strain rates. With reference to the obtained activation energies, the operative mechanisms for re-
covery in ferrite in this steel are discussed.
2 Recovery in Metals
Generally recovery models in the literature can be grouped according to the detail in the dislo-
cation structure that they consider. In the most simple approach, recovery is modelled via the re-
duction of an overall dislocation density e.g. [5]. In a more sophisticated model, a distinction is
made between dislocations within subgrains and those which make up the subgrain boundaries.
Recovery is then related to the dislocation density within subgrain interiors and to the subgrain
size. [6]. It is the intention in this study to model the recovery kinetics as a reduction in an over-
all dislocation density.
The model used is that due to Verdier, Brechet and Guyot [5]. The model assumes that the
internal stress relaxation is due to thermally activated dislocation annihilation and reorganisa-
tion, and therefore to plastic relaxation. The plastic relaxation rate H is assumed to be related to
the internal stress Vi by:
dV i
H E , (1)
dt
where E is Young’s modulus and t is time. By combining equation (1) with the Orowan law:

ISBN: 3-527-31007-X
176
M H Ubv , (2)
where M is the Taylor factor (taking a value of M = 2 for BCC metals [7]), U is the dislocation
density, b is the Burgers vector, v is the average dislocation velocity, and
§ Q · §V V ·
v bvD exp ¨ 0 ¸ sinh ¨ i ¸ ,
© kT ¹ © kT ¹
the authors obtained the following equation:
dV i E U b 2 vD § Q · §V V ·
exp ¨ 0 ¸ sinh ¨ i ¸ , (3)
dt M © kT ¹ © kT ¹
where vD is the Debye frequency, Q0 is the activation energy for recovery, V is the activation vo-
lume, k is Boltzmann`s constant and T is temperature in Kelvin. The dislocation density is then
converted to a stress using:
Vi M D Gb U (4)
where G is the shear modulus and a is a constant of the order of 0.3 [7]. Combining equations
(3) and (4) with G = 3E/8, the following is obtained:
dV i 64V i2 § Q · §V V ·
vD exp ¨ 0 ¸ sinh ¨ i ¸, (5)
dt 9M 3D 2 E © kT ¹ © kT ¹
Hence according to equation (5) the rate of change of internal stress with time depends on tem-
perature, an activation energy and volume, and stress. Q0 and V are the fitting parameters.
3 Experimental
The steel used for the present study had the composition 0.19 wt% C, 0.445 wt% Si, 1.46 wt%
Mn and 0.033 wt% Al.
Cylindrical samples were machined from rolled plate, with diameter 10 mm and length
12 mm. The samples were machined with the length parallel to the rolling direction.
To investigate the static recovery kinetics a Gleeble® 3500 thermo-mechanical simulator was
used. Each test comprised of three stages. Firstly a heat treatment schedule was applied. Sam-
ples were heated via electrical resistance heating (under vacuum) to 1100 °C to austenitise for 3
minutes. Then samples were cooled to 679 °C (in the two phase region) and held for 10 minutes.
At 679 °C for this steel the equilibrium phase distribution is 50 % ferrite and 50 % austenite.
Next, the samples were cooled at a rate of 5 °C/s to the desired test temperature and held for 5
minutes.
In the second stage the samples were deformed in compression. Lubrification was provided
via graphite paste. To protect against the possibility of carbon pick up during the test, tantalum
sheets were used as a protective layer between the sample and the lubricant.
177
Finally in the third stage, the recovery after deformation was monitored. This was achieved
by measuring the stress required to maintain a constant strain in the sample. After 50 minutes,
samples were water quenched at a quenching rate of around 200 °C/s.
Three series of stress relaxation tests were carried out. In the first series the effect of temper-
ature on the recovery kinetics was investigated. Test temperatures used were in the range
150 °C– 650 °C. The strain for each test was 0.15, whilst the strain rate used was 0.1 / s.
In the second series of tests the effect of strain was investigated by carrying out tests with
strains from 0.05–0.25. The strain rate was kept constant at 0.1 s–1, whilst the temperature used
for each test was 550 °C.
In the third series of tests the effect of strain rate was investigated using rates of 0.01/s–0.6/s.
The strain was maintained at 0.15 for each test whilst the temperature used was 550 °C.
In addition, multiple measurements were performed to test the reproducibility.
Finally one relaxation test was carried out with a higher strain of 0.5, using a strain rate of
0.6 / s.
4 Results
Example stress relaxation curves are shown in figure 1, for the series of tests concerning the ef-
fect of temperature.
300
250
Stress (MPa)
200
150 A
100
B
50
C
D
0
0.1 1 10 100 1000 10000
Time (s)
Figure 1: stress relaxation curves for test temperatures of: A) 450 °C, B) 500 °C, C) 550 °C and D) 600 °C. Strain
is 0.15 and strain rate is 0.1/s. Experimental data is indicated by markers and the best fit from equation (5) by
solid lines. The stress plotted is the internal stress due to dislocations i.e.Vi = Vf – Vy, where Vf is the flow stress
and Vy is the yield stress.
Firstly, considering only the experimental curves, we can see that for every temperature the
stress decreases with time. This is due only to recovery, since there is no distinctive three-stage
curve as would be observed if recrystallization had occurred. [8].
The effect of increasing temperature on the recovery kinetics is twofold. Firstly the initial
stress decreases. Secondly the average stress relaxation rate increases.
178
The first effect is due to a larger degree of dynamic recovery during deformation. The second
effect is explained by the model, i.e. the motion of dislocations can be considered to occur by
thermally activated mechanisms.
Figure 1 also shows the curves obtained using equation (5). The best fits to the data were ob-
tained by allowing both Q0 and V to vary for each separate curve. Attempts to fit the model to all
the experimental data by using a single value for Q0 and a temperature dependent value for V
did not give satisfactory results. As can be seen, there is good agreement between experiment
and model. The activation energies and volumes that gave the best fit for each temperature in-
vestigated are presented in figures 2 and 3 respectively:
As can be seen from figure 2, Q0 appears to rise with increasing temperature until 450 °C af-
ter which it remains approximately constant. The activation energy for temperatures between
150 °C and 300 °C, appears to be close to that for dislocation core diffusion (Qc = 174 kJ/mol
[9]) in D-iron. For temperatures between 450 °C and 600 °C, the experimental data is in reason-
300
250 Qs = 251 kJ/mol [9]
200
Q0 (kJ/mol)
Qc = 174 kJ/mol [9]

150
100
50
0
100 200 300 400 500 600
Temperature (°C)
Figure 2: Effect of temperature on activation energy for recovery. Data for strain = 0.15 and strain rate = 0.1/s is
shown by black markers. Data shown by grey markers corresponds to the data from table 1.
6.00
5.00
Activation volume
4.00
(x10 -28 m 3)
3.00
2.00
1.00
0.00
100 200 300 400 500 600
Temperature °C
Figure 3: Effect of temperature on activation volume. Data for strain = 0.15 and strain rate = 0.1/s is shown by
black markers. Data shown by grey markers corresponds to the data from table 1.
179
able agreement with the activation energy for lattice self diffusion (Qs = 251 kJ/mol [9]) in
Į-iron. For the temperature range 300 °C–450 °C the data suggests a transition region between
the two processes.
Thus at the lower temperatures the recovery process is dominated by movement of disloca-
tions controlled by core diffusion. At higher temperatures the motion of dislocations is control-
led by lattice diffusion.
From figure 3 it can be seen that the activation volume appears to be approximately constant
at about 3 10–28 m3 between 150 °C and 375 °C. Above 375 °C it appears to rise to a peak value
at 450 °C before decreasing with increasing temperature. Comparison of figures 2 and 3 sug-
gests that for recovery controlled by dislocation core diffusion the activation volume can be
treated as constant, but for recovery controlled by lattice diffusion the activation volume is a de-
creasing function of temperature.
The effect of strain and strain rate on Q0 and V for a fixed temperature of 550 °C is shown in
table 1. The values are also plotted in figures 2 and 3.
Table 1: Effect of strain and strain rate on recovery activation energy and volume. Temperature
of deformation and relaxation is 550 °C.
Strain Strain rate (/s) Q0 (kJ/mol) V (x10-28 m3)
0.05 0.1 217 1.6
0.15 231 2.6
0.25 228 2.6
0.15 0.01 223 2.4
0.1 231 2.6
0.6 226 2.7
0.5 0.6 236 2.2
As can be seen, for the range of values investigated, strain and strain rate appear to have only
a small effect on the activation energy and volume. For an increase in strain with a constant
strain rate of 0.1/s the activation volume increases slightly. However when the strain increases
at a constant strain rate of 0.6/s the activation volume decreased.
5 Discussion
5.1 Activation Energy for Recovery
The activation energy for recovery in ferrite has been shown to vary with temperature. At low
temperatures (150 °C–300 °C), the activation energy is close to that for dislocation core diffusi-
on whilst between 450 °C and 600 °C it is close to the activation energy for lattice self diffusion.
This result compares well with the work of Michalak and Paxton [4] who studied the recovery
of zone-refined iron during annealing between 300 °C and 500 °C. Their work suggested a
stress independent activation energy (i.e. equivalent to Q0) of around the value for lattice self
diffusion. In this study however, at temperatures of between 300 °C and 450 °C the analysis in-
dicates a contribution from dislocation core diffusion to the observed activation energy.
180
5.2 Activation Volume for Recovery
The activation volumes calculated from the experiments compare well with that reported by
Leslie [10], who found that the activation volume for pure iron deformed at room temperature
was around 3.8 10–28 m3.
From table 1 the activation volume was found to increase slightly when strain was increased
from 0.05 to 0.25. This is in contrast with the results of Verdier et al [5], who found that for the
recovery of an Al - 2.5 wt% Mg alloy, as strain increased the activation volume decreased. The
authors attributed this effect to a controlling annihilation mechanism in cell walls, because the
cell walls tighten with increasing strain.
6 Conclusions
It has been shown that the stress relaxation technique is able to reveal the recovery kinetics of
ferrite deformed over a range of temperatures, strains and strain rates. The recovery model of
Verdier et al [5] has been successfully applied to the experimental stress relaxation data. The ac-
tivation energy and volume for recovery was found to vary with temperature, whilst they varied
slightly with changing strain and strain rate. At low temperatures (150 °C–300 °C) the activati-
on energy was found to be close to that for dislocation core diffusion. At higher temperatures
(450 °C–600 °C) the activation energy agreed with that found for lattice self diffusion.
In the temperature range (150 °C–375 °C) the activation volume was approximately constant
whilst between (450 °C–600 °C) it decreased with increasing temperature.
The activation energies and volumes generally agree with previous studies on recovery of D-
iron.
7 References
[1] P. Uranga, A. I. Fernandez, B. Lopez, J. M. Rodriguez-Ibabe. Materials Science and

Engineering A. 2003, 345, 319–327
[2] N. Fujita, T. Narushima, Y. Iguchi, C. Ouchi. ISIJ. International. 2003, 43, 1063–1072
[3] H. S. Zurob, C. R. Hutchinson, Y. Brechet, and G. Purdy. Acta. Materialia. 2002, 50,
3075–3092
[4] J. T. Michalak and H. W. Paxton. Transactions of the Metallurgical Society of AIME.
1961, Vol. 221, p.850–857
[5] M. Verdier, Y. Brechet and P. Guyot. Acta. Materialia. 1999, 47, 127–134
[6] E. Nes. Acta. Materialia. 1995, 43, 2189–2207
[7] A. Bodin, PhD thesis Delft University of Technology., Grafisch Bedrijf Ponsen and Looi-
jen, The Netherlands. 2002
[8] L. P. Karjalainen. Mater. Sci. Technol. 1995, 11, 557–564
[9] J. Frost and M. F. Ashby, Deformation-Mechanism Maps., Pergamon, Oxford, 1982
[10] W. C. Leslie in Iron and its Dilute Solid Solutions (ed.:C. W. Spencer and F. E. Werner),
Metallurgical Society of AIME, Interscience, John Wiley and Sons 1963, p. 119–212
181
A Preliminary Study on the Grain Refinement in Ultra Low

Carbon Steels by Cyclic Heat Treatment
E. Bayraktar* and J. P. Chevalier**

*
SUPMECA-LISMMA/Paris, School of Mechanical Engineering, Saint Ouen-France
**
CNAM 2, rue conté 75003, Paris, and CECM–CNRS, 15, rue G. Urbain 94407 Vitry cedex, France
1 Summary
The effects of cyclic heat treatment on the refinement of grain size in ultra low carbon steels
(ULC) such as Interstitial Free (IF) and Bake Hardening (BH) steels are discussed. This ap-
proach is somewhat specific to steels since the thermal cycling used aims at producing repeated
D/J and J/D phase transformations. The effects of the different parameters of this special heat
treatment, such as maximum heating temperature, dwell time, cooling rate and the number of
the cycles on the grain size are evaluated in order to reduce grain size and hence increase the
specific strength.
2 Introduction
Ultra fine-grained steels have attracted much interest recently due to their expected high yield
stress associated with a low ductile to brittle transition temperature. This could lead to consi-
derable use of such high specific strength steels in the car industry to reduce fuel consumption
(and decrease greenhouse gas emission) whilst ensuring excellent crash resistance in case of
collision [1].
ULC steels such as Interstitial Free (IF) and Bake Hardening (BH) steels are largely em-
ployed in the outer parts of car bodies since complex deep drawn components can be made. In
these grades, carbon and nitrogen are kept at extremely low values by precipitation of carbides
and carbonitrides of added elements (Ti, Nb...). Due to the resulting texture, these steels have
very high formability (r t2) during press forming (IF, BH, steels) and display an increase in
strength after bake hardening resulting from paint baking (BH steel). This gives a better dent re-
sistance for exposed automotive panels, currently the main application of these steels [1, 2]. The
quality of these steels can be improved by refining the grain size and this is often quantified in a
constitutive equation that has the Hall-Petch form, i.e. it relating the yield strength to the grain
size [3, 4]:
Vy = Vi + kyd–1/2. (1)
where Vi is the friction stress and ky is the locking parameter

The main method reported in the literature for grain refinement is classic thermomechanical
processing. However, thermal cycling may provide an alternative method of grain refinement.
For example, grain size in copper-silver alloys can be decreased by a discontinuous precipita-
tion mechanism [5]. A similar mechanism has been reported for cobalt-base HS-21 alloy and for

ISBN: 3-527-31007-X
182
high-nitrogen austenitic heat resistant steels (21Cr 4Ni 9Mn) [6]. Using the phase transforma-
tions for grain refinement, successful results were obtained in titanium-gettered steels and in
lath martensitic steels [7, 8]. Some studies were carried out also on Maraging and Hadfield
steels [9, 10]. Furthermore, grain refinement by only cyclic heat treatment (CHT) has been ob-
tained in titanium aluminium based alloys [11–13].
More recently, other thermomechanical methods such as equal channel extrusion at a given
temperature, involving severe plastic deformation have been used to obtain ultra fine grain ma-
terials (14–15).
Here, we examine an alternative method for grain refinement involving only heat treatment.
It takes advantage of the refining phenomenon occurring during phase transformation and can
lead to size reduction by up to roughly two orders of magnitude. This study presents our recent
results obtained on the ULC steels and discusses the underlying mechanism in order to answer
to the following questions:
• Which materials parameters are relevant to obtain the grain refinement, and what are the
practical means in order to reduce the grain size during the CHT processing?
• What is the influence of the thermal cycle parameters (heating and cooling rates, dwell
time, etc..) in CHT in grain refinement?
3 Experimental Conditions
Different grades of IF and BH steel sheets provided by SOLLAC have been used in this study
(Table 1). The carbon and manganese values vary from 1.4 10–3 wt% to 5 10–3 wt% and from
100 10–3 wt% to 200 10–3 wt% respectively. The phase transformations, under thermal cyclic
conditions, have been studied on sheet specimens of thickness varying from 0.67 to 2 mm with
dimensions of 12 mm × 2 mm by means of the tests conducted on a quench type dilatometer
using a radiation oven (Adamel-DT 1000) computer controlled. This technique allows the accu-
rate determination of the transformation temperatures and the monitoring of the structural evo-
lution.
The control of the thermal cycle is obtained using a thermocouple (diameter of 0.1 mm)
welded at the centre of the specimen. Quenching is produced by a blast of helium gas.
Three different heat treatment routes were designed. In the first CHT route, eight cycles were
used with a cooling rate of 300 °C/s and a short dwell time of 1s. The temperature range for this
route was between 910 °C and 400 °C. In the second CHT route, six cycles were used with a
cooling rate of 300 °C/s and a short dwell time of 1s except for the dwell time for the first heat-
ing cycle, which is longer-see results section. The temperature range for the first and sixth cy-
cling route was between 905 °C and 400 °C and intermediate ones were those corresponding to
cycling between the Ac3 temperature and 400 °C. In the third CHT route, eight cycles were ap-
plied with a cooling rate of 100 °C/s and a dwell time ranging from 5 to 10 s. The temperature
range for cycling was between 910 °C and 400 °C. For all the three CHT routes, the heating rate
used was 100 °C/s. The values of Ac3 temperature measured for BH and IF steels are 870 °C and
885 °C respectively. Initial grain sizes for IF and BH grades were measured between 13 and 14
µm with an aspect ratio (dDL / dDT) ranging from 1 to 1.25.
183
Grain size was measured by optical microscopy by means of ”Areas” image analysis. The
same metallographic specimens are used for the micro hardness measurements under a load of
200 g.
Table 1: Tensile test results of different grades of ULC steels used in this study
Grade Vy (MPa) Longitudinal A (%) Vy (MPa) Transversal A (%)
Vm (MPa) Vm (MPa)
IF Ti (0.8 mm) 175 339 41 186 346 41
IFTiNb (0.8mm) 261 352 34 253 348 32
IF TiB (0.67 mm) 127 316 43 133 316 41.5
BH (2mm) 220 340 32 225 346 30
4 Results and Discussions
Figure 1 shows the first thermal cycling route leading to grain refinement of the steels. This was
applied on the different grades of IF and BH steels (Figure 2). These microstructures appear to
have nucleated mainly at the J grain boundaries but also in J grain interiors and then propagated
into the grains through massive transformation depending on the different mobility of different
J/Dinterfaces. The refining process is schematically shown in Figure 3. Image analysis results
obtained on these grades showed that over the 80 % of the grain population were found with a
grain size under the 6–8 µm. However, the structure obtained was not homogeneous and areas
with substantially large grains are observed. CHT with 8 cycles resulted in refinement of the
structure. Further cycling in the same conditions did not lead to further grain refinement.
Figure 1: First thermal cycling procedure for grain refinement applied to the IF steel
184
100µm
a) b)
Figure 2: Optical micrographs of a grade of IF steel: a) initial, b) 8 cycles of CHT procedure
γ
α
Figure 3: Schematic illustration of the refining process occurring in multiple cycle of CHT
The underlying principle used is simple: for these phase transformation, the number of active
crystallographic variants produced for each phase transformation is greater than one, and hence
a single parent grain should yield several transformed grains, finally leading to grain refinement
of the structure. However, the number of active crystallographic variants cannot be predicted
for the ferrite-austenite transformations, since the nature of the transformation is more compli-
cated than for example an order-disorder transformation, where all variants are equal-probable.
So, the (J o D and D o J) phase transformations introduce active crystallographic variants
leading to an increase in the number of nuclei. In practice, for thermal cycling routes to lead to
grain refinement, it is essential that nucleation should always be favoured in detriment to grain
growth, especially in the Jphase at high temperatures. In other words, there is a crucial balance
between the nucleation and the grain growth in the ULC steels, which have very high mobility
of the grain boundaries.
The second thermal cycling route (Figure 4) was applied on the same steel grades. The same
refining mechanism applies as explained in the previous example (Figure 5). However, relative-
ly long dwell time (5–10 s) was used during the first heating in the thermal cycling. This allows
essentially partial dissolution of precipitates produced during the previous treatment in BH
steels. Further cycling after six times of CHT under the same conditions introduces no variation
in the refining process.
185
1000
dwell time = 5s (only for the 1st cycle)
900
Temperature (°C)
800
700 dwell time=1s (for the 2-6th cycles)
600
500
400
300
Θmax=907°C, 6 CHT
200 Vh=100°C/S, Vc=300°C/s
100
0
0 20 40 60 80 100 120 140 160
Time (s)
Figure 4: Second thermal cycling procedure applied to the BH steel (here the data is undersampled)
100µm
a) b)
Figure 5: Optical micrographs of a grade of BH steel: a) initial, b) 6 cycles of CHT procedure
Image analysis results showed that the frequency of the fine grain population was over the
70 % with a grain size about 7–8 µm. The microstructure of BH steel obtained after the second
CHT route was more homogeneous than the IFS but regions with large grains still occurred.
In order to show the benefit of this CHT process, the yield strength can be estimated from Vik-
kers hardness measurements obtained after cycling treatment (16) according to the relation
Vy = (HVN / 3) (0.1) n'–2 (2)
186
where HVN is the Vickers’s hardness number and n' = n + 2 is the exponent in Meyer’s law
(n = strain hardening exponent).
Some of the results are shown in Table 2, and compared to values calculated using the Hall-
Petch relation. Good agreement is obtained.
Table 2: Measured and estimated values obtained on the different grades of ULC steels used in
this study
Grade As received n value Initial HVN Final HVN Estimated Estimated
Vy (MPa) measured after CHT Vy (MPa) by Vy (MPa) by Hard-
Hall-Petch ness values
IF Ti (0.8 mm) 175 0.19 90 130 300 285
IFTiNb (0.8mm) 261 0.185 125 160 350 350
IF TiB (0.67 mm) 127 0.20 110 145 350 300
BH (2mm) 220 0.20 140 170 350 360
In the third thermal cycling route, the cycling conditions are shown in the Figure 6. This
route was applied to both IF and BH steels. Here, the dwell time was increased, together with
lower cooling rate. It is known that the ULC steels are very sensitive to cyclic heat treatment,
since excessive grain growth can occur due to very low alloying content leading to mobile grain
boundaries. A large grain structure with a mean grain size about 40–50 µm was found after this
CHT route (Figure 7), as could be expected.
Figure 6: Third thermal cycling procedure for grain refinement applied to the BH steel
The effect of the austenitizing temperature (maximum) on the phase transformation was also
studied. The temperature range for this route was between 950 °C and 400 °C. By varying the
austenitizing temperature, the JoD transformation temperature may be changed. This is usual-
ly explained by an increase in austenite grain size with increasing austenitizing temperature,
causing a reduction in the grain boundary nucleation rate.
187
100µm
Figure 7: Optical micrographs of a grade of IF steel: 8 cycles of CHT procedure
However, here the experimental results show that the IF steels give an approximately con-
stant transformation temperature during the cycling heat treatment.
Cycling was also carried out on boron containing IF steels. Usually B addition is considered
to retard the transformation to polygonal ferrite in the ultra low carbon steels, however no ef-
fects were found here.
Since fast grain growth is a typical characteristic of J solution treated ULC steels, grain re-
finement by CHT may only be obtained by strictly controlling the cycling parameters (number
of cycles, dwell time in the J phase, maximum temperature, heating and cooling rates). The re-
sults show that dwell time and maximum temperature are indeed crucial for these steels. Fur-
thermore the cooling rate should be fast enough to prevent fast grain growth. The effects of
these parameters are given in the Table 3.
Table 3: Results of CHT routes of the ULC steels used in this study
Grade Initial N° of CHT Cooling Dwell Final Comment
grain size rate (°C/s) time (s) grain size
(µm) (µm)
IF Ti 14 8 300 1 8 Inhomogeneous microst-
(0.8 mm) ructure with refined grains
IFTiNb 13 8 300 1 6–8 Inhomogeneous microst-
(0.8mm) ructure with refined grains
IF TiB 13 8 300 1 8–9 Inhomogeneous microst-
(0.67 mm) ructure with refined grains
BH (2mm) 15 6 300 1 7–8 Inhomogeneous microst-
ructure with refined grains
IFTiNb 13 8 300 1 6–8 Inhomogeneous microst-
(0.8mm) ructure with refined grains
IF Ti 14 8 100 5–10 30–40 Substantial grain growth
(0.8 mm)
188
5 Conclusions
This study shows that only cyclic heat treatment without mechanical processing can achieve
grain refinement in low carbon steels. Results on ULC steels show that fine microstructures
with a grain size ranging from 6 to 8 µm can be obtained.
Three different heat treatment routes were designed in order to study the effect of the differ-
ent parameters on the refining pathway. It is necessary to strictly control the dwell time, the
maximum temperature and the cooling rate in order to prevent very fast grain growth in the
ULC steels, which have very high mobility of the grain boundaries.
For these steels the balance between nucleation and the grain growth is particularly crucial.
6 Acknowledgements
This study was supported by Reseau National Matériaux et Procédé under the project ”Aciers à
Grains Ultra Fins”, AGUF.
7 References
[1] A. Jouet, Rev. Metall./Cah. Inf. Tech., 1997, 11, 1425–1434

[2] W. B. Hutchinson, K. I. Nilsson and J. Hirsch, (Ed. R. Pradhan), TMS, 1989, 109–125
[3] R. A. Grange, Trans of ASM, 1966, 59, 27–48
[4] J. W. Morris, Jr., Proc. Int. Symp. Japan, (Ed. S. Takaki et. al.), ISIJ, 2001, 34–41
[5] W. Scharfenberger, A. Schutrumpf and H. Borchers, Z. Metallkd., 1971, 62, 540–552
[6] M. Tanaka, Z. Metallkd., 1994, 85, 6, 446–452
[7] Z. Guo, K. Sato, T-K. Lee and J. W. Morris, Jr, TMS, 2000, (ed. RS Mishra et. al.), 1–12
[8] S. Jin, S. K. Hwang and J. W. Morris, Jr., Met. Trans., 1975, 6A , 1721–1730
[9] K. Nakazawa, Y. Kawabe and S. Muneki, Mater. Sci. Eng., A, 1978, 33,1, 49–56
[10] S. W. Mahajan, G. Venkataraman and A. K. Mallik, Metallography, 1973, 6,1, 337–345
[11] J. Yang, J. Wang, Q. Xia and Y. Wang, Mater. Lett., 2000, 46, 193–197
[12] J. N. Wang, J. Yang, Q. Xia and Y. Wang, Mater. Sci. Eng., A, 2002, 329–331, 118–123
[13] J. Yang, J. Wang, Y. Wang and Q. Xia, Intermet., 2003, 11, 971–974
[14] Y. Fukada, K. Oh-Ishi, Z. Horita and T. G. Langdon, Acta mater., 2002, 50, 1359–1368
[15] M. Furukawa, Z. Horita, M. Memoto, R. Z. Valiev and T. G. Langdon, Phil. Mag., A,
1998, 78, 1, 203–215
[16] G. Dieter, Mc-Graw Hill, 3nd ed., 1987, p. 330
189
Authors
Bayraktar, E. 181 Liu, Q. K. K. 69

Blanpain, B. 87 Luo, H. 175
Bley, F. 3 Lyon, O. 34
Boucharat, N. 113
Brechet, Y. 42 Maugis, P. 3
Bruna, P. 126
Burkel, E. 145 Neumann, W. 69
Nicolas, M. 61
Chen, W. 69 Nishijima, G. 103
Chevalier, J. P. 181
Chiu, Y. -T. 122 Pardoen, T. 87
Crespo, D. 126 Perrard, F. 3
Pichler, A. 79
De Cooman, B. C. 95 Pineda, E. 126
Deschamps, A. 3, 61 Rösner, H. 113
Donnadieu, P. 3 Röthler, B. 79
Duprez, L. 95
Schumacher, G. 69
Esling, C. 103 Segers, D. 95
Servant, C. 34
Fecht, H. -J. 165 Sietsma, J. 175
Fischer, F. D. 26 Sinning, H. -R. 135, 145, 165
Smith, A. 12, 175
Gamsjäger, E. 26 Svoboda, J. 26
Glicksman , M. E. 155
Golovin, I. S. 135, 145, 165 Traint, S. 79
González-Cinca, R. 126
Guillon, I. 34 Van Caenegem, N. 95
Van der Zwaag, S. 53, 175
Hanlon, D. N. 175 Van Rompaey, T. 87
He, C. 103 Vermolen, F. J. 53
He, J. 103 Vodárek, V. 12
Hutchinson, C. R. 42 Vuik, K. 53
Ivanisenko, Y. 165 Wanderka, N. 69

Watanabe, K. 103
Jacques, P. 87 Wilde, G. 113
Jianu, A. 135, 145 Wollants, P. 87
Kirchheim, R. 19
Yang, G. 19
Lani, F. 87 Yeh, J. -T. 122
Liu, F. 19

ISBN: 3-527-31007-X
190
Zhang, T. 103
Zhang, Y. 103
Zhao, X. 103
Zuo, L. 103
191
Subject Index*
12CrMoVNb steel 12 Energy

– stacking 95
Alloy – study 69
– Al 113 Evolution, precipitation 61
– Al Zn Mg 61 Extremal principle, thermodynamic 26
– amorphous 113, 135, 145
– Co-Cu-Ni 34 Fe-C alloy 165
– Fe-C 165 FeMnSiCrNi SMAs 95
– multi-component 53 Ferrite softening 175
– shape memory 95 Finite element analysis 3D 69
Analysis, 3D 69 Formation, quasicrystals 135
Annealing, rapid 103 Full annealing 103
Boundary migration 42 Grain growth 19

Grain refinement 181
Carbide precipitation 42 Heat treatment
Carbon 79 – cyclic 181
– shape memory effect 95 – non-isothermal 61
– steel 175, 181 Hot deformation 175
Carbonitride 3 Impingement, soft 126
Coarsening 12, 69 Improvement, selective 79
Co-Cu-Ni alloy 34 Initial melt undercooling 19
Coupling, interphase 42 Interphase boundary migration 42
Creep rupture 12 Kinetics, crystallization 113, 126
Critical size, temperature dependent 122
Crystallization Low carbon steel 181
– primary 126
– kinetics 113 M23C6 12
Crystals, silicon 122 Magnetic field, high 103
Cyclic heat treatment 181 Martensitic transformation 87
Deformation Mechanical spectroscopy 135
– hot 175 Melt undercooling 19
– severe plastic 165 Metallic glasses 113
Diffusional transformation 26 Microalloyed steel 3, 42
Directional coarsening 69 Micromechanical study 87
Dissolution, particle 53 Microstructural development 26
Migration, boundary 42
Elastic strain energy 69 Miscibility gap 34
Embedded silicon nano crystals 122 Molecular dynamis, nanocrystals 122
*
The page numbers refer to the first page of the respecting article

ISBN: 3-527-31007-X
192
Multi-component alloy 53 Steel

Multiphase steel 87 – 12CrMoVNb 12
– carbon 181
Nanocrystals 122 – C-Mn 175
Nb-containing steel 42 – microalloyed 3
NbC Precipitation 3 – multiphase 87
Network structures 155 – Nb-containing 42
Ni-based superalloy 69 – selective improvement 79
Nickel, effect 12 Stefan condition 53
Nitrogen, shape memory effect 95 Strain energy, elastic 69
Non-isothermal heat treatment 61 Structures , three-dimensional 155
Nucleation, non-random 126 Superalloy 19
Superalloy, single crystal 69
Particle dissolution 53
Phase selection 113 Tempering peculiarities 165
Phase transformation 34, 145 Thermodynamic extremal principle 26
Plastic deformation 87 Three-Dimensional structures 155
Precipitation 42, 69 Ti-based alloy 135, 145
– evolution 61 Transformation
– NbC 3 – diffusional 26
Primary crystallization 126 – martensitic 87
– phase 34
Quasicrystal formation 135 – solid-solid 145
Transformation induced plasticity 79
Rapid full annealing 103 TRIP 79, 87
Recrystallization, solid-state 19
Refinement, grain 181 Ultra low carbon steel 181
Undercooling, melt 19
Selective improvement, steel 79
Severe plastic deformation 165 Zr-based alloy 135
Shape memory effect 95
Silicon
– content 79
колхоз 1:11 pm, 9/30/05
– nanocrystals 122
Single crystal superalloy 69
Size, critical 122
Small angle neutron scattering 3
Small angle x-ray scattering 61
Softening processes 175
Soft-impingement 126
Solidified superalloy 19
Solid-solid phase transformation 145
Solid-State Recrystallization 19
Spectroscopy, mechanical 135
Stacking fault energy 95

Alain Hazotte - Solid State Transformation and Heat Treatment-Wiley-VCH (2005)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Alain Hazotte - Solid State Transformation and Heat Treatment-Wiley-VCH (2005)

Uploaded by

Copyright:

Available Formats

I

Solid State Transformation

Solid State Transformation and Heat Treatment. Edited by Alain Hazotte

Société Française de Métallurgie et de Matériaux (SF2M)

Solid State Transformation, Heat Treatment (S3)

Solid State Transformation

Library of Congress Card No.: applied for

British Library Cataloguing-in-Publication Data:

Bibliografic information published by Die Deutsche Bibliothek

Printed on acid-free paper

Printed in the Federal Republic of Germany

Composition: W.G.V. Verlagsdienstleistungen GmbH, Weinheim

© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Metz, 13 October 2004

I Precipitation and Diffusion Processes ................................................................................... 1

Characterization and Modeling of NbC Precipitation in Model Microalloyed Steels................. 3

The Effect of Nickel on M23C6 Coarsening Rates in 12CrMoVNb Steels ............................... 12

Effect of Initial Melt Undercooling on Solid-State Recrystallization and Grain Growth

Modelling the Diffusional Transformation by Application of the Thermodynamic

Phase Transformations in a Co-Cu-Ni Alloy............................................................................. 34

The Coupling of Interphase Boundary Migration and Precipitation: Example of a

Advanced Models for Particle Dissolution in Multi-Component Alloys .................................. 53

Modeling of the Precipitation Evolution During Non-isothermal Heat Treatments

Elastic Strain Energy Study of Directional Coarsening of J' Precipitates in

II Martensitic and Stress Sensitive Transformations ........................................................... 77

Possibilities for the Selective Improvement of the Mechanical Properties of TRIP

Micromechanical Study of the Martensitic Transformation in TRIP-Assisted

Rapid Full Annealing under High Magnetic Field .................................................................. 103

III Crystallization................................................................................................................... 111

Crystallization Kinetics and Phase Selection in Amorphous Al-Alloys.................................. 113

Temperature Dependent Critical Size of Embedded Silicon Nano-crystals: A

Effects of Soft-Impingement and Non-random Nucleation on the Kinetics and

Mechanical Spectroscopy of Quasicrystal Formation from Amorphous Ti- and

Solid-Solid Phase Transformation of Amorphous Titanium Based Alloys............................. 145

IV Recrystallization and Grain Size Control ...................................................................... 153

Energetics of Three-Dimensional Network Structures .......................................................... 155

Tempering Peculiarities of Fe-C Alloy by Severe Plastic Deformation.................................. 165

Investigation of Ferrite Softening Processes in a 0.2 wt% C 1.5 wt% Mn Steel

A Preliminary Study on the Grain Refinement in Ultra Low Carbon Steels by

Author Index ............................................................................................................................ 189

I Precipitation and Diffusion Processes

Solid State Transformation and Heat Treatment. Edited by Alain Hazotte

Characterization and Modeling of NbC Precipitation in Model

F. Perrard*, A. Deschamps*, P. Maugis**, P. Donnadieu*, F. Bley*

Precipitation of carbides or carbonitrides containing Nb, Ti or V is extensively used to improve

2 Materials and Experimental Technique

Table 1: Chemical compositions of the steels (weight ppm)

Solid State Transformation and Heat Treatment. Edited by Alain Hazotte

2.2 Small-Angle Neutron Scattering

3.1 Influence of Temperature

3.2 Influence of Composition

4 Modeling of Precipitation Kinetics

4.1 Description of the Model

• Spherical and stoechiometric particles, embedded into a homogeneous matrix.

4.1.1 Nucleation and Growth

The nucleation rate is given by equation (3),

dR DNb X ssNb  X iNb

dR DNb X ssNb  X iNb 1 dN

4.1.2 Growth and Coarsening

­d R DNb X ssNb  X iNb

4.1.3 Transition Criterion

The transition between nucleation-growth and growth-coarsening stages is made continuous

4.2 Comparison between Model and Experiment

We can summarize all these results by Temperature-Time-Transformation (TTT) curves for

F. Perrard, A. Deschamps, P. Maugis**, P. Donnadieu, F. Bley

dR DNb X ssNb X iNb

dR DNb X ssNb X iNb 1 dN

d R DNb X ssNb X iNb

J J 0 G 'H seg RT ln X 0 (15)

'X § C2J / D C0 · § C1J / D C1D / J ·

°ci (r ,0) ci 0 for i {1,...., n}